5th International Electrical Engineering Congress, Pattaya, Thailand, 8-10 March 2017

Detection of Data Symbol in a Massive MIMO

Systems for 5G Wireless Communication
Sarun Duangsuwan1 and Punyawi Jamjareegulgarn2
Department of Electronic Engineering
King Mongkut’s Institute of Technology Ladkrabang,
Prince of Chomphon Campas, Pathio, Chomphon, 86160
Email: [email protected] and [email protected]

Abstract— The massive MIMO technique has played the most vast array of sensors, and other devices, leading to better
important role in 5G wireless communication. It is anticipated health, economic gains, and other advantages. 5G addresses not
that the new techniques employed in massive MIMO will not only only IoT deployments on a massive scale, but also applications
improve peak service data rates significantly, but also enhance previously not possible that depend on ultra-reliable and low-
capacity, coverage, low-latency, efficiency flexibility, latency communications. Although a far more fragmented
compatibility and convergence, thus meeting the focusing market than smartphones, the benefits will be so great that the
demands imposed by optimal detection. This paper presents the realization of IoT on a massive scale is inevitable. The only
optimal detection of data symbol in massive MIMO for 5G question is how, exactly, the market will evolve [9]-[10].
wireless communication. Based on the frequency non-selective
fading MIMO channel, we consider three difference detectors for 5G research and development accelerates, in early stages of
recovering the transmitted data symbols and evaluate their definition through global efforts and many proposed technical
performance for Rayleigh fading and additive white Gaussian approaches, could start to be deployed close to 2020 and
noise (AWGN). At the results, we show that the probability of continue through 2030. The trends of development are focusing
error rate (PER) performance of the detectors are significantly on propagation limitation [11]-[12], massive MIMO antennas
discussed. [13]-[14], beam-steering [15]-[16], beam tracking, dual
connectivity, carrier aggregation, small-cell architectures.
Keywords— massive MIMO; optimal detector; Rayleigh fading
These have been on research work increasingly, in particular,
channel; PER; 5G wireless communiaction
the massive MIMO provides the essential roles to enhance the
channel capacity and spectral efficiency. To understand what
the massive MIMO is, the massive MIMO or large scale
I. INTRODUCTION antenna system is a form of multiuser MIMO in which the
number of antennas at the base station is much larger than the
The standard mobile technology is in the trends of digital number of devices per signal resource. The massive MIMO
transformation, a challenging new capabilities that will benefit systems, which are equipped with tens or even hundreds of
research and development as a whole. The step from 4G to antennas. Firstly, energy efficiency can be significantly
4.5G was being lunched [1]-[2], and the industry is toward to, increased by massive MIMO system as they concentrate power
initially in the advances LTE and then in 5G, will be even on a shape direction. Secondly, system throughput can be
greater [3]. Standards that have been planned; that process gained by utilizing multi-user MIMO (MU-MIMO). Lastly,
expected will be in 2020 timeframe [4]. However, almost of massive MIMO systems are more robust than conventional
operators have demonstrated many of 5G capabilities, and also MIMO system as they offer excessive degree of freedom.
showed pre-standard networks for fixed applications as early as Although the massive MIMO can achieve orders of increase in
2017 [5]. The 5G is not to replace LTE, but in most spectral efficiency, but one of which is the practical signal
deployments will co-exist with it high speed technologies detection algorithm in the uplink due to the increased multi-
tightly integrated in a manner transparent to users [6]-[7]. user interference (MUI).
Many of the capabilities that will make 5G so effective are
appearing in advanced forms of LTE. With carrier aggregation In this paper, we discuss the performance evaluation of the
(CA) [8], for example, operators have not only harnessed the existing detection methods of data symbol in the massive
potential of their spectrum holdings to augment capacity and MIMO system. Consider a transmission channel and system
performance, but the technology is also the foundation for model that employs the multiple transmitting antennas and the
entirely new capabilities, such as operating LTE in unlicensed receiving antennas, as shown in Fig. 1. We assume that the
bands. With long-term evolution (LTE) growth in smartphones detector knows the elements of the channel matrix H perfectly.
and usage limited by population, innovators are turning their At the detector, we employ an optimal detector with the
attention to the Internet of Things (IoT), which promises maximum likelihood detector (MLD), the minimum mean
billions of new wireless connections. Enhancements to LTE square error (MMSE), and inverse channel detector (ICD) is
followed by 5G capabilities will connect wearable computers, a discussed. Simulation result show that the probability of error

