AM TRANSMITTER

by

RAJ DANIEL MAGNO

JOAQUIN TRINIDAD
DENZEL EDRICK VALENZUELA
ECE121L/E01

A Research Proposal Submitted to the School of EE-ECE-CoE in Partial Fulfilment of

the Requirements for the Degree

Bachelor of Science in Electronics Engineering

Mapúa University

April 2018
 Chapter 1

INTRODUCTION

The project aims to make a model around an AM radio transmitter that would fulfill

a range of at least 10 meters.

The AM signal created by the model must be received by a radio that will fill in as

the output device.

The prototype is an AM transmitter that can send audio signal to a radio receiver.

Using an audio input signal from an audio jack, the modulating signal can be heard in the

radio with a frequency equivalent to the oscillator’s output. The circuit can only transmit

audio signal thus, any other forms of signal like data and video are not available for

sending.
 Chapter 2

REVIEW OF RELATED LITERATURE

2.1 Resonant Frequency

The concept of resonance is used in the circuit to provide more power in the

operation of the transmitter. Since maximum power can be achieved when

resonance takes place in a circuit, resonance can either be in series connection or

parallel combination of an RLC circuit. Once the resonating frequency is achieved,

a higher power can give a longer range for the transmission of the audio signals to

the acting receivers.

2.1.1 Modulation Index

The index of modulation of an AM signal is the ratio of the voltage of the

modulating signal to the voltage of the carrier signal. A modulation level of greater

than one will result to over-modulation which will produce a distorted output. A

modulation level of less than one is an under-modulated signal while a perfect

modulation has a modulation level of equal to one.

2.2 AM Spectrum

The spectrum of the amplitude modulated signal is one important parameter

to be considered when designing a transmitter since it shows the behaviour of the

carrier signal and the two sidebands of the modulated signal. The spectrum shows

the three different signals, their voltage level and the operating bandwidth that is

essential in the design of the project for a more power-saving transmitter.

2.3 Carrier Frequency

The carrier frequency is produced using the oscillator in the circuit. Since an

AM transmitter is to be considered, the operating frequency of the carrier should

only range from 535 kHz to 1605 kHz since most AM radio receivers can only

detect signals from this frequency range.

2.3.1 Balance Modulator

The principle of a modulator that can combine the modulating signal

frequency and the modulating carrier frequency is also used in the system. In order

to form an amplitude modulated output, the balance modulator ( also known as the

multiplier circuit) will produce the carrier and sideband signal that is formed from

the combination of the two input signals. The amplitude of the modulated output

signal is varied by changing the frequency of the message signal.

2.3.2 Multilayer Perceptron Network

A multilayer perceptron is a network formed by perceptron which are simple

neurons. It is a concept introduced by Rosenblatt during 1958 [9]. A perceptron can

compute a single output from multiple real-valued inputs through the use of forming a

linear combination according to its input weight after which is put into the output via

nonlinear activation function. A single perceptron cannot perform very well due to the

limited mapping ability. A perceptron can only represent an oriented ridge-like function

no matter how it was activated. The benefit of a perceptron is that it can be used as a
building blocks for a larger and more practical structure [10]. Multilayer perceptron

(MLP) networks usually consists of source nodes which then forms the input layer,

hidden layers of computation nodes, and an output layer of nodes. The input layer will

accept patterns or actual data to the network, which transfers the data to one or more

hidden layers where the actual processing is done via a system of weighted connections.

The hidden layers then link to an output layer where the answer is output as shown in

Figure 2.2. [8]

Fig. 2.2 Multilayer Perceptron Network

2.3.3 Levenberg–Marquardt Algorithm

Also known as the damped least-squares method, was designed to specifically

tackle loss functions that takes form in a sum of squared errors, this factor makes it very

fast in training neural networks. It can be useful without the need for computing the exact

Hessian matrix in return, it will use the gradient vector and the Jacobian matrix. The first

step for using this algorithm is to calculate the Hessian approximation as well as the

gradient. The damping parameter is then adjusted in order to reduce the loss within each

iteration. This method however has its own negative issues one of which is it cannot be
applied to functions like root mean squared error or cross entropy error. This also

requires a huge amount of memory depending on the amount of data given so this is not

recommended when dealing with a big amount.

2.4. Raspberry Pi Model 3b

The Raspberry Pi is a small card shaped computer that is plugged to a device.

This little computer can process many tasks such as spreadsheets, word processing,

browsing the internet, and games, and also plays high-definition video. The 3rd

generation of Raspberry Pi is the latest and best version there is, as of this time. The new

and updated features of this generation is a 1.2GHz 64-bit quad-core ARMv8 CPU,

802.11n Wireless LAN, Bluetooth 4.1, and Bluetooth Low Energy (BLE). While also

having the previous version of Raspberry Pi 2. Its properties is a 1GB RAM, 4 USB

ports, 40 GPIO pins, Full HDMI port, Ethernet port, Combined 3.5mm audio jack and

composite video, Camera interface (CSI), Display interface (DSI), Micro SD card slot

(now push-pull rather than push-push), and a Video Core IV 3D graphics core []. The

benefit of this small device that it is user friendly and is easily programmable. In turn, can

be easily used in making the weather and gas sensors.

