Analog and Microprocessor System Design

The Weather Station Project

Peter D. Hiscocks
Department of Electrical and Computer Engineering
Ryerson University
Toronto, Canada

Email: phiscock@ee.ryerson.ca
URL: http://www.ee.ryerson.ca/˜phiscock
http://www.syscompdesign.com
Contents

Preface xi

1 Temperature Measurement 1
1.1 General Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 The Thermistor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Thermistor Electrical Specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.2 Thermistor Measuring Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.3 Lookup Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.4 Thermistor Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.5 Steinhart-Hart Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2.6 Thermistor Circuit Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.2.7 Resolution of the Thermistor Thermometer . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Engineering Tool: BASIC Program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.1 Estimating the Production Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.3.2 Thermistor Self-Heating and Thermal Runaway . . . . . . . . . . . . . . . . . . . . . . . 15
1.3.3 Plotting the Thermistor VI Characteristic . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.3.4 Engineering Tool: Generating Graphs with GNUPLOT . . . . . . . . . . . . . . . . . . . 21
1.4 The Diode Temperature Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.4.1 Diode Temperature Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.4.2 Transistor Base-Emitter Voltage Function . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.4.3 Base-Emitter Voltage Function of Temperature . . . . . . . . . . . . . . . . . . . . . . . 26
1.4.4 Base-Emitter Voltage Temperature Coefficient . . . . . . . . . . . . . . . . . . . . . . . 27
1.4.5 Interface Circuit for a Transistor Temperature Sensor . . . . . . . . . . . . . . . . . . . . 30
1.4.6 The Operational Amplifier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.4.7 The effect of Power Supply Ripple and Drift . . . . . . . . . . . . . . . . . . . . . . . . 38
1.4.8 The Effect of Bias Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.4.9 The Effect of Offset Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
1.4.10 The Effect of Offset Voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
1.4.11 Drift in Resistor Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
1.4.12 Estimate of Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
1.5 Appendix: Operational Amplifier, Static Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
1.5.1 Input Offset Voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
1.5.2 Input Bias Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
1.5.3 Input Offset Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

i
 ii

1.5.4 Power Supply Rejection Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

1.5.5 Software for the Diode Temperature Sensor . . . . . . . . . . . . . . . . . . . . . . . . . 52
1.5.6 Checking the Production Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
1.5.7 Spreadsheet Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
1.6 Integrated Circuit Temperature Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
1.6.1 National LM135 Precision Temperature Sensor . . . . . . . . . . . . . . . . . . . . . . . 58
1.6.2 Generating the Temperature-Sensitive Voltage . . . . . . . . . . . . . . . . . . . . . . . . 59
1.6.3 Amplifying the Temperature-Sensitive Voltage . . . . . . . . . . . . . . . . . . . . . . . 60
1.6.4 Integrated Circuit Temperature Sensor Interface Circuit . . . . . . . . . . . . . . . . . . . 62
1.7 Temperature-Current Transducer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
1.7.1 The National LM134 Temperature Sensor . . . . . . . . . . . . . . . . . . . . . . . . . 64
1.8 A High Resolution Temperature Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
1.8.1 Analysis of the Temperature-Frequency Converter . . . . . . . . . . . . . . . . . . . . . 67
1.8.2 Frequency Measurement by Accumulating Pulses . . . . . . . . . . . . . . . . . . . . . . 69
1.8.3 Measurement of Period . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
1.9 Temperature Measurement Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
1.9.1 Thermistor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
1.9.2 Diode-Connected Transistor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
1.9.3 Integrated Circuit Temperature-Voltage Sensor . . . . . . . . . . . . . . . . . . . . . . . 77
1.9.4 Integrated Circuit Temperature-Current Sensor . . . . . . . . . . . . . . . . . . . . . . . 77

2 Humidity 79
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
2.2 Vapour Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
2.3 Relative Humidity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
2.4 Saturation Vapour Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
2.4.1 Vapour Pressure Applied: Humidex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
2.5 The Wet-Bulb, Dry-Bulb Hygrometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
2.5.1 Computerizing Wet-Bulb, Dry-Bulb Measurements . . . . . . . . . . . . . . . . . . . . . 85
2.5.2 Wet-Dry Bulb Calculation: An Example . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
2.6 Dew Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
2.6.1 Dew Point Calculation: An Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
2.6.2 Dew Point Calculation by Computer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
2.7 Electronic Measurement of Humidity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
2.8 A Capacitance-Humidity Sensor Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
2.8.1 Sensor Specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
2.8.2 Oscillator Measuring Circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
2.8.3 555 Timer Design Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
2.8.4 Engineering Proverb: Do it the easy way . . . . . . . . . . . . . . . . . . . . . . . . . . 96
2.8.5 CMOS Oscillator and Divider . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
2.8.6 Humidity Oscillator Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
2.9 Calibrating the Hygrometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
2.10 Humidity Measurement Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
2.10.1 Humidity Task State Machine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
2.10.2 Humidity Task Pseudo-Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
2.10.3 Humidity Display Routine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

www.syscompdesign.com
CONTENTS CONTENTS

2.10.4 Converting Humidity Sensor Frequency to Relative Humidity . . . . . . . . . . . . . . . 103

2.10.5 Calibration Software for the Humidity Sensor . . . . . . . . . . . . . . . . . . . . . . . . 104
2.11 Project Exploration: Resistance Humidity Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . 105
2.12 Project Exploration: Monostable Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
2.12.1 Analysis of the Astable Multivibrator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
2.12.2 Humidity Astable Schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

3 Wind Speed 111

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.2 Cup Anemometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.3 Anemometer using the Cooling Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.4 Thermal Circuits, A Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.5 Thermal Model of the Cooling Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
3.6 A Constant Power Anemometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
3.7 Constant Temperature Anemometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
3.8 Comparing Thermal Anemometers: A Morality Tale . . . . . . . . . . . . . . . . . . . . . . . . 125
3.9 The Hot Wire Anemometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
3.10 Analog Squaring Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
3.11 A Hot Wire Anemometer Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
3.12 Heated Thermistor Anemometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
3.13 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
3.14 Appendix: Simplifying King’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
3.15 Appendix: A Wind Tunnel for Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

4 Wind Direction 143

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
4.2 Accuracy and Resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
4.3 Shaft Encoder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
4.4 Synchro Resolver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
4.5 Resolver Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
4.5.1 An Aside on Circuit Design Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . 149
4.5.2 Resolver Interface as Communications System . . . . . . . . . . . . . . . . . . . . . . . 149
4.5.3 Resolver Interface Building Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

5 Air Pressure 161

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
5.2 Measuring Air Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
5.3 Air Pressure and Altimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
5.4 Electronic Measurement of Air Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
5.4.1 Sensor Specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
5.4.2 Barometer Design Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
5.4.3 Barometer Resolution and Dynamic Range . . . . . . . . . . . . . . . . . . . . . . . . . 169
5.4.4 Transfer Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
5.4.5 The Barometer Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
5.4.6 Barometric Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
5.5 Barometer Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176

www.syscompdesign.com
 iv

5.5.1 Approximate Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176

5.5.2 Accurate Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
5.5.3 Automatic Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
5.6 Reliability of the Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
5.6.1 Circuit Tolerance Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
5.6.2 Temperature Drift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187

6 Software Design 189

6.0.3 The Complexity Monster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
6.0.4 Organization of a Control Program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
6.0.5 Example Control Program: CAPCON . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
6.0.6 Example: CAPCON Main Loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
6.1 Module Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
6.1.1 Module Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
6.1.2 Module Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
6.1.3 Module Heading Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
6.1.4 Module Internal Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
6.1.5 Decisions, Decisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
6.2 Module Debugging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
6.2.1 Data Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
6.2.2 Data Structure Example: Temperature Conversion . . . . . . . . . . . . . . . . . . . . . 205
6.3 The Real-Time Interrupt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
6.3.1 Accurate Time Delays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
6.3.2 Flip A Bit: Real Time Interrupt on the 68HC11 . . . . . . . . . . . . . . . . . . . . . . . 207
6.4 State Machine Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
6.4.1 State Machine Technique 1: Output Pulse Waveform . . . . . . . . . . . . . . . . . . . . 211
6.4.2 State Machine Technique 2: Motion Detector . . . . . . . . . . . . . . . . . . . . . . . . 214
6.5 The Button Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
6.5.1 Button Switch Pseudocode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
6.5.2 Button Interface to Display Manager Task (Main Loop) . . . . . . . . . . . . . . . . . . . 222

7 Instrumentation Laboratory Outline 223

7.1 Lab Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
7.1.1 Project Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
7.1.2 Project Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
7.1.3 Microprocessor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
7.1.4 Project Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
7.1.5 Grading Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
7.1.6 Tools and Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
7.1.7 Resource Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
7.1.8 Lab Schedule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
7.1.9 Project Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
7.2 Laboratory Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
7.2.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
7.2.2 Calibration Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
7.2.3 Humidity Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229

www.syscompdesign.com
CONTENTS CONTENTS

7.2.4 Barometric Air Pressure Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231

7.2.5 Wind Speed Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
7.2.6 System Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233

Bibliography and References 235

www.syscompdesign.com
 vi

www.syscompdesign.com
List of Figures

1.1 Thermistor Resistance-Temperature Characteristic . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 Thermistor Temperature Measurement Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Lookup Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Thermistor Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5 Thermistor Divider . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.6 Temperature Measurement Block Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.7 A/D – Temperature Lookup Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.8 Thermistor Circuit Measurement Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.9 Thermistor Voltage-Current Characteristic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.10 Thermistor Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.11 Thermistor Circuit Load Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.12 Thermistor Voltage from Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.13 Diode and Diode-Connected Transistor Temperature Sensors . . . . . . . . . . . . . . . . . . . . 22
1.14 The MTS Temperature Measurement Transistor . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.15 Vbe vs IC for the MTS102 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.16 Simple Diode Current Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.17 Sensor and A-D Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.18 Inverting Amplifier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.19 Inverting Amplifier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.20 Sensor and A-D Signal Seesaw Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.21 Sensor and A-D Signal Seesaw Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
1.22 Temperature Sensor Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.23 Power Supply Rejection Ratio Equivalent Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . 39
1.24 LM324 Internal Circuitry (Simplified) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.25 Operational Amplifier With Bias Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
1.26 Bias Voltage Divider Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
1.27 Offset Current Calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
1.28 Operational Amplifier With Offset Voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
1.29 Operational Amplifier Offset Voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
1.30 Operational Amplifier Offset Voltage Application . . . . . . . . . . . . . . . . . . . . . . . . . . 47
1.31 Operational Amplifier Bias Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
1.32 Operational Amplifier Bias Current Application . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
1.33 Operational Amplifier Offset Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
1.34 Operational Amplifier Offset Current Application . . . . . . . . . . . . . . . . . . . . . . . . . . 50

vii
 viii

1.35 Op Amp Power Supply Rejection, Equivalent Circuit . . . . . . . . . . . . . . . . . . . . . . . . 51

1.36 Op Amp Power Supply Rejection, Application Circuit . . . . . . . . . . . . . . . . . . . . . . . . 51
1.37 Diode Temperature Sensor Block Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
1.38 Checking Component Variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
1.39 Diode Temperature Simulation Block Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
1.40 Integrated Circuit Temperature Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
1.41 Temperature Coefficient Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
1.42 The Effect of Varying Collector Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
1.43 Temperature Sensor Amplifier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
1.44 Temperature-Voltage Interface Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
1.45 Temperature-Current Transducer Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
1.46 Temperature-Current Transducer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
1.47 LM134 Adjustable Current Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
1.48 Conversion of Current to Voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
1.49 Schematic Diagram, High Resolution Temperature Sensor . . . . . . . . . . . . . . . . . . . . . 66
1.50 Temperature to Frequency Conversion Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
1.51 Internal Details, High Resolution Temperature Sensor . . . . . . . . . . . . . . . . . . . . . . . . 67
1.52 Block Diagram, High Resolution Temperature Sensor . . . . . . . . . . . . . . . . . . . . . . . . 70
1.53 Count-Temperature Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
1.54 Count VS Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
1.55 Final Thermistor Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
1.56 Temperature Sensor Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
1.57 Temperature-Voltage Final Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
1.58 Schematic Diagram, High Resolution Temperature Sensor . . . . . . . . . . . . . . . . . . . . . 78

2.1 Vapour Pressure Demonstration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

2.2 Temperature and Saturation Vapour Pressure of Water . . . . . . . . . . . . . . . . . . . . . . . . 81
2.3 Vapour Pressure vs Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
2.4 Wet-Dry Bulb Hygrometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
2.5 Automatic Measurement of Humidity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
2.6 Psychrometric Calculation, Data Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
2.7 Vapour Pressure Demonstration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
2.8 Capacitance vs Relative Humidity for Humidity Sensor . . . . . . . . . . . . . . . . . . . . . . . 90
2.9 LC vs RC Oscillators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
2.10 Humidity Sensor Oscillator Waveform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
2.11 Timer Circuit and Astable Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
2.12 Humidity Measurement With Variable Frequency Oscillator . . . . . . . . . . . . . . . . . . . . 98
2.13 Hygrometer Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
2.14 Salts for Humidity Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
2.15 Humidity Task State Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
2.16 Humidity Task Flow Chart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
2.17 Variable Resistance Humidity Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
2.18 Humidity Sensor using Monostables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
2.19 Astable Multivibrator Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
2.20 Humidity Sensor using Monostables, Schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

www.syscompdesign.com
LIST OF FIGURES LIST OF FIGURES

3.1 Beaufort Wind Scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

3.2 Highest Wind Speed for 1 Hour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
3.3 Rotating Cup Anemometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
3.4 Optical Encoder Disc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
3.5 Thermal Circuit Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
3.6 Thermal Equivalent Circuit, Cooling Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
3.7 Constant Power Anemometer Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
3.8 Constant Power Anemometer, Basic Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
3.9 Constant Power Anemometer Circuit, Final Version . . . . . . . . . . . . . . . . . . . . . . . . . 122
3.10 Constant Temperature Anemometer Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
3.11 Constant Temperature Anemometer Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
3.12 Constant Power Anemometer, Basic Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
3.13 Constant Temperature Anemometer Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
3.14 Measuring Lamp Time Constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
3.15 Hot Wire Anemometer Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
3.16 Cooling Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
3.17 Square Function Circuit Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
3.18 Hot Wire Anemometer with Squarer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
3.19 Lamp Voltage-Current Characteristic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
3.20 Thermistor Anemometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
3.21 Thermistor Anemometer, Output Signal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
3.22 Wind Tunnel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

4.1 Compass Rose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

4.2 Shaft Encoder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
4.3 Gray Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
4.4 Resolver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
4.5 Resolver SINE Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
4.6 Resolver COSINE Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
4.7 Resolver Stator Voltage Waveform vs Shaft Angle . . . . . . . . . . . . . . . . . . . . . . . . . . 148
4.8 Resolver Interface Block Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
4.9 Resolver Spectum Diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
4.10 Wein Bridge Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
4.11 Phase Sensitive Rectifier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
4.12 Phase-Sensitive Rectifier Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
4.13 Low Pass Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
4.14 Lowpass Filter Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
4.15 Resolver Interface Schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

5.1 Mercury Column Barometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162

5.2 Pressure Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
5.3 Aneroid Barometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
5.4 Barometer Scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
5.5 Pressure Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
5.6 Resistor Bridge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
5.7 Barometer Interface Transfer Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170

www.syscompdesign.com
 x

5.8 Barometer Interface Block Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171

5.9 Interface Block Diagram, Adding Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
5.10 Interface Block Diagram, Splitting the Gain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
5.11 Barometer Interface Schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
5.12 Second Amplifier Stage, Differential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
5.13 Second Amplifier Stage, Simplified . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
5.14 Complete Block Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
5.15 Water Manometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
5.16 Digital to Analog Converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
5.17 D-A Amplifier Transfer Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
5.18 D-A Amplifier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
5.19 D-A Amplifier Connection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
5.20 Barometric Pressure Interface, Automatic Calibration . . . . . . . . . . . . . . . . . . . . . . . . 184
5.21 Circuit Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
5.22 Amplifier Spreadsheet Model and Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187

6.1 Program Complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190

6.2 Control Program Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
6.3 Solar Telescope System Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
6.4 68HC11 Branch Instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
6.5 IF-THEN-ELSE Flowchart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
6.6 Pulse Generator Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
6.7 Motion Detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
6.8 Motion Detector States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
6.9 Pushbutton State Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
6.10 Contact Bounce . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220

www.syscompdesign.com
Preface xi

Preface

This is a book about electronic instrumentation, the design and construction of machines for the measurement of
physical quantities. It is also about Engineering: the art and science of desiging electronic machinery to perform
a particular task.
As an extended example, we shall study the measurement of the physical properties associated with meteo-
rology, the science of weather. In particular, we will design instruments for the measurement of air temperature,
humidity, pressure, direction and speed of movement. The circuit designs are not an end in themselves, though
they may be useful. Beyond these, which will change with the technology, are engineering skills which outlast
changes in the hardware and software: organizing an engineering approach to the design, weighing alternatives,
analysing and understanding the behaviour of components and systems, ensuring that performance objectives are
met.

Organization of the Text

The text first emphasises the electronic design issues – analogue and digital – that are involved in the design of
meterological sensors.
Significant material is devoted to the design of microprocessor software, an area that is difficult and error
prone for many students.
The final chapter includes a copy of the Lab Outline and typical Lab Exercises that have been used in this
course.
A complete, working weather station could be built using the ideas in this book. However, because this book
focusses on electronic design, the mechanical details and the design of some of the more ordinary elements such
as the power supply are not addressed here. Anyone reading this book is assumed to be capable of translating the
ideas into complete, working hardware.

Course Description
The material in this text was originally used in a one-semester laboratory course in microprocessor-based instru-
mentation, directed to the following objectives:
• to design an electronic system using analogue and microprocessor components.
• to design circuits requiring care in selection of components and grounding, shielding and decoupling.
• to gain experience in calibrating and verifying the performance of an analogue system.
• to design and build software for an embedded microprocessor system.
Further detail on the lab curriculum is contained in chapter 7.
 xii

The Microcomputer System

The Weather Station Project uses a 68HC11 microcomputer board developed by the author for use in this project.
The microcomputer board (The MPP Board) is available for commercial sale, along with a 100 page technical
manual. Technical details, price and availabilty are listed on the web site

http://www.syscompdesign.com

Feedback
Comments, requests for clarification and cash prizes are all encouraged and should be directed to

phiscock@ee.ryerson.ca.

Acknowledgements
Special thanks to Jim Koch for designing the MPP circuit board, testing the various circuits, and his endless
patience in helping students in the lab.
Thanks to David Smith, maintainer of xfig and transfig for his help in resolving bugs.

Production Notes
This document was prepared on a Linux system using the RedHat 7.0 release and the FVWM desktop. Text editing
was done using the joe text editor. Diagrams were prepared with the excellent xfig drawing program and
incorporated as latex, epic, pictex or eps formatted files. Images were processed through the xv image
processing program and incorporated in eps format. Graphs were generated with gnuplot and then further
massaged in xfig.
Text processing was done with the LATEX typesetting program, together with the packages graphicx, epsfig,
subfigure, epic, piclatex, latexsym, afterpage and headings. (The functions of these packages
is described in The Latex Companion.)
Document previewing was done with the xdvi previewer. When the document was completed, it was for-
matted to postscript using dvips and then further converted to pdf format using ps2pdf.
Special thanks to the individuals who have provided this excellent software for the use of the technical writing
community.

Copyright
This document is Peter
c Hiscocks, 1994. Members of educational institutions - students, faculty and staff -
are welcome to copy this material for teaching purposes, providing this copyright notice accompanies the mate-
rial. Others are welcome to copy the material in this document for their own personal use, again providing this
copyright notice accompanies the material.
When this material is being sold to the end user, distribution rights must be obtained from the author, who
may be contacted at the email address given above.

www.syscompdesign.com
Chapter 1

Temperature Measurement

1.1 General Requirements

Our intent is to design a low cost temperature sensor that can be used for weather monitoring in domestic applica-
tions. It is expected that the circuits will be used primarily by students and hobbiests. As a result, the the design
should be simple, straightforward to design and analyse, use readily available parts, and be simple to construct. It
should also be be accurate without calibration or be simple to calibrate, since users are unlikely to have access to
sophisticated equipment.
The following are specific performance requirements for the temperature measuring circuit.

Range of Measurement According to the Canadian Encyclopedia (reference [1], the air temperature extremes
to be expected in Canada and the world are:

Measurement Canadian Record World Record

Maximum 45.0o C, Midale and Yellow Grass, 58.0oC, Al’azizyah, Libya,
Saskatchewan, 5 July 1937 13 Sept 1922
Minimum -63.0oC, Snag, Yukon Territory, -89.2oC, Vostok, Antartica,
3 Feb 1947 21 July 1983

Ideally, our temperature measuring circuit should be able to cope with the maximum possible range of ambi-
ent temperature on the face of the planet earth. However, as we shall see, this may have to be compromised
in the face of practical sensor limitations.
On the other hand, to simplify things for the circuit designer, we assume that all the circuits except the
temperature sensor are kept at room temperature, which we define as +15 o C and +25 o C 1 .

Resolution The resolution of the circuit is the minimum detectable change in temperature. For a domestic ther-
mometer, a resolution of 1o C may be sufficient. For professional weather-monitoring applications, a reso-
lution of 0.1o is more likely to be required.
Initially, we shall specify the resolution of the temperature measuring system as 1 degree centigrade. We
shall also examine a technique for increasing this resolution to one tenth of a degree.
1 Those with unsympathetic landlords or heritage in the British Isles may wish to choose a lower minimum temperature.

1
 2

Accuracy The accuracy of the instrument specifies the allowable error in measurement. For example, in a case
where the temperature trend is of interest, accuracy may not be all that important. On the other hand, where
temperature readings must be compared with other instruments, the accuracy requirements are much more
stringent. The resolution of the instrument should reflect its accuracy: it does not make sense to have a
readout to six decimal places when the instrument is only accurate to two decimal places.
It is reasonable to make the accuracy of the temperature measurement circuit equal to or better than the
resolution. That is, for a temperature resolution of 1 degree, the accuracy should be +/ − 1 o C over the
entire range of operation.
Speed of Response In the measuring of climate data, air temperature changes slowly with respect to the speed of
most electronic instruments. However, in the case of a clinical thermometer, a fast response is desireable to
minimize the time required for the sensor to change from room to body temperature.
Calibration From the standpoint of manufacturing cost, it is highly desireable that the instrument can be relied
upon to meet its accuracy requirement without calibration. Calibration activities are time consuming and
therefore expensive, so more expensive components can be justified if they minimize or eliminate calibra-
tion requirements.
On the other hand, for a hobbyist, the cost of labour is not a major consideration and the incentive may be
to minimize parts cost instead.
Cost The cost should be minimized while meeting the performance specifications.

www.syscompdesign.com
Temperature Measurement 1.2 The Thermistor

1.2 The Thermistor

The thermistor (thermal resistor), is so named because it has a strong positive or negative coefficient of resistance
with temperature. Thermistors of interest in to us are NTC devices: they have a negative coefficient of resistance.
The resistance of the sensor decreases with increasing temperature.

Thermistors are attractive for temperature measuring devices because they are relatively inexpensive and read-
ily available in different physical packages and with different resistance values [10]. Their sensitivity greatly
simplifies the electronic display circuitry.

The resistance-temperature characteristic of the thermistor is approximately exponential: the characteristic of

a typical thermistor is shown in figure 1.1.

1000

100

Resistance, KΩ

10

1
-40 -30 -20 -10 0 10 20 30 40 50
Temperature, C

Figure 1.1: Thermistor Resistance-Temperature Characteristic

The data for this thermistor, derived and simplified from the data sheet in reference [10] is given in section1.2.1.
We shall refer to this information throughout the design.

www.syscompdesign.com
 4

1.2.1 Thermistor Electrical Specifications

Dissipation 250 mW, maximum

Dissipation factor 7 mW/o C
Operating Temperature Range -40 C to +150oC
o

Thermal Time Constant 11 seconds, in still air

1.2 seconds, in liquid
Available Resistance Tolerances ±10%, ±5%, ±3%
β25/85 3977
β Tolerance at 25o C 0.75 %
Drift in β at 125oC ≤0.5%

Thermistor Resistance-Temperature Characteristic

Thermistor Data, Philips Thermistor 640-6, 10KΩ

Temperature, Resistance Deviation due to Resistance,
o
C β tolerance, % KΩ
–40 2.64 328.4
–35 2.40 237.7
–30 2.16 173.9
–25 1.93 128.5
–20 1.58 95.89
–15 1.49 72.23
–10 1.29 54.89
–5 1.08 42.07
0 0.89 32.51
5 0.70 25.31
10 0.52 19.86
15 0.34 15.69
20 0.17 12.49
25 0.00 10.00
30 0.16 8.060
35 0.32 6.536
40 0.47 5.331
45 0.62 4.373
50 0.77 3.606
55 0.91 2.989

The resistance of the thermistor varies with temperature, and so the resistance of a thermistor is specified at
25o C, and known as R25 . In this case, the value of R25 is 10KΩ and the thermistor is a 10KΩ device.
On the other hand, the non-linearity of the temperature-resistance characteristic is a nuisance, though it is
possible to fabricate a thermistor sensor with linear resistance-temperature characteristic at somewhat greater
cost. It is also relatively difficult to make thermistors accurate at low temperatures.

www.syscompdesign.com
Temperature Measurement 1.2 The Thermistor

1.2.2 Thermistor Measuring Circuit

To play on the strengths of the thermistor, we shall design a very simple measuring circuit, using the thermistor
of figure 1.1.
Because of the limited accuracy of the chosen thermistor at low temperatures, we shall aim for a temperature
range of –30o C to +50o C, adequate for temperate climate zones.
A suitable circuit is shown in figure 1.2(a).

s d Vrh
..
... .........
......... ......
...
.....
R1 ..........
..
R1 A-D
s d
.........
. . .....
..

s d
.... ....
...
..
Vb Converter
Vb In
 6
...
.........

..
........

...
.....
........

...... . .
.........
. ...
Rt Vt
.....
.. Rt Vt
s d?
..... ...
..

s d
....

Vrl
m100 m100
therm-divider-a.latex therm-divider-b.latex
(a) Temperature Measuring Cir- (b) Temperature Measurement, Ratiometric A-D
cuit

Figure 1.2: Thermistor Temperature Measurement Circuit

As the temperature increases, the resistance Rt of the thermistor decreases, reducing Vt . In this arrangement,
 
Rt
Vt = × Vb
Rt + R 1
The magnitude of the supply voltage Vb directly affects the thermistor voltage Vt , so for accurate measure-
ments, the supply voltage Vb must be stable.
If the measurement is by means of a ratiometric analog-digital converter as shown in figure 1.2(b), we can
remove the effect of power supply voltage. Recall that a ratiometric A-D converter generates a result which is
proportional to the ratio of the input voltage to the two A-D reference voltages, V rh and Vrl . If Vrl is connected
to the power supply reference terminal2 (the usual case), then the A-D reading NAD is given by
Vin
NAD = × Nmax .
Vrh
In the case of an 8 bit A-D converter, the maximum output value is N max is 28 − 1 or 255.
For thermistor circuit figure 1.2(b), we may substitute 256 for Nmax , Vt for Vin and Vb for Vrh :
Vt
NAD = × 256.
Vb
2 Usually somewhat loosely referred to as ground

www.syscompdesign.com
 6

Solving for Vt and then substituting in the thermistor voltage-divider equation, we find that V b cancels out of
the equation, and we have

R1
Rt = NAD × (1.1)
256 − NAD
In other words,

• If R1 and the A-D reading are known, the value of the thermistor resistance may be calculated.

• Within certain limits (ie, assuming the circuits continue to operate) the value of the supply voltage V b has
no effect on the accuracy of the measurement. This is highly desireable for stability and accuracy of the
temperature measurement.

A useful form of equation 1.1 gives the value of the A-D converter reading N AD in terms of the resistances:

Rt
NAD = × 256 (1.2)
R1 + Rt
Now that we have a method of determining the thermistor resistance from the A-D reading, we need a way to
relate the thermistor resistance to its temperature.
There are three possible approaches to this: a lookup table, an approximate formula, and the more accurate
Steinhart-Hart approximation.

1.2.3 Lookup Table

Some thermistor data sheets ([10] for example) include a table of resistance-temperature values. The temperature
values may then be placed in a lookup table and indexed by the resistance value. For an 8 bit A-D converter, a
table with 256 entries would contain all possible temperature readings, as indicated in figure 1.3.
If the output variable can be represented by some value between 0 and 255, then one byte is required per entry
of the table. For larger table values, multiple bytes can be used for each entry.
To automate the process of generating the lookup table, it is convenient to write a small program in some
readily available computer language such as QBasic, and use that to write a disk file containing the 256 values of
the lookup table. The computer program uses the value of the input variable to index into the lookup table, which
contains the output variable.
Then the disk file may be loaded into the microprocessor source code to incorporate it into the microprocessor
program.
Unfortunately, the lookup table method does not easily allow adjustment of the calibration. In most cases, the
entire lookup table must be rebuilt for a different calibration.

1.2.4 Thermistor Equation

The behaviour of an NTC thermistor is often approximated by the equation
 
1 1
β −
RT = R25 e T To (1.3)
where

www.syscompdesign.com
Temperature Measurement 1.2 The Thermistor

Y 255 227

Output Variable

173 208

208

2 15
1 12
0 10

173 X
Input Variable
The Function The Lookup Table

Figure 1.3: Lookup Table

RT is the resistance of the thermistor at temperature T

R25 is the resistance at 25o C
β is a constant which determines the sensitivity of the thermistor to temperature
T is the temperature in Kelvin
To is 25o C in Kelvin, or 298 K.

The resistance at 25o C and the value of β are often given on data sheets.
For example, suppose the resistance of a certain thermistor is 10KΩ at 25 o C and the value of β25 is given as
3977. Calculate the resistance of the thermistor at 85o C.

 
1 1
β −
RT = R25 e T To
 
1 1
3977 −
= 10 e 358 298
= 1.068KΩ

The tabulated value for the thermistor is 1.070KΩ, so the formula is quite close in this case. However, if the
same values are used to calculate the resistance of the thermistor at −25 o C, the thermistor equation yeilds 147KΩ
and the tabulated value is 128KΩ, an error of some 15%. Obviously the thermistor formula must be used with
some caution.

www.syscompdesign.com
 8

The graph of figure 1.4 (plotted on a linear scale to emphasize the discrepancy) shows how the thermistor
equation becomes progressively less accurate at low temperatures. The squares show the actual thermistor values,
from a table in the data sheet, and the dotted line shows the values generated by the thermistor equation.
450
Thermistor Characteristic
400 Thermistor Equation
350
Resistance, Kohms

300

250

200

150

100

50

0
-40 -30 -20 -10 0 10 20 30 40 50
Temperature, C

Figure 1.4: Thermistor Equation

1.2.5 Steinhart-Hart Equation

Since a microprocessor will be doing the work of the calculations and speed is not an issue in this project,
there is no pressing reason to simplify the calculations. A more complex equation is perfectly acceptable if
it generates more accurate results. There is such an equation: the Steinhart-Hart equation, [13]which is more
accurate, especially where large changes in temperature are involved:
1
= a + b ln RT + c (ln RT )3 (1.4)
T
where

RT is the resistance of the thermistor at temperature T

T is the temperature in Kelvin
a, b and c are constants that fit the equation to the behaviour of the thermistor.
To generate a correct equation for a given thermistor, the constants a, b and c must be found by solving three
equations in three unknowns. Then the equation may be used to solve for temperature at any intervening value of
thermistor resistance.
According to reference [11]:
When −40o C ≤ T 1, T 2, T 3 ≤ 150o C and | T 2−T 1 |≤ 50o C, | T 3−T 2 |≤ 50o C interpolation
data generated by this equation will be accurate to ±0.01oC or better.
For example, to determine the constants of the previous 10KΩ thermistor for the range –30 o C to +50o C, we
might choose the following values of temperature and corresponding thermistor resistance (from the data sheet):

www.syscompdesign.com
Temperature Measurement 1.2 The Thermistor

Value Temperature, o C Thermistor Resistance

T1 –30 173.9KΩ
T2 +10 19.86KΩ
T3 +50 3.606KΩ

The three equations are then:

1
= a + b ln 173.9 + c (ln 173.9)3
243
1
= a + b ln 19.86 + c (ln 19.86)3
283
1
= a + b ln 3.606 + c (ln 3.606)3
323
Using an engineering calculator3 to solve these equations for a, b and c, we obtain

a = 2.772 × 10−3

b = 2.524 × 10−4
c = 3.042 × 10−7
Now, for example, we can calculate the temperature for a thermistor resistance of 95.89KΩ:

1
= a + b ln RT + c (ln RT )3
T
= 2.772 × 10−3 + 2.524 × 10−4 × ln(95.89Ω) + 3.042 × 10−7 × (ln(95.89Ω))3
= 3.952 × 10−3

Then
1
T =
3.952 × 10−3
= 252.9oK
= −20.1oC

From the tabulated values for the thermistor, a resistance of 95.89KΩ should correspond to -20 o Centigrade,
so the Steinhart-Hart equation generates an accurate value in this case. It turns out that this is typical and the
equation may be relied upon over the entire range of thermistor resistance.

Solving for RT in the Steinhart-Hart Equation

When the value of temperature is known and the corresponding value of thermistor resistance is to be calculated,
the Steinhart-Hart equation (a cubic) must be solved for RT . A computer program such as Matlab, Maple, or
Mathematica or an engineering calculator such as the Texas Instruments TI-85 that is equipped with an equation
solving facility will do this. For example, enter the Steinhart-Hart equation as
3 For example, the Texas Instruments TI-85 calculator has a simultaneous equation solver which makes this calculation quite simple.

www.syscompdesign.com
 10

1
= a + bx + cx3 (1.5)
T
where x has been substituted for ln(RT ).
Use the calculator to solve for x and then find RT = ex .

1.2.6 Thermistor Circuit Design

Now that we have a method of relating thermistor resistance to temperature, we are in a position to determine the
value of the voltage divider resistance R1 in figure 1.2.
As a starting point, we might make the mid-point of the temperature range correspond to the mid-point of
the A-D converter range. The temperature varies from –30oC to +50oC, so the mid-point is +10oC. At this
temperature, the thermistor resistance Rt is 19.86KΩ. Then, using equation 1.1, we find that R1 should be equal
to Rt . The nearest standard resistor value is 20KΩ, so that is chosen for R 1 . This makes R1 = 2 × R25 .
The graph of figure 1.5 shows in more detail what happens to the transfer characteristic V t /Vb vs temperature
as R1 is varied with respect to the thermistor resistance R25 .
The transfer function for this graph [16] is obtained from the equation

Vt Rt
= (1.6)
Vb R1 + R t

where Rt is given by equation 1.3. (Equation 1.3 is only approximate, as we saw in section 1.2.4, but is
adequate for this purpose. It is much easier to use than the Steinhart-Hart formula (equation 1.4) because it yeilds
the termistor resistance directly.) Plotting Vt /Vb versus temperature for various values of R1 yeilds the graph of
figure 1.5.
In an ideal world, the graph would be linear, indicating that the output voltage is a linear function of tem-
perature. It’s not possible to obtain a linear transfer characteristic for any value of R 1 , but the linear region of
the graph can be moved to different temperatures. For example, the graph is more linear at low temperatures if
R1=8×R25 . The graph is more linear at high temperatures if R1=0.5×R25. For our purposes, where operation is
between temperatures of -40 and +50o C, a good compromise is R1=2×R25 .
As the graph shows, making R1 equal to Rt in the centre of the temperature region of operation is a satisfactory
’rule of thumb’ in designing simple thermistor voltage-divider circuits.

1.2.7 Resolution of the Thermistor Thermometer

The next step is to check the resolution of the temperature measuring circuit. Figure 1.5 shows that the slope of
the transfer characteristic varies with temperature so we would expect the resolution to vary with temperature. In
order to resolve a change of 1o C, we must ensure that the output of the A-D converter changes by at least one
count for each degree change in temperature.
As a quick check, we can determine the change in A-D reading for a 5 o change at the extremes and centre of
the operating range by plugging thermistor resistance values into equation 1.2.
For example, the thermistor resistance at –30o C is 173.9KΩ. Using equation 1.2, the corresponding output
from the A-D is determined to be 228.
At -25oC, the thermistor resistance is 128.5Ω and the A-D output 220. Thus, in the region of –30 oC, a 5o
change in temperature resulted in an A-D reading change of 228 − 220 = 8, something more than 1 count per
degree.

www.syscompdesign.com
Temperature Measurement 1.2 The Thermistor

1
0.9 R1/R25=0.5
R1/R25=2.0
0.8 R1/R25=8.0
0.7
0.6
Transfer
Function 0.5
Vt /Vb
0.4
0.3
0.2
0.1
0
-40 -30 -20 -10 0 10 20 30 40 50
Temperature, C

Figure 1.5: Thermistor Divider

Repeating this at the centre of the operating range, +10o C and at the extreme hot temperature, +50o C, we have
the following results:

Change in Temperature, o C Change in A-D Reading Approximate Resolution

o
C Counts per degree
–30 to –20 8 1.6
+10 to +15 12 2.4
+45 to +50 7 1.4

Figure 1.6: Temperature Measurement Block Diagram

Throughout the entire temperature range, the A-D reading changes by at least one count per degree. Thus the
microprocessor can resolve changes of 1o C throughout the full range of temperature.
In the centre of the operating range, the A-D output changes by more than 2 steps per degree, so it is possible
to resolve at least 0.5o C in this region.
This analysis is aprproximate, and to be certain the resolution is sufficient, it should be checked throughout
its range. Fortunately, this is easily done with a spreadsheet, plotting routine such as Matlab or GNUPlot, or
interactive computer program such as some dialect of BASIC.

www.syscompdesign.com
 12

1.3 Engineering Tool: BASIC Program

The BASIC programming language has many limitations and in general should not be used to construct production
software. However, BASIC has always been popular among engineers for constructing throwaway programs –
programs that are used as an investigative tool for some technical problem.
Checking the resolution of the thermistor thermometer is a good example of a problem suited for a BASIC
program. It’s also a useful technique for generating the entries for a lookup table of A-D readings versus temper-
ature.
A BASIC program of this type is shown below, which was used to generate the values in figure 1.7.
REM This program generates a table of values
REM corresponding to A-D readings from 0
REM to 255.
REM
REM The thermistor function is represented
REM by its Steinhart-Hart coefficients.
REM Notice that resistance is in Kilohms
REM
a = .002772
b = 2.524 * 10 ˆ (-4)
c = 3.042 * 10 ˆ (-7)

REM The series resistance is 20k ohms

R1 = 20

REM Open the disk file for printing

OPEN "thermist.dat" FOR OUTPUT AS #1

REM This is the calculation loop

FOR Nad= 1 TO 255
Rt = R1 * (Nad / (256 - Nad))
Ta = 1 / (a + (b * LOG(Rt)) + (c * (LOG(Rt)) ˆ 3))
Tc = Ta - 273
PRINT #1, Nad, INT(Tc);
NEXT Nad

REM Close the disk file

CLOSE
This program calculates the value of temperature that corresponds to values of Nad from 1 to 255, and prints
them to a disk file, ’thermist.dat’. This file may then be pasted into the source code of the program that reads the
A-D converter and calculates temperature.
In practice, there are probably additional requirements of the printing format. For example, for incorporation
in 68HC11 assembly language source code, each value would be prefaced by ’FCB’. This mneumonic causes
the assembler to recognize a number as a byte-size constant that should be allocated storage. As well, negative
numbers cannot be represented directly as bytes, and would have to be converted to some other representation,
such as 2’s complement form.

www.syscompdesign.com
Temperature Measurement 1.3 Engineering Tool: BASIC Program

1 197 2 161 3 142 4 129 5 120 6 113 7 107 8 102 9 98 10 94

11 91 12 88 13 85 14 83 15 80 16 78 17 76 18 74 19 72 20 71
21 69 22 68 23 66 24 65 25 64 26 62 27 61 28 60 29 59 30 58
31 57 32 56 33 55 34 54 35 53 36 52 37 51 38 50 39 50 40 49
41 48 42 47 43 46 44 46 45 45 46 44 47 44 48 43 49 42 50 42
51 41 52 41 53 40 54 39 55 39 56 38 57 38 58 37 59 37 60 36
61 35 62 35 63 34 64 34 65 33 66 33 67 32 68 32 69 31 70 31
71 31 72 30 73 30 74 29 75 29 76 28 77 28 78 27 79 27 80 27
81 26 82 26 83 25 84 25 85 25 86 24 87 24 88 23 89 23 90 23
91 22 92 22 93 21 94 21 95 21 96 20 97 20 98 20 99 19 100 19
101 18 102 18 103 18 104 17 105 17 106 17 107 16 108 16 109 16 110 15
111 15 112 15 113 14 114 14 115 14 116 13 117 13 118 13 119 12 120 12
121 12 122 11 123 11 124 11 125 10 126 10 127 10 128 9 129 9 130 9
131 8 132 8 133 8 134 7 135 7 136 7 137 6 138 6 139 6 140 5
141 5 142 5 143 4 144 4 145 4 146 3 147 3 148 3 149 3 150 2
151 2 152 2 153 1 154 1 155 1 156 0 157 0 158 0 159 -1 160 -1
161 -1 162 -2 163 -2 164 -2 165 -3 166 -3 167 -3 168 -4 169 -4 170 -4
171 -5 172 -5 173 -5 174 -6 175 -6 176 -6 177 -7 178 -7 179 -7 180 -8
181 -8 182 -8 183 -9 184 -9 185 -10 186 -10 187 -10 188 -11 189 -11 190 -11
191 -12 192 -12 193 -13 194 -13 195 -13 196 -14 197 -14 198 -14 199 -15 200 -15
201 -16 202 -16 203 -17 204 -17 205 -17 206 -18 207 -18 208 -19 209 -19 210 -20
211 -20 212 -21 213 -21 214 -22 215 -22 216 -23 217 -23 218 -24 219 -24 220 -25
221 -25 222 -26 223 -26 224 -27 225 -28 226 -28 227 -29 228 -29 229 -30 230 -31
231 -32 232 -32 233 -33 234 -34 235 -35 236 -35 237 -36 238 -37 239 -38 240 -39
241 -40 242 -41 243 -43 244 -44 245 -45 246 -47 247 -48 248 -50 249 -52 250 -54
251 -57 252 -60 253 -64 254 -69 255 -78

Figure 1.7: A/D – Temperature Lookup Table

For the lookup table to have sufficient resolution, each entry between the temperatures of −40 to +50 o C
should change by no more than 1o C. This is in fact the case, so the thermometer meets its requirement.
Outside this range, the temperature readings jump by more than 1 o C, reflecting the fact that the transfer
characteristic, figure 1.5, is less steep at high and low temperatures.

1.3.1 Estimating the Production Accuracy

In a production setting, where units are being assembled from batches of components, we must be able to predict
the accuracy of the circuit when different resistors and thermistors of the same type are installed in the circuit.
A number of factors affect the accuracy of the thermistor measuring circuit:

• Tolerance of the thermistor resistance R25 , resistance at 25oC

• Tolerance of the thermistor resistance due to variation in the β of the thermistor
• Tolerance in the value of R1
• Accuracy of the A-D converter
• Accuracy of the equation relating Rt to temperature

To determine the worst case variation in temperature reading in a production batch, we should choose the
worst case for each of these variations and add them together. However, this may be unnecessarily stringent.

www.syscompdesign.com
 14

If the tolerances are independent and uncorrelated, then a good estimate of the worst case variation is found by
taking the RMS sum of the individual variations.

Effect of tolerance in thermistor resistance and sensitivity:

Here we assume that the variation in thermistor resistance (a constant 3 %) and the variation in sensitivity (β
tolerance from figure 1.2) can both occur together.
Then we calculate the effect on the A-D reading at the lowest, middle and highest operating temperatures.
Finally, this gives us the expected variation in A-D reading due to tolerances in the thermistor characteristics. All
resistor values are in KΩ.

Temperature, o C Nominal Rt Rt Tolerance β Tolerance Rt NAD ∆NAD

–30 173.9 +3% +2.16% 182.8 229.8
–3% –2.16% 164.9 227.4 2.4
+10 19.86 +3% +2.16% 20.55 129.2
–3% –2.16% 19.16 124.8 4.5
+50 3.606 +3% +2.16% 3.742 40.2
–3% –2.16% 3.470 37.7 2.5

Effect of tolerance in Divider Resistor R1

In this case, we assume that R1 is a resistor with 1% tolerance. Then, following a similar procedure to the above,
we have the following table:

Temperature, o C Nominal R1 R1 Tolerance Rt Rt NAD ∆NAD

–30 20 +1% 20.2 173.9 228.5
-1% 19.8 173.9 228.9 0.4
+10 20 +1% 20.2 19.86 126.4
-1% 19.8 19.86 127.7 1.3
+50 20 +1% 20.2 3.606 38.62
-1% 19.8 3.606 39.28 0.6

Analogue-Digital Converter Accuracy

The 68HC11 A-D data specifies that the absolute accuracy of the A-D is ±1 least significant digit. The value of
∆NAD is therefore 1.0 for the A-D converter.

Total Accuracy
Summing up the errors at different temperatures, we have the results shown in the following table:
We previously determined that the sensitivity of the measuring circuit varies with temperature (figure 1.6),
so it is reasonable to express the variation in A-D reading as an equivalent temperature. For example, if the
accuracy is ±4.74 counts at +10oC, and the resolution is 2.4 counts per degree, then the accuracy is approximately
2.8/2.4 = ±2.8oC. This is shown in the last two columns for equivalent Worst Case error and equivalent RMS
error.
This is not good news: a thermistor-based temperature measuring circuit using these components will have a
variation in accuracy by nearly ±3o C.

www.syscompdesign.com
Temperature Measurement 1.3 Engineering Tool: BASIC Program

Temperature, o C Worst Case Error, RMS Error, Worst Case Error, RMS Error,
A-D Counts A-D Counts Equivalent Degrees Equivalent Degrees
–30 3.8 2.6 2.3 1.6
+10 6.75 4.74 2.8 1.9
+50 4.1 2.75 2.9 1.9

Figure 1.8: Thermistor Circuit Measurement Accuracy

How could we improve the accuracy of the circuit? Most of the error is due to the variation in R 25 and β,
so these must be more tightly controlled to improve the accuracy. More tightly specified thermistors (±0.1% is
available) are available (albeit for a significantly higher cost) and might be suitable.
Alternatively, we can introduce a calibration procedure that measures R 25 and β, and compensates for their
variation. This has attractions for the amateur, but is time consuming and requires additional microprocessor
software. Before getting involved in that, it would be wise to investigate other types of temperature transducers,
as we do in the next section.

1.3.2 Thermistor Self-Heating and Thermal Runaway

When a current is passed through a thermistor, as it is in this circuit to measure the resistance, a certain amount
of heat is dissapated in the thermistor by ohmic heating. Because the thermistor has a Negative Coefficient of
Temperature, this heat will lower the resistance and further increase the device current.
There is a potential here for thermal runaway, in which the heat changes the properties of the device so that it
dissapates yet more heat until eventually the device incinerates itself.
It would be prudent engineering to ensure first that thermal runaway will not happen in this particular circuit
and second, that self-heating does not materially affect the accuracy of the temperature reading.
The voltage-current characteristic of our 10KΩ thermistor is shown in figure 1.9.
A constant resistor would appear as a straight line passing through the origin with slope proportional to the
resistance. The thermistor resistance varies with current. At an ambient temperature of 25 o C, the resistance
actually becomes negative at forward currents greater than about 7mA.
The stability of a circuit containing a thermistor depends on the characteristics of the source, which may be
indicated by a load line superimposed on the thermistor characteristic, as indicated in the examples of figure 1.10.

www.syscompdesign.com
 16

30

25

20 25o C
50o C
Voltage, V 15

10

0
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018
Current, mA

Figure 1.9: Thermistor Voltage-Current Characteristic

www.syscompdesign.com
Temperature Measurement 1.3 Engineering Tool: BASIC Program

Voltage

a
q

m50 I1 Current
thermistor-loadline-ci.epic

(a) Constant Current Thermistor Source

Voltage Voltage

E1

a q q
b

r
E2
Eoc
a

I1 I2 Current Isc Current

m50 m50
thermistor-loadline-cv.epic thermistor-loadline-rint.epic

(b) Constant Voltage Thermistor Source (c) Voltage Source, Internal Resistance

Figure 1.10: Thermistor Sources

www.syscompdesign.com
 18

The characteristic of a constant current source (figure 1.10(a) intersects the thermistor characteristic at one
and only one point so the circuit will be stable.
On the other hand, the voltage source E1 in figure 1.10(b) exceeds the maximum possible voltage of the
thermistor and so will certainly cause thermal runaway. The voltage source E2 in figure 1.10(b) is conditionally
stable. It intersects the positive resistance portion of the curve at point a, so it is a stable operating point. A second
possible stable operating point is at b.
The characteristic of a voltage source with an internal resistance is shown in figure 1.10(c). In this case, the
stability of the circuit is dependent on the magnitude of the voltage source and slope of the loadline. Since the
loadline intersects the thermistor characteristic at only one point, the values shown would result in a stable circuit.
The situation for our application circuit, with a 5 volt source and 20KΩ resistance, is shown in figure 1.11.

10

6 -30o C
+25o C
Voltage, V +50o C
4 Load Line

0
0 5e-05 0.0001 0.00015 0.0002
Current, mA

Figure 1.11: Thermistor Circuit Load Line

The load line intersects the thermistor characteristic at one point only for the thermistor at three different
temperatures so the circuit will be stable. Further, we can deduce that the self-heating effect is minimal because
the thermistor characteristics are essentially straight lines at these voltages and currents.

1.3.3 Plotting the Thermistor VI Characteristic

Determining the voltage-current characteristic of a thermistor presents an interesting challenge. The basic concept
is simple: we wish to force a current through the thermistor and measure the resultant voltage across it. However,
the current induces heat in the thermistor, which changes its resistance, changing its voltage. Unfortunately, the
resulting equations are not amenable to a simple analytical solution.
The relevant equations are as follows:
The voltage across the thermistor is simply
vt = i t RT (1.7)

www.syscompdesign.com
Temperature Measurement 1.3 Engineering Tool: BASIC Program

where vt and it are the thermistor voltage and current.

The power generated in the thermistor is likewise simply

pt = i2t RT (1.8)

The internal temperature of the thermistor is equal to the ambient temperature T A plus the heat generated
across the thermal resistance θ of the thermistor by the thermistor power P t , according to equation 1.9.

Tj = TA + θPd (1.9)
where Pd is Pt in this case.
Finally, from section 1.2.4 we have equation 1.3 for the thermistor resistance:
 
1 1
β −
RT = R25 e T To (1.10)

Unfortunately, combining equations 1.7 through 1.10 does not lead to a simple algebraic solution. The voltage
corresponding to a given current can be found iteratively, however, using the iterative approach of figure 1.12.

www.syscompdesign.com
 20

Choose starting value of it , Rt

1

v t = i t Rt
2

pt = i2t Rt
3

Tj = TA + θpt
4

 
1 1
β −
Rtnew = R25 e Tj To
5

Rt := 0.1(Rtnew − Rt )
6

m70
therm-vi-flowchart.epic

Figure 1.12: Thermistor Voltage from Current

www.syscompdesign.com
Temperature Measurement 1.3 Engineering Tool: BASIC Program

We start with a value of current and an approximate value of thermistor resistance. We then determine the
heating effect and calculate a new value of resistance. This new value of thermistor resistance is used to improve
the original estimate. We repeat the iterative loop until the values of thermistor resistance converge to a solu-
tion. Once the thermistor resistance is known the thermistor voltage is simply the current times the thermistor
resistance.
In order to ensure that the iteration converges, it may be necessary to adjust the factor by which the estimate of
thermistor resistance is improved with each iteration: a factor of 0.1 in the example shown. If the result oscillates
with each iteration, reduce this factor.

