DESIGN, IMPLEMENTATION AND OPTIMISATION OF

4X4 MIMO-OFDM RECEIVER FOR

COMMUNICATION SYSTEMS
S. T. Manju, *Sanket Dessai
Department of Computer Engineering,
M.S. Ramaiah School of Advanced Studies, Bangalore-560058
*Contact Author e-mail: sanketdessai@msrsas.org

Abstract
Multiple Input Multiple Output (MIMO) systems use more than one antenna at both ends of the communication
link. For more reliability the number of antennas at receiver should be more than at transmitter side. Over the past
decade, the use of MIMO system has rapidly gained popularity due to its enhanced performance capabilities of improved
Reliability, Spatial Diversity Gain and Spatial Multiplexing Gain. Orthogonal Frequency Division Multiplexing (OFDM)
is one of the best digital modulation schemes, where signal is divided into number of narrow band signals to obtain
spectrum efficiency and minimizing the Inter Symbol Interference (ISI). Thus, combining MIMO and OFDM technologies
will improve spectral efficiency, Link reliability, spectral gain and data rate.
In this paper, MIMO and OFDM techniques are combined to improve the receiver performance. Software
reference model for 4x4-MIMO-OFDM Transmitter for wireless communication system is designed and implemented in
MATLAB and the same design is also implemented in C for embedded platforms. The evaluation of Bit Error Rate (BER),
Signal to Noise Ratio (SNR), Peak Signal to Noise Ratio (PSNR) and Mean Square Error (MSE) performance of the
MIMO-OFDM technique based on comparative analysis of modulation schemes namely QPSK over AWGN channels is
carried out. Optimization of the system implemented in C has been carried out using Real View Development Suite (RVDS)
by considering different ARM cores namely, ARM 1176, and ARM Cortex A8. Serialised image data was used to analyse
the system performance. MSE and PSNR are computed for 4x4 MIMO-OFDM and compared with QPSK modulation.
For a color image, QPSK has PSNR of 58.6713 dB at SNR 5dB and for a binary image 78.195 dB at SNR 3 dB.
Profiling results of the system developed in C is analysed in RVDS. ARM Cortex A8 gives better performance in terms of
number of cycles estimated to execution of code and code coverage when compared with lower versions of it. After
optimisation in ARM 1176, the number of cycles reduced is 39,561. And in ARM Cortex A8 the number of cycles reduced
is 31,616. The performance of the system can be improvised by using the higher methods of sub- blocks like higher QAMs,
increase the number of sub carriers in FFT and using advanced MIMO techniques.

Keywords: 4G, Alamouti STBC, MIMO-OFDM, FFT, 16-QAM, QPSK, ARM1136, Cortex A9, RVDS
Nomenclature ISI Inter Symbol Interference
Data bit, b IFFT Inverse Fast Fourier Transform
Date rate bits per second, bps MIMO Multiple Input Multiple Output
Cycle Cycle, s MISO Multiple Input Single Output
f Frequency, Hz MIPS Million Instructions Per Second
Kbits Kilo bits, Kb=220 =1024b OFDM Orthogonal Frequency Division Multiplexing
Mbits Mega bits, Mb= 220 = 1,048,576b PSK Phase Shift Keying
IC Instruction Count QAM Quadrature Amplitude Modulation
IPS Instructions per second QPSK Quadrature Phase Shift Keying
RVDS Real View Development Suite
RISC Reduced Instruction Set Computer
Abbreviations STC Space Time Codes
ASK Amplitude Shift Keying STBC Space Time Block Codes
ARM Advanced RISC Machine STTC Space Time Trellis Code
BER Bit Error Rate SNR Signal to Noise Ratio
CC Code Coverage ZP Zero Padding
CP Cyclic Prefix
DFT Discrete Fourier Transform
FSK Frequency Shift Keying
1. INTRODUCTION
FEC Forward Error Correction
FT Fourier Transform Wireless communication is one of the thrust area of
FFT Fast Fourier Transform research and it is developing extremely faster in current
IC Instruction Count generation and is promising in future. Currently 3G
IPS Instructions Per Second standards are deployed in mobile communication