 Figure 2.4 Raspberry Pi Model 3b

2.5 Air Quality Sensors

Air quality sensors are widely available in the market, for personal and industrial

use. It can go as low as few dollars and as high as thousands of dollars depending on the

accuracy and function of the sensor. These sensors output analog and digital signals,

therefore it can be utilized to be used for common microcontrollers and microcomputers.

By passing through an analog-to-digital converter, the Raspberry Pi can also read analog

signals from these sensors since it can only accept digital signals. There are variety of

weather sensors that are able to detect temperature and humidity, dust, and pressure. For

common and affordable gas sensors, mostly NO2 and CO concentrations are gauged.

Each of the following sensors are described based on their datasheets.

2.5.1 Weather Sensors

2.5.1.1 DHT22

The DHT22 is a sensor used for measuring relative humidity and temperature. It

applies exclusive digital signal collecting technique and humidity sensing technology in

order to insure its reliability and stability. It is temperature compensated and calibrated in

accurate calibration chamber. It has a working voltage of 3.3V- 5.5V DC. The DHT22 is

shown in Figure 2.2

 Fig. 2.3 DHT22

2.5.1.2 BMP085

The BMP085 has an internal piezo-resistive sensor. It can measure accurate

pressure and temperature of the surrounding area. This sensor is sensitive to light

therefore the hole must not be exposed to any direct light source during operation. With

this sensor, the absolute altitude and pressure at sea level can be measured following the

formula shown in Figure 2.4.

Fig. 2.4 Computation of absolute altitude and pressure at sea level

 Fig. 2.5 BMP085

2.5.1.3 GP2Y1010AU0F

The GP2Y1010AU0F is a compact optical dust sensor. An infrared emitting diode

(IRED) and a phototransistor are diagonally arranged into this device. It will detect the

reflected light of dust in air. This sensor is effective in detecting very fine particle such as

cigarette smoke.

Fig. 2.6 GP2Y1010AU0F

2.5.2 Gas Sensors

2.5.2.1 MICS 2710

 MICS 2710 is a gas sensor that measures the amount of Nitrogen Dioxide (NO2)

in air. It is highly sensitive and heating the sensor will make it more accurate. The sensor

response to NO2 in air is represented in Figure 2.7.

Fig. 2.7 MICS 2710 Characteristics

Fig. 2.8 MICS 2710

2.5.2.1 MICS 5525

The MICS 5525 is a silicon gas sensor consists of an accurately micro-machined

diaphragm with an embedded heating resistor and the sensing layer on top. It has the

capability to measure the amount of Carbon Oxide (CO). The sensor characteristics is

shown in Figure 2.9.

 Fig. 2.9 MICS 5525 Characteristics

Fig. 2.10 MICS 5525

2.6 Internet of Things

The idea of Internet of Things is having a system of objects, people, places, control

devices, sensors, and computers that are interconnected and has the capability to transfer

data over the internet. Most of the system would no longer require any human

intervention such as monitoring systems. Another thing is that the IoT is able to
interconnect multiple and small devices efficiently. A simple diagram of an example of

an IoT shown in Figure 2.11. [11]

Fig. 2.11 Internet of Things

2.6.1 IoT Platform: ThingSpeak

ThingSpeak is an application platform for the Internet of Things. It allows you to

create an application around data is collected by the weather and gas sensors. It does real-

time data collection, data processing, visualizations, apps, and plugins. It’s a channel that

the Raspberry Pi sends the data to be stored. It is a channel is where you send your data to

be stored. Each channel includes 8 fields for any type of data, 3 location fields, and 1

status field. It can publish data to the channel, process the data, retrieval of the data, and

have an application that can display the important detail in a manageable form. Making

the reading of data simple and easy for most people.

 CHAPTER 3

METHODOLOGY

The study aims to develop a model of ANN with an input layer of weekday,

temperature, pressure and humidity. The output layer will be carbon monoxide and

nitrogen dioxide. Data needed can be gathered by utilizing low cost weather and gas

sensors that are widely available in the market and develop a device that has the

capability to connect and send data to the internet and will be deployed in three various

locations inside Intramuros. The data set obtained will be trained using Levenberg-

Marquardt algorithm. The device will operate 12 hours a day (6AM-6PM) and 6 days a

week (Monday-Saturday) for 2 months with an hourly interval of data monitoring. 80%

of the data will be used to train the ANN model and 20% to test it.