978-1-5090-4666-9/17/$31.00 ©2017 IEEE

5th International Electrical Engineering Congress, Pattaya, Thailand, 8-10 March 2017

Input data M Output data

N +
Receiving Detector
Transmitting

Fig. 1 An uplink communication system with the massive MIMO antennas for 5G wireless communication.

rate performance of the three detector in a Rayleigh fading

channel is most easily verified by using Monte Carlo computer
simulation. Furthermore, the simulation results are also ªh11 (t ), h12 (t ),  h1N (t ) º
examined the number of users. «h (t ), h (t ),  h (t ) »
The rest of the paper is organized as follows. Section II H « 21 22 2N » (4)
briefly introduces the system model. Section III specifies the «     »
proposed detector. The simulation results of the probability of « »
error (PER) simulation results are shown in Section IV. ¬hM 1 (t ), hM 2 (t ), hMN (t ) ¼
Finally, conclusion are drawn in Section V.

II. MASSIVE MIMO CHANNEL MODEL In this case, the signal received in substitution (2) can be
rewritten as
We briefly introduce the uplink massive MIMO system
N

¦ h (t ) s
employing N number of transmitting antennas at the mobile K
terminal (MT) to simultaneously serve K user, where M ri (t ) ij j (t ) (5)
j 1
number of the receiving antennas at the base station. In the
massive MIMO, a randomly time-varying channel is and, in matrix from, the received signal vector as
represented by the M u N matrix H (W , t ) , defined as
r (t ) Hs(t ) (6)
ªh11 (W , t ), h12 (W , t ),  h1N (W , t ) º where H is constant within the time interval 0dt dT .
«h (W , t ), h (W , t ),  h (W , t ) »
H(W , t ) « 21 22 2N » (1) In this paper, we employ the non-frequency selective fading
«     » channel model to analyze the performance characteristic of
« » massive MIMO detector system. Thus, the mathematical
¬hM 1 (W , t ), hM 2 (W , t ), hMN (W , t ) ¼ convenience at the detector can be obtained as
Suppose that the signal transmitted from K user, which has M

N transmitting antennas is s N (t ) . Then the signal received at ym (t ) ¦ r (t )  n (t )

i 1
i i (7)
the receiving antennas can be expressed as
and, thus
N f
ri (t ) ¦ ³ h (W , t ) s
j 1 f
ij
K
j (t  W )dW (2) y Hs(t )  n (8)

n [n1 (t ), n2 (t ),, nM (t )] (9)

where i 1, 2, , M . Thus, the matrix notation is
where n denotes as the AWGN channel matrix.
r (t ) H(W , t ) s(W ) (3)
III. DETECTION OF DATA SYMBOL IN MASSIVE MIMO IN 5G
where the asterisk denotes convolution, and s(W ) is an N u 1 WIRELSS COMMUNICATION
vector and r (t ) is an M u 1 vector.
Based on the frequency non-selective fading massive
As the non-frequency selective fading channel, the channel MIMO channel model described in Section II, we study with
matrix is rewritten as three difference detector methods for detecting the transmitted
data symbols and evaluate their performance for Rayleigh
fading and AWGN. Throughout this section, we assume that
5th International Electrical Engineering Congress, Pattaya, Thailand, 8-10 March 2017

the study of the detector well knows the elements of the sˆ ICD H 1 y s  H 1n (14)
channel matrix H perfectly.
Each element of the estimate ŝ is then quantized to the
A. Approximated Maximum Likelihood Detector (MLD) closest transmitted symbol value. We note that the ICD
The MLD optimum detector in the sense that it mitigates estimate ŝ is not corrupted by ISI. However, this also implies
the probability of error. Due to the additive noise in terms of that the ICD does not exploit the signal diversity inherent in the
the receiving are statistically independent and identically received signal, we will observe as
distributed (i.i.d)., and also zero-mean Gaussian, hence the
joint conditional probability density function (PDF) is a normal 1
distribution. Moreover, we refer that MLD detection of the WH HH H HH (15)
symbols based on the minimizes the Euclidean distance metric
can be obtained as IV. SIMULATION RESULT
M N
2 The PER of the three detectors in a Rayleigh fading channel
ŝ MLD ¦ y ¦h s
i 1
i
j 1
ij
K
j (10)
can be evaluated by using Monte Carlo computer simulation.
The simulation results of PER performance compared with the
average signal-to-noise ratio (SNR). In all simulation, we
consider the modulation scheme of 64-QAM, and the symbol
B. Approximated Minimum Mean Square Error Detector rate 1/2 in standard convolutional code. At the receiver, the
(MMSE) correlation based on MLD, MMSE and ICD detects signals of
The MMSE detector is a linear sub-optimum detection interest. While we provide the massive MIMO elements
^
which combines the received signal yi ,1 d i d M to form ` between the transmitting antenna and the receiving antenna up
to one hundred and twenty, with the number of user as 10 or 20
an estimate of the transmitted symbols s j , 1 d j d N ^ K
` with simultaneously. The parameters setup is shown in the Table 1.
K user. The linear combining is represented in matrix form as
TABLE1. PARAMETERS SETUP
sˆ MMSE WH y (11) Parameters Values