SAS TECH Journal 57 Volume 12, Issue 2, September 2013

system and 4G is evolving worldwide. This comparative 4. METHODOLOGY
study between 3G & 4G tells about the background and
the vision for the 4G. 3G is at this time the world’s most • Design specifications of the implementation
excellent connection method when it comes to mobile issues of the 4x4 MIMO-OFDM receiver was
phones particularly meant for mobile Internet. This arrived based on the reviewed literature and
paper deals with the Receiver design of MIMO-OFDM, study
which uses multiple antennas at receiver for fast • Design and Modelling of the 4x4 MIMO-
communication. A 4G framework has the basic keys
OFDM receiver using different
like multiplicity and flexibility.
algorithms/techniques was been carried out
The growing demand for high system capacity, using the MATLAB tool based on the
high transmission rate and broadband access is main
literature review and the study carried out
reason for development of the wireless system which
will satisfy all the above conditions. One of the • Performance estimation of the 4x4 MIMO-
solutions for all these conditions is Multiple Input OFDM receiver and its different
Multiple Output (MIMO) Orthogonal Frequency algorithms/techniques has been analyzed
Division Multiplexing (OFDM) system, which is the based on the modelled design
combination of MIMO and OFDM. The combination of • Implementation of the 4x4 MIMO-OFDM
MIMO with orthogonal frequency OFDM is the
receiver has been carried out using C and
important solution for enhancing the data rates and to
improve spectral efficiency of next generation wireless Assembly coding for embedded platforms
communication systems. Improvements in data rates can using keil/RVDS/CCS for ARM cores and
be achieved by using multiple antennas both at the DSP
transmitter and the receiver side. • Performance estimation of the 4X4 MIMO-
ODFM receiver has been carried out based on
the designed model
2. PROBLEM STATEMENT • Optimization techniques has been applied to
The available spectrum is fixed but the demand for optimize the implementation of 4x4 MIMO-
high data rate and high reliability is growing day by day OFDM receiver for the embedded platforms
to suite the more improved and attractive applications. based on the design and implementation
In this paper, we design, implement and optimize the performance estimations
4x4 MIMO-OFDM receiver for different embedded • Functionalities test has been carried out on the
platform. The idea behind developing the MIMO- optimized 4x4 MIMO-OFDM receiver
OFDM transmitter is to get the advantage of both subsystems based on the design specifications,
MIMO and the OFDM technology in achieving higher implementations specifications
data rate, spectral efficiency and reliability.
5. DESIGN AND SOLUTION PROCEDURES
3. SYSTEM REQUIREMENTS Integrating the MIMO and OFDM which consists of
blocks of demodulator, de-interleaver and viterbi
The following are the system requirement decoder FFT and cyclic prefix results a block diagram
specifications arrived for different subsystems of the 4x4 MIMO OFDM receiver shown in Figure 1. The
system receiver combines the 4x4 data transmitted with noise
added in the channel. The combined data is converted to
• MIMO system typically consists of m
parallel to remove the guard interval inserted during
transmitting antennas and n receiving
transmission. After removing guard interval fed to FFT
antennas
to remove the carrier frequency to form a continuous
• In this system channel is assumed to be a
wave. Output of FFT is converted to serial. The serial
static channel and that the channel is output is demodulated to the corresponding
known perfectly at the receiver for all the
constellation points (QAM or QPSK) to obtain binary
systems being developed
output bits. The demodulated output is de-interleaved by
• OFDM system generally consists of changing the row and column matrix. Trellis is
different sub blocks namely FFT block, calculated before decoding to assume the next coming
Cyclic prefix and parallel to serial bits, then decoded using viterbi decoding. Finally
conversion reconstructing of the decoded bits had carried out to get
• Both spatial diversity and spatial the original output.
multiplexing techniques should be
supported Error correcting in digital communication is the
major challenge to eliminate noise produced in the
In this research work channel is assumed to be a channel due to noise occurred for various conditions.
static channel and is known perfectly at the receiver for Using this error correction codes can achieve a good
all the systems being developed. performance even in bad carrier or channel due to noise

SAS TECH Journal 58 Volume 12, Issue 2, September 2013

condition. Present communication uses Forward Error integrated with MIMO-OFDM that the noise level
Correction (FEC) codes. In FEC coding, during completely reduces at 2dB.
transmission a extra redundant bits are added to the
Implemented C code is used in RVDS to run on
input given. This allows correcting the errors to a
different ARM cores and performance analysis of the
certain extent at the receiver.
different cores is studied. Profiling in RVDS shows the
performance of the codes, like bottle necks in code,
code coverage by the processor. Stack utilization of the
codes and cycle per instruction. Average time taken to
Vector Data execute a instructions. By running on the different
Removing AWGN cores and analysing it, the best core for the application
FFT Operation Converting into 4x4 Combining
cyclic prefix Noise of the 4x4-MIMO-OFDM Transmitter can select.
Paral el
In MATLAB, the code development is made taking
the color image and the binary image as shown in the
Figure 3 and Figure 4. After transmission noise is added
to the transmitted data. After increasing the SNR rate to
a sufficient level, a reconstructed image equivalent to
input is obtained.
Paral el to
Reconstructing
Serial Demodulation De-interleaver FEC decoding
Output
Convertion

Fig. 1 Block Diagram of MIMO-OFDM Receiver

The modulation technique used is digital
modulation. The digital modulation provides a better
error correction capability; transmission takes place in
security and avoids multipath fading. The digital data
provided by any application is to be modulated onto a
carrier for transmission. The data is transmitted by
adding noise. The receiver receives the 4 elements at a
time along with its complex conjugates. So at a time
four elements are received with its four multiple copies.
In C, real elements are received in one array and
imaginary elements are received in other array. The
matrix multiplication is done for each element, which
contains both real and imaginary terms. To perform
matrix multiplication, a function is created
(Receiver_Amouti). A matrix multiplication function is
created for this. The output obtained in C is compared
with the MATLAB. The 4x4 combined output is input
Fig. 2 SNR V/s BER of QPSK Integrated with
to cyclic prefix.
MIMO-OFDM

6. RESULTS AND DISCUSSIONS

This section deals with the results of the
implemented 4x4 MIMO-OFDM receiver. The results
are obtained with their output file that is generated after
successful simulation in MATLAB and C. The
performance of the system is estimated in MATLAB for
image input and checking the performance using MSE
and PSNR and in C the performance is evaluated on
different embedded platforms.

