ENGINEERING FOR RURAL DEVELOPMENT Jelgava, 25.-27.05.2016.

STUDYING DIGITAL SIGNAL PROCESSING ON ARDUINO BASED PLATFORM

Aleksandr Vostrukhin1, Elena Vakhtina2
1
Stavropol Technological Institute of Service, Russia; 2Stavropol State Agrarian University, Russia
avostrukhin@yandex.ru, eavakhtina@yandex.ru

Abstract. Digital Signal Processing (DSP) has become today an integral part of the basic skills required for
scientists and engineers of many specialties. Most courses offer only using for DSP laboratory classes the Matlab
system. This approach has the disadvantage it excludes the students acquisition of the basic skills of software
development for implementation of DSP algorithms based on real digital devices. DSP is informative
technologies in real-time and in most cases they use the closest to the hardware Assembler and C/C++
languages. To eliminate this drawback the authors propose using for DSP laboratory classes the known Arduino
platform. Nowadays, this platform is one of the most affordable budgetary and rather universal decisions
produced on popular AVR microcontrollers. It allows using of the most popular in DSP Assembler and C++
languages. As an example the structure of the laboratory lesson: Development and research of the moving
average filter in Matlab and Arduino is presented. This filter is the most common. Embodiments of the filter
developed in Matlab and ArduinoUno-based controller are given. And oscillograms of the signals at the input of
the filter and at its output are shown. Integrated using of the Matlab system and Arduino platform contributes to
a better understanding of theoretical fundamentals of DSP, and more importantly, influence development of
practical skills in design and development of microprocessor systems for DSP.

Keywords: lab work, virtual-real method, Matlab, moving average filter, microcontroller, program.

Introduction
Digital Signal Processing (DSP) is one of the most powerful technologies that will determine the
further development of science and technology in the XXI century. DSP is a set of the mathematical
methods, the algorithms, and the techniques that are used to control the signals after they have been
converted into digital form. Signal processing can solve many tasks, such as: enhancement of the
image quality, recognition and synthesis of speech, compression of data for storage and transmission,
etc. Revolutionary changes have already happened in a broad range of activity spheres:
communications and broadcasting, radar and sonar, high quality sound reproduction and image
compression, in medicine, in monitoring and control systems, etc. [1]. These changes have a direct
affect to engineering education. In this article, we will focus on improving the DSP training methods.
This is important, because DSP is one of the most topical disciplines for the modern engineer
development.
Note that most DSP courses have offered using for lab works only the program Matlab. Virtual
lab, has a lack excludes students acquisition of practical software development skills to implement
DSP based on real-programmable systems (PS). DSP is a real-time information technology. In most
cases for implementation DSP based on real-PS C/C++ and Assembler programming languages are
used. Popularity rating of these languages is above Matlab [2]. The purpose of this article is to show
the ability of using Arduino platform for real lab works on DSP.
Arduino platform (hereinafter simply Arduino) contains an extensive assortment of modular
hardware and software with an open source license, continues to develop and is supported by
sufficient amount of literature [3-5]. Arduino is the most affordable budget solution, implemented on
the popular AVR microcontrollers and allows using C/C++ and Assembler languages the most
requested for DSP.
The DSP study process should be divided into two stages. The first stage is studying DSP
algorithms using computer-aided design (CAD) systems, for example, the Matlab program [6]. The
second step implementation of DSP algorithms developed in Matlab on the base, e.g., Digital Signal
Processor. However, Digital Signal Processor is quite complicated PS for beginners to learn DSP [7].
Therefore, the authors suggest for the initial stage of DSP studying to use Arduino as a simplified
analog of the Digital Signal Processor. Arduino hardware is constructed on AVR microcontrollers of
Atmels company [3-5]. It is a well-known microcontroller family, on the basis of which the
Microprocessor Technique is studied in many universities [8-11]. In addition, Arduino software is
supported by such program environments as Matlab, AVR Studio 7 (for 8-bit and 32-bit AVR
microcontrollers) and new Windows 10 operating system [12-14].