where W is an M u N weighting matrix, which is selected Modulation Type 64-QAM

to minimize the mean square error (MSE) N Transmitting and M receiving antennas 100 and 120 elements

> @
Number of users, K 10 and 20
Eª s  WH y º
2 2
wW E e (12) Number of symbol rate and code rate 1 Gbps and 1/2
«¬ »¼

E>@
The performance results in Fig. 2 and Fig. 3 illustrate that
where ˜ denotes the expectation operator, the minimization the MLD detector exploits the optimum receiver, its
of w W leads to the solution for the optimum weight vector performance is achievable in massive MIMO which both
N=100, M=100 and N=120, M=120 respectively. The proposed
w1 , w2 ,  , wN as linear MMSE and ICD detectors, achieve an error rate that the
decreases inversely as the average SNR raised, we note that the
wj R yy1rs K , y , j 1, 2,  , N (13) MMSE detector outperforms the ICD detector is about 2 dB,
j although both achieve the same two study cases of the number
where R yy >
E yy H @ HR ss H H  N 0 I is the
transmitting and receiving massive antennas.
Fig. 2 shows the PER performance between the theoretical
M u M auto-correlation matrix of the received signal vector method in massive MIMO, where N=100, M=100 and the
H
> @ >
y , R ss E ss , E nn
H
@
N 0 I . When the signal vector detection of data symbols by using MLD, MMSE, and ICD
detector. As the results of K=10 users, the theoretical massive
has uncorrelated, zero-mean components, R ss is a diagonal MIMO can achieve the PER performance is lower 10-7 as the
average SNR about 40 dB. In addition, the data symbols which
matrix. Each component of the estimate ŝ MMSE is quantized to are detected by using ICD detector, MMSE detector, and MLD
the closet transmitted symbol value. detector has the PER performance better respectively. In the
adjustable to K=20 users, we note that the PER performance
C. Approximated Inverse Channel Detector (ICD) increases more than 10-6, herein because the number of
The ICD also forms an estimate of s by linearly combining symbols and code rates might be corrupted.
^ `
the received signal yi ,1 d i d M . However, in this case, we In the Fig. 3, we can see that the theoretical massive MIMO
set M N and select the weighting matrix W so that the still belong to the theorem of spectral efficiency when has more
inter-symbol interference (ISI) is completely eliminated; that number of antennas, the channel capacity is perfectly detected.
is, W
H
H 1 and, the s detector as
5th International Electrical Engineering Congress, Pattaya, Thailand, 8-10 March 2017

0
performance results have been discussed. By the simulation
10
Theoretical method N=100, M=100 results show that the 5G technology, which has a data symbol
ICD rate up to 1 or 10 Gbps in the future, the PER performance
MMSE have to explore under 10-8 in practice theoretically. Therefore,
10
-2 MLD we claim that our simulation results are accomplished by using
Probability of error (PER)

MLD, MMSE, and ICD detector when have more order the
K=20 number of transmitting and receiving antennas than one
10
-4 hundred and twenty as the massive MIMO system.

REFERENCES
10
-6 [1] L. H. Trinh, F. Ferrero, R. Staraj and J. M. Ribero, “Mobile phone
K=10 antenna for 2G, 3G and 4G standards,” Intenational Conference on
Anvanced Technologies for Communications (ATC), Oct. 2013, pp.
195-203.
-8
10 [2] S. Sudha, N. Sasirekha, “Design of smart antenna array of 2x2 MIMO
0 5 10 15 20 25 30 35 40
Average SNR (dB)
for 4G-LTE,” in International Conference on Wireless Communication,
Signal Processing and Networking, Mar. 2016, pp.1611-1615.
[3] N. Khalid and O. B. Akan, “Experimental throughput analysis of low-
Fig. 2. Performance of MLD, MMSE and ICD detectors with N=100, M=100 THz MIMO communication channel in 5G wireless network,” IEEE
Wireless Communication Letters, vol. 5, pp. 616-619, Sep. 2016.
0 [4] J. Hansryd, “5G wireless communication beyond 2020,” in European
10
Theoretical method N=120, M=120 Solid-State Circuits Conference (ESSCIRC), Sept. 2015, pp. 1-3.
ICD [5] T. O. Olwal, K. Djouani and A. M. Kurien “A survey of resource
-2 MMSE management toward 5G radio access networks,” in IEEE
10 MLD Communications Surveys & Tutorials, April 2016, pp.1656-1686.
Probability of error (PER)