1.3.4 Engineering Tool: Generating Graphs with GNUPLOT

Plotting of engineering equations is greatly simplified with a computer or programmable graphic calculator. One
of the common programs for generating graphs is GNUPLOT, a public domain utility with a large number of
features. GNUPLOT is available for a large number of machines, ranging from DOS computers to engineering
workstations.
GNUPLOT may be used interactively from a command line or may read a file of instructions. The GNUPLOT
file to generate the graph of figure 1.5 is shown below:

#List of commands to plot thermistor divider curves

set terminal latex # direct graphical output to a file
set output "thermistor-divider.latex" # establish name of the file
set xlabel "Temperature, C" # label x axis
set ylabel "Transfer Function" # label y axis
plot [-40:55]
1.0/(1.0+(5.0/(10.0*(2.71818**(3977.0*(1.0/(273.0+x)-1.0/298.0)))))
title ’R1/R25=0.5’ with lines, \

1.0/(1.0+(20.0/(10.0*(2.71818**(3977.0*(1.0/(273.0+x)-1.0/298.0))))
title ’R1/R25=2.0’ with lines, \

1.0/(1.0+(80.0/(10.0*(2.71818**(3977.0*(1.0/(273.0+x)-1.0/298.0))))
title ’R1/R25=8.0’ with lines \
set terminal x11 # show output on terminal
replot # redo the plot

This command file contains instructions to plot the function with output first to a disk file ”thermistor-
divider.latex” (for incorporation in another document) and then to the X windows screen. The # character pre-
ceeds comments, which are ignored.

www.syscompdesign.com
 22

1.4 The Diode Temperature Sensor

A forward biased silicon diode or diode-connected transistor (figure 1.13), operated at constant current, will
exhibit a change in voltage of some -2.2mV per o C. In contrast to the non-linear behaviour of the thermistor, this
constant coefficient makes the diode a very attractive temperature sensor.

s
 
 
Vd ∝ T Vce ∝ T

m100
diode-sensor.epic

Figure 1.13: Diode and Diode-Connected Transistor Temperature Sensors

However, there are some design issues that must be dealt with to make a practical temperature measuring
circuit.
In the first place, the 8-bit A-D converter of the microprocessor requires 5/255 = 19 mV per step when
operated from reference voltages of +5 volts and ground. The diode 2.2 mV/ o C rate of change of voltage with
temperature is much less than this, and must be amplified. Second, the forward biased diode has a standing voltage
of some 600 millivolts which must be subtracted out from the measurement. Together, these requirements may be
met with a simple operational amplifier circuit.
Before launching into a design of suitable circuitry, it is appropriate to have a careful look at the characteristics
of the diode-connected transistor to determine whether it will make an accurate temperature sensor.

1.4.1 Diode Temperature Characteristics

Nelson-Jones [17] and Henderson [18] describe analogue thermometers using silicon diodes. In both cases, the
authors found that the temperature coefficient was only approximately 2.2mV per o C and depended on both the
type of diode and the the forward voltage. This implies that the diode must be selected for the desired characteris-
tics and/or that the instrument must have facilities for compensating for the different coefficients. Obviously, both
these are undesireable if they can be avoided.
According to Sachse [19], page 106):

The forward current across any p-n junction consists of three major components: diffusion current
or drift current, generation recombination current, and surface leakage current. Each of them is
determined by a different mechanism and temperature dependence results in complex compound
behaviour. Verster [20] has shown that any semiconductor material operated as a transistor simplifies
these relationships since only the diffusion component of the emitter is under consideration, while
the generation recombination and surface leakage currents are drained away as base current.

This implies that the diode-connected transistor will have a more predictable temperature characteristic. In-
deed, Motorola used to manufacture a series of transistors, the MTS102-MTS105 [21] which were specifically

www.syscompdesign.com
Temperature Measurement 1.4 The Diode Temperature Sensor

intended for temperature measurement. (Cover of the MTS102 Data Sheet is shown in figure 1.14) 4. Although
the MTS102 is no longer available, the other small signal transistors will have similar specifications.

Figure 1.14: The MTS Temperature Measurement Transistor

1.4.2 Transistor Base-Emitter Voltage Function

The classic equation relating the forward voltage to forward current in a forward biased diode or connected
transistor is  
qVbe
Ic (T ) = Is (T ) e nkT (1.11)
4 Figures 1.14 and 1.15 are Motorola
c Inc, 1981.

www.syscompdesign.com
 24

where
T is the absolute temperature in o Kelvin
Ic is the collector current
Is is the reverse saturation current
q is the electron charge, 1.6 × 10−19 coulomb
Vbe is the base-emitter voltage
n is a process constant which depends on the construction technique, typically ≈1
K is Boltzmann’s constant, 1.38×10−23 joule/o K
If we have access to values of the base-emitter voltage vs collector-emitter current for a particular transistor,
we can solve for the values of n and Is for that transistor. In this example, we shall use the graph of figure 1.15,
adapted from the Motorola data sheet for the MTS102.

Figure 1.15: Vbe vs IC for the MTS102

Calculating the Process Constant n

Assuming that the temperature is constant for this calculation, Ic (T ) and Is (T ) can be written simply as Ic and
Is .

www.syscompdesign.com
Temperature Measurement 1.4 The Diode Temperature Sensor

Taking the natural log of both sides of equation 1.11, we have

qVbe
ln Ic = ln Is +
nkT
For two different points on the Vbe Ic characteristic we have

qVbe1
ln Ic1 = ln Is +
nkT
qVbe2
ln Ic2 = ln Is +
nkT
Subtracting the two equations,
q
ln Ic2 − ln Ic1 = (Vbe2 − Vbe1 )
nkT
or
Ic2 q
ln = (Vbe2 − Vbe1 )
Ic1 nkT
Two suitable data points from the graph of figure 1.15 are 0.1 mA and 1.0 mA, a decade appart in current.
Then we have
Ic1 , 0.1 mA Vbe1 , 600 mV
Ic2 , 1.0 mA Vbe2 , 660 mv
Substituting these values in equation 1.12, we have:
1.6 × 10−19
ln 10 = (0.660 − 0.600)
n × 1.38 × 10−23 × 298

Solving for n,
n = 1.014

Solving for the Saturation Current Is

We are now in a position to substitute a pair of Vbe , Ic values in equation 1.11, along with the newly calculated
value of process constant, in order to solve for the saturation current.

 
qVbe
Ic (T ) = Is e nkT
1.6 × 10−19 × 0.6
 
−23 × 298
= Is e 1.014 × 1.38 × 10
Solving for Is , we have

Is = 1.014 × 10−14 amps

www.syscompdesign.com
 26

1.4.3 Base-Emitter Voltage Function of Temperature

When using the diode-connected transistor as a temperature sensor, our primary concern is the temperature coef-
ficient of base-emitter voltage. Taking the natural logarithm of both sides of equation 1.11 and solving for V be ,
we have

nKT Ic (T )
Vbe (T ) = ln (1.12)
q Is (T )
where the collector current and saturation current have both been shown as functions of temperature T to
emphasize that they change when the temperature changes. In other words, we cannot treat I c and Is as constants
and differentiate this equation with respect to temperature T to determine the temperature coefficient.
However, if the base-emitter voltage is known at some reference temperature T o , the base-emitter voltage at
some other temperature T may be calculated by the following [22, 23, 24]:
       
T T nkT To kT Ic
Vbe (T ) = VG0 1 − + Vbeo + ln + ln (1.13)
To To q T q Io
where the additional terms are

VG0 , the bandgap voltage of silicon, typically 1.22V

To , a reference temperature in o Kelvin
Vbeo , the base-emitter voltage at the reference temperature
Io , the reference collector current

For example, consider that a diode-connected transistor has a base-emitter voltage of 0.600 volts at a collector
current of 0.1 mA and temperature of 20o C. What is the base-emitter voltage at 100o C?
The values of the parameters are:

VGO = 1.22 volts

T = 273 + 100
= 373o K
To = 273 + 20
= 293o K
Vbeo = 0.600 volts
n = 1.014
k = 1.38 × 10−23
q = 1.6 × 10−19
Ic = 0.1 × 10−3
Io = 0.1 × 10−3

Substituting these in equation 1.13, we have

www.syscompdesign.com
Temperature Measurement 1.4 The Diode Temperature Sensor

       
T T nkT To kT Ic
Vbe = VG0 1 − + Vbeo + ln + ln
To To q T q Io
   
373 373
= 1.22 1 − + 0.600
293 293
−23
 
1.014 × 1.38 × 10 × 373 293
+ ln
1.6 × 10−19 373
1.38 × 10−23 × 373 0.1 × 10−3
 
+ ln
1.6 × 10−19 0.1 × 10−3
= −0.333 + 0.763 + 0.0078 + 0.0
= 0.437 volts
This agrees with the base-emitter vs temperature value shown on the data sheet for the MTS102 transistor.
Notice that the third term contributes only a very small amount to the final value. The fourth term has no effect
because the collector current is equal to the reference current.

1.4.4 Base-Emitter Voltage Temperature Coefficient

The first two terms of equation 1.13 are linear in T and contribute most of the change in base-emitter voltage
with temperature. The third term contributes a small non-linear value, and the fourth term may be ignored if the
collector current is constant with temperature.
Consequently, if we assume that the diode is driven from a constant current source, the temperature coefficient
of base-emitter voltage may be determined by differentiating the first three terms of equation 1.13.
From equation 1.13, the differential of the first two terms is
   
T T
Vbe ' VG0 − VG0 + Vbeo
To To
then
dVbe VG0 Vbeo
= 0− + (1.14)
dT To To
The differential of the third term is a bit more involved: We wish to find
 
d nkT To
ln
dT q T
We begin with the identity
duv dv du
= u +v
dx dx dx
and substitute
nkT
u =
q
 
To
v = ln
T

www.syscompdesign.com
 28

to obtain
 
d nkT To nkT d To To d nkT
ln = × ln + ln × (1.15)
dT q T q dT T T dT q
Using the identity
d 1 du
ln u =
dx u dx
we have that
d 1 d To
ln To T = ×
dT To /T dT T
T To
= ×− 2
To T
1
= −
T
We also have that
d nkT nk
=
dT q q
Finally, substituting in equation 1.15, the differential of the third term is given by
 
nkT 1 To nk nk To
− × + ln × = ln −1
q T T q q T
Putting this result together with the linear coefficient terms, we have the complete expression for the rate of
change of Vbe with temperature:
 
dVbe VG0 Vbeo nk To
=− + + ln −1 (1.16)
dT To To q T
For example, the data sheet for the MTS102 temperature sensor transistor gives V beo as 0.595 volts at 25o C.
Determine the temperature coefficient of Vbe .
Substituting VGO =1.22V, Vbeo =0.595, To =298o and T =298o Kelvin in equation 1.16, we have
1.22 0.595 1.014 × 1.38 × 10−23
 
dVbe 298
= − + + ln − 1
dT 298 298 1.6 × 10−19 298
= −4.094 × 10−3 + 1.997 × 10−3 − 0.087 × 10−3
= −2.184 mV/o C
The value from the data sheet is -2.265 mV/o C, which is within 3% of the calculated value.
Plugging temperature values of -40o C and +50o C into equation 1.16, we can calculate the transistor tempera-
ture coefficients at the extremes of operation:
Temperature Coefficient of Temperature Error related to
coefficient at 25o C
–40 -2.163 mV/o C -0.1%
+25 -2.184 mV/o C +0.0%
+50 -2.191 mV/o C +0.3%

www.syscompdesign.com
Temperature Measurement 1.4 The Diode Temperature Sensor

Evidently the temperature coefficient of base-emitter voltage may be relied upon to be essentially constant over
the temperature range we intend to use it.

Sensor Current Source

In section 1.4.4, we assumed that the diode would be driven from a constant current source, and that the current
through the diode would not change with temperature. Ideally, then, our circuit should match these assumptions.
The simplest possible circuit is a resistor from a positive voltage source as shown in figure 1.16.

Rd

 s
s
 Temperature
Sensor

m100
simple-diode-source.epic

Figure 1.16: Simple Diode Current Source

If the voltage across the transistor sensor is more or less constant at 0.6 volts, then if the supply voltage is +5
volts, the voltage across the resistor will be more or less constant at 4.4 volts and the current through the sensor
will be more or less constant as well. Now, as we know, the voltage across the sensor does vary, so we have to
determine whether the variation is enough to affect the accuracy of the sensor voltage.
First of all, we should calculate the value of the series resistor Rd :
Vcc − Vbe
Rd = (1.17)
Ic
For the purpose of calculating the resistance, assume Ic is 100µA and Vbe is 0.6 volts. Then

5V − 0.6V
Rd =
100 × 10−6 amps
= 44KΩ
Now we need to determine the worst case variation in current due to changes in the value of V be . From
figure 1.17, the worst case change in Vbe is to 0.742 volts, which occurs at a temperature of –40oC. Repeating the
calculation of equation 1.17 for Vbe = 0.742 volts, we find that Ic ≈ 96µamps. This value is only approximate
because its change further affects the assumed value of Vbe . However, as we shall show, the effect is slight enough
to be neglected.
Now, how much of an effect does a change of transistor current from 100 µamps to 96 µamps have on the
value of Vbe ? From equation 1.13, the current affects the base-emitter voltage according to the expression

www.syscompdesign.com
 30

 
kT Ic
∆Vbe = ln
q Io

where Ic is the collector current and Io is the original reference collector current. Thus the change in Vbe due
to change in collector current is

 
kT Ic
∆Vbe = ln
q Io
1.38 × 10−23 × 233 96 × 10−6
= −19
ln
1.6 × 10 100 × 10−6
= −0.803 mV

This is not enough to materially effect our original assumption of 0.742 volts for V be , so our assumption of
Ic =96µamps is unchanged.
The temperature coefficient of the transistor sensor is in the order of -2.2 mV/ o C, so this change of -0.8 mV
in Vbe is equivalent to approximately a third of a degree C. For greater accuracy, a different current source would
be required.

1.4.5 Interface Circuit for a Transistor Temperature Sensor

We are now in a position to design the signal conditioning circuitry to match the output of the temperature sensor
to the input of the microprocessor A-D converter.
The change in voltage across the transistor sensor is small compared to the input signal required to actuate the
A-D converter. The change across the sensor is about 2.2 millivolts per degree C. Each step of the A-D converter
output requires a change of 5 volts/256 steps = 19.5 millivolts per step. Moreover, the sensor has a standing
voltage of about 600 millivolts across it. Simply amplifying the voltage across the sensor will generate a signal
outside the range of the A-D converter.
Consequently, the interface circuit must both amplify the change in voltage across the temperature sensor and
subtract the standing voltage.
The interface circuit couples the sensor output signal to the A/D input signal, so the first step is to establish
their ranges.

Range of the sensor signal

For the measurement of ambient temperature in Canada, a suitable range might be -40 o C to +60o C. The total
change in temperature is thus a nice round number: 100o C.

Range of the A-D signal

The 8 bit A-D converter has a total range of 28 = 256 steps. A suitable match between the sensor and the A-D
would be to assign 2 A-D steps per degree of temperature change. This would then use up 200 of the 255 steps,
leaving a margin of 27 steps at each end of the range, and the input to the A-D converter ranges between step 27
and step 227. This arrangement provides a resolution of 2 A-D steps per degree, or 0.5 o C per A-D step, quite
acceptable for a domestic thermometer.

www.syscompdesign.com
Temperature Measurement 1.4 The Diode Temperature Sensor

From the data sheet for the MTS102, the temperature coefficient is -2.265 mV/ o C and the value of Vbe is
595mV at 25oC.
Using these specs, at −40o C, the sensor voltage is

(−40 + 25) × −2.265 + 595

= 742 millivolts

At +60o C the sensor voltage is

(60 − 25) × −2.265 + 595

= 515 millivolts

The corresponding voltages at the input to the A-D converter are: At step 27 ( −40 o C):

27/255 × 5
= 0.529 volts

At step 227 (+60oC):

227/255 × 5
4.451 volts

These signal swings are summarized in figure 1.17.

Voltage A-D Count

5 255

.. ..
4.451V .. .. .. .. 227
m70 4
diode-swings.latex

3.93 200
volts Steps
2

1
–40o C 0.742V
......
...... 0.227V .. .. .. ..
0.517V .. .. 27
+60o C 0.515V
0

Sensor Output A-D Converter Input

Figure 1.17: Sensor and A-D Signals

To design a suitable amplifier circuit, consider the operational amplifier circuit of figure 1.18(a).

www.syscompdesign.com
 32


 6




6 H
H 

H
H
H Ag


H 
vi 0 0 vo
 H
 H
r r e
Ri Rf  H
H

? 
 H
H
HH H
H
m HH
– H
H
vi
 vo H
H ?

+
 - -
r e
Ri Rf

m70
inverting-seesaw-a.epic m90
inverting-seesaw-b.latex

(a) Inverting Amplifier (b) Inverting Amplifier Seesaw Diagram

Figure 1.18: Inverting Amplifier

The inverting operational amplifier circuit has a voltage gain of

vo Rf
=− volts/volt (1.18)
vi Ri
We can represent this on the see-saw diagram as shown in figure 1.18(b). Point A on the diagram represents
the virtual earth point at the inverting input to the op-amp. As the voltage gain of the circuit is increased, point
A moves to the left so that a smaller change in vi causes a larger change in vo . Also notice that a positive input
results in a negative output and vice-versa, as expected from an inverting amplifier.
Now consider the offset inverting amplifier shown in figure 1.19(a).
The effect of the offset or bias voltage Vb is to move the whole see-saw diagram up by an amount Vb volts,
as shown in figure 1.19(b). By comparison with figure 1.17, this is exactly the configuration required to obtain
the combination of amplification and bias required to condition the temperature sensor signal output for the A-D
converter input.
The resulting seesaw diagram for the temperature sensor is shown in figure 1.20.
If the diagram is carefully done to scale, it is possible to read the values of A v and Vb directly off the diagram.
However, we shall use an algebraic approach, developing an expression for the transfer function of the offset
inverting amplifier.
Refer to figure 1.21.
As usual, the rules to keep in mind for negative feedback operational amplifiers are
1. Feedback action keeps the inverting and non-inverting terminals at essentially the same potential.
2. No current flows into the input of the operational amplifier.

www.syscompdesign.com
Temperature Measurement 1.4 The Diode Temperature Sensor


 6




H
6 H 

H
H
H Ag


H 
vi vo
 6
 H
H

r r e
Ri Rf 
 H
H
? 
 H
H
HH vb H
H
m HH
– H
H
 ?
vi vo H
H
+
  ?
0 0

r e
Vb
 Ri - Rf -
m70
offset-inverter-a.epic m90
offset-inverter-b.latex

(a) Offset Inverting Amplifier (b) Offset Inverter Seesaw Diagram

Figure 1.19: Inverting Amplifier

A-D
5 Input
m70 Voltage
sensor-seesaw.pictex
.... ..
..... 4.451V
.....
.....
.....
.......
.
.
. 4
.....
.....
.....
.....
..
.......
.
.....
.....
.....
.....
...
......
.....
..... 3
.....
Sensor .....
..
......
.
..
.....
.
Output .
..... 3.93
.....
.....
Voltage ..... volts
...
......
.....
.....
..... 2
.....
...
......
.
.....
...
......
.
.....
.....
.....
.....
. . . ......
....
..... 1
.....
.....
....... 0.742V ....................................................
.
........... ................................................
0.227V ..
.. .
. .. .
.. ..
... .
.. ..
. ...
....
.. .
...
....... ................................................
.......
0.515V ..... . . ................................... .....
..
..... ..... . 0.517V
...
..... ...
... .
...
.
0
.............

Bias Voltage Vb
0.673V

Figure 1.20: Sensor and A-D Signal Seesaw Diagram

www.syscompdesign.com
 34

Ri Rf
s s h
Vb
... .... ... ....
.... .... ... .... ... .... .... ... .... ...
- -
.... ... .... ...

ii i
HH o 6
– HH
H
H

vi vo

+


Vb
s h
?

m100
offset-inverter-large.latex

Figure 1.21: Sensor and A-D Signal Seesaw Diagram

www.syscompdesign.com
Temperature Measurement 1.4 The Diode Temperature Sensor

Referring to figure 1.21, we can proceed as follows:

By rule 1, the voltage at the inverting terminal is equal to Vb . Then the input current ii is
vi − V b
ii = (1.19)
Ri
and the output voltage vo is
vo = V b − i o Rf (1.20)
By rule 2, no current can flow into the op amp inputs and so io must equal ii . Substitute ii for io in equation
1.20 and then substitute equation 1.19 for ii , and we have
 
vi − V b
vo = V b − Rf
Ri
Grouping terms with Vb and vi , we have
   
Rf Rf
vo = 1+ Vb − vi (1.21)
Ri Ri
This equation says that the output voltage is composed of two components: the first is a non-inverting gain
times the bias voltage Vb , and the second is the inverting gain times the input signal, vi 5
Now, we could plug two pairs of vi , vo values into equation 1.21 and solve two equations in two unknowns to
determine Vb and the ratio Rf /Ri . Once Rf /Ri is known, one or other of these resistors can be chosen and the
other determined.
However, more understanding may be obtained by inspecting the seesaw diagram, figure 1.20. Notice that
the change in op amp output voltage is entirely determined by the change in input voltage and the inverting gain
Rf /Ri . The bias voltage simply shifts the whole diagram up or down the page. With that observation, we can
solve very simply for Rf /Ri :

Rf
Avinverting = −
Ri
∆Vo
= −
∆V
 i
volow − vohigh

= −
vihigh − vilow
 
0.517 − 4.451
= −
0.742 − 0.515
= 17.3

Now we can substitute 17.3 for Rf /Ri along with a pair of vi vo values in equation 1.21, and solve for Vb .

vo = Vb (1 + 17.3) − 17.3vi
5 The same result could be obtained using the superposition theorem. Determine first the output due to v with V replaced by its internal
i b
resistance (a short circuit). Then determine the output due to Vb with vi shorted. Then add these two outputs together to get the output due to
both Vb and vi .

www.syscompdesign.com
 36

Substitute vi = 0.742 and vo = 0.517, in equation 1.22 and solve for Vb to find that Vb = 0.673 volts. This
coincides with the value shown for Vb on figure 1.20.
We now know the ratio of Rf /Ri , and we still have to calculate their actual values. Their actual values depend
on the operational amplifier bias current and is calculation is deferred to section 1.4.8.

Sourcing the Bias Voltage

A tentative final circuit for the temperature sensor is shown in figure 1.22.

A-D Converter

+5 volts f r r f V
rh

Rd

r r r f V
Ri Rf
AD NAD
HH
r HH
r
Temperature –
R1 H

Sensor
+ 

R2 Vb

Gnd f r r f V
rl
m80
diode-sensor-circuit.epic

Figure 1.22: Temperature Sensor Circuit

The bias voltage Vb is created by a voltage divider from the +5 volt supply. However, there is a potential
problem with this. According to the geometry of the seesaw diagram figure 1.20, any change in the bias voltage
Vb will have a dramatic effect on the A-D input voltage. More precisely, equation 1.21 indicates that changes in
Vb will appear at the output of the amplifier magnified by a factor of 18.48.
However, we found in section 1.2.2 that deriving the reference voltage V RH for the A-D converter from the
same source as the bias voltage caused the A-D reading to be immune to drift or ripple of the supply voltage. Let
us see if the same effect applies here.
The A-D reading R is given by
VAD
NAD = × 255 (1.22)
VRH
where

NAD is the A-D output, 0 to 255

VAD is the input voltage to the A-D converter
VRH is the reference voltage of the A-D converter, the positive supply of the system (+5 volts).

www.syscompdesign.com
Temperature Measurement 1.4 The Diode Temperature Sensor

The output of the op amp is the input to the A-D converter, so we can write V AD for vo in equation 1.21.
   
Rf Rf
VAD = Vb 1 + − vi (1.23)
Ri Ri
Now, if the bias voltage Vb is derived from a voltage divider connected to the voltage source V RH , then it will
be some fraction K of that voltage. For example, in our case Vb has been calculated to be 0.741 volts, and VRH is
+5 volts, K would be 0.741/5 = 0.148. Then we can write

Vb = K × VRH (1.24)
Substituting K × VRH for Vb and equation 1.23 for VAD in equation 1.22, we have

   
255 Rf Rf
NAD = K VRH 1 + − vi (1.25)
VRH Ri Ri
   
Rf Rf
= 255 K 1 + − vi (1.26)
Ri Ri

Since VRH and Vb disappear, the A-D reading is independent of the A-D reference voltage and op-amp bias
voltage. However, the A-D reading is still dependent on the stability of the coefficient K. A stable value of K
should not be difficult to achieve since K depends on the ratio of two resistors.
The actual values of R1 and R2 depends on the operational amplifier parameters and is deferred to section
1.4.8.

1.4.6 The Operational Amplifier

The real skill of engineering is designing to a budget, a maxim that applies to the power supplies in this circuit. If
+/-15 volt power supplies were available, we could simply drop in any operational amplifier and the circuit would
work. However, in our case, we have a single +9 volt unregulated supply, +5 volts regulated and -10 volts at a few
milliamps. Ideally, we should design a circuit that works from these voltage supplies.
The primary limitation is the restricted input and output voltage range of a conventional operational amplifier
such as the 741. When this type of op amp is operated from power supply voltages of +15 V and -15V, the input
and output voltages are typically limited to a range of +/-12 volts. In other words, the input and output voltages
cannot go closer than to within 3 volts of the supply rails. If the op amp is operated from +9 volts and ground, its
input and output would be then limited to the range +3 to +6 volts. In fact, because it is designed to work from
+/-15 volt power supplies, a conventional op-amp may not work at all at the lower supply voltages.
Recognizing this problem, the semiconductor manufacturers have provided us with operational amplifiers
such as the National LM324 which are designed to work from a single low voltage power supply such as +5 volts.
For the LM324, the input voltage can extend from ground to V + − 1.5 volts. In our circuit, both inputs of the
op amp are held at 0.741 volts by negative feedback action, and so the inputs of the LM324 can accomodate this
voltage.
Referring to figure 1.17, we see that the output signal must range between 0.431 volts and 4.35 volts. The spec
sheet for the LM324 shows that the output can go to within half a volt of the positive power supply rail for output
currents of less than 10mA. In our circuit, the output of the op amp drives the input to the A-D converter, which is
effectively an open circuit load. Thus we can expect an adequate output swing when the op amp is operated from
a +5 volt regulated supply.

www.syscompdesign.com
 38

In addition to its ability to operate from a single supply, the LM324 is also attractive for this application
because it is low cost (under $1 in single quantities), includes 4 op-amps per package, and is readily available
from a variety of suppliers.

1.4.7 The effect of Power Supply Ripple and Drift

Modern operational amplifiers have a high degree of Power Supply Rejection Ratio, or PSRR. That is, ripple and
drift on the power supply lines appears as a much smaller equivalent signal at the input to the operational amplifier.
If the PSRR is large, the amplifier has a high degree of immunity to voltage changes on the power supply lines.
The PSRR of an op amp is defined as

∆Vps
PSRR(db) = 20 log (1.27)
∆Vips

where

∆Vps is the variation in power supply voltage ∆Vips is the equivalent variation in input voltage due
to variations in supply voltage

The question immediately at hand is whether the LM324 op-amp has sufficient power supply rejection to
ignore ripple on the power supply line. If it does, we can use the +9 volt unregulated power supply voltage as a
supply for the op-amp. To determine this, we must develop a suitable circuit for analysis of the effect of variations
in the power supply voltage.
In this case, the first step, shown in Step 1 of figure 1.23, is to replace all the voltage sources by their internal
impedances (short circuits).
In step 2, the circuit is rearranged.
In step 3, resistors R1 and R2 have no current flowing through them, so they may be removed. R d is bypassed
by a short circuit, so it may be removed. The power supply circuit is shown explicitly, with its steady component
Vs and its varying component ∆Vps .
Step 4 shows the equivalent input voltage vips , due to the variation in supply voltage ∆Vps . The vips voltage
source could be shown in series with either input: the result is the same. Putting it in series with the inverting
input is a little easier to analyse.
This final equivalent circuit is a non-inverting amplifier configuration, so the output voltage due to power
supply variations is:
 
Rf
vops = vips 1 + (1.28)
Ri
Combining equations 1.27 and 1.28 we have
 
∆Vp s Rf
vops = 1+ (1.29)
P SRR Ri

Now for an example: If the power supply ripple is 1 volt peak-peak and the power supply rejection ratio is 65
db (the worst case value given on the LM385 data sheet), calculate the ripple in the op-amp output signal.
In section 1.4.5 we found that Rf /Ri = 17.3 volts/volt. As well, using equation 1.27 a PSRR of 65 db is
equivalent to a PSRR ratio of:

www.syscompdesign.com
Temperature Measurement 1.4 The Diode Temperature Sensor

d q q d

Rd

q q q d q q q d
Ri Rf Ri Rf

d d
+5 volt
H H
q q
Supply
– H – H
H H
d d
R1
Temperature
 
 
Sensor + Rd +

d q q d d q q q q d
R2 Vb R2 R1

Gnd m50 Gnd m50

psrr-equivalent-a.epic psrr-equivalent-b.epic

(a) Step 1 (b) Step 2

q q d q q d
Ri Rf Ri Rf
∆Vsupply

H
– H
H
– H
H H
 
Vsupply Vos

 
+ +

q q q d q d
Vis

m50 m50
psrr-equivalent-c.epic psrr-equivalent-d.epic

(c) Step 3 (d) Step 4

Figure 1.23: Power Supply Rejection Ratio Equivalent Circuit

 
65
P SRR = 10 20
= 1778 volts/volt

Now use equation 1.29 to find vos :

 
∆Vp s Rf
vops = 1+
P SRR Ri
1.0
= (1 + 17.3)
1778
= 10.3 millivolts peak-peak

Each 19.6 millivolt change in the output is equivalent to 0.5oC, so 10.4 millivolts is equivalent to a ripple in
the reading of about 0.25oC.
A range of line voltage stability is +/-10%, so we should expect an unregulated 9 volt power supply to vary by

www.syscompdesign.com
 40

+/-0.9 volts for a total variation of 1.8 volts. Obviously, this will induce too large a change in the output voltage
of the op-amp, so the op-amp power supply must be regulated.
Incidentally, the typical value of PSRR for the LM324 is 100db, a PSRR factor of 10 5 . With this PSRR, the
output voltage swing due to a power supply variation of 1 volt would be 0.18 mV, which is acceptable. A designer
who lashed up the circuit and tested it in the lab might well happen to use an LM324 with a typical value of PSRR,
thereby finding an acceptable output variation due to power supply fluctuations. One has to expect in a production
run, however, that some LM324’s will be found with a minimum value of PSRR and they will not work correctly.
There are two important lessons from this:

1. Always use the worst case specification in design. Never, never design with a typical value.

2. A working model of the circuit is a necessary but not sufficient condition for the circuit to work in quantity.
A lab prototype is an exercise to test the design theory. If the prototype doesn’t work, the theory is obviously
faulty. If it does work, the circuit will work in production only if production units can function with a full
range of component tolerances.

Finally, we note that the power supply rejection ratio of the 741 op-amp, a conventional design, has a power
supply rejection ratio of 86 db(minimum). This is an order of magnitude better than the minimum PSRR of the
LM324. Perhaps a lower PSRR is something one has to accept with single-supply op-amps.

1.4.8 The Effect of Bias Currents

The input stage of the LM324 is a bipolar PNP differential amplifier, as shown in figure 1.24.

e
r r
Vcc

m m
r r
Inverting e r eOutput
e r
Input

r r
Non-inverting

r
Input

r r r
m100
opamp-internal.epic
e

Figure 1.24: LM324 Internal Circuitry (Simplified)

As a result, a small amount of base current, the bias current will flow out of the input terminals of the op-amp.
For operational amplifiers with FET inputs, the bias current is in the order of 100 picoamps, which is small enough
to be neglected in most cases. In the case of the LM324, the bias current is 250nA (maximum), which is large
enough to have an effect on the operation of the circuit. Fortunately, the effect of bias current can be minimized by
reducing the size of the resistances through which the bias current flows. However, reducing resistances increases
the power consumption of the circuit and adds to the heat dissapation, so resistors should be as large as possible.

www.syscompdesign.com
Temperature Measurement 1.4 The Diode Temperature Sensor

r r e
Ri Rf

r
m H
r
– H
i−
b HH
m + 
 vob

r
i+
b 

e r r r e
R1 k R2

Gnd m70
opamp-bias-current.epic

Figure 1.25: Operational Amplifier With Bias Current

A op-amp circuit model that incorporates the effect of bias current is shown in figure 1.25. As in figure 1.18,
we have simplified and combined the external circuit to determine the effect of the bias currents on the output
voltage.
The voltage at the non-inverting input terminal of the op-amp is

v + = i+
b (R1 k R2 ) (1.30)

The op amp drives the inverting input terminal to the same voltage. Then the current i i through Ri is

v−
ii = (1.31)
Ri
v+
= (1.32)
Ri
= i+
b (R1 k R2 ) (1.33)

If i+ +
b is equal to ii , then if = 0 and vob = v . The practical implications are that

• The value of R1 k R2 should be kept to a minimum to minimize the voltage generated by the bias current.

• The value of Ri should be made equal to R1 k R2 to minimize the value of vob .

Putting numbers to this, let us choose the acceptable output due to the op-amp bias currents at equivalent to
one quarter of a degree centigrate, or about 0.5 mV. For the LM324, the bias current is 250 nA maximum. Then
R1 k R2 should be

0.5 × 10−3
R1 k R 2 = (1.34)
250 × 10−9
= 2KΩ (1.35)

www.syscompdesign.com
 42

In section 1.4.5, we determined that the divider ratio should be 0.148. Then

R2
= 0.148 (1.36)
R1 + R 2

Solving equations 1.35 and 1.36 for the resistor values, we obtain

R1 = 13.5KΩ
R2 = 2.35KΩ

Since these exact values are not available and since the bias voltage should be adjustable, we use a poten-
tiometer in the center to adjust the voltage exactly to 0.741 volts, as shown in the figure 1.26.

13KΩ

13.5KΩ 0.849V

5V 5V 500Ω
0.741V
2.35KΩ 0.691V
0V 2K2Ω
m100
0V
final-divider.epic

Figure 1.26: Bias Voltage Divider Design

1.4.9 The Effect of Offset Current

The offset current is the difference between the two input bias currents. In the case of the LM324, the bias currents
can differ by as much as 50 nanoamps. In the previous analyis of the effect of bias current, we assumed that the
two currents were equal. If they are not, we will show that the offset current flows through R f and causes a change
in the output voltage of Ios × Rf volts. As a result, the output voltage more sensitive to offset current when the
amplifier gain is large.
The appropriate equivalent circuit is again figure 1.25. This time, however, the two bias currents i + b and ib
−

are no longer equal.

To develop an expression for vo , we call R1 k R2 an equivalent resistor Req , and develop the following
equations:
The bias current out of the non-inverting terminal generates a voltage given by

v + = i+
b Req (1.37)

www.syscompdesign.com
Temperature Measurement 1.4 The Diode Temperature Sensor

ii if

r r e
Ri Rf

r
m H
v−

r
– H
i−
b HH
m 
 vo

r
+
i+
b 
v+

e r r r e
Req

Gnd m70
opamp-offset-current.epic

Figure 1.27: Offset Current Calculation

The op-amp forces the inverting and non-inverting inputs to be at the same potential, so

v+ = v− (1.38)

The current through Ri is given by

v+
ii = (1.39)
Ri
At the virtual earth point, the currents are given by

if = i i − i −
b (1.40)

The output voltage vo due to the bias current is given by

vo = v + + i f Rf (1.41)

Combining equations 1.37 through 1.41, to solve for v o in terms of the resistances and bias currents, we have

Req Rf
vo = i + +
b Req + ib − i−
b Rf (1.42)
Ri
If we make Req equal to Ri as suggested in section 1.4.8, then we can simplify equation 1.42 to

vo = i + + −
b Req + (ib − ib )Rf (1.43)
Substituting the offset current ios for i+ −
b − ib , we have

vo = i +
b Req + ios Rf (1.44)

www.syscompdesign.com
 44

This value for output voltage is the same result we obtained in section 1.4.8 for bias current, augmented by
the term ios Rf .
In our application, the bias current is 250 nanoamps maximum and the bias current 50 nanoamps maximum.
Req and R1 are each is 2KΩ and Rf is 34.96KΩ. Then the worst-case output voltage due to bias and offset
currents will be

vo = i+
b Req + ios Rf
= 250 × 10−9 × 2000 + 50 × 10−9 × 34.96 × 103
= 2.25 × 10−3

This is at the output of the op-amp and input of the A-D converter, where each degree of temperature is a
change of 38mV. Thus the voltage created by the bias and offset currents is a small percentage of one degree.
In applications where the operational amplifier must be extremely stable, the drift of offset current with tem-
perature may be of concern. For the LM324, the offset current drift specification is 10 picoamps/ o C. In our case,
an operating temperature range for the op-amp of 25oC +/-10oC would result in a change of 0.200 nanoamps. In
our design, this is small enough to be ignored.

1.4.10 The Effect of Offset Voltage

The input stage of an operational amplifier consists of a differential pair of transistors. The base-emitter voltages
of the two transistors are not perfectly matched. The difference between the two base-emitter voltages creates an
effective input voltage, called the offset voltage. In this section, we examine the effect of the offset voltage, and
the drift of offset voltage with temperature, on the accuracy of the circuit.
The equivalent circuit of an op-amp with offset voltage, in our simplified circuit for analysis, is shown in
figure 1.28.

ii if

r r e
Ri Rf

v−
H
– H
HH

vo


+
vos

e r r e
Req

Gnd m70
opamp-offset-voltage.epic

Figure 1.28: Operational Amplifier With Offset Voltage

The analysis of the circuit is quite similar to previous work:

www.syscompdesign.com
Temperature Measurement 1.4 The Diode Temperature Sensor

Since there is no current into the non-inverting terminal of the op-amp, the voltage at the non-inverting input
v + is equal to the offset voltage vos .
v + = vos (1.45)

The op-amp negative feedback action drives the inverting input terminal to the same voltage as the non-
inverting terminal.

v+ = v− (1.46)

The current through Ri is

v+
ii = (1.47)
Ri

Since, in this analysis, there is no current into the inverting input terminal of the op-amp, the current through
Rf is equal to Rf .
if = i i (1.48)

Finally, the output voltage vo is equal to the voltage at the inverting input terminal plus the voltage across the
feedback resistor.
vo = v − + i f Rf (1.49)

Putting equations 1.45 to 1.49 together, the output voltage is

 
Rf
vo = vos 1+ (1.50)
Ri

In words, the output voltage is equal to the offset voltage times the non-inverting gain of the op-amp circuit.
In our case, the non-inverting gain is 18.48 and the offset voltage of the LM324 is +/-9mV, so the output due to
the offset will be
 
Rf
vo = vos 1 +
Ri
= +/ − 9 × 10−3 × 18.48
= +/ − 166mV

At first glance, this is somewhat alarming, since it represents a temperature change of +/-9 o C. However, since
Vbe appears in series with the adjustable bias voltage source Vb , it is possible to compensate for the offset voltage
by adjusting the bias in the opposite direction.
However, the drift in op-amp offset voltage with temperature is not so easily disposed of. The LM324 offset
voltage drift is specified as 20µvolts per o C. Over a 20o C operating temperature range for the op-amp, the offset
voltage will then change by 400µvolts. Using equation 1.49, we determine that the total drift in voltage at the
input to the A-D converter, will be 7.4mV, equivalent to about half a degree centigrade.
Sneaking a longing glance at the typical figure for offset voltage drift for the op-amp (7µvolts per o C, a third
of the worst-case value), we decide that this is marginally acceptable.

www.syscompdesign.com
 46

1.4.11 Drift in Resistor Values

To this point, we have focussed on the drift characteristics of the operational amplifier. However, even the resistors
in circuit do change their value with temperature. Carbon film resistors have a typical temperature coefficient of
-250ppm (parts per million). Metal film resistors are much more stable, at around +/-20ppm.
In our circuit, the critical arrangements of resistors are R1 and R2 , which establish the bias voltage, and Ri
and Rf , which establish the voltage gain of the op-amp. Referring to equation 1.26, we can see that in both cases
the key quantities are the ratio of two resistor values. If the two resistors change by the same percentage amount,
there is no effect on their ratio, and the circuit is unaffected.

1.4.12 Estimate of Accuracy

1.5 Appendix: Operational Amplifier, Static Errors

In this section, we summarize the operational amplifier characteristics that may cause errors at zero and low
frequencies. The definitions are from reference [27].

1.5.1 Input Offset Voltage

Definition
That voltage which must be applied between the input terminal resistance through two equal resistances to obtain
zero output voltage.

Equivalent Circuit (figure 1.29)

... ...
. .. ....
. ..
... ...
.....
.. .....
. .....
... ...
Vos ....... .. .....
.......
... ........ ..... ..
- ....... .
... - ....... .. .....
.......
....... ..... ..
... ...... ........
.......
.. ....... ....
. + ................ . ..... .
... + ........
. ...
....
....
... ....... ..... .
.....
.....
.... .....
....
... .. .
. .. ...
...... ...

Figure 1.29: Operational Amplifier Offset Voltage

Notes
• Some operational amplifiers include nulling terminals which can be connected to an external pot to reduce
the offset voltage to zero.

• Offset voltage appears at the output multiplied by the voltage gain of the amplifier, so it is especially
significant for high-gain voltage amplifiers.

www.syscompdesign.com
Temperature Measurement 1.5 Appendix: Operational Amplifier, Static Errors

Application Circuit (figure 1.30)

r
Ri Rf
. .. . .
....... .................... .....................
.. . . .
..... .......
.. ..
... ...
. .. ....
... . ..
...
.....
... .....
.....
.. ...
. Vos ....... .. .....

r
.......
... ....... ..... ..
.......
... - - ....... ... .....
.......
Vi ....... ..... ..
... ...... ... .......
....... .....
....... ....
.... + ............... . ..... . ....
...+ ...... .... .....
....... ...
.
... ....... ..... ..
. .....
.....
... ...
.. Vo
....
... .. .
. ..... ...
... . ..... .....
.... ....
....... ........
. . . . . . . . .
.... .... .... .... .... .... .... .... ....

Figure 1.30: Operational Amplifier Offset Voltage Application

The output voltage for this circuit is

 
Rf Rf
Vo = V i + Vos 1 + (1.51)
Ri Ri
 
Rf Rf
The term Vi R is desired: Vos 1 + R is an error term.
i i

1.5.2 Input Bias Current

Definition
The average of the two input currents.

Equivalent Circuit (figure 1.31)

Notes
• Most BJT operational amplifiers have much higher bias currents than JFET or MOSFET operational am-
plifiers. Bias current creates an effective input offset voltage proportional to the resistances in the input
leads.
• The effect of bias current may be cancelled by equalizing the resistances in the two input leads.

Application Circuit (figure 1.32)

The output voltage for this circuit is
 
Rf Rf
Vo = V i + Ib (Rf k Ri ) 1 + (1.52)
Ri Ri

www.syscompdesign.com
 48

r
m H
-

r
– H
i−
b HH
m 


r
+
i+
b 
+

m70
opamp-ibias-equivalent.epic

Figure 1.31: Operational Amplifier Bias Current

The second term in the equation is the bias-current induced error term, proportional to bias current, magnitude
of the input resistances, and the voltage gain.

1.5.3 Input Offset Current

Definition
The difference between the two input currents.

Equivalent Circuit (figure 1.33)

Notes
• Offset current may be minimized be selecting an opamp with small bias current. The usual problem is not
the magnitude of offset current but its change with temperature, which appears in the output as voltage drift
with temperature.

Application Circuit (figure 1.34)

The output voltage for this circuit is
 
Rf Rf
Vo = V i + Ios (Rf k Ri ) 1 + (1.53)
Ri Ri

As in the case of bias current, the second term in the equation is the offset-current induced error term, propor-
tional to offset current, magnitude of the input resistances, and the voltage gain.

1.5.4 Power Supply Rejection Ratio

Definition
The ratio of the change in input offset voltage to the change in power supply voltages producing it.

www.syscompdesign.com
Temperature Measurement 1.5 Appendix: Operational Amplifier, Static Errors

b r r b
Ri Rf

r
m H
r
– H
Vi
i−
b HH
m + 
 Vo

r
i+
b 

b r r b
m70
opamp-ibias-application.epic

Figure 1.32: Operational Amplifier Bias Current Application

Equivalent Circuit (figure 1.35)

Notes
• Power supply rejection is an issue when the power supply shows a large variation or large ripple content.
The change in power supply voltage may show up as drift in the output voltage of the op amp.
• Power supply rejection is usually expressed in decibels, so it must be converted to a voltage ratio before
use.

Application Circuit (figure 1.36)

The output voltage for this circuit is

P SRR = 20 log10 P SR (1.54)

where PSR is the rejection ratio in volts/volt, PSRR is the ratio in decibels.

Vs
Vps = (1.55)
P SR  
Rf Rf
Vo = Vi + Vps 1 + (1.56)
Ri Ri

The second term in the equation is the power supply induced error term, proportional power supply ripple,
inversely proportional to the power supply rejection, and proportional to the voltage gain.

www.syscompdesign.com
 50

r
m H
-

– H
ios
HH


+

m70
opamp-ios-equivalent.epic

Figure 1.33: Operational Amplifier Offset Current

b r r b
Ri Rf

r
m H
– H
Vi
ios
HH

Vo


+

b r r b
m70
opamp-ios-application.epic

Figure 1.34: Operational Amplifier Offset Current Application

www.syscompdesign.com
Temperature Measurement 1.5 Appendix: Operational Amplifier, Static Errors

mVs
Vps
HH
- - HH Vcc


+
+

Figure 1.35: Op Amp Power Supply Rejection, Equivalent Circuit

b r
Ri Rf

mVs
Vps
HH
HH r b
- -
Vi

Vcc


+
+

Vo

b b
r

Figure 1.36: Op Amp Power Supply Rejection, Application Circuit

www.syscompdesign.com
 52

1.5.5 Software for the Diode Temperature Sensor

The temperature sensor interface is responsible for amplifying and level shifting the diode voltage so that it is a
good fit to the range of the microprocessor analogue-digital converter. Once the diode signal is converted to a
binary number, the microprocessor software is responsible for generating the appropriate display readout.
Diode voltage is essentially a linear function of temperature, so the relationship between A-D reading and
display readout is also linear, and this simplifies the design of the temperature software.

A-D Converter Display Software

Reading
Transistor
Temperature
Sensor M
Temperature +27 Deg C
Analogue Interface
B
Display Readout

A-D Reading

Figure 1.37: Diode Temperature Sensor Block Diagram

If we are only constructing one unit, we could generate a lookup table in which the index into the table is the
A-D reading and the display value is the table entry. A-D readings can have 256 different values, so this fixes the
length of the table. To save work, the table may be generated using a BASIC program on a P/C, and then edited
into the assembly language source code.
If we are constructing a small number of these units, the analogue section would be adjusted, using poten-
tiometers, so that each analogue interface has the same voltage gain and offset. Then the same lookup table could
be used for each unit.

Calibration in Volume Production

Now consider that we are producing a large number of these units on an assembly line. Potentiometers in the
analogue circuit are undesireable because the pot and the labour to adjust the pot are both expensive. If possible,
we would like to use fixed resistors in the analogue interface section and adjust the software to suit. This is all the
more appealing because the 68HC11 microprocessor has EEPROM (Electrically Eraseable Programmable Read
Only Memory). Calibration constants can be written into EEPROM at the time the unit is manufactured, and they
become essentially a permanent part of the program.
Picture the following automatic calibration setup:

• Thermometer units are delivered by conveyer line to the calibration station. At the station, a calibration
computer plugs into the 68HC11 microprocessor. The computer touches the diode temperature sensor with
a hot probe of known temperature and notes the HC11 A-D reading. It then repeats the process with a cold
probe of known temperature.

• The calibration computer now knows two points on the function relating A-D reading and sensor tempera-
ture. The calibration function is linear, so the calibration computer can calculate the slope M and offset B
constants of the calibration function, figure 1.37.

www.syscompdesign.com
Temperature Measurement 1.5 Appendix: Operational Amplifier, Static Errors

• Finally, the calibration computer writes the M and B constants into the EEPROM of the 68HC11 micro-
processor. The thermometer unit is calibrated.

Transfer Function Calculation

The software to display temperature must solve equation 1.57.

Td = M × NAD + B (1.57)
where Td is the displayed temperature reading, M and B are the calibration constants, and N AD is the A-D
reading.
This could be accomplished using integer arithmetic: the equation is not particularly complex and the range
of numbers is limited so overflow and underflow are not likely to be a problem. However, in subsequent sections,
we will make use of the temperature value to calculate dew point and humidex readings, and floating point
calculations are much easier to use. To be consistent, then, it probably makes sense to use floating point routines
to solve this equation.

www.syscompdesign.com
 54

1.5.6 Checking the Production Design

The fact that a prototype of this circuit works is, as mathematicians put it, a necessary but not sufficient condition
that the design is correct. If the prototype fails, the design is flawed. If it works, the design may or may not be
satisfactory.
To ensure that hundreds of this temperature measurement circuit can be produced and will work reliably, we
must do a proper engineering design. This is the work that distinguishes a production design engineer from a
hobbiest or scientist who needs one working circuit. The design must work in spite of the expected variations in
component specifications.
Resistors, transistors, op-amps – all vary in their specifications and we may have gotten lucky in the prototype
circuit, using components with values favorable to the design. In a production situation, we must assume that the
components will vary through their worst case values.
There are many computer-based tools available to help with the analysis of electronic circuits: reference [97]
has an overview. However, the computer spreadsheet is readily available on computers these days, and easy to use.
Once the spreadsheet formulae are entered, the spreadsheet becomes an electronic model for the circuit behaviour
and component values may easily be tweaked to determine their effect on the circuit, figure 1.38. As well, one can
simulate circuit operation over the full specified temperature range (-40 to +60 o C) without access to a temperature
testing chamber of this range.

Component Values

Sensor Displayed
Temperature Temperature
Ts Td

Figure 1.38: Checking Component Variation

In the case of this particular circuit, the software calibration process will take care of minor component vari-
ations. A more serious concern is that the amplifier may saturate or cutoff (the output may approach too closely
either Vcc or ground), in which case it will cease to function correctly. If possible, we’d prefer to use 5% resistors
because of their lower cost, compared to 1% resistors, so an important question is whether the circuit will function
correctly with 5% resistors.
There are 4 resistors in the circuit. The amplifier may saturate or cutoff at either high or low temperature.
Thus, even in the case of this simple circuit, ignoring possible variations in the sensor temperature coefficient and
offsets in the operational amplifier, there are 8 possible situations to examine. Fortunately, the spreadsheet makes
it simple.

1.5.7 Spreadsheet Model

A model for the diode temperature sensor circuit and software is shown in figure 1.39.

www.syscompdesign.com
Temperature Measurement 1.5 Appendix: Operational Amplifier, Static Errors

Sensor Temperature
Ts
.... ....
.
Reference Sensor Voltage ....
.
....
Vbe25 = 0.600 Temperature
Vbe = Vbe25 + Vbe T C · (Ts − 25)
Sensor
Sensor Temperature Coeff ....
....
Vbe T C = −2.2 mv/o C

....
.
R1 ....
Bias R2
Sensor Voltage
Vb = R1 +R2 · Vcc Vi = Vbe
Voltage ....
R2 ....

Bias Voltage
Vb
.... ... .... ...
. .
....
.
Rf .... Amplifier
Rf Rf
....
VAD = Vb (1 + Ri ) − Vi Ri
.
Ri ....

.... ....
.

....
. VAD A/D
VRU = Vcc .... NAD = · 256
VRU Converter

.... ....
.
....
.
M ... Software
....
TD = M · NAD + B
.
B ....

Temperature Display .... ...

.
TD

Figure 1.39: Diode Temperature Simulation Block Diagram

The first step in using the spreadsheet is to enter explanatory headings (the left column in table 1.1). Next,
give the cells descriptive names, if possible. (Unfortunately, as in the case of the Microsoft Excel spreadsheet,
the exact name may have to be some variant of what we have been using. For example, the name R1 is illegal for
some reason, and we have to use something like R_1 instead. However, this is better than the generic name for a
cell, like B13.)
Next, the enter the model formulae into the spreadsheet as shown in right column of table 1.1.
It may be useful to set the format specification for individual cells on the spreadsheet. For example, if the
temperature shown on the LCD with one decimal point, the corresponding cell (last row, second column, table
1.1) should be set to the same number format.
Once the formulae are in place, the first tests should be used to determine that the model is producing credible
results. (Our first attempt uncovered an arithmetic error in the original design. Whoops.)

www.syscompdesign.com
 56

Temperature Sensor Simulation

Description Constants and Formulae
Power Supply Voltage, volts 5

Temperature Sensor
Sensor Temperature Ts , Degrees C 60
Sensor Reference Temperature Tref , Degree 25
Vbe at 25 Degrees, Vbe25 , volts 0.595
Sensor Temperature Coef, VbeT C , mv/Deg C -2.265
Sensor Voltage Vs , Volts +Vbe25 + (Ts − Tref ) × (VbeT C /1000)

Bias Supply
R1, Ohms 13000
R2, Ohms 2200
Bias Voltage Vb +Vcc × (R2/(R1 + R2 ))

Amplifier
Input Resistor Ri, Ohms 6800
Feedback Resistor Rf, Ohms 120000
Amplifier Output Voltage Vo , Volts +Vb × (1 + (Rf /Ri)) − Vs × (Rf /Ri)

A/D Conversion
Output Count, NAD IN T ((Vo /Vcc ) × 256)

Software
Function slope constant M 0.5
Function offset constant B -52
Displayed Temperature Td M × NAD + B

Table 1.1: Spreadsheet Simulation Formulae and Data

When the spreadsheet is finally debugged, the model can be used to generate some useful results. A typical
calculation run is shown in table 1.2.
This table has been modified so that the resistors R1,R2,Ri and Rf are at their extreme tolerance values (±1%)
and the temperature is at its maximum specification, 60o C. The software constants M and B have been adjusted
so that the system is calibrated: the Displayed Temperature tracks the Sensor Temperature. Most importantly,
the model shows that the amplifier output voltage is 4.705 volts under these conditions, within its allowable limit.

