19-20 Kasım 2020

ONLINE

Digital Filter Design Based on ARDUINO and Its

Applications

Muharrem Çelebi 1
1
Kartepe Vocational and Technical Anatolian High School, Kartepe, Kocaeli.
muharrem.celebi@kocaeli.edu.tr

Abstract—Nowadays, due to the rapid development of digital other frequencies to the output. It is called as notch filter and
platforms, digital signal processing has become realizable into used to suppress mains noise [1].
small size processors. The purpose of this study is to test the Digital filters are divided into two groups according to the
filtering process on both PC-based platform and embedded impulse response. These are finite impulse response (FIR) and
system. The first goal of this study is to compare the results of the infinite impulse response (IIR). FIR filters have a linear phase
filtering process obtained in MATLAB and ARDUINO response, so the group delay is the same for each input
environments. For the second purpose, filtering process will be
frequency. IIR filters have a nonlinear phase response and
tried in real-time on ARDUINO platform with the produced
filter coefficients. Findings obtained as a result of the study are
hence the group delay varies according to the applied
presented and discussed. frequency. Since FIR filters are always stable, it is possible to
use them as an adaptive filter. The filter coefficients of FIR
filters are many, so the process loads increase. However, IIR
Keywords— FIR; IIR; filtering; MATLAB; ARDUINO filters have a few filter coefficients, so, they have process loads
low level [1].
I. INTRODUCTION
The purpose of digital signal processing is to perform on
digital systems instead of operations with analog circuits. In
this way, it performs the same task based on software, without
the need for electronic materials or replacement. This method
ensures that the design is simple and upgradeable. For
example, an analog filter made with RC elements, it is
necessary to change the materials to change the filter's cutoff
frequency. But if the same operation is to be done in digital
filtering, only the filter coefficient should be changed. Despite
the advantages of digital filters, analog filters are used in many
areas, such as cross-over circuits in speaker cabinets. Fig. 1. FIR filter block diagram
The filters are systems where the desired signals are
transferred to the output and undesired signals are suppressed.
In filter design, if the materials used are made with elements
such as R, C or L, it is defined as an analog filter. If filtering is
done in a digital environment such as PC or microprocessor, it
is called digital filter.
Various filter designs were realized. In design, the
keyword is the cut-off frequency which means is decision
point. Low-pass filters (LPF) are filters that allow the
frequency values below the cut frequency to pass, and
suppress the frequency above the cut frequency. High-pass
Fig. 2. IIR filter block diagram
filters (HPF) are filters that allow frequencies above the cut
frequency to be transferred to the output, and not frequencies In the Fig. 1 and Fig. 2 above, the xn and yn values
below the cut frequency. The band-pass filter (BPF) has two represent the input signals and the output signals in a discrete
cut-off points. This filter transfers the frequency values at domain. bn and an represent filter coefficients. As can be seen
between two cut-off frequencies to output and other in Fig. 1, the FIR filter multiplies at the past values of input
frequencies are damped. The band-stop filters (BSF) block the signal multiplied by bn filter coefficients then the result of
frequencies between the two cut-off frequency and transfer the each multiplication is added. In Fig. 2, the left side of the
block diagrams is similar to the FIR filter, in addition to,

978-1-7281-8073-1/20/$31.00 ©2020 IEEE

263
 19-20 Kasım 2020
ONLINE

multiplied at the past values of output signal by an filter

coefficients then the result of each multiplication is added.
�

�(�) � � �(�) ∗ �(� − �) (1)

���
� �

y(n) � � �(k) ∗ x(n − k) � � �(j) ∗ y(n − j) (2)