[6] E. Ayanoglu, “5G today: Modulation technique alternatives,” in

-4 International Conference on Computing, Networking and
10 Communications (ICNC), Feb. 2016 pp. 1-5.
K=20 [7] B. Naeem, R. Ngah and S. M. Hashim, “Understanding the challenges
-6
towards implementation of the 5G wireless networks,” in International
10 Symposium on Technology Management and Emerging Technology
(ISTMET), Aug. 2015, pp. 223-226.
-8 K=10
[8] P. Diamantoulakis, K. Pappi, S. Muhaidat, G. Karagiannidis and T.
10 Khattab, “Carrier aggregation for cooperative cognitive radio
networks,” IEEE Trans. on Vehicular Technology, vol. PP, pp. 1-1, Dec.
2016.
-10
10 [9] G. P. Fettweis, “5G and the future of IoT,” in European Solid-State
0 5 10 15 20 25 30 35 40
Average SNR (dB) Circuits Conference (ESSCIRC), Sept. 2016, pp. 21-24.
[10] J. Marcos and C. Brito, “Trends in wireless communications towards 5G
networks-The influence of e-health and IoT applications,” in
Fig. 3. Performance of MLD, MMSE and ICD detectors with N=120, M=120 International Multidisciplinary Conference on Computer and Energy
Science (SpliTech), July 2016, pp. 1-7.
In other words, the PER performance results are acceptable [11] Md. M. Ahamed and S. Faruque “Propagation factors affecting the
significantly. Observing that the PER result is lower 10-8 as the performance in 5G millimeter wave radio channel,” in IEEE
average SNR at 40 dB. We also focus on the proposed International Conference on Electro Information Technology (EIT), May
detectors that the MLD detector is still to the optimum receiver, 2016, pp. 728-733.
the linear detector both MMSE and ICD detector can also [12] G. R. MacCartney, J. Zhang, S. Nie and T. S. Rappaport “Path loss
achieve the PER performance result, irrespective whether of models for 5G millimeter wave propagation channels in urban,” in IEEE
Global Communications Conference (GLOBECOM), Dec. 2013, pp.
increase K=20 users. 3948-3953.
It is interesting to analyze the computational probability of [13] E. Yaacoub, M. Husseini, and H. Ghaziri, “An overview of research
error of the three detectors in massive MIMO. We note that the topics and challenges for 5G massive MIMO antennas,” in IEEE Middle
East Conference on Antennas and Propagation (MECAP), Sept. 2016,
complexity of data detection by using MLD detector depends pp. 1-4.
on growing up the number of symbols in the signal [14] V. Jungnickel, K. Manolakis, W. Zirwas, B. Panzner, V. Braun, M.
constellation, whereas the MMSE and ICD detector has a Lossow, M. Sternad, R. Apelfrojd and T. Svensson “The role of small
complexity with the channel matrix of the number of antennas. cells, coordinated multipoint, and massive MIMO in 5G,” IEEE
Communication Magazine, vol. 52, pp. 44-51, May 2014.
[15] Y. Wang, H. Wang and G. Yang “Design of dipole beam-steering
CONCULSION antenna array for 5G handset applications,” in Progress in
In this paper, by successfully exploiting the massive MIMO Electromagnetic Research Symposium (PIERS), Aug. 2016, pp. 2450-
2453.
in the trend of 5G wireless communication, we have studied
the detection of data symbols based on an existing signal [16] Z. Cao, Q. Ma, A. B. Smolders, Y. Jiao, M. J. Wale, C. W. Oh, H. Wu,
A. Marcellus and J. Koonen “Advanced integration tecgniques on
processing method. The MLD detector is the optimal receiver, broadband millimeter-wave beam steering for 5G wireless networks and
herein, the comparison of three different of the PER beyond,” IEEE Journal of Quantum Electro., vol. 52, pp. PP, Dec. 2016.