Results

Out of this simulation, we found that 1% resistors and an LMC660 op amp will work in the circuit. On the other
hand, 5% resistors would not work: the amplifier saturated and cut off under high and low temperature extremes.
We also learned that the LM358 bipolar single supply amplifier will not work correctly in the circuit: its output

www.syscompdesign.com
Temperature Measurement 1.5 Appendix: Operational Amplifier, Static Errors

Temperature Sensor Simulation Constants and

Description Calculated Results
Power Supply Voltage, volts 5

Temperature Sensor
Sensor Temperature Ts , Degrees C 60
Sensor Reference Temperature Tref , Degree 25
Vbe at 25 Degrees, Vbe25 , volts 0.595
Sensor Temperature Coef, VbeT C , mv/Deg C -2.265
Sensor Voltage Vs , Volts 0.516

Bias Supply
R1, Ohms (-1% tolerance) 12870
R2, Ohms (+1% tolerance) 2222
Bias Voltage Vb 0.736

Amplifier
Input Resistor Ri, Ohms (-1%) 6732
Feedback Resistor Rf, Ohms (+1%) 121200
Amplifier Output Voltage Vo , Volts 4.705

A/D Conversion
Output Count, NAD 240

Software
Function slope constant M 0.48
Function offset constant B -55
Displayed Temperature Td 60

Table 1.2: Spreadsheet Simulation Data and Results

voltage limits at +3.5 volts.

One final word. The spreadsheet simulation is convenient and a useful tool for exploring different scenarious.
However, it is only as accurate as the original model. In this example, we have totally ignored the effect of the
op-amp offset voltages and currents. If these effects are significant, they must be accounted for in the model.
Put another way, the spreadsheet or other simulation complements but is not a substitute for a working proto-
type, the necessary but not sufficient condition of a well-engineered circuit.

www.syscompdesign.com
 58

1.6 Integrated Circuit Temperature Sensors

The diode-connected transistor is a useful temperature sensor by itself but it may be enhanced with additional
circuitry. The integrated circuit temperature sensor incorporates this additional circuitry on chip to

• Make the device less dependent on process variations

• Amplify the output signal to a more useable level

• Compensate for the non-linearity of the temperature-Vbe characteristic

The result is a temperature sensor which is less likely to require calibration, requires less or simpler external
circuitry, and is more accurate over its operating range. The cost of the integrated circuit temperature sensor is
in the order of $3.50 vs a transistor temperature sensor at $0.35, approximately an order of magnitude. Thus,
for all but the most cost-sensitive applications, the integrated circuit temperature sensor is an attractive choice for
measuring ambient temperature.

1.6.1 National LM135 Precision Temperature Sensor

Throughout this section, we will use the National LM335AZ, a variant on the LM135 series of devices, as an ex-
ample. The LM335AZ behaves as a temperature dependent zener diode in which the zener voltage is proportional
to the absolute temperature at a rate of 10mV/oK.
The performance specifications of the National LM335AZ are as follows:

Temperature Range –40 to +100oC

Operating Current Ir = 1.0 mA
Parameter Condition Min Typ Max Units
Uncalibrated Temperature Error 25o C, – 1 3 o
C
o
Uncalibrated Temperature Error Tmin < Tc < Tmax – 2 5 C
o
Calibrated Temperature Error Tmin < Tc < Tmax – 0.5 1 C
Non-linearity – 0.3 1.5 %

From the graph of Calibrated Error vs Temperature, it would appear that the error increases in a linear fashion
over the temperature range. Thus, it would appear that an adjustment of scaling factor in the external circuit
would improve the accuracy to that of the sensor non-linearity. As well, the accuracy of the sensor improves
around normal ambient temperatures.
The basic application circuit for this device is shown in figure 1.40(a). A potentiometer may be added as
shown in figure 1.40(b) to calibrate the device at 25o C.
At 25o C, the sensor will sustain a voltage of

Vsensor = Tabs × 10−3

= (273 + 25) × 10−3
= 2.98 volts

To bias the device the recommended current of 1mA from a power supply of +5 volts, the series resistor should
be

www.syscompdesign.com
Temperature Measurement 1.6 Integrated Circuit Temperature Sensors

R1
s s h
R1
s h
R2
Output
LM335AZ 10KΩ 10mV/o K
Output
LM335AZ s s h
s h
10mV/o K

m100 m100
icts-basic.epic icts-calibrate.epic Calibrate for 2.982mV at 25o C

(a) Basic Temperature Sensor (b) Sensor With 25o C Calibration Pot

Figure 1.40: Integrated Circuit Temperature Sensor

Vcc − Vsensor
R1 =
Isensor
5 − 2.98
=
1 × 10−3
= 2.02KΩ

1.6.2 Generating the Temperature-Sensitive Voltage

Consider the circuit of figure 1.41.
A single diode connected transistor Q1 is driven from a current source I c . An equal current is fed to a triple
of parallel diode-connected transistors, Q2 ,Q3 and Q4 .
Taking the logarithm of both sides of equation 1.11 and solving for the base-emitter voltage V be , the voltage
across Q1 is given by  
kT Ic
Vbe = ln (1.58)
q Is
If the transistors Q2 , Q3 and Q4 are identical, then the current Ic will split equally among them and each
transistor will conduct one-third of the total. As a result, the base-emitter voltage will be reduced by some
amount, as indicated in figure 1.42.
Then, using equation 1.58 again, the base-emitter voltage V be0 is
 0
kT I
Vbe = ln c (1.59)
q Is
 
kT Ic 1
= ln × (1.60)
q 3 Is

www.syscompdesign.com
 60

r
 
Ic
 Ic

r r
r r r r
∆Vbe

0
Vbe
Q1 Q2 Q3 Q4

r r r
m100
icts-diode.epic
Ic0 Ic0 Ic0

Figure 1.41: Temperature Coefficient Circuit

The voltage difference ∆Vbe between the two base-emitter voltages is then

   
kT Ic kT Ic
∆Vbe = ln − ln (1.61)
q Is q 3Is
 
kT Ic Ic
= ln − (1.62)
q Is 3Is
 
kT Ic 3Is
= ln × (1.63)
q Is Ic
kT
= ln 3 (1.64)
q
We have the happy event that the collector current Ic and the saturation current Is cancel out of the equation
and the value of ∆Vb e is solely dependent on the absolute temperature.
In practice, ten transistors might be paralleled, in which case at 25 o C
kT
∆Vbe = ln 10 (1.65)
q
1.38 × 10−23 × 298
= × ln 10 (1.66)
1.6 × 10−19
= 58.2 mV (1.67)
Expressed as a coefficient, this is
58.2 × 10−3 V
∆Vbe = (1.68)
293oK
' 0.2 mV/o C (1.69)

1.6.3 Amplifying the Temperature-Sensitive Voltage

The 0.2mV/oK signal is much too small for most practical applications and must be substantially amplified before
being used. The circuitry of the LM158 has two purposes: it multiplies the signal by a factor of 50 to 10mV/ oK

www.syscompdesign.com
Temperature Measurement 1.6 Integrated Circuit Temperature Sensors

Ic

Ic
3

0
Vbe Vbe
m80
vbe-curves.epic

Figure 1.42: The Effect of Varying Collector Current

and equalizes the currents in the two sets of transistors.

A block diagram of the basic circuit is shown in figure 1.43.

+ r r

r
HH
r f
HH
r
-
H

+ 
Ic Ic0 
r
26R

r r r ∆Vbe Vo = 50∆Vbe
r
R

r r r
···

r r r r f
23R

m80
icts-amplifier.epic

Figure 1.43: Temperature Sensor Amplifier

www.syscompdesign.com
 62

The voltage ∆Vbe is developed between the base of a single transistor on one side and a collection of 10
parallel transistors on the other side.
The collector resistors Rc translate the two collector currents into voltages, which are then compared in the
operational amplifier. If they are not equal, the output of the op-amp increases or decreases the voltage across the
R-49R resistor string. This change in base-emitter voltage of the transistors changes the collector currents in such
a direction as to minimize the original difference in current.
When the system settles down, the voltage difference at the input terminals to the op amp must be close to
zero, and the value of ∆Vbe must be such as to make the two collector currents equal. Because of the divider
action of the resistor string, the output voltage must be 50 × ∆Vbe .

1.6.4 Integrated Circuit Temperature Sensor Interface Circuit

The output of the integrated circuit voltage sensor is 10mV/oC, which must be amplified to provide a sufficient
input signal to the A-D converter. The requirements of this circuit are similar to that of 1.4.5 for amplifying the
signal from a diode-connected transistor sensor.
Designing for the same temperature range (-40oC to +60o C) and redesigning that circuit to accomodate the
LM335 temperature sensor, we have the result shown in figure 1.44.

Ratiometric
A-D Converter

+5 volts f r r f V
rh
R1
30K
Rsensor
2K2 Ri Rf

r r r f V
30K6 120K
AD NAD
HH +5V
– HH
LM335AZ H
Rcal R3 + 
10KΩ 5K  LM358

R2

fr r r r f V
36K
Gnd rl

Calibrate for 2.982mV at 25o C

m80
icts-voltage-circuit.epic

Figure 1.44: Temperature-Voltage Interface Circuit

www.syscompdesign.com
Temperature Measurement 1.7 Temperature-Current Transducer

The calculation of the components values of this circuit follows the development of section 1.4.5 and is left as
an exercise.

1.7 Temperature-Current Transducer

When we design a high-resolution temperature measuring circuit, it will turn out to be useful to have a temperature
sensor with output in the form of a current, rather than voltage, proportional to temperature.
In section 1.6.2, we showed that the difference between base-emitter voltages of two groups of transistors is
proportional to absolute temperature when the two groups of transistors are operated at equal collector currents.
The temperature-current transducer concept is a modification of this basic idea, as shown in figure 1.45 [30].

s
 
 
Ic Ic

s s

s
Q2 Q3 Q4
s
Q1 s s

R ∆Vbe
s

m100
current-temperature-concept.epic

Figure 1.45: Temperature-Current Transducer Concept

Transistor Q1 and transistor group Q2, Q3 and Q4 are operated at the same collector currents and so have
different base-emitter voltages. The difference between these two voltages now appears across resistor R. Since
this voltage difference is proportional to temperature, the current in R must also be proportional to temperature.
Voila, temperature to current transducer.
A simple method of ensuring that the two collector currents are equal is shown in figure 1.46. The current
mirror Q5, Q6 ensures that I1 is equal to I2 .
To develop an equation for the output current, consider the voltage loop comprised of Q1, Q2 and R. From
equation 1.64, the voltage VR across the resistor R will be

kT
∆Vbe = ln N (1.70)
q
where N is the number of parallel connected transistors, 3 in this case.

www.syscompdesign.com
 64

f
r

r
r
Q5 Q6

f
I1 I2

r r

r 
Q2 Q3 Q4

r f
Io ∝ T
r r
Q1

r
R ∆Vbe

f
Io ∝ T

m80
current-temperature-sensor.epic

Figure 1.46: Temperature-Current Transducer

The voltage across resistor R is also given by

+vR = I2 R (1.71)

Combine equations 1.70 and 1.71 and we have

1 kT
I2 = ln N (1.72)
R q
Since I1 and I2 are equal, then

Io = 2 × I2 (1.73)
2 kT
= ln N (1.74)
R q
From equation 1.74, we can see that the output current Io is proportional to absolute temperature T and that
the current scale factor may be adjusted by adjusting resistor R.

1.7.1 The National LM134 Temperature Sensor

The LM134 [31] is primarily intended to be used as an adjustable current source. As shown in figure 1.47, once
the voltage across the device exceeds about 1.2 volts, the LM134 conducts a current which is independent of the
voltage.

www.syscompdesign.com
Temperature Measurement 1.7 Temperature-Current Transducer

Io ∝ T
f

f
LM 134

f
Rset

s
m80
lm134-circuit.epic

Figure 1.47: LM134 Adjustable Current Source

The value of current is given by

(227µV /o K)(T )
Iset = (1.75)
Rset
where T is the temperature in o K.
A practical range of Iset is between 100 µA and 1mA. If Rset has a suitably low temperature coefficient (less
than 20 ppm), the data sheet indicates that the temperature will be within 1% of its correct value.
The temperature dependent current may be converted to a voltage with a series resistance, as shown in figure
1.48.

f

f
LM 134

f Rset

R1 Vo ∝ T

m80
lm134-voltage.epic

Figure 1.48: Conversion of Current to Voltage

www.syscompdesign.com
 66

1.8 A High Resolution Temperature Sensor

To this point, all the temperature sensors have had a resolution and accuracy in the order of half a degree centi-
grade. If it is feasible, a resolution of a tenth of a degree would be useful, and so we now examine a suitable
design.
The design of the high resolution temperature sensor begins with the realization that the 8 bit A-D of the
microprocessor is capable of a resolution of only 1 part in 256. Over a span of -30 o C to +50oC, a range of 80o C,
this permits a maximum resolution of 80/256 ≈ 0.3 degrees, less than we need for ’high resolution’. However,
the microprocessor also has a pulse accumulator and three 16 bit timer capture registers that can be used to
measure a frequency or time interval. If the temperature can somehow be converted into a varying frequency, the
microprocessor can resolve the temperature to much better than one part in 256.
The key to converting the temperature to a time interval is the voltage to frequency converter. It turns out to
be convenient to use a temperature to current sensor (LM134, described in section 1.7.1) to generate an electrical
current proportional to temperature. The V-F converter then converts this signal to a train of pulses, where the
frequency is proportional to the input current. The schematic of an application circuit, redrawn from [34] is shown
in figure 1.49.

V+

r r
$R {pu}$
22K
8

r r 1 r f fout ∝ T emperature
6 3
10Hz/o K
Cf LM331 Rt
100nF 7 6.81K ± %
r r

2
5
LM234
 Rset
4
2.32K

r
±1% 12K

r
Ct
Rs 10nF
5K

r r r r f
m80
hrts-schematic.epic

Figure 1.49: Schematic Diagram, High Resolution Temperature Sensor

www.syscompdesign.com
Temperature Measurement 1.8 A High Resolution Temperature Sensor

The output of the circuit is 10Hz/o K, so the output at various temperatures would be as shown in figure 1.50.

Temperature, o C Temperature, o K Frequency, Hz Period, µSec

–30 243 2430 412
+25 298 2980 336
+50 323 3230 309

Figure 1.50: Temperature to Frequency Conversion Results

1.8.1 Analysis of the Temperature-Frequency Converter

The equivalent circuit diagram for the temperature-frequency converter of figure 1.49 is as shown in figure 1.51.

r2 e m
+
Iref LM131

e e
e e
Rs 7

r
Set Reference 12K
Vref Current

5K +

r
e r
22K

e
Comparator
1 HH
H
r6 e
3
A1 H
+
Ii –
  1.1Rt Ct

m e
LM234 Monostable + f ∝ Ii

Cf Rt

r
2.32K 100nF 5 6K8 Set
±1% Pulse

r
Ct Width
10n

m70
tf-block.epic

Figure 1.51: Internal Details, High Resolution Temperature Sensor

The reference current is summed directly in capacitor Cf , the filter capacitor. The temperature sensor dis-
charges current out of the filter capacitor at a rate proportional to the absolute temperature, causing V f to ramp
downward.
The comparator compares the voltage on Cf with the reference voltage, 1.9 volts. When the voltage on C f
drops below Vref , the comparator changes state, triggering the monostable. The monostable pulse appears at the

www.syscompdesign.com
 68

output on pin 3 and causes the current switch to divert current into the filter capacitor. This causes the voltage
on Cf to ramp upwards. Feedback action ensures that the charge removed by the temperature sensor is exactly
balanced by the charge added during the monostable pulse interval.
The transfer function of the current-frequency converter may be shown to be:
fo Rs 1
= Hz/amp (1.76)
ii Vref 1.1Rt Ct
The output current of the LM134 temperature sensor is
Kt Tk
it = (1.77)
Rset
where
Kt = 227 µvolts
Tk is the temperature in Kelvin
Rset is as shown in figure 1.51.
Substituting equation it from equation 1.77 for ii in equation 1.76, we have the transfer function of the high
resolution temperature sensor, in Hz/Kelvin.
 
fo Kt R s 1
= Hz/Kelvin (1.78)
Tk Rset Vref 1.1Rt Ct
As a check on the operation of the temperature-frequency converter, we will plug the known quantities, includ-
ing a temperature of 25oC and the corresponding output frequency (10 × (273 + 25) = 2980 Hz), into equation
1.78. Solving for the value of the adjustable resistor Rs should yield a value between 12 and 17KΩ if the circuit
can be adjusted properly.

227 × 10−6 Rs
 
1
2980 = 293 (1.79)
2320 1.9 1.1 × 6810 × 101 0−9
Solving for Rs , we obtain
Rs = 14.6KΩ
which is within the range of adjustment of Rs .
For this value of Rs , then, the transfer function of the temperature-frequency converter will be 10 Hz/Kelvin,
as promised in the original application note.

Design Concept
Deriving the transfer function for this application circuit is a lot of work. Why bother when the value is given on
the application note?
Two important reasons:
First, it sometimes happens that the application circuit simply won’t work as advertized, or was designed for a
particular set of circumstances that don’t apply to your situation. By working through the analysis, you can check
that the original assumptions of the designer were correct and will transfer to your design problem.
An excellent way to get burned is to use someone else’s circuit without careful checking that it will work.
Second, the transfer function specifies the interdependencies of the various components in the circuit. In this
particular example, the value of the capacitor Cf is not important to the value of the transfer function. On the

www.syscompdesign.com
Temperature Measurement 1.8 A High Resolution Temperature Sensor

other hand, the value of the timing capacitor Ct does show up in the transfer function, so it is critical that it be a
stable capacitor near the value specified. A good circuit designer understands which components are important,
and which can be taken for granted or substituted.
In the next section we’ll have a look at how these frequencies would be measured by a microprocessor.

1.8.2 Frequency Measurement by Accumulating Pulses

The microprocessor contains a pulse accumulator, an 8 bit counter which may be configured to increment each
time a pulse occurs. It may also be be read or written to by machine instructions, and may be set up to generate
an interrupt each time it overflows from 255 back to 0. The pulse accumulator is ideally suited to measuring the
frequency of a pulse waveform.
To use the pulse accumulator in a frequency counter application, the pulse train to be measured is counted in
the pulse accumulator. At the end of a time interval the processor reads the accumulated total and divides it by
the elapsed time to determine the frequency. If the elapsed time is one second, then the frequency is simply the
accumulated count.
Because the pulse accumulator is only an 8 bit counter and the maximum count will exceed 255 for satisfactory
resolution, the counter must extended by counting accumulator overflows with an interrupt service routine. The
pulse accumulator interrupts will occur at 1/256 of the input frequency so the resultant processor computing load
is not excessive.
The timebase for the frequency counter is another interrupt at the timebase interval, which might be 1 second 6 .
This interrupt reads and displays the total count, and then clears the pulse accumulator to start another measure-
ment cycle. Notice that the accuracy of the frequency measurement can be increased with a longer timebase at
the expense of a slower update rate of the frequency reading.
In summary then:

• The temperature sensor and voltage-frequency converter of figure 1.49 are used to generate a pulse wave-
form of frequency 10 times the temperature in degrees Kelvin.

• In the processor, a 1 second real-time-interrupt (RTI) is used to establish the frequency counter timebase.
At the beginning of a measurement cycle, the RTI service routine clears the 8 bit pulse accumulator and the
byte used to count accumulator overflows.

• The pulse accumulator counts the incoming pulses. Every 256 pulses, the pulse accumulator generates an
accumulator overflow interrupt. The service routine for this interrupt simply increments the overflow count
byte.

• At the next 1 second real-time interrupt, the RTI service concatenates the current pulse accumulator count
and the overflow count into a 16 bit count of pulses during the 1 second RTI interval.

• A display routine subtracts a count of 2730 and displays the result as the temperature in tenths of a degree
centigrade.

6 In practice, the real time interrupt would more likely be in the range of 1/60 second. A one-second interrupt would then be derived from

the fast interrupt service routine. Every 60 interrupts, the RTI would determine the number of accumulated pulses. The real-time interrupt on
the ubiquitous P/C has is at the bizarre value of 1/18.206 seconds, which complicates accurate timekeeping.

www.syscompdesign.com
 70

1.8.3 Measurement of Period

As an alternative to frequency measurement using the pulse accumulator, it is also possible for the microprocessor
to use the input capture facility to measure the time between pulses (the period of the pulse train), and then relate
this value back through the various transfer functions to determine the original sensor temperature.
In this case, however, the voltage to frequency converter circuit must be modified. Using the input capture
method, each time a pulse occurs, the microprocessor timer hardware captures the corresponding 16 bit time value
and generates an interrupt. The interrupt handler uses this time value to determine the elapsed time since the last
interrupt, thereby obtaining the period of the pulse train. The free-running counter that serves as the master timer
in this system has a maximum frequency of 1MHz, and so the input capture readings have a resolution of 1
µsecond.
The last column of figure 1.50 shows that the range of measured periods is from 412 µseconds to 309
µseconds. The change in count will be 412 − 309 = 103, far less than the 800 count difference that is needed for a
resolution of 0.1o over 80o C. Thus we can use the application circuit as a basis for our design, but the component
values must be recooked.
A block diagram of the system, with the various transfer functions, is shown in figure 1.52.

1
to ∝ Tk
1MHz

Clock


i ∝ Tk 16 Bit Counter
N1 N1

f s
Temperature Voltage-Frequency
Sensor Converter Load
Tk 16 Bit Register
LM134 LM131

f
m80
hrts-block.epic

Interrupt
Flag Microprocessor

Figure 1.52: Block Diagram, High Resolution Temperature Sensor

The temperature transducer produces a current proportional to absolute temperature. The voltage-frequency
converter generates a pulse train at frequency proportional to temperature or, what is saying the same thing, at
a period inversely proportional to temperature. Each time a positive edge occurs in the pulse train, the micro-
processor transfers the 16 bit count value to a 16 bit register and generates a processor interrupt. The processor
determines the elapsed time between the two edges (the period of the pulse train) by subtracting two successive
count values N1 and N2.
Our task is to determine a suitable value for the V-F components by relating the temperature to the micropro-

www.syscompdesign.com
Temperature Measurement 1.8 A High Resolution Temperature Sensor

cessor count value. We previously determined that the transfer function of temperature-frequency was given by
equation 1.79. Then the equation for period is simply the inverse of this,
 
Rset Vref 1
to = 1.1Rt Ct seconds (1.80)
Kt R s Tk
The value of the count Nc accumulated in this interval to is equal to the product of the counter master clock
frequency fc times the interval.

Nc = f c t o (1.81)
Substituting for to in equation 1.80 and solving for the count value, we have
 
Rset Vref 1
Nc = f c 1.1Rt Ct (1.82)
Kt R s Tk
This equation effectively consists of a constant term times the reciprocal of absolute temperature:
1
Nc = ( a constant ) (1.83)
Tk
This tells us that the relationship between absolute temperature and measured interval will be in the form of a
hyperbola, figure 1.53.
As a result, the number of counts per degree of temperature change is variable and depends on the temperature.
This is not welcome news, for it means that the microprocessor will need to do a calculation (specifically, a
division operation) to relate the measured count back to the value of the temperature.
The graph also indicates that we must ensure that the resolution is sufficient, that is, that the slope of the
count-temperature curve is never less than 10 counts per degree. To do this, we must determine an expression for
the slope of the count-temperature curve and relate that to the component values of the v-f converter.
The slope of the count-temperature curve is found by differentiating equation 1.80. (To minimize writing, all
the constant terms have been bundled into one constant Kvf ).

 
dNc d 1
= Kvf (1.84)
dT dT Tk
−1
= Kvf (1.85)
Tk2
From figure 1.53, the minimum value of the slope occurs at the maximum temperature, 50 o C, or 323o K. The
minus sign can be thrown away, since it is the absolute value of the slope we are interested in. (Whether the count
increases or decreases with increasing temperature is not important, since the microprocessor software can deal
equally well with either case.) Putting Tk equal to 323, and dN/dT equal to 10, we can solve for the value of
Kvf :
−1
10 = Kvf
3232
from which
Kvf = 1.043 × 106
Now we can expand Kvf back into the full equation and solve for the unknown quantities:

www.syscompdesign.com
 72

Measured
Interval

0
0 Temperature
m70
count-temperature.epic

Figure 1.53: Count-Temperature Curve

Rset Vref
fc 1.1Rt Ct = 1.043 × 106 (1.86)
KT R s

The values of the voltage-frequency converter reference voltage V ref and the current sensor constant KT are
fixed and cannot be changed. A review of the data sheet and applications information for the v-f converter reveals
that the reference current is never changed significantly from its value of 135 µamps, so the resistor R s that sets
the reference current should probably be left at 14.1KΩ. For similar reasons, the temperature sensor resistor R set
should not be changed significantly.
The microprocessor count frequency fc can be set to 1MHz, 250KHz, 125KHz or 62.5KHz. The best resolu-
tion for a given pulse period will occur at the higher count frequency, or in other words, a higher count frequency
will allow a shorter measurement period for the same resolution. A shorter period is desireable because a larger
period requires larger timing resistor and capacitor, and components tend to become less accurate and stable at
larger values. The best choice, then, is the 1MHz count frequency.
We are left with Rt and Ct , which determine the period of the voltage-frequency converter charge pulse T p .
In summary, the known quantities are:

www.syscompdesign.com
Temperature Measurement 1.8 A High Resolution Temperature Sensor

fc 1 × 106 V
Rset 2320 Ω
KT 227 µvolts/o K
Vref 1.9 volts
Rs 14.1KΩ
Plugging these quantities into equation 1.86, we have
2320 1.9
1 × 106 −6
1.1Rt Ct = 1.043 × 106 (1.87)
227 × 10 14100
from which

Rt Ct = 757 × 106 (1.88)

The original value of timing capacitor Ct was 10nF. Choosing that value again, we have for the timing resistor
Rt

757 × 106
Rt = (1.89)
10 × 10− 9
= 7.57KΩ (1.90)

Substituting all the known values in equation 1.80, we have

1
Nc = 1.147 × 106 (1.91)
Tk
This is the key equation relating microprocessor count to absolute temperature. The microprocessor would
solve for absolute temperature by dividing the measured count into the equation constant, using floating point
arithmetic.
A graph of the equation over the range -30o C to +50o C is shown in figure 1.54.
A tangent at the temperature of 50o C can be seen to have a value of approximately 10 counts per degree,
which confirms the analysis.

www.syscompdesign.com
 74

4800
Count vs Temperature
4600

4400

4200
Count
4000

3800

3600

3400
-30 -20 -10 0 10 20 30 40 50
Temperature, Centigrade

Figure 1.54: Count VS Temperature

www.syscompdesign.com
Temperature Measurement 1.9 Temperature Measurement Summary

1.9 Temperature Measurement Summary

In this chapter, we studied a variety of electronic ambient temperature devices: the thermistor, the diode-connected
transistor and two integrated circuit temperature sensors: one with a voltage output and the other with a current
output.
The final circuit of each, with a brief summary of its properties, is given below.

1.9.1 Thermistor
It is hard to imagine a simpler measurement circuit than that of the thermistor temperature sensor circuit shown
for a 10KΩ thermistor in figure 1.55. The sensitivity of the thermistor to temperature makes it possible to connect
the thermistor directly to the A-D converter without further amplification.

s d Vrh
Ratiometric
R1 A-D
s d
20KΩ Converter
Vb In

+5V  Rt Vt
s 10KΩ d Vrl
m100
therm-final.epic

Figure 1.55: Final Thermistor Circuit

The non-linearity of the thermistor makes it necessary to use a lookup table or floating-point mathematical
calculation to relate the A-D converter output back to the temperature of the thermistor.
A useful equation for the behaviour of the thermistor is equation 1.3.
 
1 1
β −
RT = R25 e T To (1.92)

This equation is easier to use but not accurate as accurate as the Steinhart-Hart equation, 1.4:

1
= a + b ln RT + c (ln RT )3 (1.93)
T
The Steinhart-Hart equation is significantly more accurate than the standard thermistor equation at tempera-
tures below zero o C.
For an inexpensive thermistor, the ultimate accuracy of the measuring circuit was only ±3 o C, not all that
accurate.

www.syscompdesign.com
 76

In general, thermistors become progressively more inaccurate at low temperatures and are best used at tem-
peratures above 0o C. More accurate and linearized thermistors are available but at greater cost than other sensors
of comparable accuracy.

1.9.2 Diode-Connected Transistor

The base-emitter voltage of a diode connected transistor changes at a a predictable and linear rate of about -2.2
millivolts/o C. It must then be amplified to generate a useful swing for the A-D converter input. The final circuit is
shown in figure 1.22.

Ratiometric
A-D Converter

+5 volts f r r f V
rh
R1
13K
Rd
44K Ri Rf

r r r f V
6K85 120K
AD NAD
HH +5V
– HH
H

R3 
 LM358
r
+
Temperature 500  
Sensor
MTS102
R2

f r r r f V
2K2
Gnd rl
m80
diode-sensor-final.epic

Figure 1.56: Temperature Sensor Circuit

Although any transistor can be used as a sensor, a transistor characterized specifically for temperature sensing
(the MTS102) is available at low cost from Motorola Semiconductor.
Because of the large gain required of this circuit, the behaviour of the op-amp is a particular concern. In
particular, the offset voltage, offset current, bias current and (if operated from an unregulated supply) the power
supply rejection ratio must all be checked to ensure the amplifier drift is within acceptable limits.
For this circuit, the accuracy should be within 1 degree of the correct temperature over a range of -30 to +50 o C.
The resolution, limited by the 8 bit A-D converter, is approximately 0.5 o C.
In general, the base-emitter voltage at some temperature may be predicted using equation 1.13 and a knowl-
edge of the base-emitter voltage at a reference temperature. The slight non-linearity of the base-emitter voltage
characteristic with temperature may be compensated by a squaring circuit, in which case it is possible to obtain
an accuracy of ±0.1oC.

www.syscompdesign.com
Temperature Measurement 1.9 Temperature Measurement Summary

1.9.3 Integrated Circuit Temperature-Voltage Sensor

For integrated transistors of different emitter areas operated at the same collector current, the difference between
the base emitter voltages of the two transistors is proportional to temperature. This technique cancels out the
leakage current induced non-linearity of a single diode-connected transistor. The integrated circuit can also incor-
porate an on-chip amplifier, reducing the requirements for the external amplifier.
The final circuit is similar to figure 1.56. With an sensor output signal of 10mV/ oK compared to the 2.2mV/o/K
of the diode-connected transistor, less amplifier gain is required to drive the A-D converter.
Some test text Ratiometric
A-D Converter

+5 volts
.......
..... ....
........
r r ......
..... ....
........ Vrh
..
...........
R1 .....
............
.....
..
..........
30K ....
Rsensor .....
............ ..... ...
.... .
2K2 ..... Ri Rf ...

r r r
...
30K6 120K .
...
.. .. .. .. ....... . ....
...................... ...................... ..... ..... ....
.. .. .. .. ...... VAD ...
.
...
....
.
.... NAD
. ....
....... ...
.......
.......
.......
+5V .
...
........ .
– ....... ...
.......
....... .
LM335AZ ...
...... . .....
.
.......... .... ........
.. . ....
..
Rcal ........... R3 ......... + ................
..... ..... ............ .... .......
.
............ .... .. ... .... ..... ........ LM358
10KΩ ........ .. ... ... 5K .... .......
............ ..... . ... ..
..... .
.... ..
.
.
.
.
. .
.
. R2 ...........
. .....

r r r r
. ............
.
. 36K ....
.....
.
... . .......
...... ..... . ..... .....
Gnd ......... .
.
.
...... Vrl
.
.
.
.
.
.
Calibrate for 2.982mV at 25o C
m80
icts-voltage-final.epic

Figure 1.57: Temperature-Voltage Final Circuit

As a result, the amplifier contributes less error than it did in the diode-connected transistor sensor amplifier.
The worst-case temperature error circuit is primarily dependent on the temperature sensor, which is specified to
be about ±0.5o C over the range -40 to +50o C.

1.9.4 Integrated Circuit Temperature-Current Sensor

Using similar techniques to the Temperature-Voltage sensor, the Temperature-Current sensor generates a current
proportional to temperature. The current may be converted into a voltage by a resistor. In section we showed how
the current could be converted into a frequency with a voltage to frequency converter and then measured to 0.1%
resolution using the microprocessor counter circuits.
The final high resolution temperature measuring circuit is shown in figure 1.58.

www.syscompdesign.com
 78

V+

r r
$R {pu}$
22K
8

r r 1 r f fout ∝ T emperature
6 3
10Hz/o K
Cf LM331 Rt
100nF 7 6.81K ± %
r r

2
5
LM234
 Rset
4
2.32K

r
±1% 12K

r
Ct
Rs 10nF
5K

r r r r f
m80
hrts-schematic.epic

Figure 1.58: Schematic Diagram, High Resolution Temperature Sensor

www.syscompdesign.com
Chapter 2

Humidity

It’s not the heat, it’s the humidity. Everyone

2.1 Introduction

Water evaporates into the atmosphere, making it more or less humid. The degree of humidity is important for
human comfort: if the air is too dry, typical of winter in Canada, the body becomes susceptible to infections,
and sore throats and colds take up residence in the respiratory tract. As well, conditions of low humidity support
the buildup of electrostatic charge, visible in the familiar winter zapping of electrostatic charge that occurs after
dragging our feet across a synthetic carpet.
In the summer, if the air becomes too wet the human body has difficulty using evaporation to cool itself.
Elevated temperatures and humidity are an unpleasant combination for any kind of human activity.
The amount of moisture that can exist in the atmosphere increases with temperature, and so an absolute
measurement of water in the atmosphere is not very useful. A more helpful measure is the Relative Humidity, H r ,
a measure of the amount that the air is saturated with moisture. For example, an H r of 100% indicates that the air
is saturated and no further evaporation can occur. If the temperature then increases, the air will be able to contain
more water and it will no longer be saturated. If the temperature is lowered to the dew point, condensation or fog
will occur as the evaporated water forms into droplets.
The Hair-hygrometer is commonly used to measure humidity. This device depends on the fact that human
hair or hair from the tail of a horse is hydroscopic (absorbs moisture) and expands and contracts with changes in
humidity. This small movement is then mechanically amplified with a lever mechanism to drive a dial mechanism
or recording pen. The hair hygrometer is not particularly accurate. However, it does have a fast response time and
is convenient to read. Most hygrometers seen in peoples homes are of this type.
Readers may recall another charming domestic hygrometer: A model house is equipped with two figures,
one with and one without an umbrella. Both figures were mounted on a bar that rotated horizontally, moving the
figures in and out of doors in the house. As the humidity changed, one or the other figure would emerge from the
model house.

79
 80

2.2 Vapour Pressure

When liquid water is in contact with air, some water molecules have sufficient energy to leave the liquid and
become water vapour, the process of evaporation. At the same time, molecules of water vapour are re-entering
the liquid water by condensation. If the water vapour is saturated, the rate of evaporation is balanced by the rate
of condensation and the two processes are in equilibrium.
The concept of vapour pressure is nicely illustrated by an experiment with an mercury barometer [39]. Water
is introduced into the vacuum at the top of a mercury barometer column. As the water evaporates, the level of
the mercury column falls. The magnitude of the decrease in column height is equal to the vapour pressure of the
water vapour. In effect, the pressure of the water vapour is forcing the column downwards.
If more water is introduced, the column will fall further. However, at some point, liquid water forms and
co-exists with the water vapour. The region above the liquid water is now saturated and the mercury column has
fallen by an amount equal to the saturated vapour pressure e SAT of the water, figure 2.1 .

........ ........
.........
Vacuum .........
Water Vapour
. .
..... Water ...... Saturated Vapour Pressure
.
....................... .....
... of water
......
0.74” Mercury at 21.1o C
Mercury
30 inches 29.26 inches
Column ......... (1016 mB)

..... .....
.. ..
vapour-pressure-a.pictex
Mercury Reservoir ......... 60% vapour-pressure-b.pictex
60%

(a) Before Introduction of Water (b) After Introduction of Water

Figure 2.1: Vapour Pressure Demonstration

Further introduction of water has no effect on column height. However, if the temperature of the water
is raised, the water molecules will have more energy and the saturation vapour will increase, more water will
evaporate, and the column will fall again. Thus eSAT is dependent on temperature.
As implied by the experiment, vapour pressure may be measured using the same units as atmospheric pressure:
inches of mercury, or milliBars (mB).
It is important to note that the presence of air is not required and has no effect on the evaporation of liquid
water into water vapour. The notion of water vapour dissolving into the air is attractive but inaccurate.

2.3 Relative Humidity

The relative humidity is simply the ratio of the current water vapour pressure to the saturation water vapour
pressure. Put another way, relative humidity is the current water vapour pressure expressed as a fraction of the
largest possible vapour pressure [40]. As an equation, this is:

www.syscompdesign.com
Humidity 2.4 Saturation Vapour Pressure

eA
Hr = 100 × (2.1)
eSAT
where

Hr is the relative humidity in percent

eA is the vapour pressure of the water in the air in milliBars
eSAT is the maximum (ie, saturated) vapour pressure, in milliBars, at
the current ambient temperature.

2.4 Saturation Vapour Pressure

The saturation vapour pressure eSAT is the maximum vapour pressure of water that can exist at a given temper-
ature. It is a fixed and known function of temperature, so if the relative humidity or water vapour pressure are
known, the other variable, relative humidity or water vapour pressure, may be calculated.
A table of values of saturation vapour pressure vs temperature, adapted from reference [41, Appendix 7], is
shown in figure 2.2.

Temperature, o C eSAT , millibars

-10 2.83
-5 4.55
0 6.12
+5 8.72
+10 12.27
+15 17.05
+20 23.43
+25 31.78
+30 42.56
+35 56.25
+40 73.43
+45 94.66
+50 120.61
Note: Air pressure is 1016 millibars.

Figure 2.2: Temperature and Saturation Vapour Pressure of Water

Notice that the saturation vapour pressure increases with temperature. For a constant relative humidity, with
increasing temperature a larger quantity of water can be evaporated in the air.
Fitting a curve to the table of figure 2.2 yeilds equation 2.2, which can be used to interpolate between the
temperatures of figure 2.2.

eSAT = A4 · T 4 + A3 · T 3 + A2 · T 2 + A1 · T + A0 (2.2)

where

www.syscompdesign.com
 82

140

120

100 A 4 x4 + A 3 x3 + A 2 x2 + A 1 x + A 0

Saturation 80
Vapour Pressure
eSAT ,
millibars 60

40

20

0
-10 0 10 20 30 40 50
Temperature, o C

Figure 2.3: Vapour Pressure vs Temperature

eSAT is the saturated water vapour pressure, in millibars

T is temperature in degrees C
A4 = 5.33 × 10−6
A3 = 1.92 × 10−4
A2 = 0.0149
A1 = 0.446
A0 = 6.088
As an example, the saturated vapour pressure of water at a temperature of 21.1 o C is 25.02 millibars.

2.4.1 Vapour Pressure Applied: Humidex

As anyone who has lived though an August heat wave in Toronto can attest, the combination of high ambient
temperature and high humidity is most uncomfortable. The body has difficulty cooling itself when humidity is
high, so the apparent temperature is much higher than the actual temperature. The Humidex is an attempt to
measure this effect.
The Humidex formula is given by:
5
H = Tc + (eA − 10) · (2.3)
9
where

H is the humidex reading, given in apparent temperature (o C)

Tc is the temperature in o C
eA is the ambient vapour pressure

www.syscompdesign.com
Humidity 2.5 The Wet-Bulb, Dry-Bulb Hygrometer

(This formula and many others relating to humidity are listed in the Humidity FAQ, reference [51].)
The Humidex is a quantity that a microprocessor can calculate from temperature and relative humidity.

Example
The current temperature is given as 22o C and the relative humidity 85%. What is the Humidex reading?

Solution
First, substitute the current temperature T into equation 2.2 to determine the saturation water vapour pressure
eSAT in millibars.

eSAT = A4 · T 4 + A3 · T 3 + A2 · T 2 + A1 · T + A0
= (5.33 × 10−6 × 224 ) + (1.92 × 10−4 × 223 ) + (0.0149 × 222 ) + (0.446 × 22) + 6.088
= 26.39 millibars

Then use the relative humidity formula, 2.1, to determine the ambient vapour pressure e A .

eA
Hr = 100 ×
eSAT

so
Hr
eA = × eSAT
100
83
= 26.39
100
= 21.9 millibars

Finally, substitute for Tc and eA in the humidex equation 2.3 to determine the humidex reading.

H = Tc + (eA − 10)5/9
= 22 + (21.9 − 10)5/9
= 28.6

The humidex reading1 , or apparent temperature, is 28.6oC.

2.5 The Wet-Bulb, Dry-Bulb Hygrometer

A common instrument for measuring relative humidity is the wet-bulb, dry-bulb hygrometer, figure 2.4. This
hygrometer is also known technically as a psychrometer [38].
1 This formula appears to be consistent with the method used by Environment Canada, since their quotations of humidex correspond to the

results obtained with this formula

www.syscompdesign.com
 84

.... ....
... ... ... ...

ppp
ppp
ppp
ppp
ppp
ppp ppp
ppp ppp
ppp ppp
pp ppp
ppppp ppp
Wet Thermometer

ppp ppp
Dry thermometer

ppp pppp
ppp
(Dry Bulb) (Wet Bulb)

ppp ppp
ppp ppp
ppp ppp
ppp ppp
ppp ...p....
........ pp ......
p .. p ..
. p .
... p ...... ...... .... ...
.. .. ......
...
.... . .......... ..
.................. ......................
.. . ..
.. ............ Wick
.. .
. ............
.. ... ..........................
... ...
. .
.. ...
.. ..
.
............. ..
............ ....................
. . .
...
...
... . ...
Water.......Resevoir
......... ..... ... .......... .
............................................................................................................................. ...............................
.... ....................... .. .
... .. .. .
...
... ... .. .
............ ...................................................................................

Figure 2.4: Wet-Dry Bulb Hygrometer

The psychrometer consists of two thermometers. One measures the air temperature directly. The other is
cooled by a porus muslin wick. The wick is moistened and evaporates water at a rate dependent on the relative
humidity, cooling the wet-bulb thermometer. The rate of cooling, and hence the difference between the two
thermometer readings, is a measure of the relative humidity.
The rate of cooling of the wet bulb is very dependent on the movement of air past the wet bulb, and some
psychrometers are equipped with fans, or handles so they can be swung through the air (the so-called sling
psychrometer). Even so, the psychrometer is only about 5% accurate.
The speed of response of the psychrometer depends on the speed of response of the thermometers, some tens
of seconds at least. For accurate readings of humidity, both thermometers must have the same time constant [38].
For a manual determination of humidity, the observer uses a chart or special slide rule to relate the ambient
temperature and wet-dry temperature difference to relative humidity. For automatic calculation by microprocessor,
the system is shown in figure 2.5.
When one temperature sensor is already in place to to measure ambient temperature, it is a simple matter
to add a second temperature sensor, wick, water reservoir and fan. For a long-term installation, however, this
arrangement has some disadvantages: the fan is subject to clogging with debris or mechanical breakdown, the
reservoir must be refilled regularly, and mineral deposits will build up in the reservoir and wick unless distilled
water is used.
On the other hand, the operation of a psychrometer can be analysed from basic physical princibles, which is
reassuring where accuracy is required.

www.syscompdesign.com
Humidity 2.5 The Wet-Bulb, Dry-Bulb Hygrometer

Dry
Temperature
Sensor
..........
. A/D Input 1
.... ....
......... ..........
. Humidity Readout
Fan. Wet
.
...... Temperature
... ..
...... Sensor
......
.
... ... ..........
... ...
...
.... ..
.....................
. A/D Input 2
..........................
..................
. ...
Water............... .....................................
.
...
.

Resevoir

Microprocessor

Figure 2.5: Automatic Measurement of Humidity

2.5.1 Computerizing Wet-Bulb, Dry-Bulb Measurements

To automate the measurement of relative humidity from a psychrometer, we need a mathematical method of
determining humidity from from wet and dry bulb temperature measurements.
The vapour pressure in the air may be determined from the wet-bulb/dry-bulb hygrometer according to the
following formula from reference [40]:

eA = eW − A · PA · (TA − TW ) (2.4)
where
eA is the vapour pressure of the water in the air, millibars
eW is the saturated vapour pressure of water at the temperature of
the wet bulb, TW .
A is a constant, 6.66 × 10−4 (or 5.94 × 10−4 if the wet bulb is
covered with ice)
PA is the air pressure, in millibars, typically 1016 mB.
TA is the dry bulb (air) temperature, o C
TW is the wet bulb temperature, o C
In order for this formula to apply, the wet bulb must be in an air flow of at least 3.6 meters/sec, so the wet bulb
must be fan cooled.

2.5.2 Wet-Dry Bulb Calculation: An Example

Determine the relative humidity if the dry bulb thermometer is reading 15.5 o C and the wet bulb 12.8oC.

1. To find the saturated vapour pressure eW , substitute the wet bulb temperature TW , 12.8oC, in equation 2.2:

eW = eSAT
= 5.333 × 10−6 × 12.84

www.syscompdesign.com
 86

+ 1.924 × 10−4 × 12.83

+ 0.0149 × 12.82
+ 0.444 × 12.8
+ 6.088
= 14.79 millibars

2. Now apply Wet-Dry Bulb formula 2.4 to determine the actual vapour pressure e A of the water in the air:

eA = eW − A · PA · (TA − TW )
= 14.77 − 6.66 × 10−4 × 1016 × (15.5 − 12.8)
= 12.94 millibars

3. Next, use equation 2.2 again, this time substituting the dry bulb temperature, 15.5 o C, to determine the
saturation vapour pressure eSAT at the dry bulb temperature.

eSAT = 17.61 millibars

4. Finally, use equation 2.1 to calculate the relative humidity:

eA
Hr = 100 ×
eSAT
12.94
= 100 ×
17.61
= 73.5%

Checking against a psychrometric table, we find that the table shows 73% relative humidity for these same
wet-bulb/dry bulb temperatures.
The data flow diagram for this calculation is shown in figure 2.6.

2.6 Dew Point

Relative humidity is a measure of the quantity of water vapour present compared to the maximum water vapour
possible. The maximum quantity of water vapour that can exist in the air decreases with temperature (figure 2.3).
Thus, if a sample of air is cooled without removing any water vapour, the relative humidity will increase.
As the air is cooled, at some temperature the relative humidity will reach 100% and moisture will begin to
form as fog or dew. The temperature at which condensation appears is known as the dew point of the air sample.
For example, suppose a 20o C sample of air contains moisture at a relative humidity of 63%. If the air is cooled
to 12o C, moisture will begin to form on the surfaces of the container. Then the dew point of that air sample is
12o C. As a result, relative humidity at some temperature and dew point temperature are two different ways of
quantifying the amount of water vapour in the air. One may be determined from the other.

www.syscompdesign.com
Humidity 2.6 Dew Point

Wet Bulb Temperature

..........
eW
TW .......... fe (TW )

.........
Dry Bulb Temperature ..........
.......... ......... eA = eW − A · PA · (TA − TW )
TA
PA .......... ..........

..........
eA ..........
.......... eSAT .......... Hr = 100 e Hr
fe (TA ) SAT

fe (T ) = 3.36 × 10−6 T 4 + 2.92 × 10−4 T 3 + 0.139T 2 + 0.443T + 6.123

Figure 2.6: Psychrometric Calculation, Data Flow

2.6.1 Dew Point Calculation: An Example

Suppose the ambient air temperature is 25o C and the relative humidity 74%. Calculate the dew point temperature.

Solution
Step 1 Use the current value of temperature to determine the saturation vapour pressure e SAT from figure 2.2,
figure 2.3 or equation 2.2.
From equation 2.2, the value of saturation vapour pressure at 24 o C is 31.78 millibars.
Step 2 Use the current value of relative humidity Hr and saturation vapour pressure eSAT to determine the
current value of water vapour pressure eA from equation 2.1.
Substituting the value of eSAT determined in step 1, and 74% for Hr in equation 2.1, we have:
eA
74% = 100
31.78
Solving for eA , we have
eA = 23.5 millibars

Step 3 Use this value of eA and figure 2.2 or figure 2.3 to determine the temperature at which this vapour pressure
is the saturation vapour pressure. This temperature is the dew point.
Referring to figure 2.3 again, the air will be saturated with moisture at this vapour pressure when the
temperature cools to about 20o C.

2.6.2 Dew Point Calculation by Computer

The computer solution for dew point temperature proceeds exactly as steps 1 and 2 of the previous example. For
step 3, we need a mathematical relationship to relate eSAT to the corresponding value of temperature T . We
could find T by substituting for eSAT in equation 2.2 and solving for T. However, it’s easier to determine another
polynomial to fit the inverse data, that is determine fT where:
T = fT (eA )

www.syscompdesign.com
 88

Unfortunately, fitting a polynomial to the data is not very successful. Even a fourth order polynomial is in
error by 10% or so. However, recognizing that eSAT is approximately an exponential function of temperature,
then a plot of temperature vs the natural log of eSAT will yeild a graph with much less curvature and be easier to
fit2 . Trying this out, we find that a second order polynomial can predict temperature to within one or two percent,
which is what is needed.
The result:
T = A2 · [ln(eSAT ]2 + A1 · [ln(eSAT ] + A0 (2.5)
where
T = temperature in o C
eSAT = saturation vapour pressure, millibars
A2 = 1.073
A1 = 9.642
A0 = -21.06
For example, to find the temperature corresponding to a vapour pressure of 23.5 millibars:
First, find

ln(eSAT ) = ln(23.5)
= 3.157

Now, substitute this value in equation 2.5

T = A2 · [ln(eSAT ]2 + A1 · [ln(eSAT ] + A0
= 1.073 × 3.1572 + 9.642 × 3.157 − 21.06
= 20.07oC

which corresponds to our previous result for dew point.

2 Notice the similarity of this approach to the Steinhart-Hart relationship, equation 1.4 on page 8

www.syscompdesign.com
Humidity 2.7 Electronic Measurement of Humidity

2.7 Electronic Measurement of Humidity

We described in section 2.5.1, page 85, how the Wet-Dry Bulb Psychrometer may be adapted to the electronic
measurement of humidity [40]. This device requires significant care and feeding, and so is not a popular choice
for unattended, long term operation.
Dew point temperature may be measured with the type of instrument described in reference [52]. Then relative
humidity may be calculated from the ambient temperature and dew point temperature, as shown in section 2.6 on
page 86.
For the direct measurement of humidity, both variable resistance and variable capacitance sensors have been
used. Reference [43] describes a home-made sensor that uses the change in resistivity of ordinary table salt
(sodium chloride) to measure humidity.
Several commercial sensors rely on the change in the dielectric constant of a capacitor to measure relative
humidity. References [45], [44] and [50] show the use of a commercial sensor based on gold-plated plastic film.

2.8 A Capacitance-Humidity Sensor Circuit

In this section, we will spend some time describing a humidity measurement circuit based on the Philips Series
691 humidity sensor, which is inexpensive and readily available.

2.8.1 Sensor Specifications

The electrical specifications of the sensor, adapted the data sheet of reference [45], are shown in figure 2.7:

Electrical Specifications
Operating Humidity Range 10 to 90% R.H.
Operating Temperature Range 0 to 85o C
Capacitance at 25o C, 43% R.H., 100KHz 122pF ±15%
Frequency Range 1 KHz to 1MHz
Temperature Dependence 0.1% R.H./o C
Response Time (to 90% of indicated R.H.
in circulating air):
Between 10 and 43% R.H. 3 minutes max
Between 43 and 90% R.H. 5 minutes max
Typical Hysterysis (10% R.H to 90% R.H
and 90% R.H. to 10% R.H. 3%
Maximum voltage 15V
Storage Temperature Range -25 to +85o C
Current price $10.10
Part number 691-90001

Figure 2.7: Vapour Pressure Demonstration

The change of capacitance with humidity is shown in figure 2.8.

www.syscompdesign.com
 90

150

145

140

135
Capacitance,
130
pF
125

120

115

110
0 20 40 60 80 100
Relative Humidity, %

Figure 2.8: Capacitance vs Relative Humidity for Humidity Sensor

The equation of this function is

 1.4 !
Hrel
Cs = C12 0.985 + 0.34 (2.6)
100

where

Cs is the capacitance of the sensor C12 is the capacitance at a relative humidity of 12%
Hrel is the relative humidity in %

A number of important points may be determined from the data sheet information:

Operating Humidity Range The operation of the sensor is not specified below 10% or above 90%, so we’re on
our own in those regions. The curve probably continues in a smooth fashion into those areas, but we cannot
complain to the vendor about accuracy.