��� ���

Singh et al. [2], used the IIR filter to filter the ECG signal Fig. 3. Experiment connection - 1
and remove the noise. Filter coefficients were produced using
Matlab's FDA Tool and used the ATmega16A processor as the In this study, two experimental connection are used. The
processor platform. Rahmatillah and Karim [3], designed a IIR first experiment is offline filtering and the second experiment
Notch filter to reject 50 Hz power line noise. The algorithm is real-time filtering. Fig. 3 shows the structure of the
were embedded to ARDUINO Due board. Al-Busaidi and experimental connection - 1. First, a signal is artificially
Khriji [4], have builded a circuit that measures the ECG generated in the MATLAB environment. The sampling period
signal. Low-pass, notch ve band-pass filters were implemented is chosen as fs = 3000 Hz. The components of this signal have
in this structure. It was used ATmega32 processor. Tan and been added to 500Hz, 1000Hz and noise. 300 points of this
Jiang [5], presented teaching material for university students. artificial signal produced are embedded in ARDUINO. After
In this material, for digital filter implementatirons, they this process, the filter coefficients are embedded in the
underlined, sampling time, ADC tehcnique. they used 68HC12 ARDUINO. All codes and coefficients are executed in
microcontroller and assembley and C languages. Varshney ARDUINO board and the results are sent to MATLAB
and Tiwari [6], they designed finite impulse response (FIR) environment.
filter and implemented it in hardware as ATmega32 processor.
Vostrukhin and Vakhtina [7], performed moving avarage filter
using MATLAB and ARDUINO. Chiagunye et al. [8], in their
study, low-pass, high-pass and band-pass digital filter were
designed a circuit which is based on PIC16F877A
microcontroller. Sharifi et al. [9], in their paper, designed an
algorithm, for speech recognition based on AVR
microcontroller. Runge et al. [10], in their work, they
improved a system using FPGA for FIR filter. Mohammed
and Ishak [11], created digital filter based on dsPIC30F6010a
for motor driver. Koswatta and Karmakar [12], have builded a
Fig. 4. Experiment connection - 2
moving average filter work on PIC18F452 microcontrroler. It
was used filtering RFID tag reader. Gargava et al. [13], Fig. 4 shows the structure of the experimental connection -
established a system of Brain Computer Interfaces (BCI). In 2. The ARDUINO has an ADC inbuilt, however it doesn't has
this system, EEG signals were filtered, then extracted the DAC module inside. That is why, DAC circuit made using
features. Pujari et al. [14], compared the performance of FIR R2R resistor is soldered and connected with PORTD. In the
filter onto two digital systems. This systems are input circuit, sampling is done using the A0 pin. At the output,
microcontroller and FPGA. an 8-bit output value is generated via PORTD. Pull-up and
pull-down connections are performed on the A0 pin with two
II. MATERIAL AND METHODS 10Kohm resistors. In this way, the input signal is kept constant
at 2.5V DC level. In addition, with the 1uF capacitor, the
In this study, FIR_LPF, IIR_LPF, FIR_HPF, IIR_HPF ARDUINO is insulated with the signal generator. In sampling,
experiments were carried out for 4 filters. These filters were it is carried out in 10-bit resolution. However, filtering
calculated by the MATLAB program's filterDesigner tool. analysis and DAC operations are performed at 8-bit resolution.
With the export command, bn and an coefficients are produced Each FIR filter and IIR filter are designed with a 5th
and transferred to the ARDUINO environment. MATLAB- degree filter. Sampling frequency is decided as 3000Hz. The
ARDUINO can easily communicate with a single cable. This parameter values of the FIR filter are fpass = 750Hz, fstop =
allows data and results to be transferred quickly. 1000 Hz. For the IIR filter, fc = 750Hz is tuned. The response
of the filters obtained in accordance with these values are
presented in Fig. 5.

264
 19-20 Kasım 2020
ONLINE

prescaler values into ADCSRA registers. The measured fs

values are presented in Fig. 7.

Fig. 7. Maximum sampling frequency

When above Fig. 7 is examined, the minimum and

maximum sampling intervals of ARDUINO are displayed. It
can sample between 333kHz and at least 8.9kHz. To provide
Fig. 5. FIR and IIR filter response the fs = 3000 Hz we need, the delay cycle is performed with
the while loop when each sample is obtained. The codes used
III. EXPERIMENTAL RESULTS
for this process are displayed in Fig. 8.
Fig. 6 depicts the result graphics obtained after the first
experiment. In MATLAB environment, a test signal has been
artificially generated. Then, this sign was first filtered in
MATLAB environment and illustrated sequentially in the right
column. After this process, the test signal and filter
coefficients are embedded in ARDUINO. The results obtained
after this process are sent and illustrated in MATLAB
environment. When the filtering results are compared with, the
results of the operations performed in both MATLAB and
ARDUINO environment give similar results.

Fig. 8. Sampling and delay code

Another important point is that the processing time of the FIR

and IIR library used should be calculated. Improved
ARDUINO-Filters library [15] was used for filtering. The
results of these measurement times are presented in Table 1.

TABLE I. PROCESSING TIMES OF FILTERS

FIR_LPF IIR_LPF FIR_HPF IIR_HPF
145 µs 251 µs 146 µs 249 µs

If the filter degree is increased, the processing time to the

Fig. 6. Experiment - 1 filtering results
filter increases. This increases the total Ts duration and the Fs
value decreases. The filtering process is insufficient in
In the second stage, real-time filtering process is started. reducing the degree of filter. Therefore, in this study, 5th order
One of the most important problems here is the correct filters were designed for each filtering process.
adjustment of the fs value of the processor. In the ARDUINO Table I. When examined, the highest processing time is
command list, there is a command that directly adjusts the fs about 250 µs IIR filter. In addition to this time, the sampling
value. With ADCSRA register, the sampling rate can be process, output_PORTD and delay_time must be included in
adjusted. As a starting point, the sampling frequency values of processes. For this reason, the total processing time is about
this register should be determined. Therefore, 100 sampling 300 µs. When this time is reversed, the maximum real-time fs
were performed on a 100-element array. Before and after this is 3300 Hz. As a result, fs = 3000 Hz was chosen.
sampling process, time measurement was done using the
micros () command. Then, the actual fs value was calculated
by measuring the elapsed time in between two time points.
This process was calculated for ADPS0, ADPS1, ADPS2