Figure 2 shows the SNR versus BER plot for Input Image Image with noise Noise reduced
QPSK and QPSK integrated with 4x4 MIMO-OFDM. at SNR=1db
at SNR = 5db
As shown in the graph, the bit error is reduced for
QPSK integrated with 4x4 MIMO-OFDM receiver. It
proves that the MIMO system is reliable and reduces the Fig. 3 SNR V/s BER of MIMO-OFDM for Color
Images
bit error rate. The error is very low using QPSK

SAS TECH Journal 59 Volume 12, Issue 2, September 2013

 coverage by 3.27% and the uncovered code is 26.34%.
The code coverage by functions is as follows full
coverage is 32.90%, partial coverage by 40.65% and the
no coverage is 26.45%

Input Image Image with noise Noise reduced

at SNR=1db
at SNR = 5db

Fig. 4 SNR V/s BER of MIMO-OFDM for Binary

Images
Figure 5 shows the profiling graph for the ARM
Cortex A8 core. Left hand side top shows the over
analysis section. The overall analysis section gives the
information about the, total time, execution count and
sampling information, number of estimated cycles is
2,75,845 and average instructions per second is 74,244.
The code coverage pie charts are located in the bottom
left and it gives a look at percentage of instruction and
functions executed in the code during the execution in
ARM Cortex A8 core. The code coverage by instruction
pie chart shows graphically the percentage of executed
assembly instructions in ARM Cortex A8 core. In that
graph green represents the fraction of entirely executed
Fig. 5 RVDS Profiling for ARM Cortex A8
instructions, red represents uncovered instructions, and
yellow represents partially covered assembly Figure 7 shows the plot of the ARM CortexA8 and
instructions. The 2nd pie chart is like that of the code its optimized coverage based on instructions, average
coverage by instruction chart, but shows code coverage instructions and the cycles. Blue bars for the ARM
function by function. For ARM Cortex A8, its 72.14% CortexA8 and the red bar for the optimised code on the
percent instruction is completed fully that is shown in ARM CortexA8. Also the various analysis studies were
green and 3.04% in yellow which is partially used and also carried out using the ARM1176 and its
24.82% in red color which is not used at all. And code observations that the blue bar presents the ARM1176
coverage by function is 32.26 % fully executed so in and red bar presents the optimised code on the
green color, and 41.29% yellow and 26.45% in red ARM1176.The observed results form the Figure 8 say
respectively. The ARM Cortex A8 took 3, 89, 702 300000 instructions, 140000 average instructions and
cycles to execute the code. the estimated cycles obtained are 370000.In case of the
optimised code for the ARM1176 it is observed 275000
6.1 Optimised Results instructions, 130000 average instructions and the
Figure 6 shows the optimised profiler window estimated cycles of obtained are 130000.300000
of RVDS for ARM Cortex A8. The numbers of instructions, 140000 average instructions and the
instructions are 252,201 average instructions per second estimated cycles of 355000.In comparison between all
110,087 and number of cycles estimated are ARM CortexA8 and ARM1176 As shown in the Figure
359,086.The process involve in the tool of RVDS of 5 7 and 8 the number of cycles taken to execute the code
threads and its consumes it’s on space of memory and on ARM CortexA8 is less compared with
execution time are reduced compared to profiled shown ARM1176.Hence it is said that the ARM Cortex A8
in Figure 5. processor is the better for MIMO-OFDM applications as
embedded platforms.
The first pie chart shown in Figure 6 shows the
code coverage by instruction, where green colour
indicates full usage of the code, yellow colour indicates
partially used and red indicates some of the instructions
not used at all. The profiler shows the code coverage by
instructions of full coverage by 70.39%, partial

SAS TECH Journal 60 Volume 12, Issue 2, September 2013

 Fig. 8 Bar Graph for the ARM1176 Performance

7. CONCLUSIONS
The following are the conclusions drawn in the
analysis and studies carried in this paper:

• MIMO-OFDM provides higher data rate

whereas MIMO with QAM reduces the
BER
• ARM Cortex A8 is a suitable platform to
develop MIMO-OFDM receiver sub-
system

Future work involves improving the overall

performance of the system and the sub-blocks can be
improved as follows:

Fig. 6 RVDS Profiling for ARM Cortex A8 with • Using the higher order FFT to increase
Optimisation the number of subcarriers, thereby
increasing the data rates
• Using the higher order QAMs to
improve the spectral efficiency and
improving the data rate by using proper
channel estimation
• The MIMO used in this project gives a
reasonable data rates, improves diversity
and reliability. Using the advanced
MIMO techniques gives a better
performance

Fig. 7 Bar Graph for the ARM CortexA8

Performance

SAS TECH Journal 61 Volume 12, Issue 2, September 2013