Operating Temperature Range The sensor won’t work below the freezing temperature of water (which is rea-
sonable) but is not damaged until the temperature drops below –25 oC. Probably best to reserve it for indoor
use.

Capacitance at 25o C and 43% R.H. This value of capacitance is rather small, as capacitors go. Moreover, look-
ing at figure 2.8, it doesn’t change by much either: 110pF to 150pF. The measuring circuit will have to be
adapted to these small values. Furthermore, the tolerance given for the base value is a rather large ±15%.
Calibration will be required.

www.syscompdesign.com
Humidity 2.8 A Capacitance-Humidity Sensor Circuit

Frequency Range A capacitive sensor measurement may be converted to frequency or period by using the ca-
pacitive sensor as the timing element in an oscillator. This specification indicates the allowable range of
oscillator frequency.

Temperature Dependence Assuming that the sensor will be in an environment when the temperature change is
20oC, the temperature coefficient of 0.1%/o C represents a total total reading change of 2%. If possible, this
change should be compensated for.

Response Time If you breath gently into a hair hygrometer, the meter will indicate an increase of humidity, with
a time constant in the order of a few seconds. The capacitive sensor is much slower, with a time constant in
the order of minutes.

Hysterisis When the sensor is subjected to a series of low humidity, high humidity and low humidity, the final
reading may be expected to be in error by 3%. This is a rather large error, but it would be most important
in applications such as the humidity sensor for a clothes dryer. In that case, the humidity would suddenly
go from a low value to a high value when wet clothes were stuffed into the dryer. The humidity would then
drop gradually as the clothes are dried.
For our application, such drastic changes in humidity should be rare 3 , and hopefully the hysterisis effect
will not be a major problem.

The overall impression, then, is of a sensor with a small change in capacitance and relatively long time
constant. It should be compensated for temperature and must be calibrated for humidity, and the expected accuracy
should be in the region of ±2%.

2.8.2 Oscillator Measuring Circuits

There are two fundamental methods of converting the variation in sensor capacitance into an electrical signal. The
capacitor may be used as one of the frequency-determining elements in an oscillator, the method discussed in this
section. In the next section, we discuss an alternative technique.
In section 1.8, we showed how the current output temperature sensor could be converted into a variable fre-
quency or period waveform. The microprocessor could then relate this frequency or period back to the temperature
of the sensor.
The same technique is to be used here, with the difference that the variation of the capacitor is used to de-
termine the frequency of oscillation. The question is then what type of oscillator is most suitable. There are an
enormous number of possible oscillator circuits, so it we must develop some overriding princibles to narrow down
the field. To begin with, the timing circuits of oscillators depend either on inductor-capacitor circuits networks or
resistor-capacitor networks.
Figure 2.9 shows a comparison of these two oscillator types.
Because large value inductors are inconvenient to make, , LC oscillators tend to be radio frequency devices
(ie, frequency of oscillation greater than 1MHz). The frequency of operation is dependent on the square root of
the capacitance, so a doubling of capacitance changes the frequency by a factor of 1.41. On the other hand, RC
oscillators tend to operate at frequencies below 1MHz and the frequency is a function of the capacitance directly.
Comparing these two, the RC oscillator appears to be more suitable. The humidity sensor is specified to
operated at frequencies below 1Mz. and the extra sensitivity will be useful.
3 However, the humidity in a household rises quite drastically during cooking hour. This is most noticeable in the winter months when the

humidity is low to begin with.

www.syscompdesign.com
 92

Inductor-Capacitor Oscillators Resistor Capacitor Oscillators

fosc > 1MHz fosc <1MHz
1 1
fosc ∝ √LC fosc ∝ RC

Figure 2.9: LC vs RC Oscillators

Now we have to choose a suitable RC oscillator. Again, because there are many possible RC oscillator designs,
let us attempt to narrow down the field with some general observations.
First off, we must decide between measuring the frequency or the period of the oscillator waveform. The
frequency is proportional to 1/C and period is proportional to C: the choice of period simplifies the calculations.
Furthermore, referring to the resources of our microprocessor, we have three input capture ports suited to period
measurement, and only one pulse accumulator, suited to frequency measurement. It would make better sense to
use the input capture port if we can. Period measurement would seem to be the correct choice.
Next, we need to get an estimate of a suitable period for the oscillator waveform. The reasoning goes some-
thing like this:

• The transducer capacitance varies from 150pF to 110pF, for a total variation of 40pF and a minimum
capacitance of 110pF. Put another way, the capacitance ranges from its minimum value C to 1.36C as the
humidity varies from 0% to 100%.

• The period of the oscillator waveform is proportional RC, so it will vary the same way as the capacitance,
from a minimum value of T to a maximum of 1.36T.

• The part that we are interested in is the varying part of T. For a resolution of 1% R.H., the varying part
of T must be long enough to contain 100 discrete values. Now the microprocessor timing circuits have a
resolution of 1µsecond (at best), so 0.36T must correspond to 100µseconds. Then the non-varying part of
the period, 1.0T, must be 277µseconds in duration. All of this is shown in figure 2.10.

0% R.H.

100% R.H.

1.36T

277µsec 100µsec min

m70
humidity-waveform.epic

Figure 2.10: Humidity Sensor Oscillator Waveform

• The period of an RC oscillator is given by something like T = KRC, where K depends on the circuit, but
might be 1.4 or 2.2. For simplicity, we’ll assume that K is 1: this will get us into the computational ballpark.
Then if RC=277µseconds and C=110pF, the resistance is 2MΩ.

www.syscompdesign.com
Humidity 2.8 A Capacitance-Humidity Sensor Circuit

• Now the final step: with the timing resistance in the order of 2MΩ and an operating voltage of 5 volts, we can
expect the timing current to be in the order of 2.5µamps. This is a very small value of timing current. It is not
all that different from the leakage current of timing devices or the bias current of conventional operational
amplifiers. The bias and leakage currents are liable to interfere with the timing current sufficiently to affect
timing accuracy.
The timer design example given in section 2.8.3 confirms this effect.

2.8.3 555 Timer Design Example

An obvious candidate for any astable oscillator is the 555 Timer. It is simple to apply, readily available, and
inexpensive. As we shall show, however, the small oscillator currents make the 555 Timer a poor choice for this
particular application.
A simplified diagram of the internals of the timer is shown in figure 2.11(a). The timer consists of a flip flop
which is actuated from two comparators. The thresholds of the two comparators are established at 2/3V cc and
1/3Vcc by a string of three resistors. The inputs to the two comparators are known as threshold and trigger.

www.syscompdesign.com
 94

e
Vcc

e
r r eOutput
HH
r CU H
Threshold

2/3Vcc 
Flip

eDischarge
Flop

e HH
r C
Trigger H
L
1/3Vcc 

r r
m70
eGnd
555-timer-circuit.epic

(a) Timer Integrated Circuit

r
e
+5Volts

r
Rcharge

r e
r r e
HH
r C
H
U

Rdischarge Flip

e
Flop

r e HH
r C
H
L


r r
Ctiming

e
r m70
555-astable.epic

(b) Timer Astable

Figure 2.11: Timer Circuit and Astable Configuration

www.syscompdesign.com
Humidity 2.8 A Capacitance-Humidity Sensor Circuit

There are two outputs from the timer: a class B totem pole output and an open collector discharge output.
Now consider the astable oscillator circuit of figure 2.11(b). The timing capacitor C T first charges towards Vcc
through resistors Rcharge and Rdischarge . When the capacitor voltage reaches the upper comparator threshold,
the comparator causes the flip-flop to change state, activating the discharge transistor. The timing capacitor then
discharges toward ground through resistor Rdischarge . When the capacitor voltage reaches the lower comparator
threshold, this comparator causes the flip-flop to change back into the original state, deactivating the discharge
transistor. The cycle then repeats.
The voltage across the timing capacitor assumes a sawtooth-like shape 4 , oscillating between 1/3Vcc and
2/3Vcc.
With a little work, one can show that the period of the cycle is given by

T = 0.693(Rc + 2Rd )CT (2.7)

The exact proportions of Rc affect the duty cycle of the output square wave, which is not important. We’re
only interested in the period, so we might simply make 2Rd = RC . Then

T = 1.38Rc CT (2.8)

Now let us calculate a suitable value for Rc , using equation 2.8 and the known values of 110pF for C T and
277µsec for T.

T
Rc = (2.9)
1.38CT
= 1.8M Ω (2.10)

We set 2Rd = Rc so Rd = Rc /2 = 912KΩ.

To this point, we have assumed that no current flows into the threshold or trigger comparator inputs. Let us
compare the charge and discharge currents with the leakage currents of these inputs.
During charging, the voltage across the charging capacitor varies from 1/3V cc to 2/3Vcc . The currents are
then as tabulated below:

Vc Ic
1/3Vcc 1.22µamps
2/3Vcc 0.615µamps

The input current for the ordinary (BJT version) 555 timer trigger input, is given as 0.9µamp maximum. This
exceeds the charging current of the capacitor as it approaches its highest voltage, so the astable might not function.
On the other hand, there is a CMOS version of the 555 for which the trigger current is given as 10pA (10 ×
10−12 amps). However, though significantly superior in a number of respects to the BJT version, the CMOS
version is less commonly available. Murphy’s law being what it is, it is a dead certainty that someone will attempt
to substitute a BJT 555 timer in the circuit. Insidiously, some BJT 555’s will probably work, leading to a lot of
head scratching over the ones that don’t function.
4 The sides of the sawtooth are actually somewhat curved because the charge and discharge functions are exponential.

www.syscompdesign.com
 96

2.8.4 Engineering Proverb: Do it the easy way

With a few assumptions this back of the envellope analysis established that the oscillator circuit could be prob-
lematic because of the small timing currents. This is not to say that it cannot be done: there are low current timers
and op-amps available. However, it is a suggestion that it might be worthwhile to attempt to outflank the problem
rather than tackling it head-on.
In general, this analysis illustrates that the most obvious solution is not always the best. A good engineer can
step back from a solution and become a critic, finding the flaws in the proposed design. A good engineer can also
come up with several different design approaches. Hopefully, one of the designs poses less risk than the others,
and becomes the final choice.
Engineering can be an adventure. If the wrong choice is made at the outset, and the wrong design chosen, it
can take a long time and many resources to back out and start over again. Moreover, it may take some substantial
exploration to identify which solution is the best choice.
At the preliminary design stage, identify several possible solutions and try to choose the one with the least
risk of failure.

2.8.5 CMOS Oscillator and Divider

The next design springs from the realization that
• the RC oscillator needs to operate at high frequencies to minimize the oscillatior resistance, and therefore
the problem of leakage currents.
• we could lengthen the effective period by dividing down the frequency (and hence multiplying up the
period) in a cascade of flipflop dividers.
As is the often the case, such inspiration is precipitated by the availability of a suitable part. In this case
the suitable part is a CMOS integrated circuit, type CD4060 14-Stage Ripple-Carry Binary Counter/Divider and
Oscillator.
The 4060 consists of a CMOS oscillator section followed by a 14 stage binary divider chain. A high level on
the reset input stops the oscillator and forces all the flip-flops into the zero state. All of this is available for a parts
cost of about one dollar.
The oscillator period is given by
T ≈ 2.3RT CT (2.11)
The frequency, of course, is the inverse of this:
1
fosc ≈ (2.12)
2.3RT CT
In our case the oscillator capacitance CT is the humidity sensor. The value of resistor RS is not critical. It
helps stabilize the oscillator frequency and is set to a value between 2R T and 10RT . The data sheet indicates that
a maximum value for these resistors is 1MΩ, so we might set RS at 120KΩ and RT at 56KΩ. Then substituting
110pF and 150pF for the minimum and maximum values of C T in equation 2.11, we obtain maximum and mini-
mum oscillator frequencies of 70.6KHz and 51.8KHz. These are well within the maximum frequency capability
of the oscillator, 100KHz.
The minimum and maximum frequencies correspond to periods of 19.3µsec and 14.2µsec. We wish the
difference between these two periods, 5.1µsec, to be multiplied up to at least 100µsec to obtain a resolution of
1 part in 100 by our microprocessor timer circuits. The minimum binary multiplication factor is therefore 32, or

www.syscompdesign.com
Humidity 2.8 A Capacitance-Humidity Sensor Circuit

25 . To be on the safe side, we might choose to use the Q7 divider output, which gives a period multiplication of
27 = 128. Then as the humidity varies from 0 to 100%, the period of the divider output waveform will vary from
2470µsec to 1818µsec, for a difference of 652µsec. The resolution will be 1 part in 652, or about 0.15%, much
more than required.
Notice that we have to take care that the waveform period does not exceed 2 16 − 1 µseconds, since this is the
maximum measuring period of the 16 bit microprocessor input capture registers.

Oscillator Tolerance
Now we come to an interesting conundrum. According to [46] (page 522),

’Our experience is that the internal oscillator. . . (of the 4060). . . has poor frequency tolerance and (in
some HC versions) may malfunction.’

Prompted by this warning, we examine the data sheet for the RCA version of the 4060 (the CD4060B) and find a
minimum and maximum specification for oscillator frequency stability under certain test conditions. Good.
The data sheets for the Motorola version of the 4060 (the MC1460B) show the oscillator stability under
’Typical RC Oscillator Characteristics’. Oscillator performance is not guaranteed. As Bob Pease says in [47],
’Typical could mean that they once saw a device with those characteristics’. Not so good.
The data sheet for the National version, the CD4060BCN, does not show any specifications for oscillator
performance. Uh-oh.
Obviously, there is a certain amount of risk associate with CMOS oscillators. This is also fueled by the
knowledge that the logic threshold in CMOS, which we tend to visualize as around 1/2V cc may and does vary
from 1/3Vcc to 2/3Vcc between devices.
However, reference [48] indicates that the period of oscillation of a CMOS oscillator will vary as tabulated
below:

Threshold, % of VDD Oscillator Period

33 2.25RC
50 2.14RC
66 2.22RC

This variation is in the order of 5%, much less than the humidity sensor tolerance. The system must be
calibrated to allow for the 15% tolerance on the humidity sensor, so this tolerance on the oscillator frequency is
quite acceptable.

Oscillator Temperature Stability

The stability of the oscillator with changes in temperature is much more of an issue than the original frequency
tolerance. The tolerance can be calibrated out, but a variation with temperature presents more of a difficulty.
The Motorola data sheet shows a graph of oscillator frequency vs temperature in which the oscillator frequency
changes by about 1% over a 25o C range of temperature, equivalent to 400ppm/oC. Reference [48] shows that
oscillators of this type have a variation of 700ppm/oC and about 1% per volt of supply change. All of these values
are quite acceptable.
This is not an iron-clad case for oscillator stability and it certainly wouldn’t satisfy a flinty-eyed mil-spec
inspector, but it is enough reassurance to encourage us to try this design.

www.syscompdesign.com
 98

Sensor Temperature Stability

The humidity sensor responds to temperature at a rate of 0.1%/o C. If the expected range of ambient temperature
is such this represents an unacceptable error, it is necessary to temperature compensate the sensor. This circuit
is not simple to compensate in hardware. However, if the microprocessor is aware of the temperature reading, it
may compensate by adding a software correction to the reading of humidity.

2.8.6 Humidity Oscillator Circuit

The final humidity measurement circuit based on a variable frequency oscillator is shown in figure 2.12.

e6
Microprocessor
MC14060B Input Capture

HH e r HH e r r HH e r
Q7

  

r r r VDD Pin 16

e11 e10 e9 e12

Clock Out 1 Out 2 Reset VSS Pin 8

RT

r
56K
@
RS @
CT
120K 110pF–150pF
m70 Philips Humidity Sensor, 961-90001
humidity-oscillator-final.epic

Figure 2.12: Humidity Measurement With Variable Frequency Oscillator

2.9 Calibrating the Hygrometer

When certain salts are dissolved in water to the saturation point, and the resultant solution placed in a closed
container, the solution will maintain a predictable humidity in the atmosphere of the container[38, 45]. The
humidity sensor may be lowered into the atmosphere of a large jar containing the saturated salt solution in the
bottom (figure 2.13) to expose it to a known humidity.
The table of figure 2.14 shows suitable salts for the calibration, together with the calibration humidity at
various temperatures.
If two different salt solutions are available, the humidity sensor may then be calibrated at two points.

www.syscompdesign.com
Humidity 2.9 Calibrating the Hygrometer

To
.
readout.
.. ..
...... ......

p ppp ppp
pp pp
p p p p ppp pp p p p p pp p ppp p
ppp p p p p p p pp pp
pp pp p p pp
ppp pp
Test
................
... .. .........
.... .. ......... Circuit
... .. .........
.........
................. ...

......... Jar
.........
.......

Controlled Humidity Atmosphere

Salt
Solution

pppp pp
................
........................................

p pp p p p p p p
pp p p pp p p p p pppp

m80
humidity-calibration.pictex

Figure 2.13: Hygrometer Calibration

Formulation Name 15o C 20o C 25o C

KN O3 potassium nitrate 95 94 93
N aCl sodium chloride (table salt) 75 75 75
M g(N O3 )2 .6H2 O magnesium nitrate 53 52 52
K2 CO3 potassium carbonate 44 44 43
M gCl2 .6H2 O magnesium chloride 33 33 33
LiCl lithium chloride 13 12 11

Figure 2.14: Salts for Humidity Calibration

www.syscompdesign.com
 100

2.10 Humidity Measurement Software

In this section, we discuss the software for measuring the period of the humidity signal.
A general strategy for the measurement and display of humidity is as follows:

• As described in section 2.8.2 previously, the humidity sensor generates pulses at periods between 1818 and
2470 µsec, corresponding to relative humidity between 0% and 100%.

• The pulses are connected to Input Capture #1 of the 68HC11 microprocessor. This input is configured to
generate an interrupt on each positive edge. With each interrupt, the 68HC11 16 bit free running timer,
which is counting at a rate of 1 musec per count, is transferred to the input capture #1 latch. This may be
regarded as timestamping the occurence of the edge.

• The interrupt service routine notes the time T1 of the first edge and the time T2 of the second edge and then
finds T 2 − T 1 to determine the pulse period in µseconds.

• The display manager routine converts the pulse period into a humidity value and writes it to the display.

2.10.1 Humidity Task State Machine

The humidity task must cycle through a series of actions in an organized fashion. Notice that the pulse mea-
surement task is called whenever a positive edge occurs in the input waveform, or hundreds of times per second.
It’s up to the measurement task (interrupt service routine) to perform the correct action (which might be to do
nothing) each time it is called. A state machine organization organizes this nicely, with the following states (see
figure 2.15):

0: Dormant

Received Request, recorded T1

1: Second Edge

Recorded T2, measurement complete

2: Complete

Measurement Accepted by Main Routine

hum-state.epic, m70

Figure 2.15: Humidity Task State Diagram

www.syscompdesign.com
Humidity 2.10 Humidity Measurement Software

Dormant The humidity task has completed the last measurement cycle and handed over results to the display
manager. If the Request Flag is cleared, there is no request to take a measurement, then the task simply
exits.
However, if the Request Flag is set, the task has received a request for a new measurement. It therefore
notes time T 1, clears the request flag (since it is now honouring the request), changes to the Second Edge
state and exits.
Second Edge When the processor enters the humidity task and the state is Second Edge, it records the time as
T 2. Then it calculates the pulse period T = T 2 − T 1, raises the Ready Flag to signal the display manager
that a new humidity period measurement is ready and changes to the Complete state.
Complete When the display manager reads the value of the time interval T = T 2 − T 1, it clears the Ready Flag
to indicate that it has accepted the reading. The next time an interrupt occurs, the humidity task is called
into state 3, Complete. It then detects that the Ready Flag is cleared and so changes the state back to the
Dormant state.

The humidity task and the main routine are essentially cooperating tasks that communicate with each other by
the setting and clearing of software flags.
It’s not immediately obvious that a state machine is necessary for organizing the interrupt service routine.
However, imagine that the system is to be expanded to measure three pulse width input signals. Each interrupt
service routine must then have some way of keeping track of the state of the measurement. A state machine is the
right way to accomplish this.
The flow chart of the humidity pulse measurement task (interrupt service routine) is shown in figure 2.16.

2.10.2 Humidity Task Pseudo-Code

The humidity interrupt service routine is structured as a state dispatcher, that selects various state handlers, and
the four state handler routines themselves.

H_STATE: Byte {Current state of humidity task}

H_REQUEST: Byte {Measurement request flag: }
{SET by display manager, CLEARED by humidity task}
H_READY: Byte {Measurement ready flag}
{SET by humidity task, CLEARED by display manager}
H_TIME1: Word {First Edge Timestamp, microseconds}
H_TIME2: Word {Second Edge Timestamp, microseconds}
H_PERIOD: Word {Pulse width, microseconds}

Begin
Case H_STATE Of {Dispatch to the appropriate state handler routine}
0: H_DORMANT {Waiting for measurement request}
1: H_SECOND_EDGE {Waiting for second edge}
2: H_COMPLETE {Completed measurement, waiting for display manager
to acknowledge}

{State 0 Handler Routine:}

H_DORMANT: If H_REQUEST is SET

www.syscompdesign.com
 102

IRQ
Set Request Flag

0m 1mSecond Edge 2mComplete

Request Humidity
Dormant

HHH
Measurement 
HH
H
No 
HH Request? H  Measurement H
HH Accepted? 
No

HH HH
Yes Yes

HHH
Clear Request Flag Record T2  Measurement H No
HH Ready? 
HH
Record T1 Find T=T2-T1 Yes

Read Humidity
Set Ready flag

State:=1 State:=2 State:=0 Clear Ready Flag

RTI

Humidity Task Main Routine

Figure 2.16: Humidity Task Flow Chart

Then
H_REQUEST := 0 {Acknowledge request}
H_TIME1 := TIMESTAMP1 {Save timestamp}
H_STATE := 1 {Move to next state}
Exit
Else
Exit

{State 1 Handler Routine:}

H_SECOND_EDGE: H_TIME2 := FREE_RUNNING_COUNTER {Save Timestamp}
H_PERIOD:= H_TIME2-H_TIME1 {Calculate Period}
H_STATE := 2 {Move to next state}
Exit

{State 2 Handler Routine}

www.syscompdesign.com
Humidity 2.10 Humidity Measurement Software

H_COMPLETE: If H_READY = CLEARED {If reading has been taken}

Then { by main routine }
H_STATE := 0 { then move to state 0 }
Exit
Else
Exit

Exit: Return from ISR

End

2.10.3 Humidity Display Routine

The display manager
• requests a humidity reading by setting the H_REQUEST flag.
• watches the H_READY flag until it becomes set, indicating that new humidity data is available
• retrieves the new humidity value from the humidity measurement task
• maps the humidity frequency count to a value of relative humidity
• displays the humidity value on the LCD
Notice that the main routine may treat the humidity task as an independent machine. When it is requested to
make a measurement, it will eventually complete a measurement and signal that it is ready. No interventions are
required beyond setting the Request Flag (ie, making a request), copying the time value to a safe location, and
clearing the Ready Flag (ie, acknowledging that a new value has been accepted).
In pseudocode:
H_BUFFER: Word {Storage for humidity frequency count}

Begin
H_REQUEST := SET {Request a new humidity reading}

If H_READY = SET
Then
H_BUFFER := H_COUNT {Get the new reading}
H_READY := CLEAR {Signal that the reading has been received}
CALL H_CONVERT {Convert frequency to humidity}
CALL DISPLAY {Display the new reading}
End

2.10.4 Converting Humidity Sensor Frequency to Relative Humidity

To relate measured period to humidity, we begin by solving for RH in equation 5.1:
 0.714
1 Cs
Hr = 100 − 2.89 (2.13)
0.34 C12

www.syscompdesign.com
 104

where Hr is the relative humidity in percent

Cs is the sensor capacitance in farads
C12 is the sensor capacitance at 12% RH, 113 pF
The period of oscillation is given by equation 2.11. Solving for the timing capacitance C T , we have
T
CT = (2.14)
2.3RT

where CT is the timing capacitance in farads

RT is the timing resistance in ohms
T is the period of oscillation in seconds
Substituting CT for CS in 2.13 above and we have
 0.714
T
RH = 100 − 2.89 (2.15)
0.782RT C12

The measured period at the input to the microprocessor, which we will call T m , is increased by the digital
divider chain to 7 octaves above the oscillator period T . Then T m = 27 T . Substituting Tm /27 for T, 56 × 103 for
RT and 113 × 10−12 for C12 , we have the final relationship between relative humidity and period:
 0.714
T
Hr = 100 − 2.89 (2.16)
633 × 10−6

2.10.5 Calibration Software for the Humidity Sensor

The reference capacitance C12 may vary by as much as ±15% (section 2.8.1), and the oscillator frequency by an
additional ±2.5%. Thus the sensor reading may vary by a total of ±17.5% from unit to unit. We would like to
minimize the hardware cost by compensating for this variation in the humidity software.
The tolerance may be incorporated into the humidity-frequency equation as a tolerance factor K t which varies
from by 1.0 ± 0.175.
 0.714
1
RH = 100 − 2.89 (2.17)
Kt 633 × 10−6 f
To cancel out the effect of the tolerance constant Kt , we multiply the measured frequency by an adjustment
constant Ka , where Ka = 1/Kt.
 0.714
1
RH = 100 − 2.89 (2.18)
Ka Kt 633 × 10−6 f
To ensure that the adjustment has sufficient range to cancel out the tolerance, we shall assume K a varies from
0.8 to 1.2.
To calibrate the system, put the humidity sensor in an atmosphere of known humidity and adjust the constant
Ka until the display is showing the currect reading. Then install the adjustment constant K a in the processor
EEPROM.
When the processor is running through its floating point calculation of humidity, it will read the EEPROM-
installed value of Ka and use it to calculate a calibrated humidity display.

www.syscompdesign.com
Humidity 2.11 Project Exploration: Resistance Humidity Sensor

2.11 Project Exploration: Resistance Humidity Sensor

Reference [43] describes a very simple hygrometer based on the use of a hygroscopic salt, lithium chloride. This
salt absorbs moisture from the air and changes resistance in proportion to the moisture content.
A block diagram of the system and a sketch of the sensor arragement are shown in figure 2.17.
The resistance of the sensor is measured using an AC current to prevent the salt from disassociating. In the
original circuit, the oscillator and difference amplifier are each one NPN transistor and the low pass filter simply
a moving-coil current meter. The power supply is a 4.2 volt battery, unregulated.
Giannelli describes the construction of the sensor as follows:
A 1/2 inch wide strip of fiberglass cloth (about the same weight as a handkerchief) is mounted
as shown (in figure 2.17(b)). After clamping both ends in the brass plates, the weave threads are
removed from the cloth to improve response time. The completed sensor is dipped into a solution of
lithium chloride (a mound of salt about the size of a dime dissolved in a tablespoon of water). After
soaking the sensor, shake off the excess solution and allow the sensor to dry.

The Project
Redesign the circuit for operational amplifiers in the oscillator, difference amplifier, rectifier and low-pass filter
sections. Some suggestions on the redesign:

Oscillator A Phase Shift, Wien Bridge or Twin Tee oscillator circuit would work satisfactorily in this circuit. Pu-
rity of the output waveform is not terribly important, but amplitude stability could be an issue. If the power
supply is regulated, the amplitude of the oscillator can be limited simply by clipping in the operational
amplifier.
Difference Amplifier A single, four resistor difference amplifier is probably sufficient for this function. The
input to the difference amplifier may be AC coupled, since the bridge error signal will be an AC voltage.
The resistances should be chosen so that the diff amp does not significantly load the bridge circuit.
Rectifier A half-wave or full-wave precision op-amp rectifier, or a synchronous rectifier, could be used. A full
wave rectifier simplifies the design of the low pass filter.
Low Pass Filter A simple RC low pass filter on an operational amplifier active filter (such as a single-pole Sallen-
Key low pass) could be used here. Better filtering is obtained with the op-amp filter, and since op-amps are
probably available, this would be a better choice. You should ensure that the 400Hz ripple is less than one
step on the A-D converter.
Power Supply Ideally, the unit will operate from a single 5 volt regulated supply, stolen from them microproces-
sor logic. A negative voltage at currents in the order of one or two milliamps is available, or the existing 5
volt power supply could be split to provide a ±2.5 volt supply.

Several useful building blocks are shown in the circuit diagram for a syncho based wind direction interface,
figure 4.15 on page 159.

www.syscompdesign.com
 106

Humidity 3k9
Sensor

1k
100n Maximum Adj
Low
Difference Pass
250k Rectifier A/D Converter
400Hz Minimum Adj Amplifier Filter
Oscillator

2k5 200

(a) Block Diagram

2.1"

0.75"

Clamping
Bolt Brass contact pad, 4 req’d
Salt-impregnated Fiberglass
Wiring
Lug Printed Circuit Board
Material

(b) Resistance-Humidity Sensor

Figure 2.17: Variable Resistance Humidity Sensor

www.syscompdesign.com
Humidity 2.12 Project Exploration: Monostable Interface

2.12 Project Exploration: Monostable Interface

Application notes for the Philips capacitance humidity sensor [10] show the simple circuit of figure 2.18.

Co Co + ∆C
. .
... ...
.... ....
... ...
. .

r
........
.... ......... ...... ....
Astable .
....
Monostable .
.......... ......
.
....
........
.... ....

Figure 2.18: Humidity Sensor using Monostables

The astable oscillator is designed so that the period of the output waveform is proportional to a fixed capaci-
tance Co , chosen to be equal to the base capacitance of the humidity sensor. The monostable is synchronized to
the astable so that it produces pulses at the same time. The pulse width of the monostable output is proportional
to the capacitance of the humidity sensor, Co + ∆C.
The two pulses are subtracted so that the final output pulse width is proportional to the difference between
the fixed pulse width and the variable pulse width. The width of this final difference pulse is then proportional to
∆C, which is in turn proportional to the relative humidity.
The pulse width could be measured directly by the timer circuitry of a microprocessor. Alternatively, since
the pulse is of fixed and known amplitude, it may be averaged in a low pass filter to obtain a voltage proportional
to the width, and hence the humidity. This voltage could then be measured in the microprocessor A-D converter.

2.12.1 Analysis of the Astable Multivibrator

The basic circuit of the astable multivibrator is shown in figure 2.19.

www.syscompdesign.com
 108

U1 U2

r
....... .......
....... .......
....... ...... ....... .......
....... .. ....... ..
..... . ..... .
....... ....... ....... .......
....... .......
....... .......

r
R. .
.....................
. .

C.
...
.....
...
.

(a) Oscillator Circuit

U1 U2

r
........ ........
.......
....... .......
0V .......
....... .......
....... .. ....... ..
................ ................
....
.......... ....
..........

r
.... ....
R
.
.....................
.. .
+5V C. +5V
.
...
.....
...
.

(b) After Turnon

U1 U2

r
........ ........
.......
....... .......
+5V .......
....... .......
.... . .... .
....... .. ....... ..
....... ....... ....... .......
.......
. .......
.
....... .......

r
R
. ..
.............. ...
.. ..
0V C. 0V
...
.....
...
.

(c) After Transition

. ... ...
5 ...
...
...
...
... ....
.... ....
4 ....
..... ....
.....
...... ......
....... ......
........
3 ...........
..............
..
...
...
................
.
.
........
2 .......
......
.....
.
.....
.
1 ..
...
...
...
0
........... T1 . .
.................... T2 .
..........

...
....
Circuit Turn On

(d) Astable Waveform, Input to U1

Figure 2.19: Astable Multivibrator Operation

www.syscompdesign.com
Humidity 2.12 Project Exploration: Monostable Interface

After Turn On
To understand the operation of the astable, consider that it has just been turned on, and that the supply voltage is
+5 volts. The output from U2 could be 0 or +5 volts: let’s assume it’s +5 volts. The capacitor C is uncharged, as
yet, so the voltage across it is zero. Voltage cannot change instantaneously across a capacitor, so when the right
end of the capacitor is lifted up to +5 volts, the left end must go to +5 volts as well. The complete situation, just
after the power is applied, is as shown in figure 2.19(b).

During T1
These are CMOS gates, so no significant current flows into their inputs. However, the resistor R has a voltage
across it, so current flows through it, gradually discharging the capacitor. The left end of the capacitor now moves
in an exponential discharge toward ground.

Transition to T2
Eventually, at the end of a time which we’ll call T1, the voltage at the input to U1 reaches the threshold voltage of
the gate, about +2.5 volts or so. At that point, the state of U1 changes: its input is seen as below threshold (logic
0), so its output goes up to logic 1, +5 volts. This forces the output of U2 immediately down to zero volts. The
astable has changed state, and the voltages are as shown in figure 2.19(c).

During T2
Now the whole process runs in reverse. Current flows from the output of U1, through the resistor, changing the
capacitor. The input to gate U1 moves in an exponentially charging curve back up to the voltage threshold. When
it reaches the threshold, the whole cycle begins anew.
Assuming that the threshold voltage is at half the supply voltage, it may be shown that the intervals T1 and T2
are equal, and the total period T of the waveform is approximately

T = 0.693RC (2.19)
where R and C are the values of the timing resistance (ohms) and capacitance (farads).

2.12.2 Humidity Astable Schematic

The complete schematic of the Humidity Astable circuit is shown in figure 2.20.
Two CMOS quad 2-input NOR gate packages are used, type CD4001 or equivalent. U1A, U1C, and U1D are
wired as inverters. U1C and U1D are the free-running oscillator, which produces a square wave at about 10KHz.
U1A and U1B form a monostable, the pulse width of which is proportional to the humidity sensor capacitance.
Gate U2A performs the subtraction operation, so that its output width is proportional to relative humidity.
The original reference suggests that C1 be composed of two 47pF and one 22pF capacitors, all with large
positive temperature coefficients of capacitance. This will help to compensate for the temperature coefficient of
the humidity sensor.
The peak value of the output pulse is equal to the +5 volt supply voltage, and the duty cycle approaches 25%
at 100% relative humidity. The average value of the output pulse would then approach 1.33 volts, and an amplifier
stage would be required to drive the A-D converter to full scale.

www.syscompdesign.com
 110

r
.
.......... .. Ch
R1 ...
.. ...
............ .......... Humidity Sensor ..............................
.......... .....

r r
470k ....
..
.
...
...
......
.....
110 - 150 pF ...

r
... ..
...............................
.... U2B .....

r
... ......
..... .............................. . .
...
.. ... ... ...... .....
..
.........
. ..... ...........................
.. ... ...
.. U1A .......

r
.. ...
. . ..
. .
... ..... .... U1B .....
.............................. .
. .
... ..............................

r
. .. ....
..... ... ......
.
..................... . ............................. ... .....
...

r
... ...... ..
... ..... Output PW ∝ RH .. ..
.
.. ...
.. ....
.. .. U2C .....
.
.... U2A ..... .... ...
.......
.
..
.....
... ...............................
..

r
.. .
....
...........................

r r r
............................... ............................... .............................

r r
... ...... ... ...... ... ......
... .... ... .... ... .....
.. ... .. ... ... ...
.. .. ..
.... U1C ...... .... U1D ...... .... U2D ......
. . . . . .
. ....
....
......
.
. ....
....
......
.
.....
...
........
.......................... .......................... .. ..........................
C1 ... C2
...
. ........ ........
R2 .......... ..... ..... .......... .....
..
.............
110pF ...
. 3-40 pF . . ..
..... .... ....
470k

r r
..
....

C3
......
...... ...... 68 pF
U1, U2: CD4001 Quad NOR gate
. .
.... .... ....

Figure 2.20: Humidity Sensor using Monostables, Schematic

www.syscompdesign.com
Chapter 3

Wind Speed

Who has seen the wind?

Neither I nor you:
But when the leaves hang trembling,
The wind is passing through.
Who has seen the wind?
Neither you nor I:
But when the leaves bow down their heads,
The wind is passing by.
Christina Rossetti
[4]

3.1 Introduction
It is helpful to relate wind velocity to its effects using the Beaufort Scale [2] shown in figure 3.1.
Based on this scale, the wind speed measuring device, or anemometer, should have a range of zero to
120Km/hr or so in normal use. If the anemometer is to be capable of measuring record wind speeds, then the
rather more demanding maximum speed requirements are as shown in figure 3.2, from reference [3].
The highest wind speeds on the face of the earth occur in tornadoes, which contain wind velocities in excess of
480Km/hr. Tornado wind velocities are measured by doppler radar or estimated from the extent of the destruction.
Since tornadoes destroy buildings, it is probably unrealistic to expect our anemometer design to operate under
tornado conditions .
Over the years, a great many devices have been used to sense wind velocity. We shall look at two measuring
instruments, the cup-anemometer and cooling effect sensor.

3.2 Cup Anemometer

The cup anemometer usually consists of three conical or spherical cups mounted on a rotating shaft. The speed of
rotation is proportional to wind speed. The sight of such an anemometer, whirling in the breeze, is synonymous
with weather measurements.

111
 112

Beaufort Name Kilometers Effect on Land

Number per hour
0 Calm Less than 1 Calm: smoke rises vertically
1 Light Air 1–5 Weather vanes inactive: smoke drifts with air
2 Light Breeze 6–11 Weather vanes active: wind felt on face: leaves rustle
3 Gentle Breeze 12–19 Leaves and small twigs move: light flags extend
4 Moderate Breeze 20–28 Small branches sway: dust and loose paper blow about
5 Fresh Breeze 29–38 Small trees sway: waves break on inland waters
6 Strong Breeze 39–49 Large branches sway: umbrellas difficult to use
7 Moderate Gale 50–61 Whole trees sway: difficult to walk against wind
8 Fresh Gale 62–74 Twigs broken off trees: walking against wind very difficult
9 Strong Gale 75–88 Slight damage to buildings: shingles blown off roof
10 Whole Gale 89–102 Trees uprooted: considerable damage to buildings
11 Storm 103–117 Widespread damage: very rare occurance
12–17 Hurricane more than 117 Violent destruction

Figure 3.1: Beaufort Wind Scale

Canadian Record: 201.1 Km/hr, Cape Hopes Advance

(Quaqtaq) Quebec, 18 Nov 1931
World Record: 371 Km/hr, Mt. Washington
New Hampshire, 12 April 1934

Figure 3.2: Highest Wind Speed for 1 Hour

An example design, constructed from found parts by Jim Koch, is shown in figure 3.3.

www.syscompdesign.com
Wind Speed 3.2 Cup Anemometer

Plastic Measuring Spoon

(modified)

Plastic Disk

Motor Shaft

2.0 in 7.0 in
(5.1 cm) (17.8 cm)

2.0 in
(5.1 cm)

DC Generator
(Edmund Scientific #37590)

Figure 3.3: Rotating Cup Anemometer

www.syscompdesign.com
 114

Although almost any design of cup anemometer will work after a fashion, wind tunnel tests on cup anemome-
ters have shown that an optimum design will include the following features[55, 54]:

• The rotating part of the anemometer should be as light as possible to start at low wind velocities and respond
quickly to wind gusts. (In contrast to this, the British Casella anemometer type W/1254/2 the rotating part
weighs in at over 1Kg, somewhat excessive in these days of lightweight plastics. On the other hand, in the
tradition of British Robustness, it could probably survive a major hurricane.)

• The anemometer should have three cups. Two cups won’t start rotation reliably and more than three simply
add to the mass of the rotating part.

• The cups should ideally have a conical shape but spherical is also acceptable. The edges of the cups should
preferably be beaded (rounded) to reduce turbulence.

• For the most linear response of rotation speed with wind velocity, the cup diameter should be approximately
half the total wheel radius. In other words, the entire spoke should be cup, and the stalk length should be
close to zero. In the design shown in figure 3.3, the ratio of cup diameter to wheel diameter is 2/7 = 0.285,
below optimum, so we would expect a non-linear function of rotational speed versus wind velocity.

A lightweight anemometer with these design features may be expected to have a threshold wind speed of
0.8Km/hr (0.5mph) and respond to a step in wind speed with a first order time constant of 0.4 seconds.

Cup Anemometer Output Signal

In meterological applications, it has been customary to gear down the rotation of the anemometer and count the
number of turns over an interval of time. Some anemometers are equipped with a cam-operated switch which
closes briefly each time the anemometer completes a certain number of revolutions. Others include an odometer-
type turns counter1.
Counting the number of turns of the anemometer disc is the mechanical equivalent of integration, in which
case the readout is a measure of the average distance moved by the air in the time interval. This is useful to
meterologists, who are interested in the movement of weather systems. However, if one is interested in the
instantaneous velocity, as we are, some other readout method must be used.

Anemometer Distance Factor

If the circumference and output scale factor of the cup anemometer are known, then the distance the air has moved
in a give period of time may be calculated.
The key coefficient is the anemometerDistance Factor, which is the ratio of the circumference of the anemome-
ter wheel to the distance moved in one revolution of the anemometer. The distance factor may be thought of as
the slippage of the anemometer cups through the air which is required to generate torque in the anemometer
mechanism. For an ideal anemometer, the distance factor would be 1.0.
For example, suppose an anemometer has a diameter of 14 cm and rotates at 13 RPM for every Km/r of wind
speed. What is its distance factor?
1 Negretti and Zambra, manufacturers of one such instrument, suggested that the anemometer could be mounted on a mast and the turns

counter read with binoculars. This arrangement might be rather inconvenient during a force 10 gale, the very situation in which one is most
interested in wind velocity.

www.syscompdesign.com
Wind Speed 3.2 Cup Anemometer

The circumference is

S = πD
= π × 14 cm
= 43.9 cm
= 0.439 metres

The velocity constant for the anemometer is 13RPM per Km/hr. Then, in one minute, a point on the circumference
will move

S0 = 0.439 × 13
= 5.7 metres

In one minute, at 1 Km/hr, the air moves

S 00 = 1000 metres/60minutes
= 16.67 metres

The distance factor Fd is the ratio of S” to S’, or

Fd = f rac16.675.7
≈ 3 metres/metre

Thus the wind moves three metres for every metre of movement of the anemometer perifery.

Tachogenerator Output
A DC generator (or DC motor driven as a generator) generates a terminal voltage proportional to its rotational
speed, so it is useful in generating an electrical voltage proportional to wind speed. The electrical circuit is
simplicity in itself: the tachogenerator (tachometer-generator) output is connected to a voltmeter and the dial
reading of the voltmeter redrawn to read wind speed.
This works best with a moving-coil movement meter rather than a digital voltmeter: the inertia of the meter
movement averages out the ripple and commutator noise of the generator.
When the generator voltage is to be read by an A-D converter as part of a digital voltmeter or microprocessor,
an RC low pass filter will be required between the generator and A-D converter. The cutoff frequency must
be chosen so that the is effective at the lowest ripple frequency whithout materially affecting the average A-D
reading. A low frequency cutoff corresponds to a large time constant in the step response, and so an effective low
pass filter may slow the transient response unacceptably.
For example, if the step response of the anemometer to a wind gust is to have a time constant of 0.4 seconds,
then the lowpass filter should have a time constant a small fraction of this, or less than 40 milliseconds. This may
time constant not be effective in blocking noise from the generator.

www.syscompdesign.com
 116

Pulse Frequency Output

A tachogenerator has the virtue of simplicity if the output of the anemometer is to be displayed on a moving coil
meter. However, tachogenerators have brushes that become progressively more noisy and eventually wear out.
The tachogenerator signal voltage also requires analogue-digital conversion if the ultimate destination is a the
digital domain.
When a microprocessor or other digital system is to process the tachometer output, it is convenient to convert
the rotation of the anemometer into a variable frequency pulse train. The anemometer is directly coupled to an
optical encoder, a transparent disc with alternating opaque and transparent areas. The disk is positioned between
an LED and a phototransistor. As it rotates, the opaque stripes on the optical encoder interrupt the light beam,
causing a pulse waveform to be generated at the output of the phototransistor (figure 3.4).

Photo-interrupter

Axis of rotation
Photodiode (light emitter)

Phototransistor (light detector)

Figure 3.4: Optical Encoder Disc

The output pulse frequency is directly proportional to the rotational speed of the anemometer and the number
of stripes on the disk.
From this point, the usual methods of pulse frequency measurement by microprocessor may be applied to
determine the pulse frequency. For example, the anemometer pulses may be used to generate a processor interrupt.
The interrupt service routine increments an anemometer count variable and, once per second, the main routine

www.syscompdesign.com
Wind Speed 3.3 Anemometer using the Cooling Effect

reads the number of accumulated pulses to determine the current anemometer frequency and wind speed.
Accumulating the total count of anemometer pulses over a longer period of time may be used to determine the
distance of movement of the air mass.
There is nothing particularly special about the value of the number of stripes on the encoder wheel, and the
number may be reduced for simpler construction. It is not a good idea to base the measurement of wind speed
on the period of the anemometer pulse train, since the anemometer will completely stop in calm air and the pulse
period will go to infinity.

3.3 Anemometer using the Cooling Effect

We know from our earliest experience that hot food can be cooled off by blowing on it, that is, heat is removed
more quickly from a substance when the air in contact with the substance is in motion. This is the cooling effect,
and it is some function of the air speed. A measurement of the cooling effect is then an indirect measurement of
wind speed.
There are two methods of measuring the cooling effect:
Constant Power Input A temperature sensor is heated with constant power input and the temperature measured.
If the air speed increases, the temperature of the sensor decreases.
Constant Temperature A temperature sensor is held at a fixed temperature above ambient by a feedback system.
As wind speed increases, the heat loss from the temperature sensor increases. As a result, the power input
to the sensor must increase to hold the sensor at the setpoint temperature. A measure of the power input is
thereby an indicator of the wind speed.
A cooling effect sensor has no moving parts and may therefore be expected to be more reliable than the
rotating-cup anemometer. On the other hand, it turns out that the cooling effect is a non-linear function of wind
speed. The non-linearity makes the cooling effect sensor very sensitive to air movement at low air velocities and
less sensitive as the wind speed increases.
As well, the thermal inertia of the temperature sensor tends to slow the response of a cooling effect sensor, so
the sensor must have low thermal mass.

3.4 Thermal Circuits, A Review

Before launching into the details of the cooling effect sensor, we review of thermal equivalent circuit concepts.
Using this approach, we can model thermal behaviour with electric circuits familiar to electrical engineers.
Fourier’s Law of heat transfer states that the flow of heat through a heat conducting substance, due to thermal
conduction, is proportional to the temperature gradient across the conductor.
This may be stated as

P = k∆T

where

P is the power (rate of flow of energy) in watts

∆T is the temperature gradient (temperature difference) between the ends of the conductor,
degrees C
k is the constant of proportionality, known as the thermal conductivity, watts per degree C

www.syscompdesign.com
 118

For our purposes, it is convenient to rewrite this equation as

1
P = ∆T (3.1)
θ
where θ is the inverse of k, and is called the thermal resistance, degrees C per watt. This form of the equation
resembles Ohm’s law, where voltage is similar to temperature gradient, power is similar to electrical current,
and thermal resistance is similar to electrical resistance. For an electrical engineer, this helps to provide an
intuitive understanding of thermal behaviour. For example, if the thermal resistance is increased for a given power
disappation, the temperature gradient will increase.
Like voltage, the temperature gradient is measured between two points. In many cases, the reference temper-
ature (similar to voltage ground) is the ambient temperature, TA .

Example
A certain power transistor is disappating 15 watts and has a thermal resistance to ambient of 3.2 oC/watt. Calculate
the temperature rise above ambient of the junction.
Solution
The equivalent circuit is shown in figure 3.5.

Tj

15 watts m 3.2o C/W ∆T

m80
thermal-circuit-example.epic

H
H
TA

Figure 3.5: Thermal Circuit Example

Rewriting equation 3.1 to solve for the temperature gradient, we have

∆T = P ×θ
= 15 × 3.2
= 48o C above ambient

For example, if the ambient temperature is 40o C, then the transistor junction is at 40 + 48 = 88o C.

3.5 Thermal Model of the Cooling Effect

In 1914, Louis Vessot King of McGill University showed that the heat loss from a heated wire in moving air is of
the form √
P = (A + B U)∆T watts (3.2)

www.syscompdesign.com
Wind Speed 3.6 A Constant Power Anemometer

where
√
A and B u are dissapation constants, watts/◦ C
U is the air velocity (we use U rather than V to avoid confusion with Voltage
∆T is the temperature above ambient
This suggests that the thermal equivalent circuit consists of two parallel thermal resistances as shown in figure
3.6, where θS = 1/A and θD = 1/B.

Tj H
H
r

P m θS θD
√ ∆T
U

r
m80
king-equivalent.epic

Figure 3.6: Thermal Equivalent Circuit, Cooling Effect

One of these resistances is the previously identified thermal resistance θ that has been renamed θ S , to indicate
that this is the static thermal resistance, independent of wind velocity. The other thermal
√ resistance is a constant
θD (for dynamic), divided by the square root of wind velocity U . (The quantity θ D / U has the units degrees-C
per watt, so θS has different units than θD ).
If the wind velocity U is zero, the dynamic branch resistance goes to infinity and the current through that
branch is zero. As the wind velocity increases, the dynamic branch resistance decreases and provides more of a
path for the thermal power, decreasing the temperature differential.
Notice that the wind factor is the square root of the wind velocity. The effect is to make a small wind velocity
very noticeable, with decreasing effectiveness at higher velocities.
At very small wind velocity, the dynamic effect predominates, and the thermal resistance may be approximated
by

θ
θT ≈ √D (3.3)
U
(3.4)

The explanation is given in an appendix, page 139.

3.6 A Constant Power Anemometer

The conceptual circuit diagram of a constant power anemometer is shown in figure 3.7.
In this circuit, the feedback action via the operation amplifier is such as to force the two transistors to have
equal collector-emitter voltages and equal collector currents, thereby forcing them to dissapate the same power.

www.syscompdesign.com
 120

Vcc r r
......... .........
... ...
R1 ............ ............ R2

r
.. ... .......
.... ... .......
.......
........

r
- .......
.......
.......
.
U1 ...................
. ..
+ ............
.....
........

r
.......
Q1 Q2
.. .......
....... ..

r
.......... ....
..
. .. . .. .
...
...
.. .. ....... .......
.. ... .. .... ....
... .......
...... ...... . Output Voltage eo
......
W ... (negative with increasing wind)
...... .......
..... ... ... ... .......
... .... ...
............. . . .
.... .... ....

Vee r

Figure 3.7: Constant Power Anemometer Concept

In calm air, the thermal resistances of the two transistors are equal, forcing their junction temperatures to be equal
as well.
Now suppose that Q1 is cooled by moving air. Then its base-emitter voltage will tend to increase, causing it
to conduct less base current, causing it to conduct less collector current. A lower collector current I c1 results in
a higher voltage at the inverting input to the op-amp. The op-amp responds by lowering the base voltage of Q2,
lowering the collector current of Q2 until it is again equal to the collector current in Q1.
Putting it another way: the negative feedback action of an op amp forces the inverting and non-inverting
terminals to equal voltages. In this case, the negative feedback action forces the two collector resistors R1 and
R2 to have the same voltage across them and consequently the same current through them. This ensures that the
collector currents and voltages of the two transistors are forced to be equal, and therefore that they dissapate equal
power. The feedback signal eo is a measure of the cooling effect on Q1, and therefore the wind velocity.
A very simple circuit that demonstrates this concept is shown in figure 3.8.

Vcc q q q
+9V
. .
......... .........

q
R1 ............. .
............ R2

q
. . .......4
.... .... .......

q
56Ω 56Ω 2 - ..........741
.......
....... 6
......
U1 ............... ......
....... Output Voltage eo
+ .......
3 ............... ..........
. (+1.5 to -7.5V)
.
............ R4

q
Q1 Q2 7 .
.... 24KΩ
2N4401 2N4401
. .......
....... .

q
.
.......... ..
..... ..... ..... ................ ........
... ..... .
.......... R5
....... .... . .
..... ........... 100Ω
W .....
.......... R3 .......
.
............ 110Ω, 1/2W
..... ..... ..... ..... ..... .....

q q
.....