265
 19-20 Kasım 2020
ONLINE

real-time process, setting the Fs value incorrectly, or not being

able to set it, will result in false results. There is no need for
such an adjustment in the experiments carried out in the PC
environment.
Fig. 9. Total processing time The results obtained in this study showed that when
MATLAB and ARDUINO results are compared, they give
However, as seen in Fig. 7, although the sampling values similar results. As a result of the results obtained in the second
are high, the fs value decreases due to the inclusion of other part, the digital filters has been successfully realized in real-
processing times (filtering time, output_PORTD etc.). time. High sampling frequency can be reached with faster
The sine signals, whose amplitude is 4.2Vpp and the ARDUINO platforms.
frequency values are 500Hz, 1000Hz and 1400Hz
respectively, are produced from the signal generator. These
signals are applied to the A0 pin, respectively. The filtered REFERENCES
sign obtained was exported by PORTD with DAC circuit. [1] S. Ertürk, "Digital Signal Processing", Birsen Publishing, Istanbul,
Both input signs and filtered signs are displayed in Fig. 10. Turkey, 2005.
[2] N. Singh, S. Ayub and J. P. Saini, " Design of digital IIR filter for noise
reduction in ECG signal", 5th International Conference and
Computational Intelligence and Communication Networks IEEE, pp.
171-176, 2013.
[3] A. Rahmatillah and Ataulkairm, " IIR digital filter design for powerline
noise cancellation of ECG signal using arduino platform", International
Conference on Physical Instrumentation and Advanced Materials,
Journal of Physics, 2017.
[4] A. M. Al-Busaidi and L. Khriji, "Digitally filtered ECG signal using
low-cost microcontroller", International Conference on Control,
Decision and Information Technologies (CoDIT), 2013, pp.258-263.
[5] L. Tan, J. Jiang, "Teaching Digital Filter Implementations Using the
68HC12 Microcontroller", American Society for Engineering Education
(ASEE), 2011.
[6] V. Varshney and M. Tiwari, "Realization of an FIR filter using
ATMEGA32 microcontroller", IEEE International Conference on
Emerging Trends in Computing and Communication Technologies
(ICETCCT), 2017.
[7] A. Vosrukhin and E. Vakhtine, " Studying Digital Signal Processing on
Arduino Based platform", Proceedings of the 15th International
Scientific Conference on Engineering for Rural Development, Jelgava,
Latvia, 2016.
[8] T. Chiagunye, I. Somtoochukw and E. A. Okereke, "Designing a
Microcontroller-Based Low-Pass, High-Pass and Band-Pass Digital
Filter with Op-Amp", Journal of Multidisciplinary Engineering Science
and Technology (JMEST), Vol. 2, issue 5, 2015.
[9] O. Sharifi, A. Zaeri and E. Ghafarioun, "Design and Implementation of
Simple Speech Recognition System Based-on Digital Filters in AVR
Fig. 10. Oscilloscope images of signals
Microcontroller", Majlesi Journal of Multimedia Processing, vol.
1, 2012.
12 pictures seen in Fig. 10 have been added. In each
[10] C. R. Runge, M. Cerqueira and F. J. Arnold, " Introducing
picture, the input signal and output signal are drawn. The top programmable logic devices in physics laboratories: a practical guide for
signal is the input sign, the bottom signal is the output sign. the implementation of experiments", Revista Brasileira de Ensino de
500 Hz, 1000Hz and 1400Hz filtered signs are displayed from Fisica, 2020.
left to right. FIR_LPF, IIR_LPF, FIR_HPF, IIR_HPF results [11] M. F. Mohammed and D. Ishak," Improved BLDC motor performance
are listed top to below in this field. with digitally filtering back-EMF using dsPIC30F microcontroller",
IEEE Student Conference on Research and Development (SCOReD),
When examining Fig. 10, in LPF processes, when it will 2009.
go towards high frequency, the output amplitude value [12] R. Koswatta and N. C. Karmakar, " Moving average filtering technique
decreases. Similarly, in HPF processes, the output amplitude for signal processing in digital section of UWB chipless RFID reader",
decreases as we go towards low frequencies. When FIR and Asia-Pacific Microwave Conference, 2010.
IIR filters are compared among themselves, they are 5th [13] P. Gargava, K. Sindwani and S. Soman, "Controlling an arduino robot
using Brain Computer Interface",3rd International Conference on
degree in both filters. However, the ratio of the IIR filter to the Reliability, Infocom Technologies and Optimization, 2014.
compression of the signals is better. [14] S. Pujari, A. Yeotkar, V. Shingare, S. Monin and B. Kokare,
"Performance analysis of microcontroller and FPGA based Signal
Processing a case study on FIR filter design and implementation",
IV. CONCLUSION International Conference on Industrial Instrumentation and Control
(ICIC), 2015.
In filter design, the exact determination of the Fs value is
[15] website : https://github.com/tttapa/Filters
important for the accuracy of the processes. However, in a

266