236
ENGINEERING FOR RURAL DEVELOPMENT Jelgava, 25.-27.05.2016.

Materials and methods

As an example, implementation in Matlab and Arduino moving average filter, which is the most
common in DSP, is considered. We rely on a generally accepted approach used in the implementation
phase of research and development (R & D) during designing PS. This approach with regard to lab
classes is: from virtual model to real sample. Didactic effectiveness of this approach was discussed
in detail in the works [15-17]. In this example, the virtual model is created in Matlab, while real PS
on Arduino.
Studying of the moving average filter is based on its principle of operation by averaging a number
of points from the input signal. This is expressed by equation (1) from the work [1]:
M 1
1
y[i ] =
M
x[i + j] ,
j =0
(1)

where x[] input signal;

y[] output signal;
M number of points used in the moving average;
i sample number (index) of the output signal;
j number of averaged sample.
This equation only uses points on one side of the output sample being calculated. For example, in
the 4 point moving average filter, point 27 in the output signal is given by (2):
( x[27] + x[28] + x[29] + x[30])
y[27] = . (2)
4
Sensor signals of automated process control systems are influenced by various interferences
called noise. Noising of a useful signal is determined by:
signal-to-noise ratio, which is equal to the mean divided by the standard deviation;
coefficient of variation (CV), which is defined as the standard deviation divided by the mean,
multiplied by 100 % [1].
Students task: to evaluate effectiveness of the moving average filter for noise suppression by the
CV. As initial data the filter order and input signal parameters are used.
For solving this task a student must obtain an input signal with given parameters of digital noise.
Digital noise is an important concept in both electronics and DSP [1]. Generation of digital noise is
realized by the random number generator (RNG). There is a special function for this task in most
programming languages. So, in C/C++ language of Arduino the expression: X = random (0, 255)
performs appropriation of variable X of a random value with each new execution of this command.
This random value will be located in the range from 0 to 255 and has an equal probability of
appearance within these limits.
Step 1 Matlab-based (virtual) part of the lab work. The model of the system for forming the
input signal samples, which is shown in Fig. 1, is developing in the Simulink (included in Matlab)
software. Uniform Random Number (URN) block is a RNG with uniform distribution. Rounding
Function block is intended for rounding off numbers, generated by URN, to integers. Scope unit is
designed for visualization of a signal applied to its input.

Fig. 1. Model of the system for forming the input signal samples
Range generated numbers from 0 to 255 are set in URN settings. This range is convenient because
the maximum decimal number 255 is equivalent to the maximum value of 8-bit binary code. To store
this code, you can use 8-bit registers, which simplifies the task solution while implementing DSP
algorithms in Assembler.

237
ENGINEERING FOR RURAL DEVELOPMENT Jelgava, 25.-27.05.2016.

Fig. 2 shows an input signal consisting of 30 samples. The values of the input signal samples are
read visually from the oscillogram of the Scope block (Fig. 1).

Fig. 2. Filter input signal

Let us imagine the signal as a vector consisting of 30 samples in the form of one-dimensional
matrix from 10 columns by 3 samples in each. This matrix is a part of the Program 1, which is
presented in Fig. 3. Matlab processes each sample (matrix element) by columns, starting with the first
sample in the following sequence: x[1] = 170, [2] = 251, [3] = 141, [4] = 60 x[30] = 241.
% FINITE IMPULSE RESPONSE (MOVING AVERAGE FILTER)
% This program is designed to process 30 samples of the input signal by the 4 point moving
% average filter
x = [170 60 129 111 225 156 220 95 28 7
251 162 3 126 84 203 114 64 116 35
141 56 204 32 153 155 176 30 36 241];
for i = 1:26 % Loop through each sample in the output signal
y(i) = 0; % Fill sample of the output signal with null values
for j = 0:3 % Loop for calculation of the next sample
y(i) = y(i)+ x(i+j);
end % Loop end
y(i) = y(i)/4; % Calculate the average value of the next output signal
end % Loop end
y % Output of values of the output vector on monitor
i = [1:26]; stairs(i,y); % Build the graph of the vector of output samples
Fig. 3. Program 1
The result of the Program 1 is the output signal vector consisting of 26 samples and shown in
Fig. 4. The value of the 27th sample is not calculated, because this calculation requires the 31st sample
of the input signal, which is absent.
Matlab has functions meant to work with a vector argument. To calculate the average values the
function (mean), standard deviation function (std) are used. These functions could be utilized for
processing the input and output filter signals. As a result, we obtain for the input signal the average
value i 119.4 and its standard deviation i 72.2. For the output signal o 115.4 and
o 39.9, respectively. Then the CV of the signals: at the filter input Ci = i / i =
= (72.2 / 119.4)100 % 60 % and at the filter output Co = o / o = (39.9 / 115.4)100 % 35 %.
Evaluating effectiveness of the filter allows the noise reduction factor, which shows in how many
times the noise at the filter output is less than the noise at its input. In this case, the noise reduction
factor by the filter that was realized in Matlab: Cm = i / o = 60 / 35 1.7.
Step 2 Arduino-based (real) part of the lab work. The experiment involved two controllers
Arduino Uno. The first controller is required to generate the input signal with the set parameters, the
second controller for implementing the filter. ATmega328 Microcontroller, on which Arduino Uno
controllers are implemented, has the supply voltage Vcc = 5 V. The 8-bit DAC with R-2R structure is
connecting to the parallel ports of each controller. An example of DAC construction and its connection
to the Arduino Uno controller is in the resource [18]. Input and output filter signals are controlled by
the DSO-2090 USB dual-channel oscilloscope.