Vee
-9V

Figure 3.8: Constant Power Anemometer, Basic Circuit

www.syscompdesign.com
Wind Speed 3.6 A Constant Power Anemometer

In the original circuit, the output voltage of the op-amp is tied directly to the base of a transistor, and so the
output voltage will change only a few millivolts as transistor Q1 is cooled by moving air. In the practical circuit
of figure 3.8, a voltage attenuator between the output of the op-amp and the base of Q2 to forces the output signal
eo to change by larger amount to generate the same feedback signal. If the gain of the attenuator is 1/K, then the
change in output is multiplied by the factor K. For figure 3.8,
R5
K =
R4 + R5
100
≈
24000
so
1/K = 240
The change in the base-emitter voltage, which is proportional to the cooling effect, is multiplied by a factor of
240 as the output voltage.
As shown, the circuit of figure 3.8 has several limitations:
• The power dissapated in R3 is rather wasteful. A transistor constant current sink would be more efficient
and provide better common mode rejection in the bargain.
• The circuit requires two power supplies. A single supply would be more convenient.
• The output voltage runs from 0 volts to -7 volts. For input to a microprocessor A-D converter, a range
between 0 and +5 volts is preferred.
The circuit of figure 3.9 incorporates these refinements. The availability of a low cost, quad operational
amplifier, the LM324 makes the design much easier. The cost of an op-amp, which used to be in the tens of
dollars, is now below $0.20 each, which means they can be used with wild abandon in the circuit design.

The Complete Circuit

Amplifier U1A and Q3 establish the voltage of VR1 across R3, thereby setting the constant current to 75mA, as
in the basic circuit. In this case, however, the voltage across Q3 and R3 is about 3.4 volts, so the power loss in the
constant current sink is about a third of the previous circuit in the 110Ω resistor. The collector impedance of Q3
is also much higher than 110Ω, which will improve the common mode rejection ratio of the circuit.
An ’artifical ground’ is created at the junction of R6 and R7, allowing a single supply to be used. The ground
node of the previous circuit is then moved to the output of op-amp U1D, which buffers the voltage divider to
supply a lower source impedance for this ’ground’ to the ensuing circuit. For this to work correctly, of course, the
currents into and out of the new ground node must not exceed the capabilities of the operational amplifier, a few
milliamps.
The voltage across R5 is translated to a ground referenced signal by the differential amplifier U1C and its
associated resistors. Notice that the feedback gain of this circuit is reduced to 120, to keep the output signal less
than 5 volts.

Design Review
Now that we have a design for the circuit, it is important to check that static (DC) requirements are met. For
example, we must ensure that the transistors are not cutoff or saturated, and that the DC conditions into the
op-amps allow them to work correctly.

www.syscompdesign.com
 122

q q q q q
R1 .... R2 U1: LM324N
....... ...........

q
..... .
............ 56Ω Q1, Q2, Q3: 2N4401
56Ω ............

q q
. .......4
. .... .......

q
6 - ................ .
....... 7 R9 10k R10 ....
....... . . . .
R6 .... R8 ........... U1B ............... ..................... ..................... .....
...... ..... . . ....
11k .............. 1k ........... W + ..............
. R4 10k .
. ..... .. .
.... ......... ....... . 5 ......... ..
. . .
........

q
. .
..... .....
... . ........... 12KΩ

qq q
...... .........
11 .
12V .......
. ....
12 + ................ R13 Q1 Q2
....... 14 . .......
.......

q
....... .
13 U1D ............... .......................
......... ..
.
- ............. . ........
.. 100Ω
....... ........... R5
.......

q
.
...........
R7 .... ....... .. 100Ω
...... ....... ....
5600 .............. 3 + ................
....... 1 R14 Q3 .......
..... ....... .
.......................

q
2 U1A ............... .........
......
.......
.
- ............. . 10 + ................

q
.....
... 100Ω ....... 8
.. .......
....... U1C ............... .....
- ..............
R3 9 ............... Output Voltage eo
.
.
.... VR1
.............. .......... 33Ω (0 to +4V)

q q
. .
1N5221 ............
.
....

q q q q q q
2V R11 R12
. . . .
..................... ..................... ......
. .
10k 10k
. . . . . .
.... .... .... .... .... ....

Figure 3.9: Constant Power Anemometer Circuit, Final Version

Starting with the information provided in the text, you should be able to determine the voltages and currents
in the circuit of figure 3.9. As it says on the memo paper of a colleague: ”I’ve got a good idea: you do this instead
of me.” Work through the following list of questions about the circuit.

1. Assuming that the current out of the R6-R7 voltage divider tap is negligible compared to the divider current
(an assumption that we’ll confirm in a moment) what is the voltage at the base of transistor Q1?

2. What is the voltage across R3? What is the tail current of the differential amplifier Q1-Q2?

3. The voltage across transistor Q3 must be sufficient to ensure that it is not saturated. Calculate V ce3 .

4. A reasonable minimum value of β for the 2N4401 is 50. Estimate the base current of Q1. Does this confirm
our original assumption about the voltage divider?

5. Assuming that there is no cooling effect in Q1, calculate the collector voltage for Q1 and Q2.

6. Determine the power dissapation in Q1 and Q2.

7. The thermal resistance of the 2N4401 is given on the data sheet as 357 o C/watt. Calculate the junction
temperature, degrees centigrade above ambient, of Q1 and Q2. The data sheet maximum is 135 o C. Is this
likely to be exceeded in warm ambient conditions?

8. The output differential amplifier should not load the source that is driving it. By eyeballing the circuit,
without calculations, make an argument that this is in fact the case.

9. Resistors R9, R10, R11 and R12 are not critical in absolute value, but should match each other as closely
as possible. Why? What would be the effect if they are not matched?

www.syscompdesign.com
Wind Speed 3.7 Constant Temperature Anemometer

10. If the op amp common mode input voltage (the average voltage at the two inputs) exceeds its allowable
range, the op amp will cease to work correctly. The common mode input voltage range of the LM324 op
amp is given on its data sheet as 0 to Vcc − 1.5 volts. Does this circuit meet that requirement? What does
this requirement tell you about the range of values for resistors R1,R2?
11. The output voltage range of the LM324 is 0 to Vcc − 2.0. Is this a problem for any of the op amps in this
circuit?
12. Estimate the total power consumption of the circuit, in milliwatts.
13. The LM324 op amp was chosen for this circuit because its output voltage can go to zero volts. The output
of the 741 op-amp, in contrast, is limited to moving with 1.5 volts of either supply rail. In which of the
three op-amp positions of figure 3.9 would a 741 not operate correctly?
14. An interesting way to view this circuit: Transistors Q1 and Q2 form the differential input to an operational
amplifier, configured as an inverting amplifier, in which R4 is the feedback resistor and R5 the input resistor.
The offset voltage of this new op-amp is enhanced by differentially heating the input transistors (actually,
heating them both and cooling one of them). The offset voltage is amplified, as always, by the gain of
the op-amp circuit.Redraw the circuit, including as many components as possible in this new operational
amplifier.

3.7 Constant Temperature Anemometer

The constant power anemometer relies on the changing temperature of the wind sensor. Heating and cooling the
sensor takes time, and the time constant of the previous circuit, with the TO92 (plastic) case 2N4401 transistor as
a sensor, is about 18 seconds. This is far short of the ideal, which is about 0.4 seconds for a lightweit whirling
cup anemometer. As a consequence, the constant temperature sensor will not follow rapidly varying winds, such
as gusts.
As an attempt to speed up the reaction time of the sensor, let us turn to a constant temperature scheme. Here,
the sensor is maintained at a constant temperature above ambient, and the power required to do this is some
function of the cooling effect and the wind speed. Because the temperature sensor does not have to change
temperature, thermal inertia should be less and the instrument should respond more quickly.
As in the case of the constant temperature circuit, we will use the base-emitter voltage to signal the junction
temperature of the device. The basic scheme is shown in figure 3.10.
Both transistors are operated at the same emitter current, enforced by the current sources I E . Transistor Q2
is heated to a reference temperature above ambient by the fixed collector supply V ce2 . The operational amplifier
then compares the emitter voltages of the two transistors and drives the the collector of Q1 to a potential where
the two emitter voltages equalize. For example, if Q1 is cooler than Q2, its base emitter voltage will be larger.
The inverting terminal of U1 will then be more negative than the non-inverting terminal, and the op-amp output
voltage will rise until the two inputs are equal again.
The collector voltage of Q1 is now a measure of wind speed. As wind cools Q1, its collector voltage rises to
pump more power into Q1 to maintain its junction at the same temperature as Q2. At zero wind speed, the two
collector voltages are equal and then, as the wind speed rises, the voltage between the two collectors will increase.
A voltmeter between the collectors of Q1 and Q2 could be used directly to read out wind speed. However,
it is more useful to convert the floating differential signal at the collectors of Q1 and Q2 to a single-ended,
ground referenced signal which can be fed into an A-D converter of a microprocessor. As in the constant-power
anemometer, this may be accomplished with a simple differential amplifier.

www.syscompdesign.com
 124

Output Voltage eo

r r
+
.....
-
.....
.... ....

W..................
..... .......
..... ...
.. . Vce2
..........

r
Q1 .
Q2
.. .......
....... ..
.......... . ..
. . .. ....... .
.... ..... .... ..... .... .....

r
..
.......
.......
.......

r
.. ..
. .......
. -
...
.......
............ U1
........
.......
.......
....... + Vee
.......
........
... .......... ..........
.... .... .... ....
.... ... ... ... .... ... ... ...
IE IE

r
... ... ... ... ... ....
............. ............

Figure 3.10: Constant Temperature Anemometer Concept

A practical circuit is shown in figure 3.11.

r r r r
U2

r
R1 Q3 Output to Diff Amp LM317
510Ω
. .
....
....... +7V
.. ...

r r
..... ....... ... .... ......... .........
... ... ........... + -
... .. Out In
R4 ...........

r
C2 .....

r
............
220p ....
..
240Ω ....
.....
Vcc

r
C1 .. R2 R3 20V
...

r
100n .. .. .
.... .... ....
. ... .......
....
100Ω
. .
100Ω
. .
....
....... R5 Adj
.. .. .. .. ..
. .. ...................... ...................... ..........
..
.....
Q1 ......
........
.. .. .. .. ...........
...
Q2 1k1 .....
............
..
..........
.. .. ..
... . .... .... .... .... .....
................

r r r
.. ..... ............. 2k
4..................... .... .....
........... ........
...
.....
..
........
- 2
.... .....
.......

r
..
. ......
1 ........ .. .. ..
.... .... ....
....
W
r
..
........... ..... .. .. ..
....... U1A
....... ....... ..... ..... .....
.......
....... .......
....... + ....... 5
....... . .
..
..
11
........
....... 3
...
.......
....
.......
.... 7..................... +
........
...... ..... ....... U1B Vee
........ ......... .......
....... ... ..
....... - ...... D1

r
Q4 Q5 ........
....... 6 ..
1N4148 5V
.......
.
... ...
......
..
D2
R6 R7 1N4148
.. ..
.......... ..........
..... ..... ... ..
56Ω ............ 56Ω ............ ..... D3

r r r r
..... ..... ..
.... .... 1N4148

Figure 3.11: Constant Temperature Anemometer Circuit

In the practical circuit, the output current of the op-amp U1A is not sufficient to heat transistor Q1 so its output
current is augmented by the emitter follower Q3.
All feedback loops have the potential to oscillate, and this one is especially enthusiastic: a high frequency
waveform persisted at the output of U1A until C1, C2, R1, R2 and R3 were added to the circuit. Emitter followers

www.syscompdesign.com
Wind Speed 3.8 Comparing Thermal Anemometers: A Morality Tale

have a strong tendency to radio frequency oscillation, and a common fix is a small resistor in the base lead, which
is the function of R1, R2 and R3 in the circuit.
The constant voltage for the collector of Q2 is supplied by a three-terminal regulator IC, U2. An op amp
voltage regulator could have been used, but the LM317 has much better regulation than a simple op-amp circuit,
and can dissapate more power to boot.
Op-amp U1B and transistors Q5 and Q5 form the constant current sinks for Q1 and Q2 respectively. For their
currents to be equal, Q4 and Q5 must be similar and at the same temperature.
Preliminary tests indicated that the differential output voltage could go as high as 7 volts, With a fixed voltage
across Q2 of 7 volts and a 2 volt margin at the output of op-amp U1A, this requires that Vcc be at least 18 volts,
so 20 volts is specified here.

3.8 Comparing Thermal Anemometers: A Morality Tale

Once upon a time, I was commissioned to develop a microwave link between a balloon borne video camera and
a ground station, a distance of some 300 metres or so. Lashing up a couple of gunnplexer transmitter-receivers
[68, 69] in the lab was immediately successful, and in an hour or so, I could transmit video pictures across the lab.
Extrapolating from this, I thought that the communications link would be completely straightforward.
Over a longer range, however, the signal-noise ratio was terrible. After much experimentation and head
scratching, I finally did a proper signal budget calculation and discovered that I needed a low noise preamp circuit
and a high gain parabolic antenna to make the system work. The parabolic antenna had lots of gain but had to be
aimed precisely at the balloon and so, in the end, I had to abandon the microwave system for a diffent approach
altogether.
If I had done a systematic engineering analysis at the start of the project, I would have saved weeks of work.
It was an important lesson.
Now let us consider the circuts we described in sections 3.6 and 3.7. Both seem to be credible designs. If
breadboarded in the lab, both circuits seem to work correctly: the output voltage changes with wind speed.
However, it turns out that one of these circuits is much better than the other, and this is revealed by a careful
analysis of the workings of the circuits.
This highlights why it is critical that the performance of the circuit can be defined mathematically. A system-
atic analysis will reveal critical factors that may not be evident from a visual inspection or even the operation of a
prototype. In effect, the engineer must be creative at first, and then analytical to properly critique the design. This
can be a challenge. Psychologically, we don’t want to believe that our beautiful circuits have problems. However,
if we don’t find them, nature and Murphy’s law will.
On to the analysis:

Analysis of the Constant Power Anemometer

The basic circuit of the Constant Power Anemometer is repeated here for convenience:
As we’ve remarked before, the collector current of a junction transistor is given by

Vbe
Ic = I s e VT (3.5)

where

www.syscompdesign.com
 126

Vcc q q q
+9V
. .
......... .........

q
R1 ............. .
............ R2

q
. . .......4
.... .... .......

q
56Ω 56Ω 2 - ..........741
.......
....... 6
......
U1 ............... ......
....... Output Voltage eo
+ .......
3 ................ ..........
. (+1.5 to -7.5V)
.
............ R4

q
Q1 Q2 7 .
.... 24KΩ
2N4401 2N4401
. .......
....... .

q
.
......... ..
..... ..... ..... ................ .......
... ...... .
.......... R5
....... .... .
..... ........... 100Ω
W .....
........... R3 .......
.
........... 110Ω, 1/2W
..... ..... ..... ..... ..... .....

q q
.....

Vee
-9V

Figure 3.12: Constant Power Anemometer, Basic Circuit

Is is the saturation current

Vbe is the base-emitter voltage
VT is the thermal voltage, a temperature-dependent voltage about 26 millivolts at room tem-
perature
For figure 3.12 we have

Vbe1
Ic1 = Is1 e VT 1
Vbe2
Ic2 = Is2 e VT 2

Feedback action in this circuit keeps the two collector currents equal so

Vbe1 Vbe2
Is1 e VT 1 = Is2 e VT 2

For matched transistors, the saturation currents are equal and cancel. Taking the natural log of both sides, we
have

Vbe1 VT 1
=
Vbe2 VT 2

Now,

kT
VT =
qe

www.syscompdesign.com
Wind Speed 3.8 Comparing Thermal Anemometers: A Morality Tale

where
k is Boltzmann’s constant
T is the absolute temperature
qe is the charge on the electron
Substituting for VT in 3.6, we have
ηkT 1
Vbe1 qe
=
Vbe2 ηkT 2
qe
T1
=
T2

In words, the ratio of base-emitter voltages is the same as the ratio of the junction temperatures of the two
transistors.
By KVL around the base-emitter loop of Q1 and Q2:
eo
−Vbe1 + Vbe2 − = 0
Kg

Substitute for Vbe1 from 3.6 and

T2 eo
−Vbe2 + Vbe1 − = 0
T1 Kg

Solving for the output voltage eo we have

 
T2 − T1
eo = Kg Vbe2
T2
Not good. Recall that transistor T1 is cooled by moving air, transistor T2 is the ’still air’ reference transistor,
which will be at ambient temperature.
Ideally, the output should be independent of the ambient temperature, but the equation indicates that the output
voltage is dependent on the base-emitter voltage Vbe2 of the reference transistor and the junction temperature of
the reference transistor. Unless these factors can be made to cancel out, the temperature of the reference transistor
Q2 will have to be stabilized, which adds to the circuit complexity and power consumption.
Let’s see if we can do better with the constant temperature anemometer.

Analysis of the Constant Temperature Anemometer

The circuit concept for the constant temperature anemometer is shown again in figure 3.13 for convenience.
The emitter currents IE are forced to be equal by the two constant current sinks. If β is large, then the collector
currents Ic1 and Ic2 will likewise be equal.
By equations 3.5 and 3.6 above, we again have:

Vbe1 T1
= (3.6)
Vbe2 T2

www.syscompdesign.com
 128

Output Voltage eo

r r
+
.....
-
.....
.... ....

W..................
..... .......
..... ...
.. . Vce2
..........

r
Q1 .
Q2
.. .......
....... ..
.......... . ..
. . .. ....... .
.... ..... .... ..... .... .....

r
..
.......
.......
.......

r
.. ..
. .......
. -
...
.......
............ U1
........
.......
.......
....... + Vee
.......
........
... .......... ..........
.... .... .... ....
.... ... ... ... .... ... ... ...
IE IE

r
... ... ... ... ... ....
............. ............

Figure 3.13: Constant Temperature Anemometer Concept

In this circuit, feedback action forces the base-emitter voltage of the two transistors to be equal. Since the
base-emitter voltage is a known function of temperature, the two transistors must be at the same temperature.
In both cases, the thermal dissapation equations hold:

T1 = TA + θ1 PD1 (3.7)
T2 = TA + θ2 PD2 (3.8)

where
T1 ,T2 are the junction temperatures of the two transistors
TA is the ambient temperature
θ1 , θ 2 are the thermal resistances of the two transistors to ambient
PD1 , PD2 are the power dissapations of the two transistors
The temperatures can be in degrees Celcius or in Kelvins, as long as they are all in the same units. We’ll
choose Kelvins, since it is temperature in Kelvins that is used in the junction temperature equation for the BJT,
and this is consistent with equation 3.5.
Since these two temperatures are equal,

TA + θ1 PD1 = TA + θ2 PD2 (3.9)

Ambient temperature TA cancels out, which is a hopeful sign. We are interested in the thermal resistance θ 1
of the cooled transistor, which we can relate to wind speed using King’s equation, so solve for θ 1 :

PD2
θ1 = θ2 (3.10)
PD1

The power in each transistor is equal to its collector-emitter voltage times its collector current, so

www.syscompdesign.com
Wind Speed 3.8 Comparing Thermal Anemometers: A Morality Tale

PD1 = Vce1 Ic1 (3.11)

PD2 = Vce2 Ic2 (3.12)

Substitute in 3.10 for PD1 and PD2 :

Vce1 Ic1
θ1 = θ2 (3.13)
Vce2 Ic2

The collector currents cancel because they are equal, and we have that

Vce1
θ1 = θ2 (3.14)
Vce2

Referring now to the thermal model of the cooling effect, figure 3.6 on page 119, the thermal resistance of the
reference transistor is simply θs , the static thermal resistance. For our circuit,

θ2 = θs (3.15)

The total thermal resistance of the wind-cooled transistor is

θD
θ1 = θs k √ (3.16)
U

where U is the wind speed and the other terms are constants.
Applying the equation for parallel resistances:

√
1 1 U
= + (3.17)
θ1 θs θD

Substitute for θ1 from 3.17 into 3.14 and solve for the wind velocity:

2
= θD Veo

U (3.18)
θS ce2

This is very nice. The wind velocity depends on some constants and the square of the output voltage. No
temperature stabilization, beyond that inherently provided by the circuit, is required. It’s very clear the supply
voltage Vce2 must be stabilized, but that the output is independent of other circuit parameters.
Clearly, of the constant power and constant temperature designs, constant temperature is the one to pursue.

www.syscompdesign.com
 130

3.9 The Hot Wire Anemometer

In previous circuits to detect wind speed using the cooling effect, the BJT base-emitter voltage was used as the
temperature sensing element. The BJT has the advantage that its base-emitter voltage is a predictable, linear
function of temperature, about −2.2mV/o C, as discussed in section 1.4.2 on page 23.
However, transistor packages create considerable thermal mass, and this mass slows the response of the
anemometer. Ideally, an anemometer should respond to wind gusts with a time constant of about 0.4 seconds
while a test of a transistor with TO-92 case showed a time constant of 18 seconds.
In the constant temperature anemometer circuit, the time constant of the same transistor varied from 3 to 6
seconds, still an order of magnitude larger than desireable.
An alternative cooling effect sensor with lower thermal mass is a hot wire such as the filament of an incan-
descent lamp. The resistance of a wire conductor increases with temperature and this may be used to sense the
cooling effect. The time constant of a 6 volt, 100mA lamp was measured in the circuit of figure 3.14 at about 150
milliseconds, much less than any transistor package.

Current, mA

r
....
.......... .....
. .. .... . ....
.................... .... ......... .... ....... ......
. .. ... .
..............
70 .......
........
...........
..........
.... ......
50Ω 60
...................
.... ..
...
Function ...............
50
.
Generator ..........
.. ...................
............ 10Ω 40 ...........
...
..... ........
.......
......
......... 30
.....
....

r
.....
.
.
.. 20
.......
10
.
..........

Time, 50msec/div

Figure 3.14: Measuring Lamp Time Constant

A suitable circuit for the operation of the Hot Wire anemometer is shown in figure 3.15.
Transistor Q1 provides power to the resistor bridge, one branch of which is the hot wire. The operational
amplifier senses any voltage between its input terminals and drives transistor Q1 appropriately until the voltage
difference disappears, and the bridge is balanced. For equal values of R 1 and R2 , the bridge will balance when
Rw = R 3 .
Step by step:
If wind cools the resistive sensor, its resistance will decrease. This causes the voltage at the non-inverting terminal
of the op-amp to decrease and the output of the op-amp to increase. The increasing voltage at the output of the
op amp increases the output voltage at the emitter of Q1 , increasing the current flow through the elements of the
bridge, which increases the heat in the resistance wire Rw , causing its resistance to increase again. The net effect
is that Rw = R3 , even when wind is cooling the resistance wire. However, the voltage across the bridge will
increase when the resistance wire is cooled, and this may be used as a signal to measure the cooling effect.
Because the circuit maintains the resistance wire at a constant resistance, it also maintains it at a constant
temperature, so this is a type of constant temperature anemometer.

www.syscompdesign.com
Wind Speed 3.9 The Hot Wire Anemometer

+Vcc

.......
..
Q1
r
. ..
........

........

r
...
......

.. ..
........ ........
R1 .................. R3 ..................
r
.......
.... .... ........
.......
...

r
+ ....................
.....
U1 ..................... Output Signal
- ................
....
.......
.......

. . ..
........ ........ ........
R2 .................. Rw........................ .............
.
.... ..... .... ........ ........

r
. .... ..... .....
.... ..... ...
...
W .......
.. . .. . .. .
.... .... .... .... .... ....

Figure 3.15: Hot Wire Anemometer Circuit

Detailed Analysis
The temperature Tw of the wire is typically maintained just below incandescence, about 1000 oC, much greater
than the ambient temperature. Then the power dissapation Pw is given by

Tw − T A
Pw = (3.19)
θ
Tw
≈ (3.20)
θ
where

Pw is the power dissapation in the wire, watts

Tw is the temperature of the wire, degrees C
TA is the ambient temperature, degrees C
θ is the thermal resistance from the wire to ambient, o C/watt
When the air is moving, the thermal resistance is given by
√
1 U
= (3.21)
θ θD

where

U is the wind velocity

θD is the dynamic thermal resistance constant
Substituting for θ from 3.21 into 3.20 we have

www.syscompdesign.com
 132

Tw √
Pw = U (3.22)
θD

Now that we have the wire power dissapation related to wind speed, we need to relate the wire power dis-
sapation in the output voltage. In a normal resistor, the current is directly proportional to voltage. The power
dissapated in the resistor is then given by the familiar

V2
P = (3.23)
R

In an incandescent lamp, the resistance of the filament increases as the applied voltage increases, so the current
does not increase directly with the voltage. However, we’ll ignore this effect for the moment and deal with it later.
Assuming that equation 3.23 is correct, then, the voltage Vw across Rw is

r
Vw = Pw (3.24)
RW

If R3 is chosen to have the same resistance as Rw , the output voltage eo is simply twice Vw . Substitute for Pw
from 3.22 and we have

r
Pw
eo = 2 (3.25)
Rw
s√
Tw U
= 2 (3.26)
θD Rw
 1/2
Tw
= 2 U 1/4 (3.27)
θD Rw

The bad news here is that the output voltage eo is proportional to the fourth root of the wind speed, which
makes it a highly non-linear function, as shown in figure 3.16.
In figure 3.16, the values of k1 and k2 were chosen that the wind speed and output voltage each vary from 0
to 100 units, so that it is easy to compare the shape of the fourth root and square root functions. The fourth-root
function changes value dramatically at low wind speeds and then only very slightly at high wind speed. When
the output signal eo is applied to the input of an A-D converter, each resolveable increment in wind speed should
cause at least one increment in the A-D converter reading. For an 8 bit converter, because the fourth-root function
changes so slowly at high wind speed, the ability to detect small changes in wind speed becomes very poor.
The square root function changes quickly at low wind speeds, but there is still sufficient slope to the function
at higher wind speeds that the resolution is satisfactory.
To deal with the fourth root function, the designer could increase the resolution of the A-D converter or place
a squarer circuit between the wind sensor and A-D converter. Such non-linear function circuits are useful in many
applications, so we shall spend the next section describing this one.

www.syscompdesign.com
Wind Speed 3.10 Analog Squaring Circuit

100
eo = k1 U 1/4
eo = k2 U 1/2
80

60
OutputVoltage eo
40

20

0
0 20 40 60 80 100
Wind Velocity

Figure 3.16: Cooling Functions

3.10 Analog Squaring Circuit

The conceptual squaring circuit is shown in figure 3.17. First we will show that the output current I 3 is propor-
tional to the square of the input current I1 , and then how these currents are converted from and to voltages.

Output Current

. . .
.... .... ....
I3

r r
. . .
.... .... ....
VA
.......

Q1 .......
Q4
..
....... .....
... ..
.. .. ...
. . . ........ .......
I1 .... .... ....

r
....
...... ....... Q2 Q3

r
..... ......... ...
... ..
............... .. .......
....... ..

r
.......... ...
........ ... ........
.......
.......
.......
- .......
........ ...
.......
. ...... .......
U1 .................. ..... ... .. ....
... ..... ...
I4
+ ................. VB
..............
.
. .
..
......
.......
. . . . . . . . .
.... .... .... .... .... .... .... .... ....

Figure 3.17: Square Function Circuit Concept

www.syscompdesign.com
 134

The operational amplifier establishes that voltage VA is a virtual earth, and stays at zero volts. It also makes
VB sufficiently negative that all the input current I1 flows through transistors Q1 and Q2.
The squarer circuit relies on the exponential relationship between base-emitter voltage and collector current:

Vbe
Ic = I s e VT

This relationship holds over a very wide range of collector currents: at least 1000:1, so the dynamic range of
a non-linear function circuit using this princible will be similar.
Taking the logarithm of both sides and re-arranging:

Vbe = VT ln IIc (3.28)

s

Adding logarithms is equivalent to multiplication. Because the collector current of a transistor is exponentially
related to its base-emitter voltage, the sum of base emitter voltages creates a product of collector currents, an effect
that Gilbert [64] calls the translinear princible.
Summing base-emitter voltages around the loop of transistors, starting with Q1, we have

−Vbe1 − Vbe2 + Vbe3 + Vbe4 = 0

Substituting for the Vbe terms from 3.28 we have

I1 I2 I3 I3
−VT ln − VT ln + VT ln + VT ln = 0
Is Is Is Is
All the transistors are assumed to be matched and so all the saturation currents I s are equal. If the transistors
are all at the same temperature, then the VT terms are also equal. Transistors Q1 and Q2 are in series, so I1 = I2 .
Applying these simplifications, we have

I1 I3 I3
−2 ln + ln + ln = 0
Is Is Is
Taking the antilog of both sides,

I3 I4
= 1
I12
or

I2
I3 = I1
4

In this circuit, I4 is fixed, so it simply acts as a scaling factor, and the output current I 3 is proportional to the
square of the input current I1 .

www.syscompdesign.com
Wind Speed 3.11 A Hot Wire Anemometer Circuit

Vcc C4
.....
+15V

q
C2 1µ
500p ..
. . .....
.....

q q
...

q
...
R13
R6 ........... ....... ....................

q
.
2k ............. .........
R7 500k

q q q
Q5 I3 .......
.......

q
150K, 1% .....
- ................

q q
2N6533 . . ...

q
.................. ... U3 ..............
. ...
... + ................
C1 ... ... .
.
....
. Q1 ....... ......
.
.......
.
R1 100n .
... . LT1002
...
..... . . . .........

q
.. .
......... .... .... ....
500k, 1% ...............

q q
... C3 ....... Q4 ......
R3 .......... R4 R8 ......

q q q .... I4
.....
... ..... .......
....... R5 ...... 1n. Q3 ........

q
27Ω 2k. .. ..... 33k ............. . Output Signal
1W .................
.
- ................. 2k .....
...
Q2 .......
.......

q q
.. .. ..
..................
U1 ................... ........
........
......

q q
+ ............ ....... R9 . .....
.......
...... .....
....... - ................. 2k
LT1002 .. .. ..
................. .......... R10
R2 ..... U2 .............. .
...........
............. Rw + ................
. .. .... ... D1 .... ....
100k, 1% ............... .............. ......... #328 Lamp .. .. ....... 12k

q q q q q q q
....... D2 ... ...
... W......... ........ ...... 1N4148 LT1002 ..
....... ......
.... 1N4148 .
. . . .
...... .... .... ....
R11 ................... VR1

q
...
..
1k .............. .............
LT1004-1.2 . .
Zero Flow

q Resistance wire Rw is a #328 lamp with the

-1.2V
.
........
R12 ...........
.....

q Transistors Q1, Q2, Q3, Q4 are a matched array,

glass envelope removed. 3k3
......
Vee
Harris CA3046 or equivalent. -15V

q Adjust R13 to vary circuit sensitivity.

Connect substrate to -15V.

Figure 3.18: Hot Wire Anemometer with Squarer

3.11 A Hot Wire Anemometer Circuit

A complete hot wire anemometer circuit, from reference [59], is shown in figure 3.18.
There are a number of refinements in this circuit:

• The resistance wire is a #328 lamp, rated at 6 volts and 200mA. The glass envellope is removed and the
lamp is operated below incandesence.

• The outputs of op-amps U1 and U2 are both limited by 2kΩ resistors and clamped with diodes. The
clamping circuit ensures that an op-amp cannot produce the wrong polarity of voltage and damage the
circuitry it is connected to.

• Feedback capacitances around the op-amps helps to prevent spurious oscillation in the circuit.

• In the circuit analysis, the resistors in each arm of the bridge were assumed to be equal. In this circuit, they
are in the ratio of 5:1, which has the effect of increasing the voltage across the bridge and, more importantly,
any change in the voltage across the bridge.

• Resistor R6 supplies ’start-up’ current to drive the feedback loop into action.

www.syscompdesign.com
 136

• The output voltage from the anemometer sensor, at the emitter of Q6, is converted to a current by resistor
R7. The squarer circuit generates a current I3 proportional to the square of this current. Finally, op-amp U3
acts as a current-voltage converter to create an output voltage proportional to I 3 .
• Voltage regulator VR1 creates a very stable -1.2 volt supply which is used for the Zero Flow pot and, via
resistor R10, to generate the reference current I4 .
• Resistor R13 sets the overall circuit gain, output voltage per unit air speed, so it may need adjustment to set
a suitable range for the microprocessor A-D converter.
• Transistors Q1 through Q4 are a matched transistor array such as the Harris CA3046 or equivalent. Notice
that the substrate of the array must be connected to -15V to properly bias these transistors.

The Effect of Changing Wire Resistance

We previously noted that the characteristic of an incandescent lamp is not linear, but tends toward constant current
as the voltage is increased. We should check to determine what effect this has on the performance of the hot-wire
anemometer.
The voltage-current characteristic for a 6 volt, 100mA lamp is shown in figure 3.19.

0.06
1 0.436
I= 23.6 V 3 3
0.05 3
3
3
3
0.04 3
3
Lamp 3
Current, 0.03 3
mA 33
33
3
0.02 3
3
0.01

03
0 0.5 1 1.5 2
Lamp Voltage, V

Figure 3.19: Lamp Voltage-Current Characteristic

Fitting an equation to the data, we have that

1
I = V 0.436
23.6
The power in the lamp is the product of lamp voltage and current, so

www.syscompdesign.com
Wind Speed 3.12 Heated Thermistor Anemometer

P = V ×I
1
= V × V 0.436
23.6
1
= V 1.436
23.6
Comparing this with P = V 2 /R, it’s clear that the lamp power does not increase as rapidly with voltage as
the power in a fixed resistor. In the hot wire anemometer, this is an advantage because it tends to straighten out
the transfer function, output voltage/wind speed, of the device.
1 1
The output function is now proportional to U 3.46 rather than U 4 . After squaring, the output function is
1
proportional to U 1.46 : not quite linear, but easily linearized in software.

3.12 Heated Thermistor Anemometer

A cooling effect anemometer may be constructed using a thermistor as the temperature sensing element, in a
circuit that is very similar to the hot wire anemometer. A thermistor uses simpler construction than a hot wire and
is more rugged, so it may be more suitable for the outdoor environment.
On the other hand, a thermistor must be operated at a much lower temperature than a hot wire and so the effect
of ambient temperature is important and must be subtracted from the output signal.
As well, a thermistor has a much longer time constant than a hot wire. The typical time constant for a small
thermistor is in the order of 1 second or so. As well, the time constant is dependent in a complex manner on the
magnitude of the change in wind velocity and the bias current in the thermistor [61, 62], which may have an effect
on the calibration of the instrument.
A simple cooling effect anemometer circuit using a thermistor is shown in figure 3.20.
In contrast to the hot wire, whose resistance increases with temperature, thermistor resistance decreases with
temperature. As a result, the input terminals of the op amp in the thermistor anemometer are opposite to the
polarity of the hot wire circuit.
The value of R2 and the position of the potentiometer wiper are chosen to heat the thermistor to about 100 o C
above the ambient temperature. The appropriate level of thermistor current depends on the thermistor resistance
and its thermal resistance to ambient. (For a small value of R2, the circuit became a thermal oscillator with a
period of 8 seconds, so the thermistor must be operated as a ’constant current’ device to be stable.)
The output signal from the thermistor anemometer of reference [60] is shown in figure 3.21.
The curves fit to a function of the form

eo = ku1/8.3

which is approximately an 8th root function. Obviously, this will present even more severe problems in
dynamic range than the hot wire anemometer. One can admire the persistence of Okamoto and his co-workers in
making this work, but other approaches may be simpler.

An Engineering Proverb
Just because someone else made it work doesn’t mean you can. Check other peoples designs very carefully.

www.syscompdesign.com
 138

r r Vcc
+12V
R3
Q1 ....... 100
.. . ..
....................

r
. .. . .
TIP112 .........

.......

r
.....
....
C1
1n
...
.....
R2 ............. ...
..
...

r
..
..........
5k9 ............... .......

r
........
R1 ...
............ ....
.......
.......

r
.... - .......
.......
100k ............ .... .......
....
..
U1 ..................... Output Signal
≈ 0.8V + .............. .. +4.5V
....
....... LM324
.......

.. .
...................
RT H ....
.......... ...
..... ..... ........
..... .... ........ .........

r
. .... ..... ......
.... ..... ...
...
W ........
.. . .. . .. . . .. .
.... .... .... .... .... .... .... .... ....

r Thermistor: Stantel P23, RT H 1k6Ω at TA at 23o C

r Output voltage change ≈ 2V in gusts

Figure 3.20: Thermistor Anemometer

The Okamoto heated thermistor design is a good example. An analysis warns us that the circuit will be very
difficult to make work, and that it might be better to choose a different approach.
Another example of a difficult-to-replicate approach is shown in reference [67]. In this design, wind speed is
measured by measuring the stress in a beam that is deformed by wind drag. Unfortunately, most materials exhibit
a certain amount of hysterisis, so that the material tends to stay in a deformed state once it has been stressed. This
is not noticeable when the stress is always in the same direction, but is quite evident when the direction of the
stress is reversed.

3.13 Conclusions
• The traditional method of measuring wind speed is the rotating cup anemometer. Anemometers of this type
are not particularly accurate, especially at low wind speed. With relatively large mechanical inertia, they do
not follow wind gusts accurately.

• An alternative technique is the cooling effect anemometer, in which the thermal resistance of a heated sensor
changes with the square root of wind speed.

• A number of possible configurations are possible. The constant power anemometer using a BJT sensor
requires compensation for changes in ambient temperature, which is not particularly convenient.

www.syscompdesign.com
Wind Speed 3.14 Appendix: Simplifying King’s Law

8.5
8 3
3 3
7.5 3
3
7 3 + +
3 +
6.5 + +
Output 3
+
Voltage, 6 +
volts
5.5 +
5 5.5u0.12
4.7u0.13
4.5 3
4+
3.5
0 5 10 15 20 25
Wind Speed, metres/sec

Figure 3.21: Thermistor Anemometer, Output Signal

• The constant temperature anemometer using a BJT automatically compensates for ambient temperature. Its
output is proportional to the square root of wind speed. Transient response is relatively slow, similar to a
rotating cup anemometer.

• The hot wire anemometer has very fast transient response and ignores changes in ambient temperature. With
a squaring circuit, its response to wind speed lies somewhere between linear and square root functions.

• The heated thermistor anemometer is dependent on ambient temperature and highly nonlinear, proportional
to the eighth root of the wind speed. Some sort of complex linearization circuit would be required with this
approach.

3.14 Appendix: Simplifying King’s Law

The form of King’s law governing the heat loss from a filament in a current of air is given in his original paper
[58] as

p
H = κT + 2 (πκsσa)U 1/2 T

where

www.syscompdesign.com
 140

H is the heat loss, in watts, per unit length of the filament

κ is the thermal conductivity of the fluid, 5.66 × 10−5
T is the temperature of the filament, in degrees C
s is the specific heat of the fluid per unit mass, 0.171 calorie
σ is the fluid density, 0.00193
a is the diameter of the filament, in centimetres. Measuring a miniature lamp, we find a
diameter of about 7.6 × 10−3 cm.
U is the fluid velocity in cm/sec.
If the dynamic term is to be at least ten times the static term, then

2 (πκsσa)U 1/2 T
p
≥ 10κT

Substituting numeric values in equation 3.29, we have

√
3.96 × 10−4 U ≥ 5.66 × 10−4

and

U ≥ 2 cm/sec
≥ 0.072 km/hr

The dynamic thermal resistance, due to the movement of air, will predominate if the air velocity exceeds 0.072
km/hr, so for all practical purposes this will always be true.

3.15 Appendix: A Wind Tunnel for Calibration

A variable speed wind tunnel is useful in testing wind speed sensors. An arragement that is useful for cooling
effect sensors is shown in figure 3.22 on page 141.
Speed control of the vacuum cleaner is obtained by supplying it from a variable transformer. In our case,
we used a vacuum cleaner with a nameplate motor rating of 11 amps, and it generated wind speeds up to 89
miles/hour in the 2” diameter tunnel2
The wind speed sensor to be tested, such as a transistor or hot wire, is mounted on its circuit board, on the
end of few centimetres of stiff wire. The circuit board is then placed below the tunnel, and the sensor offered up
through the tunnel access hole. Should the sensor come loose during a test, it may be recovered from the canister
of the vacuum cleaner.
The manometer shows the difference, in inches of water, between the air pressure at the static port and the
dynamic port. To test that the manometer is operating correctly, operate the wind tunnel at a moderate speed and
2 Alas, after several hours of running at low speed, the vacuum cleaner motor began to smoke badly and then expired. It turns out that the

manufacturer relies on a high-speed flow of air across the motor to cool it. When cooled with a blast of air, the wire used to wind the motor
can be of a thinner guage, thereby reducing the manufacturing cost of the motor. However, driven at low speeds, the motor overheats. If you
build a wind tunnel like this, you’ll need a fan such a variable speed squirrel cage impeller that is designed to run at low as well as high speeds.
In this application, squirrel cage fan is probably superior to an axial fan because it can develop more back-pressure. Of course, if you have a
budget, you can buy a small wind-tunnel.

www.syscompdesign.com
www.syscompdesign.com

Wind Speed
Manometer scale
(mm)

Manometer tubing
5/15" OD,
3/16" ID

Water-soap solution

Shop-vac 1 1/4" flexible hose

(6 feet)
Variable Transformer, 15 Amp 2 1/4 " OD
Shop-Vac solid plastic pipe
Static pressure connection
Flush with inside of tunnel
Wood or Aluminum
Bell mouth entry

Dynamic port connection

Attach hose to fin, face 1 cm Dia
entry upstream. Access Hole
for Sensor
Figure 3.22: Wind Tunnel

Vacuum Cleaner
(3 HP Shop Vac, 11 Amp)

Inlet

3.15 Appendix: A Wind Tunnel for Calibration

 142

then block the opening where wind enters the tunnel. After some oscillations, the manometer should show zero
height difference between the two columns. The pressure in the tunnel is well below ambient under this condition,
but it should be equal at the dynamic and static ports, and the manometer should reflect that.
Under normal operating conditions, the height difference h in the manometer is proportional to air speed, as
follows [6], page 124:

P = ρw gh (3.29)

where
P is the pressure in dynes
ρw is the density of water, 1 gm/cm3
g is the gravitational constant, 981 cm/sec2
h is the height difference between the two manometer water columns, cm
This is balanced by the pressure of the air velocity impinging on the dynamic tube opening:

1
P = ρa V 2 (3.30)
2
where
ρa is the density of air, 1.225 × 10−3 gm/cm3
V is the velocity of the air in cm/sec.
Equating these two equations and solving for the air velocity, we have
s
2rhow gh
V = (3.31)
ρa
r
2 × 1 × 980 × h
= (3.32)
1.225 × 10−3
√
= 1264 h V in cm/sec, h in cm (3.33)
√
= 12.64 h V in m/sec, h in cm (3.34)

In case you don’t have your conversion tables handy, 1 metre/sec is equal to 3.6 kilometers per hour or 2.237
miles per hour.
As an example, we found a manometer height difference of 9.4 cm at the maximum speed of the wind tunnel.
This corresponds to:

√
V = 12.64 h (3.35)
√
= 12.64 9.4 (3.36)
= 38.8metres/sec (3.37)
= 140kilometers/hour (3.38)
= 86.7miles/hour (3.39)

www.syscompdesign.com
Chapter 4

Wind Direction

When the wind is in the east,

Tis neither good for man nor beast;
When the wind is in the north,
The skilful fisher goes not forth;
When the wind is in the south,
It blows the bait in the fishes mouth;
When the wind is in the west,
Then it is at the very best
Anonymous, quoted in J.O.Halliwell,
Popular Rymes, 1849

4.1 Introduction
In the latitudes between 30o and 60o , warm air from the tropics meets cold air from the artic. The intermingling
of these air currents causes very changeable weather, in contrast to the stable weather of tropical regions.
Changing weather manifests itself as a depression, a low pressure air system. As a result of the rotation of
the earth, the Coriolis effect causes winds in the Northern hemisphere to circulate around the depression in an
anticlockwise direction. Thus the rule,
if you stand with your back to the wind, your left arm points in the direction of the depression.
The direction of the wind is of interest not only to meteorologists in forecasting the weather, but to aircraft
and boats.

4.2 Accuracy and Resolution

A traditional division of the compass is shown in figure 4.1. The 360 degrees of the compass are divided into 16
segments, each labelled as shown in the diagram.
For a readout on the LCD of our microprocessor, we could display the appropriate segment name (NNE, for
example) and/or the wind heading in degrees. The resolution required of the direction sensing device is rather
low: one part in 16.

143
 144

North
NNW NNE

NW NE

WNW ENE

West East

ESE
WSW

SW SE

SSW SSE
South

Figure 4.1: Compass Rose

The wind direction system must include a windvane, some sort of paddle that is free to rotate with the wind
direction. In nautical environments, the main requirement of the windvane is that it not be carried off during a
gale nor seize up from corrosion.
The position of the windvane is converted into an electrical signal by a rotary transducer of some description.
The electrical signal is then processed and displayed on the computer.

4.3 Shaft Encoder

In one possible approach, the 16 points of the compass rose are directly encoded into 4 bits by an encoder of the
sort shown in figure 4.2 (reference [7]).
As the encoder shaft rotates, it moves a pattern of code tracks between the light emitting diodes and the
phototransistors. At any given rotation angle, the some phototransistors are illuminated and others dark, according
to by a 4 bit code. The code creates a pattern of high and low signals on the four output lines, which the external
circuit can interpret as an angle of rotation.
A four-bit binary code is the obvious one to use, but not the best choice. Consider that the disk tracks are
binary coded and the position is such that the disk is at the edge of a transition from 0111 to 1000. It is unlikely
that all four phototransistors will switch at once, and false states such as 0110 are likely to occur between 0111
and 1000. This difficulty can be avoided by using the Gray or reflected code shown in figure 4.3.
The name reflected code derives from the algorithm to generate the code sequence:

1. Write 0 followed by 1 in the first column.

2. Reflect this about a horizontal line, so that the next entries in the first column are 1 and 0.

3. Put zeros in front of the first half of this group, and 1’s in front of the second half.

4. Reflect the table so far, and again put 0’s in front of the first half and 1’s in front of the second half.

www.syscompdesign.com
Wind Direction 4.4 Synchro Resolver

To Wind Vane

Light Emitting Diodes

Transparent Disk
Phototransistors

Code Wheel

Code Track

Figure 4.2: Shaft Encoder

5. Continue until the table has 16 entries.

The result is a code such that only one bit changes at any transition. In other words, the detected code will
belong to one group or the other, but not some erroneous value between the two.

4.4 Synchro Resolver

The resolver is a electro-mechanical component widely used in commercial and military avionics equipment
for the measurement of shaft angle. It looks much like a small motor, but is actually more closely related to
transformers.
The resolver consists of three coils, a rotor attached to the shaft of the resolver and two orthoganal stators
attached to the frame. The terminals of each coil are brought to the outside world. An AC voltage supply,
typically 118volts RMS at 400Hz, is connected to the rotor winding. As the shaft rotates, the coupling between
the rotor and the two stators changes, changing the magnitude of the output voltage of the stator windings. The
voltage induced in a stator winding is at a maximum when the rotor and stator windings are aligned, and zero
when they are at right angles to each other, figure 4.4.
The equations for the two stator voltages are:
es1 = er Kr sin θ (4.1)

www.syscompdesign.com
 146

0000
0001
0011
0010
0110
0111
0101
0100
1100
1101
1111
1110
1010
1011
1001
1000

Figure 4.3: Gray Code

r
........
... r
.... .... r
Stator #2
..........................
. ........ ..... .... ....
. ..... .......
. ...................
... ......... .......................
.......................... ...........
... ..
. .
.. ... ...... ...

r
... ........... .......

r
..
......................... .............. ....... ....

r
. ..... .......... ...... ....
. . .. ... ......
....... ... ...
....

Rotor ....
Stator #1
...
....
....
....
...
θ ..................
...
...... .
............

Figure 4.4: Resolver

es2 = er Kr cos θ (4.2)

where
er is the rotor AC supply voltage
Kr is a constant for the resolver, typically 0.3
es1 , es2 are the voltages induced in the two stator windings
θ is the angle between the rotor and stator S1
The magnitude of the AC stator voltage follows the sine function through a full 360 o of shaft rotation, figure
4.5.
For the first 180o of rotation, the sine function and stator voltage are positive, so the AC voltage induced in
the stator is in phase with the rotor voltage. For the second 180o of rotation, the sine function and stator voltage
are negative, implying that the AC voltage induced in the stator is inverted with respect to the phase of the rotor
AC voltage.

www.syscompdesign.com
Wind Direction 4.4 Synchro Resolver

0.8

0.6

0.4

0.2

Stator Voltage
0

-0.2

-0.4

-0.6

-0.8

-1
0 90 180 270 360
Shaft Angle

Figure 4.5: Resolver SINE Output

For any given voltage on this waveform (except for those at a maximum or minimum point) there are two
possible angles. To unambiguously determine the shaft angle, we need the cosine output of the resolver as well,
figure 4.6.
1

0.8

0.6

0.4

0.2
Stator Voltage

-0.2

-0.4

-0.6

-0.8

-1
0 90 180 270 360
Shaft Angle

Figure 4.6: Resolver COSINE Output

In this system, the microprocessor reads both the Sine and Cosine voltages. The shaft angle is then

er Kr sin θ
tan θ =
er Kr cos θ
so
sin θ
θ = tan−1 (4.3)
cos θ
(Some care must be taken in calculating θ at the angles 90o and 270o , where the value of the cosine approaches
zero and the tangent function approaches infinity.)

www.syscompdesign.com
 148

Figure 4.7 shows what the stator voltage waveform would look like if the rotor were to be rotated so that one
rotation corresponded to 20 cycles of the AC excitation voltage. 1
A careful examination of that figure shows that the stator voltages in the 0-180 o segment and 180-360o seg-
ment equal in magnitude but out of phase. A suitable electronic circuit can decode the magnitude and phase to
determine the angle of the rotor.

0.8

0.6

0.4

0.2
Stator Voltage

-0.2

-0.4

-0.6

-0.8

-1
0 90 180 270 360
Shaft Angle

Figure 4.7: Resolver Stator Voltage Waveform vs Shaft Angle

The resolver is an attractive device to use for shaft angle measurement. Little torque is required to rotate the
shaft and the output function is an accurate measurement of the shaft angle.
For a low-cost consumer grade weather station, the resolver is a rather expensive device to use for wind
direction measurement. As an avionics component, the unit cost is in the order of $50 (and up) each. However,
resolvers and their near-relatives, the synchro-transmitter are commonly available on the surplus market for a very
modest price. For one or two custom-built units, it may be a practical approach.
Now we need a method of converting the resolver signals into something that can be digested by the micro-
processor. A suitable interface circuit is described in the next section.

1 To obtain the waveform shown in figure 4.7, the resolver shaft would have to be rotated at exactly the correct frequency, which is all but

impossible to do in practice.

www.syscompdesign.com
Wind Direction 4.5 Resolver Interface

4.5 Resolver Interface

4.5.1 An Aside on Circuit Design Methodology
It is customary in academic texts to begin an explanation with a general description of the matter to be studied, and
then (if space permits and the reader is fortunate) develop examples or show the principles realized in hardware.
The general description may be difficult to comprehend at first reading, but is eventually useful in extrapolating
to new applications and different implementations of the general princible.
Unfortunately, this misrepresents the process of electronic circuit design: a mixture of brainstorming, theoret-
ical analysis, exploration of the subject literature, reading of manufacturer’s catalogues, and circuit experiments,
all in no particular order. Many useful insights come from the contemplation of circuit design failures, a subject
that is almost never discussed in the literature.
Nontheless, once the thing works, we organize the results in an attactive sequence of theory, circuit and results,
as if we planned it that way in the first place.
The circuit at hand is a good example. The building blocks came from a variety of sources, a circuit lash-
up was made to work after a certain amount of tinkering, and only then did the author develop a conceptual
framework around the final circuit. The moral? Do not feel that your process of arriving at a circuit design must
proceed in a linear fashion from theory to working hardware. Life is not that neat.

4.5.2 Resolver Interface as Communications System

Some insight into the configuration of the resolver interface can be obtained by regarding the resolver as a modu-
lator in a communications system, in which the carrier is the 400Hz excitation of the resolver and the information
signal is the rotation of the shaft. The corresponding block diagram (for one channel) is shown in figure 4.8.

Ec sin(ωc t) · sin(Θ)

m
Shaft Angle Θ

Demodulator
(Phase Sensitive Rectifier)

m r c t)
ec = Ec sin(ω Ec2
sin Θ
2

400Hz
Carrier Oscillator Modulator Low Pass Filter
(Resolver)
Ec2 Ec2
2
sin Θ − 2
cos 2ωc t
Reference Signal

Figure 4.8: Resolver Interface Block Diagram

The amplitude modulated waveform, output from the resolver, is shown in figure 4.7. It is given by the
equation
Ec sin(ωc t) · sin(Θ) (4.4)
where

www.syscompdesign.com
 150

Ec is the peak value of the carrier voltage

ωc is the carrier frequency in radians per second
Θ is the shaft angle of the resolver
In spectral terms, the original 400 Hz carrier is shown in 4.9(a). With a 10 Hz modulation signal (resolver
shaft rotation of 600 RPM2 ), the amplitude modulated carrier is as shown in figure 4.9(b). In communications
terms, the amplitude modulated carrier is known as a Double Sideband Suppressed Carrier (DSB) signal.