Figure 3.1 Basic ANN Modelling Flowchart

B. Proposed ANN Model

This figure shows the basic structure of the ANN model. ANN is consists of three

layers: the input layer, hidden layer and the output layer. In the input layer, the nodes

accept the data from the external data set and it will be transmitted to the hidden layer. In

the hidden layer, physical connections is not visible in the network because the actual

processing is done thru the system of weighted connections. The last layer is the output

layer wherein the processed data from the input layer up to hidden layer is being yielded.

Fig 3.2 Multilayer Perceptron Network – Levenberg Marquardt Algorithm

C. System Design

The figure shows the basic process on how the ANN works in the system. In the

input selection phase, the predictor variables are being selected based on the available

data. In this case, the chosen input variables are the weather sensors such as temperature,

pressure and humidity sensors. Then, the training data sets will be gathered. Training data

sets are raw data that should be transformed into a numerical representation wherein the
deep-learning algorithm could understand. The acquired data will be tested and validated

to ensure the accurateness of the statistics.

Fig 3.3 System Flowchart of the ANN Model

D. Device Development for Air Quality Monitoring

A Raspberry Pi model 3b will be used for the study as it has already the capability

to connect to the internet without any additional modules compared to the commonly

used ATmel-based microcontroller.

I. List of Components required for a single operating device.

1. Raspberry Pi model 3b

- It is the microcomputer that will be reading data from the sensors and will

send it to the cloud-based platform.

2. DHT22

- Sensor for reading temperature and humidity.

3. GP2Y1010AU0F

- Sensor used for dust amount level.

4. MICS 2710

- Sensor used for Nitrogen Dioxide.

5. MICS 5525

- Sensor used for Carbon Monoxide.

6. BMP085

- Sensor used for Pressure.

7. MCP3008

- ADC for Sensor to RPi connection.

8. Pocket Wi-Fi
 - Used to provide internet access to the RPI to allow it to transmit data to

the cloud-based platform.

9. Power Bank with Solar Panel

- Power source for the device.

Figure 3.4 Block Diagram of the Air Quality Monitoring Device

For the air quality monitoring device, the signal from the sensors will be received

and decoded by the Raspberry Pi. The weather and dust sensors would require an ADC

converter since the signal are in analog and needed to be converted to a digital signal in

order for the RPi to read. The RPi would be powered by a power bank with a solar panel.

The RPi is connected to the Pocket Wi-Fi for it to be able to send data to the cloud.

E. Hyperbolic Tangent Transfer function (Hidden Layer)

Fig3.5 Tan-Sigmoid Transfer

Function
 Hyperbolic tangent transfer function is very much related to the bipolar sigmoid

function because its output ranges from -1 to +1. This transfer function or sigmoid

transfer function is mostly used in neural networks because it is faster and its graph

depends generally on its speed rather than its exact shape of the transfer function. It is

good for the ANN hidden layer because even if the argument of the function is

substantially negative, it will generate a value closer to -1 thus, the transfer function will

maintain the learning since its output ranges from -1 to +1.

 REFERENCES

[1] Macas, T. “Metro Manila’s air quality even worse this year- data”. Retrieved from

http://www.gmanetwork.com/news/lifestyle/healthandwellness/509142/metro-manila-s-

air-quality-even-worse-this-year-data/story/, 2015

[2] United States Environmental Protection Agency, “Criteria Air Pollutants”, Retrieved

from https://www.epa.gov/criteria-air-pollutants, April 3, 2017

[3] Environment and Climate Change Canada “About the Air Quality Index” Retrieved

from http://www.ec.gc.ca/cas-aqhi/default.asp?lang=En&n=065be995-1, June 7, 2016

[4] Environment Protection Authority Victoria, “Australian air quality forecasting

system”, Retrieved from http://www.epa.vic.gov.au/air/aaqfs/, December 22, 2006

[5] United States Environmental Protection Agency, “Air Pollution: Current and Future

Challenges” Retrieved from https://www.epa. gov/clean- air-act-overview /air-pollution-

current-and-future-challenges, 2017

[6] Committee on Environment and Natural Resources, “Air Quality Forecasting”, 2001

[7] Artificial Neural Networks Technology, “What are Artificial Neural Networks?”

Retrieved from http://www.psych.utoronto.ca/users/ reingold/courses/ai/cache/neural2.

html

[8]"A Basic Introduction To Neural Networks", University of Wisconsin, 2007. [Online].

Available: http://pages.cs.wisc.edu/~bolo/shipyard/neural/local.html/.

[9] Panchal, G., Ganatra, A, et al. “Behaviour Analysis of Multilayer Perceptrons with

Multiple Hidden Neurons and Hidden Layers”,2011

[10] Honkela, A. “Multilayer Perceptrons” 2001

[11] Rose, K. et al. The Internet of Things: An Overview Understanding the Issues and

Challenges of a More Connected World. 2015

Quesada, A., “5 Algorithms to Train a Neural Network”, Retrieve from

https://www.neuraldesigner.com/blog/5_algorithms_to_train_a_neural_network,

2017