238
ENGINEERING FOR RURAL DEVELOPMENT Jelgava, 25.-27.05.2016.

Fig. 4. Filter output signal

The first controllers RNG is tuned to generate a random number between 0 and 255. Each
random number is output by this controller in the form of 8-bit binary code to a parallel port. If binary
code that is equivalent to 255 is fed into the port, then the maximum voltage level 5 V on the DAC
output is generated. The minimum voltage level 0 V is generated at the DAC output when binary
code of zero is fed to the input port. If the value of a random number is changed by one unit, then
DAC output voltage is changed to the value determined by the expression:
Vcc / 255 = 5 / 255 19.6 mV.
The signal at the DAC output of the first controller, i.e. at the filter input, is displayed by the CH1
channel in Fig. 5. The signal at the DAC output of the second controller, i.e. at the filter output, is
displayed by the CH2 channel in the same figure.

Fig. 5. Oscillogram of signals: CH1 - the filter input signal; CH2- the filter output signal
The following Fig. 6 shows the program 2 fragment for a 4 point moving average filter that is
realized on the Arduino Uno controller.
We form the sample values of the input and output filter signals according to scillograms and
process them in Matlab. As a result, we obtain for the input signal of the filter the mean value
i 2.7 V and the standard deviation i 1.4 V; for the output signal, respectively 2.5 and o
0.5 V. We define the CV for the input and output signals of the filter: Ci = i / i = (1.4 / 2.7)100 %
52 %;

239
ENGINEERING FOR RURAL DEVELOPMENT Jelgava, 25.-27.05.2016.

o = o/o = (0.5/2.5)100 % 20 %. In this case, the noise reduction factor by the Arduino-based
filter is Ca = Ci / Co = 52/20 2.5.
{ // Loop Start
x1 =
x2; // Send the second sample into a memory cell of the first sample
x2 =
x3; // Send the third sample into a memory cell of the second sample
x3 =
x4; // Send the fourth sample into a memory cell of the third sample
x4 =
analog Read(ADC0); // Read from ADC the conversion result of the next sample
// and send its value into a cell memory of the fourth sample
x4 = x4/4; // Convert 10-bit binary code to an 8-bit
y = (x1+x2+x3+x4)/4; // Calculate the average value of the last four samples
PORTD = (y); //and output this value to the parallel port PD
} // Go to the top of the loop
Fig. 6. Fragment of Program 2

Results and discussion

The important thing is that in both cases we have the noise reduction factor close to a definite
value. It is known that the noise reduction factor of the moving average filter is equal to the square
root of the samples number involved in the averaging. For example, a 100 point moving average filter
reduces the noise by a factor of 10 [1]. Consequently, for our 4 point filter the noise reduction factor
should in ideal case be 2. This enables to conclude that both study steps of the moving average filter
simulation and experiment are correct.
This pedagogical project of implementing the virtual-real method to organization of DSP
laboratory classes using Arduino is in its infancy. There is only one study Arduino platform as the
means of DSP training [7]. It is not unusual, because, on the one hand, the DSP discipline was formed
as a separate course of radio engineering faculties in Russia only in the last decade, on the other hand,
the developing Arduino software has just begun to adopt as training means.