2 In practice, of course, the resolver shaft is connected to a wind vane and does not normally rotate at this rate. As a result, the modulation

frequency is very nearly zero Hz.

www.syscompdesign.com
Wind Direction 4.5 Resolver Interface

Amplitude

0 400 Hz Frequency
resolver-carrier.epic
70%

(a) 400 Hz Carrier Spectrum

Amplitude

0 390 Hz 410 Hz Frequency

resolver-modulated.epic
70%

(b) DSB Modulated Carrier

Amplitude

0 10 Hz 810 Hz Frequency

resolver-modulated.epic
70%

(c) Demodulated Signal

www.syscompdesign.com

Figure 4.9: Resolver Spectum Diagrams

 152

To extract the angle information from this waveform, it is multiplied in the demodulator by the carrier fre-
quency,

ec = Ec sin(ωc t)

The result is a signal containing two frequencies. The first frequency is at twice the carrier and a second,
consisting only of the modulating signal
Ec2
sin(Θ)
2
is at or near zero frequency. The corresponding spectrum is shown in figure 4.9(c).

Ec2 E2
sin Θ − c cos 2ωc t (4.5)
2 2
The signal of interest is the modulation signal, which may be separated from the twice-carrier-frequency signal
by a low pass filter.
The output signal is then the modulation signal

Ec2
sin Θ (4.6)
2
The corner frequency of the low pass filter must be set low enough that the twice-carrier signal is sufficiently
attenuated. On the other hand, the modulation signal becomes a low frequency sine wave (moves away from zero
frequency) when the shaft is rotated, and this should not be attenuated by the low pass filter.

4.5.3 Resolver Interface Building Blocks

The resolver interface consists of the following functional blocks:

Carrier Oscillator
A source of 400Hz sine wave alternating voltage is required to excite the resolver rotor winding. A common sine
wave oscillator, the Wein Bridge oscillator, is shown in figure 4.10, reference [74].
The resistor/capacitor bridge network R1 C1 R2 C2 determines the frequency of oscillation ω o . If R1=R2=R
and C1=C2=C, then it may be shown that
ωo = 1/RC (4.7)
For oscillation to occur, the loop gain around the amplifier and its feedback network must have a magnitude
of unity and a phase shift of zero degrees, the so-called Barkenhousen criterion. At frequency ω o , it turns out that
the phase shift through the bridge network is 0o and the gain 1/3 volts/volt, so the amplifier must have a gain of
3 volts/volt to generate unity loop gain overall. If the amplifier gain is slightly larger than 3, the amplitude of the
oscillation will grow until the amplifier clips the waveform. In effect, this lowers the gain until it is exactly 3,
and the amplitude of the waveform stabilizes. The clipping action of the amplifier results in a slightly distorted
sine wave. This would not be acceptable in an audio application, and some more elaborate feedback stabilization
system would be required. In this application, some distortion is acceptable and a reasonable tradeoff for the
simplicity of the circuit.
Resistors R3 and R4 define the closed loop forward gain of the op-amp circuit and are proportioned to provide
a gain of slightly more than 3 to ensure that the oscillator will start reliably.

www.syscompdesign.com
Wind Direction 4.5 Resolver Interface

R1

C1
-

+
R3
C2 R2

R4

Figure 4.10: Wein Bridge Oscillator

Demodulator
An analogue multiplier could be used as the demodulator for the resolver output. However, an analogue multiplier
is a relatively complex device and it is worthwhile to look for a simplification. A useful approach is to convert the
reference waveform to a square wave of value ±1, at the same frequency as the reference. Multiplication is then
equivalent to switching the polarity of a signal, and the phase sensitive rectifier of figure 4.11 may be used.
In the frequency domain, the square wave reference signal may be regarded as a Fourier series containing
frequencies at the fundamental, and with decreasing amplitudes, at odd harmonics of the fundamental. Multi-
plication in the demodulator by harmonics of the fundamental will generate product frequencies at the sum and
difference of the input signal and the various harmonics of the reference. However, these product frequencies do
not occur at zero frequency so the low pass filter will remove them and the low pass filter correctly extracts the
zero-frequency resolver angle signal. In other words, a square wave reference signal and a polarity switcher will
correctly demodulate the resolver signal.
Referring to figure 4.11, the operation of the phase sensitive rectifier is follows: Consider initially the part of
the cycle for which the reference signal causes the switch S to be open.
• Because no current iref flows into the non-inverting terminal of the op-amp, the voltage at that terminal is
equal to the input voltage.
• By negative feedback action, the voltage at the inverting terminal must be equal to the voltage at the non-
inverting terminal.
• Then the input current ii must be zero, the feedback current if must be zero, there are no voltage drops in
any of the resistors, and the output voltage is equal to the input voltage.
• In other words, the voltage gain through this interval is +1.
Now consider second half of the cycle, for which the reference signal causes the switch S to be closed.
• The non-inverting terminal is connected to ground, so its voltage is zero.
• By negative feedback action, the voltage at the inverting terminal must be equal to the voltage at the non-
inverting terminal, so it also is zero.

www.syscompdesign.com
 154

iin if

r r b
R R

HH
b
Product Voltage Out
- HH
b r r 
R

+

b
Signal Voltage In

e
iref
b
e S

b
Reference Voltage In

Signal 0 Signal 0

Reference 0 Reference 0

Output 0 Output 0

Signal and Reference Signal and Reference

In Phase Out of Phase

Figure 4.11: Phase Sensitive Rectifier

• With the inverting terminal at zero volts and equal input and feedback resistors, the amplifier is configured
as a inverting amplifier of gain −1.
In other words, the phase sensitive rectifier has a gain of plus or minus 1, depending on whether the reference
switch is open or closed. The waveforms of figure 4.11 show the output signal when the input waveform is in
phase with the reference. If the input and reference are out of phase, then the output voltage will be inverted.
The effect of phase sensitive rectification on the signal from the resolver may be seen by comparing figure 4.7
on page 148 with the waveform at the output of the phase-sensitive rectifier, figure 4.12 on page 155.
Passing this waveform through a low-pass filter removes the high frequency components, leaving the envellope
of the waveform, a signal proportional to the sine of the resolver shaft angle. In the frequency domain, for a fixed
shaft angle, the PSR output, a full wave rectified signal, is given by the Fourier series (ref [5]):
  
2 4 cos 2ωt cos 4ωt cos 6ωt
e(ω) = Ec − + + ··· (4.8)
π π 1·3 3·3 5·7
The average (DC) term in the series, the voltage passed by the low pass filter, is
2
edc = Ec
π

www.syscompdesign.com
Wind Direction 4.5 Resolver Interface

0.8

0.6

0.4

0.2
Stator Voltage

-0.2

-0.4

-0.6

-0.8

-1
0 90 180 270 360
Shaft Angle

Figure 4.12: Phase-Sensitive Rectifier Output

The remaining AC terms contribute to the ripple at the output of the low pass filter.
The switch in figure 4.11 must conduct current in both directions, so a single junction transistor or FET will
not work properly. The 4066 Quad Bilateral Switch is a better choice. It conducts in both directions and may be
controlled by a single 5 volt signal. The on resistance of the switch is given as 1050Ω maximum, so the other
resistors in the phase sensitive rectifier should be much larger than this.

Low Pass Filter Design

The low pass filter extracts the average value of the full wave signal, and supplies some gain to make up for losses
in the system.
The circuit shown in figure 4.13 is suitable. The transfer function of the circuit shown in figure 4.13 is
 
eo (jω) Rf 1
=− (4.9)
ei (jω) Ri 1 + jωRf Cf
This describes a voltage gain function for which the low frequency gain is R f /Ri volts/volt and the gain
decreases at 20 db/decade above the cutoff frequency ωo , where
1
ωo = radians/sec (4.10)
Rf Cf
The amplitude response of the filter is shown in figure 4.14.

www.syscompdesign.com
 156

Cf

c r r c
Ri

Rf
ei HH eo
c c
13 -
HH14
+

12 

Figure 4.13: Low Pass Filter

Determining the Required Gain

The gain in the final stage is required to set the peak positive and negative values of the voltage at the input to the
A-D converter, as the angle Θ moves between +90o and −90o. The shaft angle modulates the carrier amplitude,
but the output swing is less than the amplitude of the carrier wave for two reasons:
• the transfer gain of the resolver is Kr , less than 1 (equation 4.2 on page 4.2). In the prototype, K r was
found to be 0.3 volts/volt

• the positional information is found by averaging the modulated output waveform, and the average of a
full-wave rectified waveform is 2/π of the peak value, equation 4.8.
Putting these two together, the peak output swing is
2
∆Vo = Kr × × E c × Kf (4.11)
π
where

Kr is a constant for the resolver, typically 0.3

∆Vo is the output voltage swing of the amplifier-filter stage, volts
Ec is the peak value of the carrier voltage, volts
Kf is the voltage gain of the amplifier stage, volts/volt
We might choose a voltage swing into the A-D converter ∆Vo of 4 volts p-p, 2 volts peak. The carrier voltage
swings more or less between the supply rails at 0 and +5 volts, so E c is 2.5 volts. Solving for the required gain
Kf in equation 4.11, we have

2, 0
Kf =
0.3 × π2 × 2.5
= 4.19 volts/volt

Arbitrarily choosing Ri as 100kΩ, the feedback resistor must then be at least 4.19 × 100kΩ = 419kΩ.

www.syscompdesign.com
Wind Direction 4.5 Resolver Interface

-5

-10

-15

-20
Amplitude, db
-25

-30

-35

-40

-45
0.01 0.1 1 10 100
Frequency, ω/ωo

Figure 4.14: Lowpass Filter Response

Designing the Low Pass Filter

Now we need to calculate the value of the low pass filter feedback capacitor C f . The position of the filter cutoff
frequency ωo should be set low enough that the 800Hz demodulation signal does not appear in the signal to the
A-D converter.
From the Fourier series for a full-wave rectified sine wave, equation 4.8, the most significant AC term is
 
4 cos 2ωt
e2f = Ep − · (4.12)
π 1·3

This describes a signal at a frequency of twice the carrier, 800Hz, and amplitude E p volts peak. In this case,
using the same reasoning used to develop equation 4.11, we can determine that E p = 2.5 × 0.3 = 0.75 volts peak.
Substituting in equation 4.12 for Ep and the equation of the AC component is

e2f = 0.424 cos 2ωt

The low pass filter should reduce the peak-peak amplitude of this sine wave to well below the amplitude of
one A-D converter step, which is

5 volts/256 steps = 0.0195 volts/step

We might choose an acceptable peak-peak ripple amplitude of one half this, about 9.7 millivolts. To determine
the corresponding value of ωo , we can begin by manipulating equation 4.9. For frequencies much higher than ω o
(ie, on the -20 db/decade section of the amplitude function in figure 4.14), equation 4.9 becomes

www.syscompdesign.com
 158

 
eo (jω) Rf 1
≈− (4.13)
ei (jω) Ri jω/ωo

The magnitude is given by

   
 eo  Rf 1
 = (4.14)
 ei  Ri ω/ωo

Substitute the following values:

eo = 9.7 millivolts
ei = 0.424 volts
ω =800 Hz (5 × 103 radians/sec)
Rf =430kΩ
Ri =100kΩ

Solving for ωo , we find that

ωo = 11.34 radians/second, (1.8Hz)

Now substitute for ωo and Rf in equation 4.10 and solve for Cf , from which we find that

Cf = 205 nanofarads

Using a value of 220nF in the circuit, the ripple into the A-D converter measured approximately 8mV p-p.
Notice that the angle signal sin Θ will be attenuated by the low pass filter if the shaft rotation speed is faster
than 1.8 Hz. If this is a problem, one would choose a 2nd order filter which has a sharper rolloff and can therefore
accomodate a higher cutoff frequency.

Power Supply Splitter

In previous circuits, we have taken care to fit the design to the available power supply, 0 and +5 volts. In this case,
we take a different approach, and generate a bipolar supply, ±2.5 volts.
This may be accomplished by splitting the power supply at its halfway point, using a voltage divider. The
centre point then becomes a pseudo-ground: effectively a local ground for the devices in this circuit.
If the current from the positive supply is equal to the current to the negative supply, the current in the pseudo-
ground will be zero. However, that is unlikely to be the case, and a buffer amplifier is required to supply the
difference current to ensure that the pseudo-ground stays halfway between the positive and negative supply rails.
An operational amplifier could be used to buffer the voltage divider. However, the maximum output current
for an op-amp is in the order of some tens of milliamps, which may not be sufficient. In this circuit, an audio
power amplifier, the National LM386, has been used.
The LM386 will work from power supplies as low as 4 volts, automatically biases its output at (approximately)
half the supply voltage, and can source or sink in the order of 250 milliamps, all for a cost of under a dollar. It
must, however, be decoupled with the capacitors shown in figure 4.15 to prevent spurious oscillations.

www.syscompdesign.com
Wind Direction 4.5 Resolver Interface

Resolver Interface Circuit Diagram

The complete circuit diagram of the resolver-based wind direction interface is shown in figure 4.15. The Cosine
channel is simply a copy of the Sine channel.
One minor detail that is not discussed elsewhere: the output of each amplifier-lowpass filter stage is connected
to its A-D converter input via resistor R10, 3kΩ. The input of the A-D converter is very high resistance, so there is
effectively no voltage drop across R10. However, the LMC660 is easily destabilized by load capacitance, and the
A-D converter presents a capacitive load to the op-amp. Resistor R10 separates the output of the op-amp from the
A-D input capacitance and prevents the op-amp from oscillating. Of course, the op-amps should all be decoupled
near their power supply pins, using a capacitor in the range of 10nF or so.

Low-pass Filter and Amplifier

Phase-sensitive Rectifier
C4 100n
R5 R7 R8 R10
13k 13k 82k 3k0

400Hz R9 To A-D Input

R1 5Vpp 9 - 430k
3k9 R6 8 SINE Channel Output
V+ 13k U1C
+
10 13
C1 +0.75v -
14
10n 4 1 U1D
3 - Stator 13 U2A 0 +
1 12
U1A S1 4066
+
R2 2 R3
C2 11
2
10n 3k9 2k2 Rotor

R4
1k1 Stator
S2 C4 100n
R5 R7 R8 R10
V- 13k 13k 82k 3k0
400 Hz Oscillator
R9 To A-D Input
R11 Resolver 9 - 430k
C3 R6 8 COSINE Channel Output
100k U1C
13k +
2n2 10 13
5 -
-
7 14
U1B 1 U1D
R12 + 13
+
U1 National Semiconductor LMC660 6 U2B 12
100k
U2 Motorola MC14066 4066
U3 National Semiconductor LM386 5 Vpp 2

400Hz Reference
+5VDC Supply
V+

6
2
-
To A-D Input
5
C5 3 U3
+
10n C6
4 Pseudo-common
10n

V-
5 VDC Common

Figure 4.15: Resolver Interface Schematic

www.syscompdesign.com
 160

Software
The microprocessor reads the two A-D channels and then subtracts the pseudo-ground voltage to determine the
values of the sine and cosine functions. It then applies equation 4.3 to determine the resolver shaft angle θ. Notice
that the arctan function will report an error if the angle is sufficiently close to 90 or 270 degrees that the floating
point values overflow.

Calibration Procedure
Providing the angle function is symmetrical about the zero point, it is possible to calibrate the system at the angles
90o and 270o, points where the angle function is maximum. The calibration device, a human or robot, rotates
the wind vane until the A-D voltage is at its positive maximum. This is the 90 o angle. The human or robot then
rotates the wind vane to obtain a minimum input voltage. This is the 270 o angle. The computer can now determine
intermediate voltages and their corresponding angles.
With most wind vane devices, the unit must be carefully aligned so that North on the case of the transducer is
actually pointing to geographic north. With a microprocessor controller, it is possible to have the software add a
correction angle β to the reading from the wind vane. The wind vane is caused to point to some known heading
and the correction angle incremented or decremented until the true heading is obtained. This would be particularly
convenient where the wind vane is inaccessible for adjustment, at the top of a mast, for example.

www.syscompdesign.com
Chapter 5

Air Pressure

There is a sumptuous variety about the New England weather that compels the stranger’s
admiration - and regret. The weather is always doing something there; always attend-
ing strictly to business; always getting up new designs and trying them on the people
to see how they will go. But it gets through more business in spring than in any other
season. In the spring I have counted one hundred and thirty-six different kinds of
weather inside of twenty-four hours. Mark Twain, 1876

5.1 Introduction
As the air surrounding the earth is heated by the engine of the sun and cooled by radiation into space, air density
differences from place to place result in the air movements we sense as winds. These winds bring us different
types of weather, so measuring the air pressure is a very important technique in the prediction of weather.
For example, a sudden drop in air pressure often signals the onset of stormy weather; high pressure signals
continuing fine weather.

5.2 Measuring Air Pressure

The classical method of measuring air pressure is the mercury barometer, a column of liquid that is supported by
atmospheric pressure, figure 5.1.
A closed tube is filled with mercury and then inverted into a reservoir or cistern of the liquid. The liquid
column will fall, forming a vacuum above its top surface, until the weight of the column is balanced by the
atmospheric pressure. Other liquids can be used, but mercury is attractive because its high density results in a
relatively compact instrument. For precise measurements, the observer must carefully determine the height of the
column above the level in the reservoir, and compensate for the temperature of the barometer.
In the resevoir, the downward pressure Pm of the mercury column is balanced by the air pressure PA .
Pm = P A (5.1)
The pressure of the mercury column is
fm
Pm = (5.2)
am

161
 162

..
... ...
.........
Vacuum
..
.....

Mercury
30 inches
Column ......... (1016 mB)

.....
..
vapour-pressure-a.pictex
Mercury Reservoir ......... 60%

Figure 5.1: Mercury Column Barometer

where
fm is the force exerted by the column
am is the cross-sectional area of the column
The column force is
fm = m m · g (5.3)
where
mm is the mass of the mercury column
g is the gravitational constant, 980 cm/sec2 in the cgs sys-
tem
The mass of the column is
m m = d m a m hm (5.4)
where
dm is the density of the mercury, 13.6 g/cm3 in the cgs system
am is the cross-sectional area of the mercury column
hm is the height of the column
Collapsing these equations, we obtain a useful expression for air pressure in terms of the column height and
density.
d m a m hm g
PA = (5.5)
am
= d m hm g (5.6)

For example, the so-called standard pressure of physics and chemistry causes a mercury column height of 76
cm. This is an atmospheric pressure of

PA = 13.6 × 76 × 980
= 1013 × 106 dynes/cm2

www.syscompdesign.com
Air Pressure 5.2 Measuring Air Pressure

Pressure Units
A variety of units of pressure have evolved over time. The bar is defined as 10 6 dynes/cm2 so standard pressure is
1.013 bars. Weather forcasts commonly quote air pressure in millibars, so standard air pressure is 1013 millibars.
In the metric system of measurement, the standard unit of pressure is the pascal, one newton/metre 2. Air
pressure is conveniently described in kilopascals, or kPa. Standard pressure becomes 101.3 kPa.
In the English system, the corresponding units are inches of mercury for atmospheric pressure and pounds per
square inch for pressure guages.
The values of standard pressure in some common units of pressure are summarized in figure 5.2.

1.0 Atmosphere ATM

1013 millibars mB
101.3 kilopascals kPa
76.0 centimetres mercury cm.Hg
160.2 centimetres water cm.H2 O
14.69 pounds per square inch PSI
29.92 inches mercury in.Hg
406.8 inches water in.H2 O

Figure 5.2: Pressure Units

The Aneroid Barometer

With careful attention to detail, a mercury column barometer can be very accurate. However, for household use
where accuracy is less critical, the aneroid barometer is more practical.

. Scale
. .........................................................
... ................................. . .
...... .. .. .................
........ .
.
.
.......
....
..
. .... ...........
......
....
Moving Pointer
Sealed .. ....
.......
Chamber .... ............
..
............ ............ Linkage and Pivot
...... ..........
...... ....
...... ..
........
....... .......
. . Flexible
....
... Membrane ...
. . . ..........................
.....
....
.
.
....
.. ...
.....................

Figure 5.3: Aneroid Barometer

As shown schematically in figure 5.3, the sensing element of the barometer is a sealed chamber is equipped
with a flexible membrane. As the atmospheric pressure changes, the gas in the chamber increases or decreases in
volume. The resultant slight movement of the membrane is mechanically amplified and causes a pointer needle
to move.
A typical aneroid barometer scale (unrolled) is shown in figure 5.4.

www.syscompdesign.com
 164

Inches
28.0 28.5 29.0 29.5 30.0 30.5 31.0

95 96 97 98 99 100 101 102 103 104 105

Kilopascals

Figure 5.4: Barometer Scale

We will use this in section 5.4.3 to determine the requirements for an electronic barometer.
It is also interesting to consider the absolute extremes of measured air pressure, to see if our instrument can
cope with them. According to the Canadian Encyclopedia [1], the air pressure extremes for Canada and the world
are:

Measurement Canadian Record World Record

Maximum 106.7 kPa, Mayo, Yukon Territory, 108.38 kPa, Agata, Siberia,
1 Jan 1984 31 Dec 1969
Minimum 94.02 kPa, St Anthony, Nfld, 87.64 kPa, Eye of Pacific Ocean
20 Jan 1977 typhoon June, 19 Nov 1975

Table 5.1: Air Pressure Records

Notice that the record high pressures tend to occur in artic regions, where the air mass is cold and therefore
dense. The record low occurred in the eye of a hurricane.
Domestic barometers cannot normally cope with these extremes of air pressures, but we should design for
them if the cost penalty is not severe.

5.3 Air Pressure and Altimetry

Air pressure decreases with height, an effect that is used by aircraft altimeters. If the barometer is sensitive
enough, a change of altitude by a known amount (an elevator ride, for example) may be used to calibrate the
barometer.
First, we need to know the air density, which is given by:
p
ρ= (5.7)
R·T
where
ρ = density of air, grams/cm3
P = air pressure, dynes/cm2 (or Kilopascals × 1000)
R = gas constant for air, 2.87 × 106 cm2 /sec2 o C
T = air temperature, o K

www.syscompdesign.com
Air Pressure 5.4 Electronic Measurement of Air Pressure

Then the change in air pressure is

∆P = g · ρ · ∆H (5.8)
where
∆P = change in air pressure, millibars
g = gravitational constant, 981 cm/sec2
∆H = change in height, cm

Example
Find the change in air pressure over a change in height of 30 metres if the air temperature is 20 oC and the pressure
1013 kPa.
Solution
From 5.7 above,
1013 × 1000
ρ =
2.87 × 106 × (273 + 20)
= 1.2 × 10−3 gms/cm3

Then, substituting for ρ in equation 5.8,

∆P = 981 × (1.2 × 10−3 ) × (30 × 1000)

= 3532 dynes/cm2
= 3.53 millibars

5.4 Electronic Measurement of Air Pressure

Electronic pressure sensors are used in great numbers in automobile engine control systems. As a result, suitable
air pressure sensors have become available at very reasonable cost.
The design shown here is based on the Motorola MPX100AP sensor, a sensor for absolute pressures between 0
and 1000 millibars1 (figure 5.6). The pressure sensor is essentially a miniature aneroid barometer. The membrane
is a thin silicon diaphram into which has been diffused a network of four resistors in a bridge configuration. The
resistors function as sensitive strain guages, changing resistance as atmospheric pressure deforms the diaphram,
figure 5.5.
The resistance of a conductor is given by
l
R=ρ (5.9)
A
where
1 We require a maximum pressure measurement of 1050 millibars, which exceeds the maximum rating of the sensor by some 5%. As we

will see in the design notes, using a sensor rated for higher pressure would halve the sensor electrical output signal and require double the
voltage gain from the sensor amplifier. Even the 5% overload is well below the maximum rating of the MPX100AP (2000 millibars), so we
are in no danger of damaging the sensor. Hopefully, its output remains linear in the 5% overload region.

www.syscompdesign.com
 166

..... Pressure Port

Wire Bond ... ......
... ......
... ........
... ........
...
.
Sensor Chip
......... ...........
.. ...........
..........
................... ...........
.... .. .... ..
..................
.
... .....
... ...
.. .................
........
... .. .
................
... ... .........
... ........
.......
... .......
.. .......
.
... .......
.......
... Cross-Section ........
......
Connecting Leads Vacuum Chamber
....
...
.....................................
... .......... ...................................................
... ........ ............ ........ ......
... ...... ......... ....... ......
... ..................... ...... .....
. . ..... ....
.. .... ...... ..... ...
.......... ..........
. .... ....
...
......... ... ...
... ..
.
.. ... ... ...
.... .... ... ...
... ...
... ... ... ...
.... .... ... ...
... .... .. ...
.. . .. ..
.. ... . ... ..
. ..
... ... ..... .... ..
.. ...
.... ... ...... ... ...
... ... .. ...
.
... ... .. ...
.
... ... ... ..
.
... ....
... ... ... ...
... .. .. ..
... .... .
.......
... ... .. ..
... ... ... ...
.... .... .... ...
..... ....... ..... ...
..... ...... ...................
...... ....... .. ..
....... ......... ....... .....
........ ................... ........................................
........... .....
.........................................

Top View

Figure 5.5: Pressure Sensor

ρ is the resistivity of the conductor

l is the length
A is the cross-sectional area

When a resistor is stretched, its length increases and cross-sectional area decreases, both increasing the resistance.
In most conductors, this effect is very slight. In the pressure sensors, the resistors are constructed of semiconductor
material that shows large changes with small deformations.
When configured as a strain guage bridge, the resistors are located so that diagonally opposite resistors in the
bridge change resistance in the same direction, either R(1 + ∆) or R(1 − ∆).
The differential output voltage is then simply V24 = Vcc × ∆.

Example

For the MPX100AP sensor, R = 500Ω and Vcc = 3V .

If the maximum differential output V24 , at full pressure, is 0.06 volts, determine the corresponding values of the
bridge resistors at full pressure.
Solution:
From

V24 = Vcc · ∆

www.syscompdesign.com
Air Pressure 5.4 Electronic Measurement of Air Pressure

3 Vcc
r
.
.............

..
..........
V24 ..
........
R(1 + ∆) -..... + ........................ R(1 − ∆)
r 4 2 ......... ..r.
............
.....
.... .
..
............ ......
.. ..
........... ...........
R(1 − ∆) ..... .....
R(1 + ∆)
............ ............
..... .....
.... ....

r
resistor-bridge.pictex
1 .. .. ..
............
80%

Figure 5.6: Resistor Bridge

so
V24
∆ =
Vcc
0.06
=
3
= 0.02

Then
R(1 + ∆) = 500 · (1 + 0.02)
= 510Ω

and
R(1 − ∆) = 500 · (1 − 0.02)
= 490Ω

5.4.1 Sensor Specifications

The key specifications for the MPX100AP pressure sensor are as follows:

Notes:
Pressure Range We will exceed the maximum pressure slightly to 105kPa. This is still well below the burst
pressure.
Supply Voltage If the three volt supply is obtained by dropping 2 volts across resistors in series with the sensor,
it turns out that the temperature drift of the sensor is substantially reduced (reference [80]).

www.syscompdesign.com
 168

Minimum Typical Maximum Unit

Burst Pressure - - 200 kPa
Pressure Range 0 - 100 kPa
Supply Voltage - 3.0 6.0 Volts
Supply Current - 6.0 - mA
Full Scale Span 45 60 90 mV
Offset 0 20 35 mV
Sensitivity - 0.6 - mv/kPa

Table 5.2: Pressure Sensor Specifications

Full Scale Span From these figures, we can determine that the sensor gain varies from 0.45 to 0.9mv/kPa.
Offset This voltage is caused by mismatch of the bridge resistors and appears as a fixed voltage at the output of
the sensor.
Supply Current This figure enables us to determine that the bridge resistors are nominally 500Ω.
Sensitivity KT This is a somewhat redundant statement of the transducer gain, which we determined from the
full scale span specification

5.4.2 Barometer Design Issues

There are a number of design challenges which need to be addressed in this system:

Power Supply The available power supply is +5 volts. Either the interface circuit operational amplifier must
operate from this or a converter must be available to generate the usual positive and negative voltages for
the op amp.
The latter approach requires a DC-DC converter and some method of ensuring that the output of the op amp
does not exceed the 0-5 volt range of the HC11 A/D converter input.
Ever mindful of cost, we’ve chosen the single supply approach. For the DC-DC converter approach, see
[87].
Amplifier Output Swing The output voltage of a single supply bipolar op amp such as the National LM324 or
Motorola MC34074 is very limited: 0.5 to 3.5 volts when operated from a +5 volt power supply. Some
CMOS op amps, such as the National LMC660CN, will produce a larger output swing. The data sheet for
the LMC660CN shows 0.2V to 4.7V for a +5 volt supply and load greater than 2KΩ, so this is a suitable
amplifier for the pressure sensor interface.
Temperature Drift A back-of-the-envellope calculation shows that the sensor amplifier will require a voltage
gain in the order of 300V/V or so. Any drift in offset voltage, bias, power supply or resistance values has
the potential for being amplified by this large gain to appear as drift in the output voltage. As well, the
sensor itself is sensitive to temperature.
All of these temperature effects must be checked to ensure that the circuit functions as a barometer rather
than a thermometer. References [78] and [79] mention the problem of temperature drift, a warning that it
must be taken seriously.

www.syscompdesign.com
Air Pressure 5.4 Electronic Measurement of Air Pressure

As well as designing for low temperature drift, we should choose the minimum voltage gain that meets the
requirements, thereby reducing the effect of resistor and voltage drifts.
Calibration The Pressure Sensor Specifications shown in Table 5.2 on page 168 show that the sensor gain can
vary over a 2:1 range, so some sort of calibration procedure will be required. The usual approach is to
provide two potentiometers: one for gain and the other for offset. Potentiometers are inherently undesireable
in a production design. The part cost is higher than a fixed resistor and a pot requires human intervention
for adjustment. It is preferable, if at all possible, that adjustments be done in software.

Subtraction of Bias and Offsets Referring to the Barometer Scale shown in figure 5.4 on page 164, the inter-
esting part of the air pressure signal is a 10kPa variation sitting on top of a 100kPa constant pressure. The
constant pressure must be subtracted at some point. As well, the output of the pressure sensor bridge is a
differential signal sitting on a half-supply common mode signal. The common mode signal must be ignored,
so the sensor amplifier must be differential and have a satisfactory common mode rejection ratio.

5.4.3 Barometer Resolution and Dynamic Range

An early and critical decision is the resolution of the barometer, in units of A/D counts per kilopascal of pressure.
We’d like a large resolution in order to detect small changes in air pressure. However, higher resolution requires
higher voltage gain from the interface and consequent greater sensitivity to a variety of nasty drift signals. Our
philosophy should therefore be to make the resolution no higher than necessary.
The face of an aneroid barometer is typically divided into 60 divisions [77] and weather broadcasts are typi-
cally given to the nearest tenth of a kilopascal. This would imply 100 steps over the 10kPa variation in air pressure.
Therefore, we might fix on 1 part in 100 as a suitable target for resolution.
A suitable dynamic range, referring to figure 5.4 on page 164, might be 95 to 105 kPa. This does not cope
with the extremes of pressure shown in table 5.2 on page 168, but will do for routine operation.

5.4.4 Transfer Function

It is useful to characterize the fixed component of air pressure, 100kPa, as P ref , which creates a fixed component
of voltage Vref at the input to the microcomputer A-D converter. The variation in air pressure is ±∆P a around
Pref , creating a variation in A-D voltage of ±∆VAD around Vref .
The value of ∆VAD is the product of the resolution, previously fixed at 100 steps, and the voltage per step,
19.5 mv/step, for a 5 volt, 8 bit A-D converter.
Then

∆VAD = 100 × 19.5 × 10−3 /2

= 1.95/2
= 0.975 volts

We’ll round this off to ±1.0 volts.

Now we can fix Vref . It must be large enough that the amplifier doesn’t exceed its maximum or minimum
output voltages. A good choice is 2.5 volts, halfway between 0 and 5 volts.
With this information, we can draw the transfer function, shown in figure 5.7.

www.syscompdesign.com
 170

A/D
Input Voltage
VAD
Volts 4

b
hhh
Barometer Interface Transfer Function



∆VAD 3

b
Vref 

∆VAD 2 
b


1

0
0 10 20 30 40 50 60 70 80 90 100 110 Input Pressure,
PA ,
∆P ∆P Kilopascals

Pref

Figure 5.7: Barometer Interface Transfer Function

www.syscompdesign.com
Air Pressure 5.4 Electronic Measurement of Air Pressure

Substituting two point coordinates in the straight line equation y = mx + b, we can solve for m and b,
determining that the transfer function is
VAD = 0.2PA − 17.5 (5.10)
where
VAD is the input voltage to the A-D converter
PA is the air pressure in kilopascals
Translating the interface transfer function into a block diagram, we have figure 5.8.

+m
to A-D input
0.2 V/KPa + 1.5 to 2.5 volts
PA
95 to 105 kPa −

17.5V
teb1.epic
70%

Figure 5.8: Barometer Interface Block Diagram

The typical sensor gain KT (table 5.2 on page 168) is 0.6 × 10−3 , so the amplifier gain must be 0.2/0.6 ×
−3
10 = 333 volts/volt as shown in figure 5.9.

+m
KT KA + A-D
PA
0.6 × 10−3 333 Volts/volt Input
−

0.2 V/KPa 17.5V

Figure 5.9: Interface Block Diagram, Adding Sensor

There are two practical problems with figure 5.9.

• The 100kPa pressure Pref tries to generate a 20 volt signal at the output of the amplifier KA . This will
saturate the amplifier, since it is operated from a +5 volt supply.
• The 17.5 volt offset is difficult to generate in a +5 volt system.
The solution to both problems is to divide the gain KA into two roughly equal stages, KA1 and KA2 , as shown in
figure 5.10.
In this case, the offset voltage VOS is +1.0 volts, easily generated from +5 volts. (Henceforth, for clarity, we
shall rename KA2 to KOS , the offset gain).

5.4.5 The Barometer Circuit

The final circuit is shown in figure 5.11.

www.syscompdesign.com
 172

+m
KT KA1 KA2 + A-D
PA
0.6 × 10−3 18.9 Volts/volt 17.6 V/V Input
−

333 Volts/volt

KA2
teb3.epic VOS 17.6 V/V
70%
1.0V

Figure 5.10: Interface Block Diagram, Splitting the Gain

www.syscompdesign.com
Air Pressure 5.4 Electronic Measurement of Air Pressure


+5V
U2

e
R3
3K3


r b
Offset Adj Test Point
Notch R4
1K
+5V 1 4
10T
R1 HH R2C R2D
H 1 r r r13
3 R5 VOS
12K14
U1A H
160 + 3 12K 4 820
 1
U2
r
10
2
 R2A
−
MPX100AP 3
100n R2G

r
12K 12K

r
HH
16 7

r r
500 500
4 2
Rg
10 + H
U1C H8

r
1K3 +5V
9 −
 R4
r
500 500

HH4 200K

r c
15
HH R2B H 14
H12K 2 r r r r U1D H
−
1 6 100n 13
U1B H 
−

7
5
11
+
5 12 11 6
12 To
+
  R2D R2F
A-D
12K 12K
R3
160
×18.9 V/V ×17.6 V/V

U1 National Semiconductor LMC660CN

U2 Motorola MPX100AP, absolute pressure, hose port
U3 Bourns 4116R-001-123 Resistor Array
All 12K resistors are part of U3. Small numbers are DIP package pin numbers.
Discrete resistors must be low temperature coefficient, Philips MRS 25F series or
equivalent.
R4 Cermet 10 turn pot, Bourns 3299-1-102 or equivalent.
Change Rg to alter gain.
R1,R3,Rg 160Ω, 1/4 watt, 1%

Figure 5.11: Barometer Interface Schematic

www.syscompdesign.com
 174

From the data sheet for the sensor, the resistors in the sensor are 500Ω each and 3 volts should appear across
the pressure sensor. Then resistors R1 and R2 are 160Ω each.
The instrumentation amplifier, U1A, U1B and U1C provides a high impedance input for the pressure sensor
signal. The voltage gain is given by
2R2
KA1 = 1 + (5.11)
Rg
and is set to 18.9 volts/volt. Somewhat arbitrarily, we have chosen R2 as 12KΩ, which makes R g
2R2
Rg =
KA1 − 1
2 × 12000
=
18.9 − 1
= 1341Ω

The nearest standard 5% value is 1300Ω.

If the gain needs to be adjusted, Rg should be changed. To simplify calibration, it should be a fixed resistor,
not variable.
In addition to providing voltage gain, this stage removes the 2.5 volt common mode sensor voltage.
The second stage, U1D, subtracts the offset and provides a final gain of 17.6 volts/volt. It could have been
implemented with a differential amplifier of gain ×17.6, as shown in figure 5.12.

r r e
+m
Ri Rf
KA2 + A-D
17.6 V/V
−
Input HH
V1 − HH
 Vo

r
+
KA2

e
17.6 V/V
1.0V V2 Ri Rf
Rf /Ri = 17.6 V/V

Figure 5.12: Second Amplifier Stage, Differential

However, if we’re willing to tweak the voltage at the non-inverting input, we can simplify the circuit as shown
in figure 5.13.
By superposition, the output voltage is
Rf Rf
kvo = − v1 + ( + 1)v2 (5.12)
Ri Ri
where Rf = 200K and Ri = 12K to obtain a non-inverting gain of 17.6V/V and an inverting gain of 16.6V/V
for this circuit. Then V1 , the offset voltage, should be set to 18/16.6 = 1.084 volts.
The final block diagram is shown in figure 5.14.
For good common mode rejection, the resistors of the differential stage U1C are from a resistor array. All
other resistors must be low temperature film, ±50ppm temperature coefficient. The offset pot should be cermet
for low drift.

www.syscompdesign.com
Air Pressure 5.4 Electronic Measurement of Air Pressure

r r e
+m
Ri Rf
KA2 + A-D
17.6 V/V
−
Input HH
V1 − HH
 Vo

+
KA2

e
16.6 V/V
1.05V V2

Figure 5.13: Second Amplifier Stage, Simplified

+m
KT KA1 KA2 + A-D
PA
0.6 × 10−3 18.9 Volts/volt 17.6 V/V Input
−

333 Volts/volt

KA2
16.6 V/V
1.05V

Figure 5.14: Complete Block Diagram

5.4.6 Barometric Software

In section 5.4.4, we established the transfer function of the system, relating the atmospheric pressure P A to the
input voltage of the A-D converter, VAD . Equation 5.10 was given as:

VAD = 0.2PA − 17.5 volts/kPa (5.13)

We may rewrite this as
VAD = KT F PA − VOS KOS volts/kPa (5.14)
where
KT F is the gain of the transfer function in volts/kilopascal
KOS is the offset gain, in volts/volt
VOS is the offset voltage, in volts

The transfer function gain KT F is the product of two components: the sensor (transducer) gain K T , and the
amplifier gain KA . Then substituting KT KA for KT F in equation 5.13,

VAD = KT KA PA − VOS KOS volts/kPa (5.15)

www.syscompdesign.com
 176

We have one final constant to consider. Ultimately, we would like to relate the A-D reading, N AD , to air
pressure. We do this with the relationship
VAD = NAD KS (5.16)
where KS is the step size of the A/D, in volts. For an 8 bit A-D converter operated from a 5 volt supply, the value
of KS is 5/256 = 19.5 × 10−3 volts.
Substituting NAD KS for VAD in equation 5.17 we have

NAD Ks = KT KA PA − VOS KOS volts/kPa (5.17)

Solving for atmospheric pressure PA we obtain the equation that the software must use to find atmospheric
pressure:
NAD Ks + VOS KOS
PA = (5.18)
KT KA
where
PA is the air pressure, in kilopascals, to be calculated by the computer
NAD is the A-D reading
KS is the step size of the A-D converter, in volts. In this system, it is 19.5 × 10 −3 volts
VOS is the offset voltage, in volts This value will depend on the offset value that is adjusted
into the hardware to set the 100 kPa output to 2.5 volts, and is set at calibration.
KOS is the offset gain, set at 16.6 volts/volt.
KT is the gain of the pressure transducer, nominally 0.6 × 10− 3 volts/kilopascal, set pre-
cisely at calibration.
KA is the gain of the amplifiers in the interface, about 333 volts/volt for this system.

The transducer gain KT and the offset voltage VOS must be determined in order that equation 5.18 contain
sufficient information that the computer program can relate A-D reading N AD to air pressure PA .

5.5 Barometer Calibration

In this section, we develop methods of calibrating the electronic barometer. We shall look at two manual methods
of calibration, and then an automatic method that eliminates the offset potentiometer and the need for human
intervention in adjusting it.

5.5.1 Approximate Method

Calculating Sensor Gain KT
In this method, set the potentiometer R4 so that the voltage into the A-D converter is within its operating range.
Then measure the offset voltage VOS at the test point shown on figure 5.11, and the current reading of the A-D
converter NAD . (You can obtain this from the microprocessor or from the voltage V AD into the A-D).
The the current value of air pressure PA may be obtained from a weather broadcast. The values of amplifier
gain KA and offset gain KOS are known, since they are determined by fixed resistor ratios.
This is sufficient information that equation 5.18 may be used to solve for the unknown variable, transducer
gain KT .

www.syscompdesign.com
Air Pressure 5.5 Barometer Calibration

Setting Offset Voltage VOS

Once KT is known, the offset voltage VOS may be adjusted to its correct value. To do this, again use equation
5.18. This time substitute the newly-found value for transducer gain K T together with the known values for step
size KS , amplifier gain KA and offset gain KOS .
We also know that an air pressure of 100kPa corresponds to an A/D input count of 125 (halfway between 0
and 255).

5.5.2 Accurate Method

The calibration method of section 5.5.1 is only approximate because it assumes amplifier and offset gain values,
based on nominal resistor values. A more accurate method of determining transducer gain is to apply a known
change in pressure ∆PA to the sensor and observe the corresponding change in A-D input voltage, ∆V AD . The
ratio of these two is the slope of the transfer characteristic:
∆VAD
= K T KA (5.19)
∆PA
A suitable apparatus for generating a known change in pressure is shown in figure 5.15. The liquid is water,
laced with red food colouring to make it visible. The tubing is flexible plastic hose available from the local
hardware store. The hose is filled with water so that it forms a U shape.

Move hose to
vary pressure

∆H

Water
Pressure Sensor Column

Figure 5.15: Water Manometer

The right side of the manometer is raised or lowered to create a height differential of ∆H. The resulting
pressure may be determined from figure 5.2 on page 163. For example, a pressure differential of 2 kilopascals
may be created by a height differential of
5
∆H = × 160.2
101.3
= 7.9 cm

www.syscompdesign.com
 178

Once the transducer-amplifier gain KT KA is determined, the current air pressure PA and corresponding A-D
reading NAD may be used in equation 5.18 to solve for the offset term VOS KOS . Finally, the offset voltage VOS
may be set as in the approximate procedure.
Once the values of the various terms in equation 5.18 are known, they may be entered into the computer
equation that displays the current air pressure.

5.5.3 Automatic Calibration

If the current air pressure is known and the pressure interface is constructed with fixed resistors, then it should be
possible for the microprocessor to read the A-D converter and determine the calibration constants automatically.
This need only be done once: the constants are written into EEPROM and are not changed unless the system is
re-calibrated.
Unfortunately, there is a problem. The large variation in sensor gain coupled with the high gain of the amplifier
section will cause the amplifier to saturate or cutoff unless the offset is adjusted correctly. In section 5.5.1 the
operator did this manually.
If the microprocessor can be provided with the means to adjust the offset so that the amplifier is operating in
its linear range, the microprocessor can determine the calibration constants for its computer program. This may
be accomplished by a D-A converter, controlled by the microprocessor, that generates the offset voltage V OS . It
turns out that modest resolution is acceptable. As a result, the D-A converter circuit is quite simple and may be
driven by a microprocessor parallel port.

The D-A Converter

The schematic of a suitable type of D-A converter, a voltage switching converter, [85] is shown in figure 5.16.

r b
2R
MSB

R
r
2R
NSB
R Vout

r
2R
LSB

b
p-dtoa.epic 2R
70%

Digital Register

Figure 5.16: Digital to Analog Converter

If the MSB and NSB are both at 0 volts and the LSB is at +5 volts, we may determine the effect on V out by
repeatedly applying Thevenin’s theorem to the divider circuit. Then V out = 0.625 volts. Similarily, the NSB
contributes 1.24 volts and the MSB 2.5 volts. If all the digital bits are set to logic 1 (+5 volts), then according to
the Superposition Theorem, Vout = 0.625 + 1.25 + 2.5 = 4.375 volts. In other words, this D-A has a resolution
of 0.625 volts and a range of 0 to 4.375 volts.

www.syscompdesign.com
Air Pressure 5.5 Barometer Calibration

In general,  
Vlogic N1 N0
Vout = N2 + + (5.20)
2 2 4
where
Vout is the output voltage of the D-A converter
Vlogic is the logic level into the D-A converter, 5 volts in this case
N2 , N1 and N0 are the MSB, NSB and LSB respectively of the binary number input to the D-A con-
verter
Now we need to determine how many bits are required in the D-A converter for the pressure amplifier circuit.

D-A Converter Resolution

We can determine the required resolution of the D-A converter according to the following reasoning:

• The D-A will generate a signal VOS that replaces the offset pot, which had a range of 0.75 volts to 1.65
volts.This is amplified by the offset gain KOS (16.6 volts/volt) to shift the amplifier output signal VAD up
or down. The total range of shift is then (1.65 − 0.75) × 16.6 = 18.3 volts. The step size of the D-A must
be a small fraction of this 18.3 volts.

• The output swing of the amplifier worst case occurs for a maximum sensor gain K T of 0.9mv/kPa. This is
amplified by the forward gain of the amplifier KA , 333. The output voltage swing for a full-scale change in
air pressure ∆PA of 10KPa is then

∆VAD = ∆PA × KT × KA
= 10 × 0.9 × 10−3 × 333
= 2.97volts
≈ ±1.5volts

• We would like to locate the output of the amplifier so that the signal never swings below 0.2 volts or above
4.7 volts. This allows a guard band of 0.75 volts above and below the output swing. We might somewhat
arbitrarily choose to be able to place the offset signal with a resolution of half this, 0.375 volts.

• The required resolution of the D-A is in then the order of one part in 18.3/0.375 = 48.8. The next larger
binary number is 26 = 64, so we require a 6 bit D-A converter.

This resolution is low enough that discrete 1% resistors (20KΩ and 10KΩ for example) may be used for the
R-2R resistor ladder.

CMOS Output Specifications

The usual approach to D-A design is to have the logic signals switch an accurate, stable reference voltage. How-
ever, the accuracy required of this D-A converter (1 part in 64, 1.5%) may be low enough that a CMOS latch may
be used to drive the ladder directly. This would greatly simplify the circuit design.
The output logic swing from CMOS logic (unlike the TTL family) is very nearly equal to the power supply
levels: 0 and +5 volts in this case. The data dook for Texas Instruments 74HC logic ([86]) shows a typical output

www.syscompdesign.com
 180

logic swing to within 1 millivolt of the supply levels. For the worst case, the output swing is still to within 10
millivolts of the supply levels. This suggests that the CMOS latch will drive the R-2R ladder with sufficient
accuracy.
Furthermore, the output resistance of the logic, worst case, is given in the data book as 50Ω. If we use 10kΩ
and 20kΩ as resistors in the R-2R ladder network, then the driving resistance of the logic device will be much
lower than the resistance of the ladder network, so resistive loading will not be a problem.
Based on these specs, a CMOS latch such as the 74HC273 can drive the 6 bit ladder network directly.

D-A Amplifier
Analysing the output of a 6 bit D-A converter as we did in section 5.5.3 , using equation 5.20 on page 179
(modified for 6 bit input) we can determine that the output of the 6 bit D-A converter ranges from 0 to 4.92 volts
in steps of 0.078 volts. The barometer interface circuit requires that the offset voltage V OS vary between 0.75 volts
to 1.65 volts, so amplification (actually, attenuation) and level shifting are required. The amplifier that provides
this level shifting and attenuation also serves to buffer the internal resistance of the D-A from its load.
A non-inverting amplifier won’t work, since the required gain is less than unity. The gain of an inverting
amplifier may be set to anything from zero up2 , and the sign inversion introduced by the inverting amplifier may
be taken care of in the software.
The transfer function of the amplifier circuit is shown in figure 5.17, from which the slope m and offset b may
be determined.

2
1.65 ..... ..... ............................. .....
.................
Transfer Function:
.................
.................
.................
.................
V os = −0.18VDA + 1.65
.................
.................
Offset Voltage .................
.................
.................
.................
VOS .................
.................
1 .................
.................
................. ..
......... ..
. ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... .............................. .....
0.75 ....
...
.
...
.
...
.
...
.
0 ....
...
0 1 2 3 4 .5
....

0 D-A Output Voltage VDA 4.92

Figure 5.17: D-A Amplifier Transfer Function

The circuit of the amplifier is shown in figure 5.18.

By comparison between the equation of the transfer function the equations for the amplifier, we have that

Rf
= 0.18
Ri

2 Well, actually, up to the open loop gain of the op-amp, to be precise.

www.syscompdesign.com
Air Pressure 5.5 Barometer Calibration

r r e
Ri Rf

HH
VDA − HH
 VOS

+

Vb e

Rf Rf
VOS = −VDA Ri
+ Vb (1 + Ri
)

Figure 5.18: D-A Amplifier

and
 
Rf
Vb = 1.65/ 1 +
Ri
= 1.39 volts

Looking back into the D-A converter output, the load sees an internal D-A resistance R DA of R ohms. This
internal resistance could be used as Ri , (figure 5.19A).

www.syscompdesign.com
 182

eTest point A

r r r e
RDA Rf

HH
− HH

VDA  +
 VOS
 
Vb e

D-A Converter

(a) D-A, Direct Connection to Amplifier

eTest point B

r r r e
RDA Ri Rf

HH
− HH

VDA   VOS

+

Vb e

D-A Converter

(b) D-A, With Ri

Figure 5.19: D-A Amplifier Connection

www.syscompdesign.com
Air Pressure 5.5 Barometer Calibration

However, in this arrangement there is no voltage signal representing the D-A output voltage by itself: the test
point A is a virtual earth. For troubleshooting purposes, it is better to provide an extra resistance and test point B
(figure 5.19B) where the output of the D-A can be measured. The voltage at B will be half the open circuit D-A
voltage VDA , but can be used to indicate that the D-A is operating correctly.

Automatic Calibration: Schematic

The complete schematic of the system, including a D-A converter for automatic calibration, is shown in figure
5.20.

www.syscompdesign.com
 184

e Test point B

r r r r
20k 10k 3k6
D7
HH
r
10k
20k H
U2A H
−


D6 72k
r 
+

r
10k
MPP 20k 1.39V
D5
U6 27k

r
Parallel Output Port 10k
From Microprocessor 20k
D4
10k
r
20k
D3
70% Offset Adj

r r b
10k
p-schematic-autocal.epic 20k Test Point
D2

b
20k VOS

+5V
R1 HH R2C R2D
3
H 1 r r
12K14
r13
U1A H
160 + 3 12K 4

 1
U2
r
10
2
 R2A
−
MPX100AP 3
100n R2G

r
12K 12K

r
HH
16 7

r r
500 500
4 2
Rg
10 + H
U1C H8

r
1K3 +5V
9 −
 R4
r
500 500

HH4 200K

r c
15
HH R2B H 14
H12K 2 r r r r U1D H
−
1 6 100n 13
U1B H 
−

7
5 5 12 11 6
12
+

 11
To
+
  R2D R2F
A-D
12K 12K
R3
160
×18.9 V/V ×17.6 V/V

Figure 5.20: Barometric Pressure Interface, Automatic Calibration

www.syscompdesign.com
Air Pressure 5.6 Reliability of the Design

Automatic Calibration: Algorithm

The microprocessor essentially mimics the manual calibration process of section 5.5.1 on page 176. The only
information it needs from the external world is the current value of atmospheric air pressure P A .

1. When the microprocessor first reads the A-D input voltage VAD , it will probably be near +5 volts or 0
volts. The processor then monitors that voltage while increasing the count into the D-A converter, which
causes the offset voltage to ramp between its maximum and minimum values. At some point, the A-D input
voltage should move to a value near the centre of the A-D input range. At this point, the micro freezes the
value in the D-A register. Because it knows the constants relating D-A count and offset voltage, the micro
now knows the value of the offset voltage VOS that moves the interface into its linear region.

2. The micro uses the current value of air pressure PA with the known values of amplifier gain KA , offset gain
KOS and the offset voltage VOS that it set in the previous step, in equation 5.18 (page 176) to calculate the
sensor gain KT .

3. The micro now adjusts the D-A output so that the offset voltage V OS is at such a value that an air pressure
PA of 100kPa would cause an A-D input voltage VAD of 2.5 volts.

4. The micro stores the current values of the interface constants K A , KOS , VOS and KT in semi-permanent
EEPROM memory. The interface is now calibrated and A-D readings can be used to calculate and display
the current air pressure.

An assembly line production would use this process. An external control computer would download a cali-
bration program and the current air pressure into the microprocessor. The calibration would take place without
human intervention.
The calibration program should also have the capability for detecting that calibration did not occur properly.
Then a failed production unit can be shunted into a reject bin for rework.

5.6 Reliability of the Design

Now that we have a circuit design, we must ensure that the circuit will work reliably, allowing for component
tolerances and the effect of temperature induced drift of the components.
For example, the sensor constant can vary over a range of 2:1. Resistors have a tolerance of ±5%. The
operational amplifiers have offset voltages which can vary by ±7mV. Can we be sure that the circuit will work
when components of the tolerance extremes are used in the circuit?
The sensor has a temperature coefficient of −0.16% per degree C, the resistors change by 250ppm (parts per
million) per degree C, and the amplifier offset voltages may change by as much as ±30µvolts per degree C. What
effect will these drifts have on the operation of the barometer, bearing in mind that the circuit is not supposed to
act as a thermometer?
We can and should build and test one or more prototypes. However, the correct functioning of a prototype
is a necessary but not sufficient condition to determine a reliable design. The fact that a prototype works merely
means that at least one version of the circuit will function. It’s no guarantee that all circuits will function.
To ensure the reliable operation of the circuit, the correct strategy is to perform an engineering analysis,
checking circuit operation by calculation and simulation. This will provide the necessary confidence to build the
circuit in quantity, and be assured that it will function under all specified conditions. Where possible, to ensure
that some massive blunder has not occurred, the calculations should be checked against the prototype. If the

www.syscompdesign.com
 186

calculations and simulation accurately predict the behaviour of the prototype, then we can have some confidence
in the predictions.

5.6.1 Circuit Tolerance Analysis

The tolerances of the circuit components raise two concerns:

• Will the circuit function, or will some voltage or current run into saturation or cutoff?

• Can the circuit calibration procedure compensate for circuit tolerances, or do we lose measurement accuracy
under some conditions?

This is potentially an unweildy problem, because of the combinatorial explosion of tolerance variables. The
parameters and equations of a spreadsheet model for the manual offset adjustment version of the interface are
shown in figure 5.21. A typical spreadsheet printout is shown in figure 5.22.

+5V
R3

Offset Adj
R4 V8

HH
Vs
V1 = Vcm + + Vso
H V3 = rV1 + I1 R2 r
2
V5 = V3 /2 R5
U1A H
+
Rs 
 R2A
−
Vso R2C R2D R2G

r
r V9 = V 7
V1 HH
H
Vs

r
+
U1C H
2


V1 −V2

r
Vs Rg I1 = Rg


−
R4
Vs V2
Vcm HH
r c
2
HH R2B H
r r r U1D H
2.5V −
H R2D R2F
U1B H 
−
Rs


+

 V4 = V2 − I1 R2
+
V2 = Vcm − Vs V6 = V 5 V7 = 2V5 − V4
2

×18.9 V/V ×17.6 V/V

Figure 5.21: Circuit Equations

The tinkering with spreadsheet model turned up the following results:

• The sensor offset Vso does not cause the amplifier to saturate but does have a dramatic effect on the output
voltage VAD . (Notice that sensor offset Vso is a property of the sensor: do not confuse it with the offset
voltage VOS of the interface.)

www.syscompdesign.com
Air Pressure 5.6 Reliability of the Design

Sensor Constant KT (data) 0.6 × 10−3 V/kPa

Air Pressure PA (data) 100 kPa
Sensor Offset Vso (data) 10 × 10−3 V
Gain Resistor Rg (data) 1300 Ω
Resistor R2 (data) 12000 Ω
Sensor Output Vs KT P A 0.06 V
V1 Vcm + V2s + Vso 2.53 V
V2 Vcm − V2s 2.47 V
I1 (V1 − V2 )/Rg 48 µAmp
V3 V1 + R 2 I 1 3.084 V
V4 V1 − R 2 I 1 1.916 V
V5 V3 /2 1.542 V
V6 V5 1.524 V
V7 2V5 − V4 1.167 V
Kos V8 (data) 1.68 V
V9 V7 1.167 V
Resistor R4 (data) 200000 Ω
R4
Output Voltage VAD V7 (1 + R 2
) − V8 R
R2
4
2.00 V
VAD
A/D Reading NAD 5 256 102 counts

Figure 5.22: Amplifier Spreadsheet Model and Results

• The offset voltage VOS may be adjusted to compensate for the effect of sensor offset voltage, but a larger
range of offset is required than that originally anticipated. The output voltage V AD is very sensitive to the
setting of VOS , so the pot R4 should be a multi-turn unit.
• A combination of large offset and high sensor gain KT cause V3 , the output of U1A, to exceed 3.5 volts.
This is the maximum output of the LM324 operational amplifier, and so an LMC660 is required.
• For low sensor gain, the change in A-D reading over the full range of air pressures (95 to 105 kPa) is over
60 counts, so the resolution is satisfactory even for low sensor gain.
These results could have been predicted from an analysis of the circuit, but the spreadsheet model makes it
easy to explore the effect of a variety of options and combinations of parameters.
A circuit simulation program such as SPICE could also be used to analyse the circuit, and is a better choice
where an accurate op-amp model is required. The spreadsheet model assumes ideal op amps. On the other hand,
spreadsheet programs are readily available and easy to use.

5.6.2 Temperature Drift

In every engineering project, there is at least one killer problem which determines success or failure. It’s important
to identify the killer problem as early as possible. In this system, the killer problem is temperature drift.
There are three evident sources of temperature drift:
Pressure Sensor Drift An analysis of the pressure sensor temperature drift [80] in the circuit of figure 5.11 shows
two competing effects: the gain of the sensor decreases with temperature, but this is partially compensated
by in increase in bridge resistance. The net result is a coefficient of −0.16% per degree C.

www.syscompdesign.com
 188

The effect on the output is a change of A-D reading of about 1.6 counts/ oC. Over a ±10o C temperature
range, this is an error of 16 counts out of a total of 100, or a 16% error: not very acceptable. Fortunately,
since it is a predictable effect, it may be compensated for by measuring the ambient temperature and modi-
fying the sensor constant.
Offset voltage drift For the LMC660, the typical figure for offset voltage drift is given as 1.3µvolts per degree
C. The spreadsheet model (or an algebraic analysis) turn up the result that this causes a drift of about 0.5
mv/o C, much less than one count of the A-D converter. This can therefore be neglected.

Resistor temperature coefficients Most of the resistors (R2A through R2H) are on the same package and so can
be expected to track in temperature. As well, the two amplifier gains are the result of ratios of resistance
(equations 5.11 and 5.12), so the temperature coefficients may be expected to cancel. Simulation of resistor
drift with the spreadsheet confirms this: the effects of resistor drift are small enough to be neglected.

www.syscompdesign.com
Chapter 6

Software Design

The design and construction of reliable control software must be approached with the same respect and care as the
design of hardware systems. It is usual to underestimate the difficulty of building a computer program. Programs
often contain defects and bugs that are poorly understood and difficult to eliminate.
In this part of the book, we provide some general approaches and specific techniques that help organize the
design of software and make the process of development more predictable and systematic.
The ideas of this section are applicable to any microprocessor, but the assembly language examples are specific
to the Motorola 68HC11 microprocessor. For information on programming the 68HC11, consult [98].

6.0.3 The Complexity Monster

Programming in machine language is difficult. Unlike Pascal and even C, the structure of a machine language pro-
gram is unconstrained and unstructured. Thus, any discipline or structure that is imposed on a machine language
program originates in the design skills of the programmer: it is not encouraged by the features of the language.
This may be construed as an argument for using a structured language such as C for program development and
indeed, C has many attractions. However, as the integrators of software and hardware, electrical engineers must
be able to work at the machine language level to develop and test hardware and to build the structures that will
support a high level language. Thus, although it is possible to minimize the use of C in program development, it
is neither possible nor desirable to avoid assembly language entirely. 1
To understand the complexity of assembly language, consider the diagram of figure 6.1.
Each block in this diagram represents a machine instruction. As the machine executes these instructions,
branch instructions and jump instructions can transfer execution to other instructions.
The first program contains two instructions, which can transfer control only to each other. The second program
contains three instructions, and there are 6 possible transfer paths. With three instructions, there are 12 paths, and
so on. In general, for n instructions, there are n × (n − 1) possible transfer paths.
In practice, only some of the instructions in a computer program transfer control, so this analysis overstates the
case. However, it is true that the number of transfer paths increases very rapidly with the number of instructions.
Now, a moderate sized assembly language control program might contain two or three thousand machine
instructions. For 3000 lines of code, there are some 9 million possible transfer paths. This enormous number of
1 At Ryerson University, after years of teaching assembly language, we we ran two semesters of microprocessor programming based on

the C progamming language. The results were unsatisfactory. Students had no understanding of the underlying hardware, and there was too
much ’magic’ in the program development process. We then reverted back to assembly language.

189
 190

m80
module-complexity.epic

Figure 6.1: Program Complexity

possible control paths will completely overwhelm the ability of human intelligence to understand it.
The Complexity Monster is particularly insidious because it is not evident for small programs. Whatever their
structure, small programs can eventually be debugged. However, as unstructured programs become larger, they
suddenly overwhelm the designer and become impossible to troubleshoot and make reliable.
How do we manage this complexity? In overview, the solution is to clump the program into groups of instruc-
tions, which are then known as software modules. Then:
• Each module has a clearly defined function, rather like a giant machine instruction, and this makes it easier
for the human mind to understand the larger structure.
• Control transfer takes place between modules rather than between individual instructions, thereby minimiz-
ing the number of possible transfer paths.
• The modules themselves are arranged in a heirarchy, in which a master routine calls the subsidiary modules.
• Each module may be individually tested as a small program, which makes its complexity manageable.
Inside modules, the transfer of control is structured into a few carefully chosen patterns.
With this approach, bugs are minimized at the outset and a systematic procedure can applied to pinpoint errors
and eradicate them entirely. Furthermore, extending the program becomes more routine and less of an exercise in
brinkmanship. (Will this last feature break the program?)
However, structure does not arrive by accident and it cannot be applied after the fact: it has to be designed in
from the beginning. In the next sections, we discuss how this process of software design may be accomplished.

6.0.4 Organization of a Control Program

Control programs are part of a larger system, accept sensor inputs, and generate the appropriate display and output
control signals. Often, a control program operates under time constraints: it must update a control output within a
certain time interval of the input signals. If the system is deterministic, that is, if its response time is predictable,
then it is known as a real-time system.
The general organization of a control program is shown in figure 6.2.

www.syscompdesign.com
Software Design Software Design

ISR 1

Initialize

ISR 2

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .

f
.... . .... .
.... ...... .... ...... . .
..... .....
. .... .
. ... .
....
. .
Read Inputs . .

f
Main .
.
.
.
....
Loop .
...
... .
. .
. .

f
. .
Actuate Outputs . .
. .... .
...
. ... .
. .
. Subroutines .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

m80
control-program.latex

Figure 6.2: Control Program Structure

The program begins with an initialization section, in which the hardware is configured and variables initialized.
Then it enters the main loop. The main loop gathers input information and then actuates outputs. In its simplest
form, the main loop is a sequence of calls to subroutines. Usually there are conditional branches that determine
which of the subroutines is actually called.
The main routine is the focus of control for the system, and it determines what processing takes place. It
should be kept as simple and short as possible.
The actual processing is delegated to a number of subroutines, which may in turn call lower level routines.
The main routine may be executed frequently enough that external events can be detected reliably by routines
called from the main loop. If an external event is not reliably detected from the main loop, it will be necessary to
use an interrupt service routine, ISR to do this. The function of interrupt service routines is maintain a record of
the occurrence of external events, not to do serious processing functions. The ISR simply communicates messages
about events to the other routines in the program.
This simple model of a control program brings us to a number of useful design concepts.

Loop Timing It’s quite easy to monitor the speed of this system. Simply cause an output signal to be generated
with each trip around the loop (flip the polarity of an output line, for example) and monitor this signal
with an oscilloscope. This is a useful indicator of the average speed of execution and, what is often more
important, of deviations caused by increases in computing load. On a more fundamental level, the lack of a
pulse indicates that the patient has expired: the software has crashed.
Decoupling Inputs and Outputs The naive way of organizing a control program is to read one input, actuate the
corresponding output, read the next input, actuate the next output, and so on.
However, it is much better to separate the operations of gathering information and actuating outputs . This

www.syscompdesign.com
 192

requires that the information gathering routines write input information into memory registers as flags or
data values. The output routines then read these flags and values when actuating outputs. The input routines
may be regarded as accumulating and saving messages, which are then acted upon by the output routines.
This decoupling of input and output processing into two distinct phases has several advantages.
• Output actions may become conditional on different combinations of input signals simply by various
boolean operations on flag variables, something that is difficult or impossible to do if the input and
output routines are intertwined.
• Debugging is simplified. By breakpointing the main loop at the end of the information gathering
phase, it’s possible to examine the complete set of system input signals. By modifying the input vari-
ables and running the output routines, its possible to check that the output routines are acting correctly.
Various routines can be disabled and enabled by patching the subroutine call in the main loop. (To
disable a subroutine call, use the monitor program to change the JSR WHATEVER instruction to three
NOP instructions.)
• The input flags are available as operator display status indicators. These have considerable debugging
value and ultimately, may be useful in impressing and entertaining clients.
Waiting Loops It is undesireable to have the processor tied up in a waiting loop, whether waiting for some input
signal to become available or in time sequencing output variables. A corollary of this is that delay loops
should be avoided. If a particular routine cannot proceed, then its state (the values of its important variables)
should be stored, the main loop resumed, and the routine checked again on the next pass around the loop.
Prelude to Multitasking In a sense, the main loop is a kind of task dispatcher and the subroutines are system
tasks2 . The main loop schedules the system, calling system tasks as they are needed. No task has the right to
tie up the system while waiting for an input signal to arrive or delay loop to complete. Tasks communicate
by sending messages to each other, in this case via shared variables. A message is nothing more than a flag
bit that is set by one task routine and read by another task routine.

It is important to keep this model of organization in mind as you develop your own system programs.

6.0.5 Example Control Program: CAPCON

To attach some reality to these concepts, we will now take a look at a project completed by the author, the Solar
Telescope Capsule Controller for the McLaughlin Planetarum in Toronto.
A block diagram the system is shown in figure 6.3.

2 We shall use the term task extensively. In this context, you may interpret the term to mean one or more subroutines that work together to

obtain a particular result.

www.syscompdesign.com
Software Design Software Design

Solar Telescope Capsule

System Controller
........
...
.. Power Supply
... Limit Switches
........

&
Calibration Detectors
......
... ... ............. ........
.... .. ... ...... ..... ... .... ...
.... ..
... ...
... ....
. ....
RS232 Link 68HC11 ....... .......
... .. ......
.. ............ .......
..... .... ..... ....
Microprocessor ....... .......
......
........... ......
..... .... ..... ....
....... . .......
.
.
Guider .
.
.
. ..
..
... .. . ... .
...
.
...... ...... ...... ......
......... .........

.
............
..... .... ....
..............
........
.... ....
Cloud and
system-diagram.pictex .... .... ....
m80
..............
.. ..
Temperature
.............
..... .... ....
..............
Sensors
.

Heaters ... ...

..... ........ ....... ........ ....
... ..

RA
.................
...
Motor .... ... ...
... ... ...
.
................
Driver ...

Dec .................
...
Motor ..... ... ....
... ... ...
.
................
Driver ...

Shutter .................
...
Motor ..... ... ....
... ... ...
.
................
Driver ...

Figure 6.3: Solar Telescope System Diagram

www.syscompdesign.com
 194

The solar telescope consists of a moveable primary mirror on the roof of the McLaughlin Planetarium. The
primary mirror is moved by two stepping motors to reflect the sun onto one of several fixed mirrors, which then
deliver an image of the sun to a viewing screen in the planetarium gallery.
The two telescope stepping motors (Right Ascension and Declination) are operated by the micro via stepping
motor translator drives located in the capsule electronics package. Each motor drive accepts step and direction
logic signals to actuate the motors.
The control system is operated by a System Controller, a 486-33 IBM compatible P/C, and a Capsule Con-
troller, a 68HC11 microprocessor. The P/C is located in the building, adjacent to the exhibit display. The micro
is located in the telescope capsule, on the roof of the building. These two computers communicate with each other
via an RS232 bidirectional serial link.
The P/C controller has sufficient computing power to perform the astronomical calculations of sun position
while the telescope is tracking3. The only cables going into the capsule are 117VAC power and RS232 serial
communications, which minimizes the number of control wires connecting the system controller to the capsule.
Under normal operation, the controller P/C issues commands to the capsule processor and displays the re-
sulting status messages. However, for debugging purposes, the capsule may be tested independently of the P/C
controller by plugging it into a source of 117VAC and sending it serial commands from a P/C terminal emulator.
The 68HC11 capsule controller responds to a repertoire of 35 commands in the following categories:

Capsule Motion Requests for single steps, move to limits, move to the calibration points, reset position counters,
inquiries of capsule position, start slew at controlled rate.
Capsule Heat Requests for temperature readings and commands to enable and disable capsule heat.
Cloud and Shutter Requests to open or close the shutter and to send the current cloud reading.
Switch Status Request for the status of all capsule switches and calibration sensors.
Housekeeping Verify capsule internal program, check processor memory, terminate to processor monitor, report
supply voltage.

3 As an interesting sidelight on the development of computing technology, the original telescope system used an 8080 based processor and

required 30 minutes to calculate a table of sun positions for the day. This new system calculates the sun position 30 times per second.

www.syscompdesign.com
Software Design Software Design

6.0.6 Example: CAPCON Main Loop

.
The main loop of the capsule controller program is shown in the listing below. (To head off any criticism
about the missing preamble to this section, it has been removed for brevity.)
You should be able to identify the following features:
Initialization The system uses three interrupts. The SCI (Serial Communication Interface) generates an interrupt
each time a message character arrives. The slew command requires that the stepping motor receive step
pulses at a constant rate, and a timer interrupt is used for this. If the system tries to execute an illegal
operation code, this causes an interrupt which generates an error message and transfers control to the Buf-
falo monitor. (Hopefully, this will not happen during normal operation.) Each of these interrupt sources
requires various setup operations, which are called from this section. Finally, the CLI instruction enables
all maskable interrupts.
The Main Loop The first thing to notice about the Main Loop is the predominance of subroutine call instructions.
In fact, the main loop is primarily subroutine calls with a few branch instructions to enable and disable
certain tasks.
Consider the instruction sequence

NO_STATUS_REQ LDAA AUTO_MESSAGE If auto status messages is enabled

ANDA NEW_SWITCH_STAT and switch status has changed
BEQ NO_OUT_MESSAGES
JSR SEND_STATUS Then send the status message
LDAA #0 and clear the ’changed switch’ flag
STAA NEW_SWITCH_STAT
NO_OUT_MESSAGES (etc)

The precise function of this group of instructions is not important, but it does serve to illustrate the concept
of ANDing two flags together to determine if a task should be executed.
Now consider the sequence

NO_NS_POS LDAA REQ_IN_TEMP If any of the temperature

ORAA REQ_OUT_TEMP
ORAA REQ_CLOUD_LEVEL or cloud data requests
ORAA REQ_VOLTAGE or 28 volt reading
BEQ NO_AD_REQ
JSR READ_AD then read the A/D converter
NO_AD_REQ (etc)

The various request flags (REQ_IN_TEMP and so on) are set or cleared by communication commands,
which are an input to the system, and are established in the first part of the Main Loop.
The logic of these instructions ensures that the A-D converter is read if there is a request for internal
temperature OR a request for external temperature OR a request for cloud level OR a request for motor

www.syscompdesign.com
 196

voltage reading. This kind of logic is simple when the various flags are available and there is a single
command for calling a particular task.
The last instruction in the Main Loop is JMP MAIN, which closes the loop.

**********************************************************************
* Capcon Initialize
*
* This is where the capcon program starts up
*
CAPCON_INIT EQU * Capcon entry point
LDS #STACK Initialize the stack pointer, compatable
with Buffalo
*
LDAA #$93 Set up the ’option’ register per Buffalo
STAA OPTION adpu, dly, irqe, cop
*
LDAA TMSK2 Set the timer prescaler to x4 rate
ORAA #%00000001 for the ’stepping’ signal
STAA TMSK2 Must happen in first 64 cycles after reset
*
JSR INIT_FLAGS Initialize control switch flags
JSR INIT_SPI Initialize the SPI parallel ports
JSR INIT_SCI Initialize SCI system
JSR INIT_BAD_OP Initialize illegal opcode interrupt vector
JSR INIT_SLEW Initialize slew interrupt vectors
CLI Enable maskable interrupts
* *********** Main Loop ************
MAIN_LOOP JSR UPDATE_INPUTS Read the input data
JSR RA_TASK Update right ascension position
JSR DEC_TASK Update declination position
JSR DOOR_TASK Update (shutter) door position
JSR UPDATE_SPI_OUT Update SPI port output data
JSR ACTUATE_SPI Send SPI data and get SPI input data
*
LDAA MSG_COMPLETE Check whether there is an incoming
message
BEQ NO_IN_MESSAGES
JSR DISPATCHER If so, handle the message
*
NO_IN_MESSAGES LDAA REQ_STATUS_MSG Check for request for status message
BEQ NO_STATUS_REQ
JSR SEND_STATUS If so, output a status message
*
NO_STATUS_REQ LDAA AUTO_MESSAGE If auto status messages is enabled
ANDA NEW_SWITCH_STAT and switch status has changed
BEQ NO_OUT_MESSAGES
JSR SEND_STATUS Then send the status message
LDAA #0 and clear the ’changed switch’ flag
STAA NEW_SWITCH_STAT
*
NO_OUT_MESSAGES LDAA REQ_EW_TRANS Check for position count request, e/w
BEQ NO_EW_POS
JSR SEND_EW_POS
*
NO_EW_POS LDAA REQ_NS_TRANS Check for position count request, n/s

www.syscompdesign.com
Software Design 6.1 Module Design

BEQ NO_NS_POS
JSR SEND_NS_POS
*
NO_NS_POS LDAA REQ_IN_TEMP If any of the temperature
ORAA REQ_OUT_TEMP
ORAA REQ_CLOUD_LEVEL or cloud data requests
ORAA REQ_VOLTAGE or 28 volt reading
BEQ NO_AD_REQ
JSR READ_AD then read the A/D converter
*
NO_AD_REQ LDAA REQ_IN_TEMP Check for request for inside temperature
BEQ NO_IN_TEMP
JSR SEND_I_TEMP If so, send it.
*
NO_IN_TEMP LDAA REQ_OUT_TEMP Check for request for outside
temperature
BEQ NO_OUT_TEMP
JSR SEND_O_TEMP If so, send it.
*
NO_OUT_TEMP LDAA REQ_CLOUD_LEVEL Check for request for cloud level
BEQ NO_CLOUD_REQ
JSR SEND_CLOUD If so, send it.
*
NO_CLOUD_REQ LDAA BREAK_FLAG Check for break to monitor
BEQ NO_BREAK
JMP BUFFALO
*
NO_BREAK LDAA REQ_EPROM_CHECK Check for request to send checksum
BEQ NO_EPROM_CHECK
JSR CHECK_EPROM
*
NO_EPROM_CHECK LDAA REQ_RAM_CHECK Check for request to check RAM
BEQ NO_RAM_CHECK
JSR CHECK_RAM
*
NO_RAM_CHECK LDAA REQ_VOLTAGE Check for request to send voltage
BEQ NO_VOLTAGE_CHK
JSR SEND_VOLTAGE
NO_VOLTAGE_CHK JMP MAIN_LOOP and do it all again

6.1 Module Design

The program Main Loop calls a number of subsidiary subroutines. One of the first challenges of the software
design, then, is partitioning the project into these various modules.
Each module has the following characteristics:

Functionality The function of a module should be describable in a few words. If it takes several sentences to
describe the function of the module, it’s probably too complex and should be further decomposed.

Communication The module should act on the machine state in a clearly defined way. That is, we should be
quite clear what does and does not get changed by a particular module. Most modules change the contents

www.syscompdesign.com
 198

of the machine registers: the accumulator, index registers, stack pointer and so on. If you want to maintain
the machine state, then you must push it onto the stack before calling the subroutine and then pop it back
off the stack at the end of the subroutine. If the module uses a buffer area of RAM for some calculations,
the information in the buffer area is destroyed by the module, and this is relevant.
The more things the module changes, the more difficult it will be to understand what the function of the
module if (when) it misbehaves. To minimize complexity, minimize what each module changes.

Entry and Exit Points Each module should have one entry point and one exit point. It may call other subrou-
tines, but the exit from this module should occur at the same point, the physical end of the module, regardless
of the processing that takes place in the module. A practical consequence of this is that a subroutine should
have one and only one RTS instruction.

Re-useability If a software module is properly designed, it should be re-useable from project to project. For this
to work, the interfaces to the module (what data goes in, what data comes out) must be carefully defined.
This is especially true for high level languages, which can be recompiled on a new target machine to
generate new assembly language code. However, even assembly language modules may be ported (adapted)
to a new machine if they are well structured and documented. For example, much of the reference material
for the 6502 microprocessor, a predecessor to the 68HC11, can be easily adapted to the 68HC11 if it is
correctly structured.

6.1.1 Module Example

The CAPCON command set includes a command to report the inside temperature of the capsule. In the CAPCON
software, this is accomplished by a heirarchy of modules:
SEND_INSIDE_TEMP is the highest level module, called from the main loop when a temperature report is
requested. It retrieves temperature sensor input voltage reading from the A-D buffer area, generates the binary
representation of the temperature, calls some routines to format the binary number suitably and then finally calls
SEND_S_HTEMP to transmit the ascii temperature message.
SEND_S_HTEMP in turn calls DO_MESSAGE to send the ascii string out the serial port.
DO_MESSAGE calls OUTSTRG, a general purpose routine which issues a string out the serial port. It then calls
OUTCRLF, which issues a carriage return and line feed character.
Both OUTSTRG and OUTCRLF use OUTCHAR, a routine that sends one character out the serial port.
This example uses a heirarchy of routines, ranging from the high level and special purpose routine SEND_INSIDE_TEMP
down to the general purpose routine OUTCHAR.
Notice the functionality of each of the routines. Each one may be described in a short sentence:

• SEND_INSIDE_TEMP: Send message string containing inside temperature value.

• SEND_S_HTEMP: Send message from the HTEMP buffer

• DO_MESSAGE: Send message out the serial port, followed by carriage return and line feed.

• OUTSTRG: Send null terminated ascii string.

• OUTCRLF: Send carriage-return, line-feed characters.

• OUTCHAR: Send ascii character.

www.syscompdesign.com
Software Design 6.1 Module Design

6.1.2 Module Description

Once the function module function is described in one-sentence, the design of the module header can proceed.
The header is extremely important. It provides three vital pieces of information:

What does this module do? This the one sentence description of the module. Additional material may explain
the function in the context of the larger program.
How do I use it? This section provides enough information to re-use the routine for some other application. In
order to do this, it precisely defines the parameters that are used and produced by the module. In technical
terms, this section defines the module interface. With a knowledge of the interface, it should be possible to
use the routine as a black box, without understanding of the internal workings of the module.
How does it work? This information is the key guide to debugging and modifying the internal operation of the
module. It may give an algorithm or reference for the method, or may set out the logic of the module in
pseudo-code. Comments in the body of the code are helpful but are no replacement for an overview of the
design in the header.

Once the heading is written, then it is a matter of translating the interface and internal description into working
code. Writing the header is part of the module design process, not something that is done after the code is written.

www.syscompdesign.com
 200

6.1.3 Module Heading Example

The 68HC11 processor generates an interrupt when it attempts to execute an illegal instruction operation code.
This is usually the result of an erroneous attempt to execute data as instructions.
The routines given below cause a bad op code interrupt to print out the location and terminate to the Buffalo
monitor.

******************************************************************
* Initialize Bad Op Code Vector
* This routine sets up the vector to point at the ’bad op code’
* interrupt handler routine.
*
INIT_BAD_OP LDAA #$7E Revector the bad opcode interrupt
STAA BAD_OP_VECTOR using the pseudo-vector location
LDD #BAD_OP_ISR
STD BAD_OP_VECTOR+1
RTS
*
******************************************************************
* Bad Op Code Handler
* This routine handles an interrupt caused by the attempted
* execution of an illegal op code.
* It generates an error message indicating the error location
* and then terminates to the monitor.
* Determining the location of the bad opcode:
* When the interrupt occurs, the processor puts the machine
* state on the stack as follows (starting at the higher
* address and working down):
* 8 PCL
* 7 PCH
* 6 IYL
* 5 IYH
* 4 IXL
* 3 IXH
* 2 ACCA
* 1 ACCB
* 0 CCR <- .X points here after TSX instruction
* <- Stack Pointer
* To display the location of the illegal op code, we must transfer
* the stack to the X register using the TSX instruction. This
* puts the stack pointer value plus one in the X index register.
* Now load the accumulator with an offset indexed load, where
* the offset corresponds to the offset of the PCH on the stack.
* Then use OUTLHLF, OUTRHLF to print it.
* Repeat for PCL, and we’re done.
*
BAD_OP_ISR LDX #EM_BAD_OPCODE Print error message
JSR OUTSTRG without crlf
TSX Find bottom of stack frame
LDAA 7,X Get PCH from stack
JSR OUTLHLF Output MSD
LDAA 7,X Get PCH again
JSR OUTRHLF and output LSD
LDAA 8,X Get PCL from stack
JSR OUTLHLF Output MSD

www.syscompdesign.com
Software Design 6.1 Module Design

LDAA 8,X Get PCL again

JSR OUTRHLF and output LSD
JMP BUFFALO Break to the monitor.

Notice there are more lines of comment in the header than there are of code. This is as it should be: the most
important comments give the big picture. For different commenting styles, it is illustrative to look through the
source code for the Buffalo monitor and MATH11 floating point package 4 As you read these listings, ask yourself
what additional information would have been helpful in the module documentation.

6.1.4 Module Internal Design

A brief word on code comments, beginning with the author’s favorite comment:

LDA BLAH Load blah into the accumulator

Now, anyone reading assembly language code is likely to know that LDA means Load the Accumulator,
so the comment is completely useless. What the reader needs to know is not what the machine does but rather
what is your intent with this instruction. Put another way, how does this particular instruction contribute to the
logic of the module?
Now contrast this set of comments5
LDA FUBAR Load FUBAR into the accumulator
CMP NURT Compare with NURT
BNE ZING If they are not equal, branch to ZING
NON_ZING etc etc

with these:
LDA FUBAR If the ZURTBLOT flag
CMP NURT is not equal to NURT then
BNE ZING do the ZING stuff
NON_ZING etc else do the NON-ZING stuff

The two sets of code are identical, but the second set of comments, while somewhat surreal, convey a much
better idea of the intent of the module. Notice as well that the second set mimic the IF-THEN-ELSE decision
structure of Pascal or C. Again, this increases readability.
Other useful decision structures are REPEAT-UNTIL, DO-WHILE, and CASE.

6.1.5 Decisions, Decisions

Decision structures (conditional branches) are particularly error prone in assembly language. Fortunately, the
68HC11 reference manual [98] includes a handy summary of branch instructions in Appendix A: Instruction Set
Details. The table makes it much easier to use the branch instructions correctly.
The table is reproduced (without the Op Codes) in figure 6.4.
4 Both listings are available on the web page of Jim Koch at Ryerson University.
5 For the benefit of readers for whom English is not their first language, I should explain that FUBAR, NURT, ZING, NON-ZING, and
ZURTBLOT are nonsense words that are used to represent machine symbolic addresses. FUBAR does, however, have another application: it
stands for Fouled Up Beyond All Repair, as in The equipment is completely FUBAR. Various other F words have been known to stand in for
Fouled.

www.syscompdesign.com
 202

Test Boolean Mnemonic Complementary Branch Comment

r>m Z + (N ⊕ V ) = 0 BGT r≤m BLE Signed
r≥m N ⊕V =0 BGE r<m BLT Signed
r=m Z =1 BEQ r 6= m BNE Signed
r≤m Z + (N ⊕ V ) = 0 BLE r>m BGT Signed
r<m N ⊕V =1 BLT r≥m BGE Signed
r>m C +Z =0 BHI r≤m BLS Unsigned
r≥m C=0 BHS/BCC r<m BLO/BCS Unsigned
r=m Z =1 BEQ r 6= m BNE Unsigned
r≤m C +Z =1 BLS r>m BHI Unsigned
r<m C=1 BLO/BCS r≥m BHS/BCC Unsigned
Carry C=1 BCS No Carry BCC Simple
Negative N =1 BMI Positive BPL Simple
Overflow V =1 BVS No Overflow BVC Simple
r=0 Z =1 BEQ r 6= 0 BNE Simple
Always - BRA Never BRN Unconditional

Figure 6.4: 68HC11 Branch Instructions

Comparisons are always between a machine register, such as accumulator A, and some memory location. In
this table, r represents the machine register and m a memory location.
For example, in the seqence
LDA FUBAR
CMP NURT
BNE ZING
NON_ZING etc
the comparison is between the value in the accumulator and the value in the memory register ZING.
Four different types of conditional branches are shown in this table: unconditional, simple, unsigned and
signed.

Unconditional Branches The BRA instruction causes an unconditional branch to the offset operand, +/-127 bytes
away. It is useful as a two byte, relocatable replacement for the JMP instruction. The reference manual
attempts to make a case for the BRN instruction, but it’s difficult to imagine where it would actually be
used.
Simple Branches These branches are determined directly by the various bits of the condition code register:
Carry, Zero, oVerflow and Negative. Each machine instruction in the reference manual contains
information on how it affects these bits in the condition code register.
Unsigned Branches These branches are based on byte comparisons where the bytes are considered to be un-
signed numbers. Thus, for example, if r contains 1111 1111 and m contains 0000 0000, then r is consid-
ered larger than m.
Signed Branches These branches consider the bytes to be 2’s complement numbers. In this case, a number is
negative if its most significant bit is set, and this will be reflected in the condition code N bit. In this case, if

www.syscompdesign.com
Software Design 6.1 Module Design

r contains 1000 0000 and m contains 1111 1111, then r is considered smaller than m because 1111 1111
is interpreted as −1 in decimal.

Branch Code Example

Generate the assembly language code for a decision between ALPHA and BETA. We wish one block of code to
be executed if ALPHA is larger that BETA. A different block of code is executed if ALPHA is less than or equal to
BETA. Both numbers are in 2’s complement format.

Q
 Q
 Q
α > β Compare QQα ≤ β
Q α : β 
Q
Q 
Q 
Q

if-then-else.epic
m80

Figure 6.5: IF-THEN-ELSE Flowchart

Solution: We should use the first line of the chart, because we are comparing signed numbers. Then the code
is

LDAA ALPHA If alpha is greater

CMPA BETA than beta, then
BGT ALPHA_IS_LARGER do the ’alpha_is_larger’code
ALPHA_IS_LEQ_BETA (etc) else ’alpha_is_less_than_or_equal’.
(some code here}
BRA CONTINUE_HERE Skip the alpha_is_larger code
ALPHA_IS_LARGER (etc)
(more code here)
CONTINUE_HERE (both branches continue here)

If we goofed and wanted to reverse the sense of the decision, we would replace BGT with its complement from
the table, BLE. If ALPHA and BETA were both unsigned variables, then the appropriate branch instruction would
be BHI.

www.syscompdesign.com
 204

Miscellaneous Hints on Branching

It is possible to do a conditional branch after any instruction, and the branch will take place based on the contents
of the condition code register. For example, after an ADDA operation, the C bit will indicate whether there was a
carry and the Z bit will indicate if the result is zero. If the operands were 2’s complement numbers, the N bit will
indicate whether the result was negative and V will indicate whether an overflow occurred.
However, if the ultimate in speed is not required, it is safest to do an explicit compare operation before each
branch instruction. For example, the Carry bit is not affected by an INCA instruction, so a conditional branch on
the carry flag cannot be used to branch after INCA. Likewise, INX (an increment on the 16 bit X index register)
only affects the Zero flag.
In other words, it is best to use a compare instruction to unambiguously set up the condition code register
in preparation for a conditional branch. For example, to check for a zero result in accumulator A after some
operation, CMPA #0 is guaranteed to work. Similarily, CPX #$FFFF will detect when the X index register has
been decremented from $0000 to $FFFF

6.2 Module Debugging

With the design constraint that a module has one entry and one exit point, it is easy to place breakpoints in the
module for debugging purposes. One breakpoint goes at the beginning of the module, to check variables at module
entry. A second breakpoint goes at the exit point of the module, to check variables at module exit. If the processor
enters the module, it must do so at the entry point and leave at the exit point. Between those two points, the
module changes the machine state, hopefully in some predictable way.
In a sense, a comparison of the input and output states provides us with a measurement of the transfer function
of the module. A transfer function is the ratio of the input and output signals, so it is essential to verify the content
of the input signals. In this context, you must know the state of the machine and memory registers when the
processor enters the module. Determine this by examining memory and register contents at the entry breakpoint.
Now suppose the module contains an error. Move the exit breakpoint backward through the module until the
code between the entry and exit breakpoints is known to work correctly. In difficult cases, this may be as little
code as one instruction. Then move the exit breakpoint forward, carefully verifying operation with each change.
It helps to be suspicious. Verify all your assumptions. Cause a bug to occur again and again, checking that the
symptoms are predictable. Write down the symptoms.
If all else fails:

Simply stare at the code until little drops of blood appear on your forehead,
then it will come to you.

6.2.1 Data Structures

A data structure is a way of organizing data. Arrays, linked lists and lookup tables are all examples of data
structures.
Procedures act on data structures, and it is often possible to tradeoff one for the other. A large procedure,
possibly with embedded data items, may be made into a smaller procedure and an associated data structure.

www.syscompdesign.com
Software Design 6.2 Module Debugging

6.2.2 Data Structure Example: Temperature Conversion

Now we reconsider the example of a thermistor temperature measuring circuit first mentioned in section 1.2.3. To
recap, the ambient temperature is some function of the microprocessor A-D reading. In section 1.2.3 I suggested
that we use the A-D reading as an index into a lookup table, and read out the value of temperature corresponding
to that A-D reading.
Now let us look at that solution more critically. As engineers, it is important to be able to propose several
solutions to every problem, and then to identify the most effective one.
Here are some possible solutions:
Single Entry Lookup Table Under this scheme, the lookup table consists of 256 entries, each one corresponding
to an A-D reading. Each table entry is a byte value representing temperature. Unfortunately, this limits the
temperature value byte to a range of 0-255. As well, we have to be able to represent negative temperatures,
so the range of readings is really something like +/-127. This is sufficient for temperature readings to the
nearest degree, but insufficient for temperature readings to a tenth of a degree. However, temperature lookup
is fast and simple. The table requires 256 bytes of storage.
Double Entry Lookup Table This is similar to the previous scheme, except that each temperature value is con-
tained in 2 bytes, for a total range of +/-32768. Obviously, this will provide sufficient resolution for tenth
of a degree resolution. The storage requirement of 512 bytes is starting to become significant. Again, how-
ever, temperature lookup is fast and simple. The two bytes of each reading could be stored in adjacent table
entries in a 512 byte table. Alternatively, there could be two, 256 byte tables, one for the high byte and one
for the low byte of each temperature value. (Notice that double byte entries still allow you to display only
256 distinct temperature readings, but each reading can have a greater display precision. One can argue
rather effectively that this type of display is misleading: it implies greater accuracy than the A-D converter
provides.)
Floating Point Calculation If we have the floating point package present in any case, it might be just as easy to
process the A-D reading through a series of floating point calculations to generate the temperature reading.
(For a description of the floating point package and its application, see Reference [100]). This approach has
several attractions:
• The storage requirement will undoubtable be less than a double entry table, and maybe even less than
a single entry table.
• The floating point routines essentially eliminate the necessity of scaling the A-D readings to fit the
table.
• If the temperature value is to be used in further calculations, having it in floating point format could
be helpful.
• The floating point package has a routine (FLTASC) which converts floating point numbers to a string
of ascii characters, making temperature easy to display on an LCD display. (You might want to convert
from exponential to fixed point format for display, however.)
On the other hand, floating point calculations will be much slower than table lookup. This is not a con-
sideration if a human is viewing the readout, but it might be important if the output is part of a high speed
control system.
Remember, you have to use discretion in interpreting the results of the floating point calculation. The results
will valid to 0.5%, limited by the original A-D reading.

www.syscompdesign.com
 206

Before we commit to one or other method, we should think ahead a bit. Suppose this device is going to be
shipped to a country where temperature measurement is still practiced in Fahrenheit. To which of these schemes
would it be easier to add a Fahrenheit readout? Using table lookup would require a second table, doubling the
table storage space. Or, the table output value could be converted using fixed point numerical calculations. Using
floating point would require an additional formula calculation. Now suppose the thermistor is changed, requiring
recalibration of the system. Could we somehow use an existing lookup table, or would it be easier to modify a
floating point formula constant?
Notice the space-time tradeoff here: lookup table methods are require lots of space and little time. Floating
point calculations require little space and lots of time. The tradeoff of space and time occurs again and again
in software design. There’s no one approach that gives the ’right’ answer every time: it depends entirely on
the application. However, it is important for the designer to be aware of the different possible methods and the
strengths and weaknesses of each.

6.3 The Real-Time Interrupt

Microprocessor control programs are commonly required to perform certain tasks at a known interval. For ex-
ample, the program may maintain an internal clock that must be reasonably accurate and updated 60 times per
second. Or there may be peripheral devices that need to be serviced at regular intervals. For example, it may be
necessary to scan a group of switches 30 times per second to determine if any one is closed.
The real time interrupt, RTI is a useful device for handling these types of tasks. The microprocessor timer
is set up to interrupt the processor at a regular rate, say 60 times per second. With each interrupt, the processor
stacks the machine state and then branches through an interrupt vector to the timer interrupt service routine. This
ISR calls the various subroutines that need to be serviced on a regular basis. When it has completed the ISR tasks,
the processor restores the machine state and returns to the point of the original interruption. The tasks called by
the real time interrupt are often referred to as background tasks, compared to the main application program, the
foreground task.
The magnitude of the background task has a direct impact on the speed of the foreground task. For example,
if the real time interrupt frequency is 60 Hz (period 16.6 millisec) and the ISR takes a total of 5 milliseconds
to complete, then the processor is spending 5/16.6 × 100 = 30% of its time servicing the background task.
This then reduces the throughput of the foreground task to 70% of what it would be if it did nothing but process
the foreground task. The background tasks must be kept as short as possible and the interval between real time
interrupts as long as possible.
For example, it would not be wise to include floating point calculations in the background task: they are too
time intensive. Nor would it be wise to set the real time interrupt interval to 1 millisecond: Unless the background
routine were very short, the processor would spend all its time executing background routine.
The background task should service its peripheral and leave the gathered data for the foreground routine to
process. For example, a switch scanner routine would leave the result of the switch scan in some memory registers,
but it would not make any decisions based on the detected switch positions.

6.3.1 Accurate Time Delays

Many tasks require accurate time delays. To provide a task with a delay timer

1. Allocate some memory bytes to contain the time delay count.

2. Allocate one byte as a timer enable flag to control whether the timer is counting or not.

www.syscompdesign.com
Software Design 6.3 The Real-Time Interrupt

3. Include a small timer service routine as part of the real time interrupt. The routine checks the timer enable
flag and, if the timer is enabled, increments it by one count.
For example, a two byte counter and 60 Hz real time interrupt would be able to time up to 2 16 ×16.6×10−3 = 1087
seconds.
To clear the timer, the foreground task should first disable the timer and then clear all the timer count bytes to
zero.
The foreground task can now use this timer as a resource. It can enable and disable the count, and can
periodically read the timer to determine elapsed time. These background timers do not add much load to the real
time interrupt, so they can be added as needed.
There is, however, the possibility of a Very Nasty Bug that you should be aware of.
Suppose the timing count is contained in two bytes, which we will call TIMERLOW and TIMERHI. Further,
suppose the two bytes contain the count $FFFF. The foreground routine now decides to do a read of the timing
count. It begins by reading the low byte, which is $FF.
However, and this is the nasty bit, before the foreground task can read the high byte, a real time interrupt
occurs. The real time interrupt increments the counter, which rolls over to $0000. The real time interrupt service
routine then completes and returns control to the foreground task.
The foreground task now reads the high byte, now $00. Putting the low and high byte readings together, the
foreground task determines a timer delay count of $00FF, which is completely erroneous.
This problem is particularly nasty because it only occurs when the real time interrupt happens to occur exactly
between foreground task reads of the two bytes. As a result, the bug appears to occur at random intervals and is
not reproducible.
The solution is simply to disable the real time interrupt while reading a multiple byte counter 6. This will cause
a brief jitter in the timing interval, but no counts will be lost. Of course, the real time interrupt should be disabled
for the shortest possible time.
Other forms of this bug can afflict the servicing of peripheral devices. It is best to have either the background
task or the foreground task service some peripheral, but not both. For example, the background task should read
various switches and the foreground task should read data provided by the background task. The foreground task
should not read the switches directly.

6.3.2 Flip A Bit: Real Time Interrupt on the 68HC11

This section shows an example of a Real Time Interrupt (RTI) initialization and service routine for the 68HC11
microprocessor.
Points to notice:
• This program sets up the real time interrupt so that the state of the output PA6 is changed with each interrupt.
When the interrupt is running correctly, the PA6 output should show a 61 Hz square wave, even when the
main loop has stopped. For example, even if the main loop terminates back to the monitor, the output
waveform should continue.
• The INITIALIZE section of the program sets the OPTION register and the stack pointer exactly the same
as the Buffalo monitor would do. These instructions are not necessary if the program is being started from
the monitor program. However, once the program is debugged, you may then point the reset vector at this
program so that it, rather than the monitor, starts after reset. In that case, it may be a good idea to mimic the
startup of the Buffalo monitor.
6 Technically, reading the two byte counter is known as a critical section, which should not be interrupted.

www.syscompdesign.com
 208

• Certain clock rates require adjusting the 68HC11 timer prescaler. The HC11 hardware requires that any
changes to the prescaler must be done within 64 cycles of reset. Thus, for example, you cannot reset to the
monitor, and then start this program, and expect the prescaler value to be changed by the instructions in this
program. If you must change the prescaler, you have two choices: (a) incorporate the prescaler instructions
is part of the monitor startup routine (you’ll have to re-assemble and re-burn the monitor program) or (b)
do without the monitor and have this program start directly out of reset. Obviously, if you can live with the
prescaler default value, this simplifies life greatly.
• Notice the CLI instruction at the end of the initialization section. The interrupt mask bit in the status register
is automatically set during reset, in order to disable all maskable interrupts. The CLI instruction clears the
interrupt mask bit in the status register to enable all maskable interrupts. Each interrupt source must also be
individually enabled to activate it.
• As shown in this example, the main loop is simply one NOP instruction. In practice, a variety of tasks would
be installed here.

www.syscompdesign.com
Software Design 6.3 The Real-Time Interrupt

*******************************************************************
* Real Time Interrupt Demonstration Program
*
* Flips bit PA6 as a background task, driven by real-time interrupt.
*
* On the use of CLI and SEI:
* The machine comes out of reset with the I bit in the CCR set, so CLI
* must be executed during initialization to enable interrupts.
* An interrupt sequence automatically disables interrupts and RTI
* automatically re-enables them. Thus SEI is not normally required.
* Interrupts and Buffalo Monitor:
* The RTI will run behind the Buffalo monitor.
* However, startup must be via a monitor ’CALL’ instruction. A ’GO’
* followed by an SWI causes the interrupts to be disabled.
* A reset disables the interrupts, of course.
* Definitions used in this example:
STACK EQU $0047 Stack, to be compatable with BUFFALO monitor
RTI_JMP_VECTOR EQU $00EB RTI vector in interrupt jump table
TMSK2 EQU $1024 TOI,RTII,-,-,0,0,PR1,PR2 Timer Mask Register2
TFLG1 EQU $1023 OC1F,OC2F,OC3F,OC4F,OC5F,IC1F,IC2F,IC3F
TFLG2 EQU $1025 TOF,RTIF,PAOVF,PA1F,0,0,0,0 Timer Flag Register2
PACTL EQU $1026 DDRA7,-,-,-,0,0,RTR1,RTR2 Pulse Accumulator Reg
PORTA EQU $1000 Data register for port A
OPTION EQU $1039 Option register (see page 3-9)
******************************************************************
INITIALIZE EQU * Program starts here
LDS #STACK Initialize the stack pointer, compatable with
* Buffalo
*
LDAA #$93 Set up the ’option’ register per Buffalo
STAA OPTION adpu, dly, irqe, cop
*
LDAA #%00000011 Set free running timer prescaler to divide by 16
STAA TMSK2 (must be done within 64 cycles of reset)
*
NOP More initialization here, if required
*
CLI Enable maskable interrupts
JSR ENABLE_RTI Enable the real time interrupt
*
MAIN_LOOP NOP Foreground task goes here
JMP MAIN_LOOP Breakpoint here to return to monitor.
******************************************************************
* Real Time Interrupt Routines
* Enable_RTI:
* 1. Set RTI rate to generate 16.38 msec real time interrupt.
* (See page 10-12 of HC11 manual)
* With 4MHz crystal, must set RTR0 and clear RTR1
* 2. Install JMP IRQ_SERVICE instruction in interrupt vector table
* 3. Enable RTI interrupt bit.
*

www.syscompdesign.com
 210

ENABLE_RTI LDAA PACTL Initialize RTI rate

ORAA #%00000001 Set RTR0
ANDA #%11111101 Clear RTR1
STAA PACTL
*
LDAA #$7E Install ’JMP’ op code
STAA RTI_JMP_VECTOR
LDD #RTI_SERVICE Install IRQ service routine address
STD RTI_JMP_VECTOR+1
*
LDAA #%01000000 Clear the RTI flag to prevent immediate
STAA TFLG2 interrupt on enabling RTI
*
LDAA TMSK2
ORAA #%01000000 Enable RTI interrupt bit
STAA TMSK2
*
RTS RTI is now running.
*
*************************************************************
* Real Time Interrupt Service Routine
* This routine is called every 16.38 millisec and simply flips
* bit PA6
* The RTI will run behind the Buffalo monitor.
*
RTI_SERVICE LDAA #%01000000 Clear the RTI flag
STAA TFLG2
*
LDAA PORTA Get port A
EORA #%01000000 and flip PA6
STAA PORTA and put it back so we see PA6 flipping
*
NOP Any other RTI tasks go here
RTI Done with Real Time Interrupt

6.4 State Machine Techniques

In many microprocessor control problems, it is useful to regard a software module as having a group of well-
defined states. The classic example is a traffic light, which has a red state, an amber state and a green state. The
software controlling the traffic light maintains a variable to indicate the current state and a set of rules that define
how the module gets from one state to another, the so-called state transition rules.
When an event occurs that is relevant to the state machine, the action of the state machine is determined by
the nature of the event and the current state. In the traffic light example, the colour of the next light depends on
the current light and a timer event.
There are a number of advantages to the state machine model:

• The design process becomes much more systematic and less ad-hoc, thereby an aid the comprehension of
a limited human intellect. This results in a more reliable design and one that is easier to troubleshoot if it
misbehaves.

• It is possible to consider all possible combinations of input conditions and current state, thereby avoiding
the unpleasant surprizes that occur when unexpected input and state combinations occur.

www.syscompdesign.com
Software Design 6.4 State Machine Techniques

• For large state machines, it is possible to incorporate the current state, next state and input conditions into a
table that is then read by the procedural code in the module. This makes the behavior easy to maintain and
modify. It also turns out that such a data-driven state machine is very fast in execution [73].

6.4.1 State Machine Technique 1: Output Pulse Waveform

In section 6.0.4, we pointed out that a control program cannot operate correctly if it contains delay loops. We now
discuss a state machine technique that avoids the use delay loops.
Consider that we wish to generate the pulse waveform shown in figure 6.6(a).