Conclusions
Arduino platform can and should be used to study the DSP course. The results presented in this
paper prove that there is an opportunity of developing real DSP systems on the base of this platform.
This fact contributes to realization of the important in educational process scientific and didactic
principle from theory to practice. The benefits of real lab work compared with virtual ones are
obvious and do not require evidence, sufficient to recall the words of Confucius: I hear and I forget, I
see and I remember, I do and I understand. Reasonable integration in lab works the virtual and real
techniques is contributing to deeper understanding of the studied issues.

References
1. Stewen W.S. The Scientist and Engineers Guide to Digital Signal Processing. Second edition.
San Diego: California Technical Publishing, 1999. 650 p.
2. TIOBE Index for February 2016 [online] [12.02.2016]. Available at:
http://www.tiobe.com/index.php/content/paperinfo/tpci/index.html
3. Igoe . Making Things Talk: Using Sensors, Networks, and Arduino to see, hear, and feel your
world. OReilly Media, 2011. 496 p.
4. .. Arduino (Projects using Arduino
controller). 2nd ed. SPb.: BHV-Petersburg, 2015. 446 p. (In Russian).
5. Blum J. Exploring Arduino: Tools and Techniques for Engineering Wizardry. Wiley, 2013. 384 p.
6. .., .., .., ..
MATLAB: . (Digital Signal Processing and MATLAB: textbook).
SPb.: BHV-Petersburg, 2014. 512 p. (In Russian).
7. Hochgraf . Using Arduino To Teach Digital Signal Processing. Reviewed Paper of ASEE
Northeast Section Conference, March 14-16, 2013, Norwich University. [online] [12.02.2016].
Available at: http://asee-ne.org/conferences/aseene/2013/index.php/aseene/aseene2013/paper
/view/235/24.
8. . . AVR. : . (AVR
Microcontrollers. Workshop for beginners: tutorial). Moscow: Publ. house Bauman Moscow State
Technical University, 2013. 280 p. (In Russian).

240
ENGINEERING FOR RURAL DEVELOPMENT Jelgava, 25.-27.05.2016.

9. .. AVR (Development of devices on

AVR microcontrollers). SPb.: Science and Technique, 2013. 528 p. (In Russian).
10. .., .. AVR
: (Introduction in Programming of AVR Microcontroller in
Assembler Language: textbook). Moscow: Publ. house Ileksa, 2010. 184 p. (In Russian).
11. .. AVR Tiny Mega ATMEL
(TinyAVR and MegaAVR Microcontrollers of ATMELs company). Moscow: Publ. House
Dodeka-XXI, 2015. 560 p. (In Russian).
12. Arduino Support from MATLAB. [online] [15.09.2015]. Available at:
http://uk.mathworks.com/hardware-support/arduino-matlab.html.
13. Arduino programming with Atmel Studio. [online] [10.10.2015]. Available at:
http://www.visualmicro.com/page/Arduino-for-Atmel-Studio.aspx.
14. Microsoft brings Windows 10 to Makers. Arduino Partnership. [online] [25.04.2015]. Available
at: https://blogs.windows.com/buildingapps/2015/04/29/microsoft-brings-windows-10-to-makers.
15. ., .
(Integration of the real and virtual in a lab experiment). Vysshee obrazovanie
v Rossii, 2008, No 6. pp. 77-81. (In Russian).
16. .. (Didactic project of lab
experiment). Pedagogicheskij zhurnal Bashkortostana, 2009, No 3 (22). pp. 147-155. (In
Russian).
17. Vakhtina E., Palkova Z. Didactic designing of Learning Objects. Proceedings of 14th
International Scientific Conference Engineering for Rural Development, May 20-22, 2015,
Jelgava, Latvia, pp. 661-668. [online] [31.05.2015]. Available at:
http://www.tf.llu.lv/conference/proceedings2015/Papers/107_Vakhtina.pdf.
18. Arduino Waveform Generator. [online] [02.12.2015]. Available at:
http://www.instructables.com/id/Arduino-Waveform-Generator/?ALLSTEPS.

241