0 Pulse Disabled

Pulse Enabled



Pulse 1 Delay Td in progress
Complete


Td Tp Delay Complete, Initialize Tp


2 Pulse Tp in progress
m100 m100
pulse-waveform.epic pulse-state-diagram.epic

(a) Pulse Waveform (b) State Diagram

Figure 6.6: Pulse Generator Example

The naive approach to this is to write two software loops, one for the delay time T d and a second for the pulse
time Tp . The processor would lower the output line, delay Td in a software loop, raise the output line, delay Tp in
a software loop, and continue.
There are a number of problems with this approach. First, software delay loops are dependent on processor
clock speed and the number of cycles in each machine instruction. As a result, they change with each new
generation of processor7 and should be avoided.
Second, each delay loop ties up the processor in unproductive looping, when it could be doing something
useful, such as servicing the rest of the main loop.
A better approach is to use hardware timers or the real-time interrupt to generate the delays. Then they
are a predictable function of clock speed. (If you change the processor crystal, all the delay times will change
proportionally, and it will be necessary to adjust the hardware timer prescaler factor.)
Consider first the real-time clock method of generating the delay.
7 DOS software that was written with suitable software delay loops for the 286 processor is completely unuseable on a 486 machine because

the 486 is so much faster.

www.syscompdesign.com
 212

As shown in Chapter 6.3, the real-time-interrupt may be used to service any routines that depend on service at
regular intervals. If this interval is 60 Hz, then the service occurs every 16.6 milliseconds.
We wish to generate a time delay of 0.5 seconds followed by a pulse of 0.1 seconds, so the pulse routine will
be called by the real-time interrupt service routine and will have two timing counters. The delay counter T d will
be preset to 0.5/16.6−3 ≈ 30. The pulse counter Tp will be preset to 0.1/16.6−3 ≈ 6. Pulse generation is a two
stage process. First, lower the output line and count down the delay timer. When it is zero, raise the output line
and count down the pulse timer. When it is zero, repeat the delay time. And so on.
A systematic way of sequencing the pulse counts is with the state machine shown in figure 6.6(b). The state
machinery is implemented with a variable PSTATE which contains the current state of the pulse generator state
machine. Each time the pulse generator routine is called from the real-time interrupt, the processor uses the value
of the current state (0, 1 or 2) to determine what it should do.

www.syscompdesign.com
Software Design 6.4 State Machine Techniques

Pseudocode for the pulse generator routine is shown below:

Data Registers: OUTPUT_ENABLED, byte 0 to disable output, 1 to enable

PSTATE, byte Current state
DELAY_COUNT, byte Delay count
PULSE_COUNT, byte Pulse count
STATE_DISPATCHER IF PSTATE=0 THEN
GOTO STATE_0_SERVICE
IF PSTATE=1 THEN
GOTO STATE_1_SERVICE
IF PSTATE=2 THEN
GOTO STATE_2_SERVICE
ELSE
SIGNAL STATE ERROR
RETURN
STATE_0_SERVICE Td := 30 Initialize the delay timer
Tp := 6 Initialize the pulse timer
IF OUTPUT IS ENABLED THEN
PSTATE=1
MAKE OUTPUT LOW
RETURN
STATE_1_SERVICE DECREMENT THE DELAY COUNTER Td
IF Td=0 THEN
PSTATE = 2
MAKE OUTPUT HIGH
RETURN
STATE_2_SERVICE DECREMENT THE PULSE COUNTER Tp
If Tp=0 THEN
PSTATE = 0
MAKE OUTPUT LOW
RETURN

There are several points to notice about this routine.

• The delay and pulse counts are both byte sized, so the maximum delay and pulse intervals would be
255 × 16.6−3 = 4.23 seconds. The resolution of the intervals is the real time clock pulse interval, 16.6
milliseconds.

• The STATE_DISPATCHER routine will be called by the real-time interrupt, 60 times per second. The
dispatcher routes control to the appropriate state service routine, based on the current state, PSTATE. It is
possible to design the state machine without the state dispatcher and current state variable, but having a
variable to reflect the current state makes this approach easy to debug. For example, the state of the pulse
generator task can be determined at any time by examining PSTATE. Read it out to a display, and you have
a winking indication of the operation of the pulse generator.

• Notice what happens if the pulse generator is disabled by changing OUTPUT_ENABLED to zero. As the
pulse generator is designed, it completes its current pulse and then discovers that output has been disabled
when it returns to state 0. This holds it in state 0 until the output is enabled again. When the output is
re-enabled, the task starts off in a known state, with the time delay part of the pulse.

www.syscompdesign.com
 214

• Because the pulse generator routine is in handled in software, it is capable of very long delays by increas-
ing the size of DELAY_COUNT and PULSE_COUNT to multiple bytes. However, generation of a pulse
waveform in software is not well suited for delays of less than some 10’s of milliseconds. In that case, the
68HC11 hardware timers should be used.

6.4.2 State Machine Technique 2: Motion Detector

Now let us consider a more complex state machine application.
This task is required to count the number of objects past a point. Two photo-transistors are used to detect the
motion of objects: humans in and out of a building, for example 8. A suitable arrangement of hardware is shown
in figure 6.7(a).

 
Lamps 

JJ
BLOB

BB BB
BB BB
A Obscured

Photo-Detectors s s
+ +
B Obscured
t0 t1 t2 t3 t4
m80 A B m80
motion-detector-hardware.epic motion-detector-waveform.epic

(a) Hardware (b) Waveforms

Figure 6.7: Motion Detector

When a photo-transistor is illuminated, the light causes collector current to flow. Assuming a correct value of
collector resistor and illumination, the transistor is saturated, and the output voltage is approximately zero.
As an object goes past the photo-transistors, it blocks them from the light sources. When the light is removed
from a photo-transistor, the collector current drops to zero and the collector voltage rises up to the supply voltage,
+5 volts in this case.
As the blob object9 travels from left to right, the output voltages of photo-transistor A and B follow the
sequence shown in figure 6.7(b).
Treating zero volts as logic 0 and +5 volts as logic 1, we can show figure 6.7 as a table:
8 A common use for this circuit is to monitor the number and direction of rotation of a motor or control shaft. Two photo-interrupters detect

the passage of holes in a disk.

9 I know, it doesn’t look like a human. Use your imagination.

www.syscompdesign.com
Software Design 6.4 State Machine Techniques

Blob Moves Right

Time A B State Code
t0 0 0 0
t1 1 0 2
t2 1 1 3
t3 0 1 1
t4 0 0 0

If the blob moves in the opposite direction, we have

Blob Moves Left

Time A B State Code
t0 0 0 0
t1 0 1 1
t2 1 1 3
t3 1 0 2
t4 0 0 0

Now the trick: treat each of these A:B combinations as a state number to obtain the state diagram shown in
figure 6.8. (Notice the order of the states: they are not sequential. They order of the states is a Gray Code, in
which one and only one bit changes with each number.)

Backward Forward
   

   
2 3 1 0

m100
direction-states.epic

Figure 6.8: Motion Detector States

The next task is to have the microprocessor monitor signals A and B and detect the passage of an object, either
forward or backward. The simplest scheme is to attach the A and B signals to two input port lines. If the speed
of the program sufficiently fast that all changes in the levels can be reliably detected, the port may simply be read
by the processor with each trip around the main loop. The current state of the two lines is then compared with the
previous state to determine which way the object is moving.
Alternatively, changes in either of the A and B signals too rapid to be detected by checking from the main
loop, may be detected by interrupt. In addition to the port input lines, the signals would be connected to the
68HC11 input capture lines IC0 and IC1. The timer edge interrupt is then set up to generate an interrupt on any
edge. Every time A or B changes, this generates a processor interrupt, which causes the processor to go to the

www.syscompdesign.com
 216

motion detector interrupt service routine. The interrupt service routine then reads the current levels of A and B
and determines the new count and direction.
Each time the photo-interrupter lines change, the direction of the object movement that caused this change may
be determined by comparing the current state of the A and B lines with their previous state. This is summarized
in the following table: (– means no change).

Direction Transition Table

Current New State
State 0 1 2 3
0 – Left Right Error
1 Right – Error Left
2 Left Error – Right
3 Error Right Left –

This table is known as a state transition table because it summarizes all possible transitions that can occur, and
their possible significance. It considers all possible inputs, it should prevent the situation where a combination of
input signals, unanticipated by the programmer, crashes the program.
With the state transition table in hand, we can now write the pseudocode. This is similar in concept to the
previous example. A variable maintains the current state and a state dispatcher routes control to different sections
of the program based on the current state.
We have one more concern to take care of, and that is knowing when to increment the blob counter. We might
decide that an increment should occur when the blob moves out of the sensing area to the right, and a decrement
should occur when the blob moves out of the sensing area to the left.
Then the increment and decrement actions occur for the following state transitions:

Count Transition Table

Current New State
State 0 1 2 3
0 – – – Error
1 Increment – Error –
2 Decrement Error – –
3 Error – – –

www.syscompdesign.com
Software Design 6.4 State Machine Techniques

With both the state transition tables in hand, we can now write the pseudocode. This is similar in concept to
the example of section 6.3. A variable maintains the current state and a state dispatcher routes control to different
sections of the program based on the current state.

Data Registers: CURRENT_STATE, byte Current state of input lines: 0,1,2 or 3

NEW_STATE, byte New state of input lines: 0,1,2 or 3
DIRECTION, byte 0=Left, 1=Right
COUNT, byte Count of objects passing through
STATE_DISPATCHER IF CURRENT_STATE=0 THEN
GOTO STATE_0
IF CURRENT_STATE=1 THEN
GOTO STATE_1
IF CURRENT_STATE=2 THEN
GOTO STATE_2
IF CURRENT_STATE=3 THEN
GOTO STATE_2
ELSE
SIGNAL STATE ERROR
STATE_0 IF NEW_STATE =0 THEN
NO CHANGE
RETURN
IF NEW_STATE=1 THEN
DIRECTION := LEFT
CURRENT_STATE := NEW_STATE
RETURN
IF NEW_STATE=2 THEN
DIRECTION := RIGHT
CURRENT_STATE := NEW_STATE
RETURN
ELSE ERROR
RETURN
STATE_1 IF NEW_STATE =0 THEN
INCREMENT COUNT
DIRECTION := RIGHT
CURRENT_STATE := NEW_STATE
RETURN
IF NEW_STATE=1 THEN
NO CHANGE
RETURN
IF NEW_STATE=2 THEN
ERROR
RETURN
IF NEW_STATE=3 THEN
DIRECTION := LEFT
CURRENT_STATE :=NEW_STATE
RETURN
STATE_2 IF NEW_STATE=0 THEN
DECREMENT COUNT
DIRECTION := LEFT
CURRENT_STATE := NEW_STATE
RETURN
IF NEW_STATE=1 THEN

www.syscompdesign.com
 218

ERROR
RETURN
IF NEW_STATE=2 THEN
NO CHANGE
RETURN
IF NEW_STATE=3 THEN
DIRECTION := RIGHT
CURRENT_STATE :=NEW_STATE
RETURN
STATE_3 IF NEW_STATE=0 THEN
ERROR
RETURN
IF NEW_STATE=1 THEN
DIRECTION := RIGHT
CURRENT_STATE := NEW_STATE
RETURN
IF NEW_STATE=2 THEN
DIRECTION := LEFT
CURRENT_STATE := NEW_STATE
RETURN
IF NEW_STATE=3 THEN
NO CHANGE
RETURN

On the basis that there are only two situations where increment and decrement occur, you might expect that
there is a simpler routine that will have the same effect. The state machine approach, however, instantly detects
any change of direction and can deal with such pathological situations as the blob moving into the sensing area
and then backing out again. If this happens, the blob should not be counted, and is not, using this routine. This
routine is also quite robust at detecting illegal transitions in the input signals. This could be handy detecting
hardware problems with the detectors.
Although there is a significant amount of code here it is a direct translation of the state transition tables into
code: there is nothing tricky or unusual. In fact, it is possible to eliminate nearly all of the conditional branch
instructions and have the program read the state transition tables directly, as discussed in reference [73].
As in the previous example, the CURRENT_STATE and NEW_STATE variables are handy in debugging: they
indicate the current and new status of the machine.
Incidentally, you will notice that this pseudo-code has a zillion RETURN statements in it. To adhere to our
one entrance, one exit rule for procedures, all of these RETURN statements should be translated into JMP EXIT,
where EXIT is the address of a final RTS instruction.

www.syscompdesign.com
Software Design 6.5 The Button Interface

6.5 The Button Interface

Eventually, we will need several buttons (for example) to
• select the display
• set the time-of-day clock
• reset minimum and maximum readings to zero
For now, however, let us start with one pushbutton that selects the current display. Each time this SELECT
button is pushed, a different display is shown.
This seems simple enough, but there are several important considerations in the design of the button software.

One Event per Push The display should change once and only once per button push. Now, the software will
likely be fast enough to detect a pushed button many times during the interval it is depressed by the human
operator. A routine that simply changes the display each time it detects a depressed pushbutton will then
change the display several times with ach button push, not what is wanted.
The button routine must therefore check for the release of the button as well as its actuation. A suitable
state diagram for the button is shown below:
'$

0 Button Released
&%


'$ Set button flag

1 Button Depressed
&%
button-state.epic
m100

Figure 6.9: Pushbutton State Diagram

Scan Rate The button must be checked sufficiently frequently to detect the most rapid button jab by a human
operator, or pressing the button must cause an interrupt. We choose to scan the button by polling software
since we will be adding switches and have a limited number of available interrupts.
The main loop cannot be used to scan the button. The execution of the main loop varies as the tasks change,
and so a minimum scan time for the pushbutton cannot be guaranteed.
On the other hand, if the switch scan routine is added to the real time interrupt service routine, the switch
may be checked at a constant rate, 60 times per second. It turns out that this is frequent enough to detect
the briefest possible actuation by a human.

www.syscompdesign.com
 220

Button Bounce When a mechanical switch closes, the contacts bounce open and closed for a few milliseconds
after the initial contact. As a result, the switch circuit generates a train of pulses as shown in figure 6.10
below.

$v o$
+5 volts
+5v
r
e
vo
0v
e
To
Pushbutton Microprocessor
Input Time
2 millisec (approx)

contact-bounce.epic=
m70

Figure 6.10: Contact Bounce

The train of pulses caused by contact bounce may erroneously be seen by the software as multiple button
pushes. A simple fix is to delay and reconfirm that the switch signal has not changed. The delay is typically 3
milliseconds or so, short enough that it may be implemented with a software loop.

www.syscompdesign.com
Software Design 6.5 The Button Interface

6.5.1 Button Switch Pseudocode

B_STATE: Byte {Keeps track of button state}
B_FLAG: Byte {Set whenever button is pushed}
RELEASED: 0 {Definition of states}
PRESSED: 1

Initialization: B_STATE := 0
B_FLAG := Cleared

{Real Time Interrupt causes entry here}

Routine: Case B_STATE Of {Sends the program to the correct state }

0: B_RELEASED { handler software}
1: B_PRESSED

B_RELEASED: If button released

Then Exit
Else
Delay 3 milliseconds
If button is (still) pushed
Then
Raise B_FLAG
B_STATE := PRESSED
Exit

B_PRESSED: If button pressed

Then Exit
Else
B_STATE := RELEASED
Exit

EXIT: Continue servicing RTI or Return from Interrupt

www.syscompdesign.com
 222

6.5.2 Button Interface to Display Manager Task (Main Loop)

The Main Loop checks the button flag B_FLAG on each cycle around the main loop. If the button flag is set, the
main loop increments a current display counter. For example, a current display of 0 might select temperature, a
current display of 1 humidity, and so on.
Then, to indicate that it has detected the button event message, the main loop clears the button flag.
If button events occur faster than the main loop can accept them (perhaps because the main loop is preoccupied
with lengthy calculations), the button task will attempt to set the button flag a second time, before it has been
cleared by the main loop. This could be used to generate an error message.

www.syscompdesign.com
Chapter 7

Instrumentation Laboratory Outline

7.1 Lab Description

The Instrumentation Laboratory is based on a single project – an instrument for measuring properties of the
atmospheric environment.
The instrument uses a microprocessor and various analogue sensors to detect and record

• Ambient temperature

• Humidity

• Wind speed and direction

• Barometric air pressure

Readings are displayed on a liquid crystal display. The software includes a facility to store and display a
history of past readings.
A minimum level project will consist of a device to measure and display three quantities, as specified by your
instructor, from the above list. Advanced students will add other sensors to the system or may investigate new
approaches to sensor design. For example, an advanced student may wish to build a wind velocity sensor that
combines the measurement of wind direction and speed.

7.1.1 Project Objectives

The student will:

• design an electronic system using analogue and microprocessor components.

• design circuits requiring care in selection of components and grounding, shielding and decoupling.

• gain experience in calibrating and verifying the performance of an analogue system.

• design and build software for an embedded microprocessor system.

223
 224

7.1.2 Project Management

This project is managed at a level known as structured design, ie, between a structured laboratory exercise and an
open ended project (as in the thesis project). Students will be provided with a project manual which will include
extensive information on the project and approaches to the design of components in the project. Since the lab
will in some areas be in advance of the lecture material, enough theoretical background will be provided that the
students can design circuitry used in the project.
Students will also be expected to have the Motorola 68HC11 Reference Manual.
Construction of project hardware and software should be done outside of lab hours. It is recommended that
lab time be used for debugging nasty problems with the help of the instructor, calibration of sensors, and demon-
strations. Students in previous sessions of the lab have spent at least 6 hours per week outside the lab on this
project.

7.1.3 Microprocessor
Emphasis in the project is on analogue circuit design, construction and verification, and on microprocessor soft-
ware design and debugging. To this end, a microprocessor kit the MPP Board will be available to purchase for
the microprocessor section of the project. After assembling the kit, the student will have a functional 68HC11
microprocessor system. An undedicated area of the board may be used for analogue signal conditioning circuitry.
Detailed specifications for the microprocessor board are given in a separate manual, [100].
The microprocessor module may be considered as a ’single chip micro’ development system. It is provided
with RS232 terminal port and monitor software so that the application program may be debugged. Eventually, a
single chip version of the micro could be used in the final product, in which the weather station software is burned
into an EPROM.

7.1.4 Project Features

The microprocessor based weather station includes the following features:

Ambient Temperature Measurement The student is required to design and build temperature measuring cir-
cuitry. The student will consider the various engineering tradeoffs in using solid state temperature sensors
vs thermistors or other devices. The final circuit will be tested against lab calibration standards. The drift
performance of the measuring circuit must be predicted.
Lessons: Temperature measurement, nonlinear correction (thermistor), amplification and offset signal con-
ditioning (solid state sensor). Use of lookup tables or floating point calculations.

Humidity Measurement The measuring circuit must determine the relative humidity of the air. This may be
accomplished using a sensor which changes dielectric constant with humidity or using an electronic hy-
grometer. Lessons: measuring frequency and converting to measured quantity.

Air Speed Measurement The air speed measurement circuit may be devised by the standard ’whirling cups’
mechanical anemometer or by means of a thermal feedback circuit (preferred because it has no moving
parts).
Lessons: Differential amplifier measurements, thermal time constants, thermal transfer effects.

Wind Direction Measurement A conventional wind vane with rotational sensor may be used to sense wind
direction.

www.syscompdesign.com
Instrumentation Laboratory Outline 7.1 Lab Description

Lessons: Interfacing to a rotating sensor (gray scale sensing or other analogue rotational sensor).

Air Pressure A solid state absolute air pressure transducer may be used to generate a (small) signal proportional
to barometric pressure. Conditioning this signal (a total variation of 1.67 mV over a typical change of air
pressure for the Motorola MPX100A) is a challenging exercise in analogue circuit design.

Lessons: small signal amplification, control of noise, chopper stablized amplifier techniques, autocalibra-
tion technique, use of lookup tables or floating point package.

Sensor Readouts The instrument must have selectable readout displays, which should read out to 1 % precision,
with (hopefully) corresponding accuracy. Calibration factors are stored in the microprocessor EEPROM.

7.1.5 Grading Criteria

The student project will be graded according to the following criteria:

• Meeting the checkpoint deadlines

• Overall integrity of the design

• Accuracy of measurement

• Ease of operator use (user interface)

• Software structure, methodology and style

• Construction technique: packaging neatness, serviceability, grounding technique

• Creativity in design and measurement technique

• Design analysis and documentation

• Quality and completeness of the final report

Marks will be assigned to different parts of the project as shown below:

www.syscompdesign.com
 226

MPP Board Completed on time 5%

Functional 5% 10%

Sensor #1 Completed on time 5%

Accuracy 5%
Construction technique 5%
Software 5% 20%

Sensor #2 Completed on time 5%

Accuracy 5%
Construction technique 5%
Software 5% 20%

Sensor #3 Completed on time 5%

Accuracy 5%
Construction technique 5%
Software 5% 20%

System Integration Completed on time 10%

Final Report Completeness and Quality 20%

Total: 100%

7.1.6 Tools and Equipment

Students are expected to have their own hand tools for assembling electronic circuits.
PAL and EPROM programmers are available.
Computers will be available in the instrumentation lab, though it is strongly recommended that the student
have access to their own computer after hours. This way, software development can proceed on evenings and
weekends, outside lab time.
Software development may be done with 68HC11 assembly language. Alternatively, students may prefer
to use the C programming language. A suitable C compiler that runs under the Windows operating system is
available from Imagecraft (http://www.imagecraft.com).

7.1.7 Resource Materials

Manuals for the microprocessor board and sensors are available in electronic form on the computers in the instru-
mentation lab.
Postscript files (suffix .ps) may be printed on the network laser printers.
Material currently available includes

• Construction and operation manual ’mpp_manual.ps’ for the Microprocessor Project Platform (MPP),
the 68HC11 microprocessor board.

• Instrumentation lab manual, ’insmain.ps’.

• Miscellaneous schematics, parts lists and assembly instructions.

www.syscompdesign.com
Instrumentation Laboratory Outline 7.1 Lab Description

• Software for the 68HC11.

You are strongly advised to obtain copies of this material and read it carefully.

7.1.8 Lab Schedule

Week # Week of Activity Checkpoint
1 Sept 6 Orientation
2 Sept 13 Construct MPP Board
3 Sept 20 Install MPP Software Demo MPP Board
4
4 Sept 27 Build Sensor #1 Hardware
5 Oct 4 Write Sensor #1 Software
6 Oct 11 Calibrate Sensor #1 Demo Sensor #1

7 Oct 18 Build Sensor #2 Hardware

8 Oct 25 Write Sensor #2 Software
9 Nov 1 Calibrate Sensor #2 Demo Sensor #2

10 Nov 8 Build Sensor #3 Hardware

11 Nov 15 Write Sensor #3 Software
12 Nov 22 Calibrate Sensor #3 Demo Sensor #3

13 Nov 29 Integrate complete software Demo Complete System

- Nov 29 Write Report Submit Report

• All checkpoint demonstrations must be completed by the ending time of the students lab on the due date.
• A Checkpoint Demonstration includes a demonstration of operational hardware, properly documented soft-
ware and a calibration report. Calibration constants must be stored in EEPROM, not as part of the applica-
tion software. Incomplete demonstrations will receive part marks. Students may be quizzed on aspects of
the project, such as the software.
• A penalty of 5% per week is applied to late projects to a minimum of half the mark for that project.
• The completed unit must be demonstrated by 5 PM on the day of the lab in week 13. No work will be
accepted after that date.
• The report must be submitted to the office by 4:30 PM on the due date. Late reports will not be accepted.
• Students are encouraged to help each other solve problems, including hunting for software bugs. However,
plagarism (copying) of software is strictly forbidden and will be treated extremely seriously.

7.1.9 Project Report

The Lab Report documents the intellectual property that you have generated during your work on the project.
There is no requirement to duplicate information, such as Ohm’s Law or the MPP Manual, which is available
elsewhere or is known to your instructor. At the conclusion of the project, you are required to submit a lab report
which includes the following information:

www.syscompdesign.com
 228

• Properly drawn and labelled schematics

• Detailed parts list

• Properly commented and documented software listings

• Calibration curves and measurement results

The lab report should observe the following guidelines. (These pointers are based on the reports from previous
iterations of this course.)

• The report must include a table of contents.

• Show calibration results, preferably in graphical form.

• Show all wiring diagrams, including connections between the analog sections and the microprocessor.

• Show a parts layout drawing with parts designations for each hardware section you constructed (not the
MPP Board).

• Include a parts list, with quantity, designation, description, supplier and price of all items. (The micropro-
cessor board may be treated as one item.)

• Schematics should be complete with designations and parts values. Any potentiometers should be labelled
with their function.

• Formulae should have all terms defined.

• Software should be organized as a heirachy of main routine and subroutine tasks (modules).

• Software modules should be short: less than one page.

• Each software module should have a header and include a pseudocode description of its operation. A header
should never go in the center of a software module.

• Document the software algorithm used to convert quantities before display.

• Software modules should be structured as ’black boxes’ with well defined inputs and outputs. Input and
output parameters must be specified in the header of each module.

• The emphasis on software documentation should be in the header. Try to specify the intent of the module,
as opposed to a blow-by-blow description of its operation.

• If a lookup table is used, the method of generating the table should be shown as part of the header docu-
mentation.

• It is bad practice to disable interrupts. Try to use mechanisms, such as state machines, that do not require
disabling interrupts.

• One and only one software module, the display manager, should write to the display. Other modules
communicate with the display manager subroutine to write to the display.

www.syscompdesign.com
Instrumentation Laboratory Outline 7.2 Laboratory Exercises

• Avoid unnecessary ORG statements in code. As much as possible, allow the assembler to determine the
memory address of code.

• Do not make statements you can’t prove. For example the statement, ”The circuit worked perfectly” is
meaningless. A better statement is ”The circuit met specifications”, where the specifications have been
defined and the actual performance measured and documented.

• Do not use more than two font types, and stick to plain fonts such as Helvetica or Times. Medieval script is
not appropriate for a technical report, even the title.

• Do not use justification unless your word processing program has full, intelligent justification. Most pro-
grams don’t, so it is better to leave the right edge of text ragged.

• The report must be in full electronic format. It is not acceptable to paste pieces of paper in the report or
hand-draw graphs.

• The report must be printed on a laser or inkjet printer. Dot matrix is not acceptable quality.

• The preferred publishing tools are LATEX, using xfig for drawings and Matlab or gnuplot for graphs. These
tools have require some learning to use, but will also be extremely useful in generating your thesis project
report. Documentation for LATEXand xfig is available on the computer network.

• Reports are expected to be the original work of the person submitting the report. Any material, such as
figures, graphs or software, which has been generated by some other party must be given credit in the
report. Failure to do this constitutes plagiarism. Material from another party that is contains mistakes,
whether attributed or not, will result in a substantial lowering of marks.

7.2 Laboratory Exercises

7.2.1 Overview
This document contains instructions for the instrumentation projects for the Instrumentation laboratory. It should
be read in conjuction with the Course Outline for the Lab, and the Lab Theory Notes.

7.2.2 Calibration Constants

Many of these circuits require calibration constants to allow the microprocessor to generate the correct reading
for a given electrical input. These calibration constants should be stored in EEPROM and marks will be
deducted otherwise. Such calibration constants should not be written into the source code program.
Notice that it is a simple matter to manually write a calibration value into EEPROM using the Buffalo monitor.
The computer program may then read the EEPROMed value during the execution of the program.

7.2.3 Humidity Sensor

Your first assignment is to develop a humidity sensor for use in the weather station. Read over Chapter 2 of the
Instrumentation System Design manual. You may use any of the following three methods to measure humidity:

www.syscompdesign.com
 230

Wet-Dry Bulb (Hygrometer) Two thermistor temperature sensors are interfaced to microprocessor A-D chan-
nels. Software reads the two A-D channels and calculates the temperature of each thermistor.
One thermistor, the dry bulb, measures air temperature. The other thermistor, the wet bulb, is cooled by a
muslin wick, dampened from a water resevoir, in a fan cooled air stream. A formula is used to calculate the
relative humidity from the two temperatures.
This approach requires some mechanical construction but the software is quite straightforward, especially
if one thermistor is already used to measure ambient temperature.

Humidity-Capacitance Sensor: Oscillator A humidity sensor is available as Philips part number 691-90001
from Electrosonic Supply for about $10. The capacitance of the sensor, which varies with relative humidity
is used to control a variable frequency oscillator. The resultant frequency or period is measured by the
microprocessor and a formula used to calculate relative humidity.
This approach requires the use of the timer input and interrupts on the 68HC11 microprocessor. The hard-
ware is very simple, but the software will require careful design and coding.
Because of the variability of the sensor, the sensor must be calibrated using atmospheres of known humidity.
At least three jars containing known humidity atmospheres will be provided. Students should measure the
oscillator frequency at these three humidity points and then fit a curve to the measured data. The sensor
response time is in the order of 5 minutes, so each measurement should allow sufficient time for the reading
to stabilize.
The sensor itself generally cannot be extended on wires because the added wiring capacitance affects the
calibration constant of the sensor. It is preferable to extend the entire oscillator circuit on wiring, and allow
that to hang in the calibration atmosphere.

Humidity-Capacitance Sensor: Charge Amplifier This approach will appeal to students who are comfortable
with hardware construction and wish to avoid interrupt programming. The variable capacitance of the
Philips humidity sensor is converted into a variable analog voltage which may be read by the microprocessor
A-D converter.
Suitable circuits are shown in the Linear Technology Linear Applications Handbook, application note 3.
(One of these circuits is shown in Operational Amplifier Circuits, Theory and Applications, by E.J.Kennedy,
Holt Rinehard and Winston.) A copy of the circuit diagrams should be available from your lab instructor.
The Linear Technology Switched Capacitor LT1043 part used in this circuit is available from Electrosonic
Supply. The operational amplifier may be the Linear Technology LT1056 or some suitable substitute such
as the National LMC660CN.

Humidity-Resistance Sensor A home-made humidity sensor, which relies on the change of resistivity of table
salt with humidity, is described in the October 1972 issue of Popular Electronics: ‘Electro-Chemical Hy-
grometer: Low Cost Unit measures Relative Humidity, by Joseph Giannelli’. This is a very simple circuit -
it uses only two transistors - but it would have to be adapted to supply a signal to the 68HC11 A-D input.

Dew Point Sensor The Peltier Cell is a semiconductor device that may be used to cool objects. (Peltier cells may
be available from Active Surplus on Queen Street). A mirror mounted on and cooled by a Peltier cell will
collect a surface of moisture (dew) from the atmosphere.
In a dew point sensor, a lamp reflects a light beam off the surface of the mirror into a phototransistor. A
simple feedback amplifier with input from the phototransistor drives the Peltier cell, such that dew is just

www.syscompdesign.com
Instrumentation Laboratory Outline 7.2 Laboratory Exercises

beginning to form on the mirror surface. The temperature of the mirror is then at the current dew point
temperature.
Relative humidity may be calculated from the ambient temperature and the dew point temperature (See the
Instrumentation System Design manual).

Deliverables
On the due date, the student should show

• A working humidity measurement circuit. The instructor will breathe heavily onto the sensor and a change
in reading should be apparent.

• Properly planned and documented software listing.

• For devices using the Philips humidity sensor, a three point calibration curve and equation.

Bonus Marks
A bonus mark of 10% will be awarded to students who use the Wet-Dry Bulb, Humidity-Capacitance Sensor:
Charge Amplifier, Humidity-Resistance Sensor or Dew Point Sensor techniques. For a bonus mark to be awarded,
the circuit must be substantially functional.

7.2.4 Barometric Air Pressure Sensor

Your second assignment is to develop a barometric air pressure for use in the weather station. Read over Chapter
5 of the Instrumentation System Design manual.
All designs will used the same sensor, the Motorola MPI100AP sensor, available from Electrosonic Supply.
(Be certain to get the ’P’ version, which has a pressure port connection, which is required for calibrating the
circuit). The interface circuit drives the microprocessor A-D converter, and suitable software determines the air
pressure.
Suitable operational amplifiers are the National LMC660CN or LM324N. The LMC660CN is a ’rail to rail’
device and can work directly from the +5 volt logic supply, but is somewhat prone to spurious oscillations,
especially with any capacitive loading. Its output should be carefully checked with an oscilloscope to ensure that
it is not oscillating.
To drive the A-D to its full +5 volt positive limit, the LM3224 must be operated from a +6.5 volt supply. This
may be derived from the raw DC input voltage to the MPP board, using a three terminal regulator (and suitable
resistors to program it for 6.5 volts) such as the National LM317T or LM317L.
The barometric sensor circuit is a high gain instrumentation amplifier. The design and parts layout should
be carefully planned. The circuit should be constructed carefully, neatly and with attention to proper analog
grounding (ask your instructor.)
The barometric sensor can be calibrated using a water manometer apparatus available in the laboratory.

Deliverables
On the due date, the student should show

www.syscompdesign.com
 232

• A working air pressure measurement circuit. The student will blow positive and negative pressure into the
sensor and a change in pressure reading should be apparent.

• A minimum of three points on a calibration plot, and the fitted equation relating air pressure and A-D sensor
reading.

• Properly planned and documented software listing.

Bonus Marks
The lab theory manual describes how the circuit may be modified for automatic calibration of air pressure. A
bonus mark of 10% will be awarded to circuits using this technique. You may wish to use the 6 bit discrete
component D-A converter shown in the notes or you may prefer to purchase an integrated circuit 8 bit D-A
converter such as the DAC0800. The DAC0800 and many other IC D-A converters require an op-amp current-
voltage converter and a negative power supply, which may be stolen from the LT1081 RS232 driver circuit.
We strongly recommend that you first get the basic sensor circuit, using potentiometer calibration, working
properly. Then add the auto-calibration circuit, carefully checking the operation of the D-A converter indepen-
dently of the sensor circuit. Finally, jumper the D-A into the offset adjust point to test for auto-calibration.
For a bonus mark to be awarded, the entire circuit must be substantially functional.

7.2.5 Wind Speed Sensor

Your third assignment is to develop a wind speed sensor for use in the weather station. Wind speed is a new sensor
design this year.
Read over Chapter 3 of the Instrumentation System Design manual.
Three wind speed sensor designs are possible:

Cup Anemometer For students who enjoy constructing mechanical devices, the Rotating Cup Anemometer may
be attractive. A high quality DC motor, such as the Edmund Scientific #37590 could be used as a tachogen-
erator. Alternatively, a pulse frequency may be generated using a segmented disk and photo-interrupter.
In either case, the Rotating Cup Anemometer must be carefully calibrated. It may be possible to use one
of the wind tunnels in the Department of Mechanical Engineering, or it may be necessary to attach the
anemometer to a vehicle. In the latter case, the anemometer must be mounted well away from the vehicle
body to avoid slipstream effects. It is also necessary to compensated for ambient wind velocity by measuring
output vs vehicle velocity in both directions on a stretch of road. (Please choose a quiet location: not the
Don Valley Parkway.)
It is to be expected that the output of a tachogenerator may require some low pass filtering to remove noise
and some type of amplitude adjustment (voltage amplifier or voltage divider attenuator) to fit the dynamic
range of the tachogenerator to the 5 volt range of the microprocessor A-D converter.
A careful graph of output vs wind speed should be measured. The distance constant should be calculated
and the time constant estimated. The method of calibration must be carefully and completely documented.

Constant Temperature: Transistor A constant temperature anemometer, using junction transistors to sense
temperature, is shown in the notes. The output from this circuit must be conditioned by a difference ampli-
fier before it is connected to the microprocessor A-D converter.

www.syscompdesign.com
Instrumentation Laboratory Outline 7.2 Laboratory Exercises

This design will require a separate +20 volt, -5 volt power supply, since the voltages are not the same as the
MPP board and current requirements are considerable. (Lab power supplies are acceptable for demonstrat-
ing circuit operation.)
It is expected that a miniature wind tunnel will be available to calibrate the circuit. A graph of output vs
wind speed should be measured. The time constant should also be measured and documented.

Constant Temperature: Hot Wire A Hot Wire Anemometer circuit is also shown in the notes. A #328 lamp
has its glass envellope removed to expose the filament. This is then used to measure wind speed.
This design requires a separate +15, -15 volt power supply, since the voltages are not the same as the MPP
board and current requirements are considerable. (Lab power supplies are acceptable for demonstrating
circuit operation.)
It is expected that a miniature wind tunnel will be available to calibrate the circuit. A graph of output vs
wind speed should be measured. The time constant should also be measured and documented.

Deliverables
On the due date, the student should show

• A working air speed measurement circuit. Air will be caused to move past the sensor and a change in wind
speed reading should be apparent.

• A detailed calibration plot with description of the calibration method, and the fitted equation relating wind
speed and A-D sensor reading.

• Properly planned and documented software listing.

7.2.6 System Integration

The final assignment is to demonstrate the weather station functioning with all three sensors operational at the
same time.
If the software for each sensor has been constructed in a modular fashion and properly documented, integra-
tion of the three sensors should be straightforward. Conceptually, the main loop should be viewed as a display
manager. The display manager determines what should appear on the display, queries the sensors, formats the
information, and builds the readout information.
The display can cycle through the three readings in sequence, or all three readings can occur at the same time,
or a pushbutton may be added to cause the display to sequence through the various readings.

Deliverables
On the due date, the student should show

• A weather station system with three sensors functional.

• Properly planned and documented software listing for the entire system.

www.syscompdesign.com
 234

Bonus Marks
A bonus mark of 10% will be awarded for additional readouts provided in the final working system. These could
be readouts for ambient temperature or wind direction, for example.
In the case of bonus marks for ambient temperature readout, an additional bonus of 5% will be awarded if the
final system shows displays of dew point temperature, and humidex reading.

www.syscompdesign.com
Bibliography

[1] The Canadian Encyclopedia

Second Edition
Hurtig Publishers Ltd.
Edmonton, Canada, 1988

[2] The World Book Encyclopedia

World Book, Inc, 1985
Wind, pp 278-279

[3] The Climates of Canada

David Phillips
Environment Canada, 1990
Available from:
Canadian Government Publishing Centre
Supply and Services Canada
Ottawa, Canada, K1A 0S9 Catalogue # EN56-1/1990E

[4] Piping Down the Valleys Wild:

A merry mix of verse for all ages
Edited by Nancy Larrick
Dell, 1968

[5] Mathematical Handbook of Formulas and Tables

Murray R. Spiegel
Schaum’s Outline Series, McGraw Hill, 1968

[6] Practical Robotics

Bill Davies
Werd Technology Inc.
Unit 35B, Suite 155
10520 Yonge Street
Richmond Hill, Ontario, L4C 3C7

[7] Build a Computerized Weather Station

Steve Ciarcia
Byte Magazine, February 1982, PP 38-68

235
 236

[8] Weather Station

Ronald M. Jackson
Electronics Now, October 1993, pp 31-38

[9] Absolute Zero and the Conquest of Cold

Tom Schachtman
Mariner Books, 1999

Temperature: Thermistors
[10] Philips Components
1990-91 Data Book
Philips Passive Components Group
2001 W. Blue Heron Blvd.
P.O.Box 10330
Riviera Beach, Fl. 33404, USA

[11] Omega Temperature Measurement Handbook

Omega International Corp
P.O.Box 22284
Stanford, CT., 06906 USA

[12] Thermistors
E.D.Macklen
Electrochemical Publications, 1979

[13] Calibration Curves for Thermistors

J.S.Steinhart and R.C.Hart
Deep Sea Research and Oceanographic Abstracts
Volume 15, 1968, pp 497-503

[14] Design Temperature Converters with Thermistor-Based Oscillators Harry L. Trietley

EDN, February 3, 1983, pp 115-118

[15] Direct-reading Converter Yeilds Temperature

James Williams and Thomas Durgavich
Electronics, April 3, 1975, pp101,103

[16] Thermistors Make Good Thermometers

C.S.Molee
Electronic Design 8, April 12, 1978, pp 90-92

Temperature: Diodes and Diode-Connected Transistors

[17] Surface Temperature Thermometer
L. Nelson-Jones
Wireless World, April 1969, pp 180-183

www.syscompdesign.com
BIBLIOGRAPHY BIBLIOGRAPHY

[18] Electronic Thermometer

A.S.Henderson
Wireless World, June 1981, pp 50-52

[19] Semiconducting Temperature Sensors and their Application

Herbert B. Sachse
Wiley and Sons, Inc, 1975

[20] The Silicon Transistor as a Temperature Sensor

T.C.Verster
Temperature, Its Measurement and Control in Science and Industry
Volume 2, Part 2, Instrument Society of America
Pittsburgh, 1973, pp 1125-1134

[21] Data Sheet, MTS102 Silicon Temperature Sensor

Motorola Semiconductors P.O.Box 20912, Phoenix, Arizona,85036, USA

[22] An Exact Expression for the Thermal Variation

of the Emitter Base Voltage of Bi-Polar Transistors
Robert J. Widlar
IEEE Proceedings, January, 1967, pp 96,97

[23] Accurate Analysis of Temperature Effects in Ic-Vbe

Characteristics with Application to Bandgap Reference Sources
Yannis P. Tsividis
IEEE Journal of Solid State Circuits, Vol SC-15, No. 6, December 1980, pp 1076-1084

[24] LM34/LM35 Precision Monolithic Temperature Sensors

Application Note AN-460
Linear Applications Handbook
National Semiconductor, 1994

[25] Transistors– A hot tip for accurate temperature sensing

Pat O’Neil and Carl Derrington
Electronics, October 11,1979, pp 137-141

[26] A Precision Silicon Transistor Thermometer

Akira Ohte and Michiaki Yamagata
IEEE Transactions on Instrumentation and Measurement
Vol IM-26, No. 4, December 1977, pp 335-341

[27] Operational Amplifier Databook, 1993 Edition

National Semiconductor Corporation, 2900 Semiconductor
Drive, P.O.Box 58090, Santa Clara, California, 95052-8090, USA

IC Temperature Sensors

www.syscompdesign.com
 238

[28] LM34/LM35 Precision Monolithic Temperature Sensors

National Semiconductor Application Note 460
Linear Applications Handbook
National Semiconductor Corporation, 2900 Semiconductor
Drive, P.O.Box 58090, Santa Clara, California 95052-8090, USA
[29] Integrated Temperature Transducers
Michael J. Riezenman
Electronics, November 14, 1974, pp 130-132
[30] Operational Amplifier Circuits, Theory and Applications
E.J.Kennedy
Holt, Rinehart and Winston, 1988, pp 498-503
[31] LM134/LM234/LM334 3-Terminal Adjustable Current Sources
Data Sheet, Linear Data Book 2, 1988 Edition
National Semiconductor Corporation
[32] An RS-232 Thermometer
Don McLane
Home Automation and Building Control
October 1995, pp71-73

Voltage-Frequency Conversion
[33] 1Hz to 30MHz Voltage to Frequency Converter
Keith Williams
Electronics and Wireless World
September, 1985, page 58
[34] LM131/LM231/LM331 Precision Voltage-to-Frequency Converters
Application Circuit, ’Temperature to Freqency Converter’
Data Sheet, Linear Data Book 2, 1988 Edition
National Semiconductor Corporation
[35] Voltage-Frequency Converters
Eugene L. Zuch
Electronics, May 15, 1975, pp 91-95
[36] Versatile Monolithic V-F’s Can Compute
Robert A. Pease
Electronic Design 25, Dec 6, 1978, pp 70-76
[37] Designing V-F Converters to Handle Bipolar Signals
Robert A. Pease
Electronics, June 7, 1979, pp 139-145

Humidity

www.syscompdesign.com
BIBLIOGRAPHY BIBLIOGRAPHY

[38] Meterological Instruments

W.E. Knowles Middleton and Athelstan F. Spilhaus
University of Toronto Press, 1953
Chapter 4, The Measurement of Atmospheric Humidity

[39] A Course in Elementary Meteorology

Meteorological Office
Her Majesty’s Stationery Office, 1962
(The Campfield Press, St. Albans)

[40] A Thermistor Hygrometer

B. Bollen
Wireless World, December 1969, pp 557-561

[41] Elements of Meteorology

Albert Miller and Jack C. Thompson
Charles E. Merrill Publishing Company, 1970

[42] Relative Humidity Guage

Anthony J. Caristi
Electronic Hobbyists Handbook, Fall 1994
pp 84-88

[43] Electro Chemical Hydrometer

Low Cost Unit Measures Relative Humidity
J. Giannelli
Popular Electronics, October 1972, pp 33-35

[44] Relative Humidity Sensor Signal Conditioner

Jim Williams
Application note 3.
Linear Applications Handbook, Volume 1
Linear Technology, 1990

[45] Data Sheet, Series 691 Relative Humidity Sensor

Philips Components
1990-91 Data Book, page 684
(see reference [10] above)

[46] The Art of Electronics

Paul Horowitz and Winfield Hill
Cambridge University Press, 1989

[47] Troubleshooting Analog Circuits

EDN Series For Design Engineers
Robert. A. Pease
Butterworth-Heinemann, 1991

www.syscompdesign.com
 240

[48] Astable and Monostable Oscillators

Using RCA COS/MOS Digital Integrated Circuits
D.V.DiMassimo and A.R.Maslowski
RCA Application Note ICAN-6466
RCA Solid State COS/MOS Integrated Circuits Data Book,1980

[49] CMOS Cookbook

Don Lancaster
Howard W. Sams, 1977
Astable circuits, pp225-235

[50] Measure, Display Relative Humidity

W. Freeman
Electronic Design, January 1990, p136-137

[51] The Humidity FAQ

sci.geo.meteorology SIG
Richard Clark
rclark@lpl.arizona.edu or rdc@gas.uug.arizona.edu

[52] Small Scale Dewpoint Humidity Sensor

A.C.Easty and S.Young
Journal of Physics E, Scientific Instruments
Volume 9, No. 2, pp106-110, February 1976

[53] Vapourware: Revealing Your Humidity

Jeff Bachiochi
Circuit Cellar INK
Issue 57, April 1995, pp92-96

Wind Speed
[54] Meterological Instruments
W.E. Knowles Middleton and Athelstan F. Spilhaus
University of Toronto Press, 1953
The Cup Anemometer, pp 143-146

[55] Instruments for Physical Environmental Measurements

J.Y.Wang
Milieu Information Service
Cup anemometers, pp 190-193

Cooling effect sensors

[56] Cupless Anemometer Has Diode Wind Sensor
J.P.Scoseria
Electronics, August 30, 1979, p 156

www.syscompdesign.com
BIBLIOGRAPHY BIBLIOGRAPHY

[57] Measuring the Wind with Hot Metal

Shawn Carlson
Scientific American, November 1995, pp 98-99

[58] On the Convection of Heat from Small Cylinders in a Stream of Fluid: Determination of the Convection
Constants of Small Platinum Wires, with Applications to Hot Wire Anemometry
Lois Vessot King
Proceedings of the Royal Society of London
Vol XC, September 1914, pp563-570

[59] Thermal Techniques in Measurement and Control Circuitry

Jim Williams
Application Note 5
Linear Applications Handbook, Volume 1
Linear Technology, 1990

[60] A Digital Anemometer

Ken Okamoto et al
IEEE Transactions on Instrumentation and Measurement
Vol 43, No.2, April 1994, pp 116-120

[61] Dynamic Behaviour of Thermistor Flowmeters

Andrea Taroni and Gianni Zanarini
IEEE Transactions on Industrial Electronics and Control Instrumentation
Vol IECI-22, No.3, August 1975, pp 391-396

[62] Experimental Results on Dynamic Behaviour of Thermistor Flowmeters

Bruno Morten, Maria Prudenziati, Andrea Taroni and Gianni Zanarini
IEEE Transactions on Industrial Electronics and Control Insntrumentation
Vol IECI-25, No.3, September 1976, pp 232-234

[63] Log Converters

Application Notes 29, 30, 31, 211,264
National Semiconductor Linear Applications Handbook, 1986

[64] Translinear Circuits, A Proposed Classification

Barry Gilbert
Electronic Letters, 9 January 1975, Vol 11, No 1, pp 14-16

[65] A New Wide Band Amplifier Technique

Barry Gilbert
IEEE Journal of Solid State Circuits
Vol SC-3, No 4, December 1068, pp 353-365

[66] Supermatched Bipolar Transistors Improve DC and AC Designs

Carl T. Nelson
EDN, January 5 1980, pp 115-120

www.syscompdesign.com
 242

[67] Analog Multipliers Determine True Wind Speed

J.A.Connelly and M.B.Lundberg
EDN Magazine, April 20, 1974, pp 69-74

[68] The Gunnplexer Cookbook

Bob Richardson
Ham Radio Publishing Group, 1981

[69] The ARRL Handbook, 1991 Edition

The American Radio Relay League

Wind Direction
[70] The Weather for a Hobby
Raymond F. Yates
Dodd, Mead and Company, 1947
Selsyn weather vane, p 85

[71] Synchro Engineering Handbook

A.R.Upson and J.H.Batchelor
CRC Press, 1965

[72] Synchro and Resolver Conversion

Geoffry S. Boyes
Analog Devices, 1980

[73] Measuring Transducer Directivity

Peter D. Hiscocks
Wireless World, August 1978, pp 61-66

[74] Microelectronic Circuits

Adel S. Sedra and Kenneth C. Smith
Oxford University Press, 1991
Chapter 12: Signal Generators and Waveform Shaping Circuits

Air Pressure
[75] Meterological Instruments
W.E. Knowles Middleton and Athelstan F. Spilhaus
University of Toronto Press, 1953
Chapter 2, The Measurement of Atmospheric Pressure

[76] Instruments for Physical Environmental Measurements

J.Y.Wang
Milieu Information Service
Mercury Barometers, pp 172-173

www.syscompdesign.com
BIBLIOGRAPHY BIBLIOGRAPHY

[77] Solid State Barometer

Sudhir K. Gupta
Radio Electronics, June 1984, pp41-44

[78] Electronic Altimeter

Frank Ogden
Wireless World, May 1985, pp 37-41

[79] Digital Altimeter

Frank Ogden
Wireless World (issue unknown, after May 1985)

[80] Temperature Compensation Methods for the

Motorola X-ducer Pressure Sensor Element
C. Swartz, C. Derrington, J. Cragg
Motorola Application Note AN-840
(included in Industrial Control Applications
Motorola Publication DL412/D)

[81] Using the Motorola X-ducer Pressure Sensor Data Sheet

H. Nagel, R. Frank
Motorola Application Note AN-919
(included in Industrial Control Applications
Motorola Publication DL412/D)

[82] Temperature Compensation, Calibration and Applications

of Motorola’s X-ducer Pressure Sensor
H. Nagel
Motorola Application Note AN-922
(included in Industrial Control Applications
Motorola Publication DL412/D)

[83] MPX Pressure Sensors Used For Switch Applications

G. Swensen
Motorola Application Note AN-962
(included in Industrial Control Applications
Motorola Publication DL412/D)

[84] Calibration Free Pressure Sensor System

M. Burri
Motorola Application Note AN1097
(included in 8 Bit MCU Applications Manual
Motorola Publication DL408/D Rev 1)

[85] CMOS DACs in the Voltage Switching Mode

Steve Stephenson
Analog Dialog, 14-1, 1980, pp 16-18

www.syscompdesign.com
 244

[86] High Speed CMOS Logic Data Book

Texas Instruments, 1984
Data Sheet, SN74HC273, page 3-239
Specifications for HC MSI Circuits, page 2-10
Designers Information, page 7-17

[87] A Solid State Barometer for the HCS II

John Morley
Circuit Cellar Ink
October 1995, pp 63-69

Rain Guage
[88] The Weather for a Hobby
Raymond F. Yates
Dodd, Mead and Company, 1947
Rain guage, p 134

Passive Components
[89] A knowledge of capacitor subtleties helps solve capacitor-based troubles
Robert A. Pease
EDN, February 16, 1989, pp 127-134

[90] Understand capacitor soakage to optimize analog systems

Robert A. Pease
Electronics+Wireless World
October 1992, pp 832-835
also in
EDN, October 13, 1982, page 125

[91] Capacitors
Ray Marston
Electronics Now, March 1993, pp 57-64

[92] Under The Hood III: Capacitors

Brian Bergeron
QST, January 1994, pp 45-48

[93] Resonant Frequencies of Capacitors

Radio Communication Handbook
Radio Society of Great Britain, 1991

[94] Capacitors: A survey of present day technology and applications

R. A. Fairs
Wireless World, December 1974, pp 510-514

www.syscompdesign.com
BIBLIOGRAPHY BIBLIOGRAPHY

[95] Resistors: A survey of present day technology and applications

R. A. Fairs
Wireless World, October 1975, pp 487-492
[96] Troubleshooting gets down to the component level
Robert A. Pease
EDN, February 2, 1989, pp 176-184
[97] Tools for Schools: Math Packages for Education
IEEE Spectrum, September 1966, pp 41-45

68HC11 Microprocessor
[98] Motorola 68HC11 Reference Manual
Motorola Publication M68HC11RM/AD,Rev 3
General applications information on the 68HC11
family of microprocessors. A ’must have’.
[99] Motorola 68HC11 E Series
Motorola Publication M68HC11E/D, Rev2.0
Contains data on the E series
of devices, including electrical characteristics.
[100] 68HC11 Microcomputer
Technical Manual
Peter D. Hiscocks
Syscomp Electronic Design Limited, 2001
www.syscompdesign.com

www.syscompdesign.com