Volume 114 No. 12 2017, 419-427 ISSN: 1311-8080 (printed version); ISSN: 1314-3395 (on-line version) url: http://www.ijpam.eu ijpam.eu Analysis of Different Detection Techniques of MIMO In Future Generation of Wireless Communication I. Sridevi mounika 1, Dinesh sharma 2, Purnima k sharma 3 1 B.Tech. (Final Year), ECE Dept., VRSEC, Vijayawada (A.P.), India 2,3 Associate Professor, ECE Dept. VRSEC, Vijayawada (A.P.), India 1 srideviindugula16@gmail.com, 2 sharma82dinesh@gmail.com, 3 purnima.kadali@gmail.com Abstract The MIMO has multiple antennas at the transmitting end and also at the receiving end. It provides an improvement in performance in terms of data rate and reliability of the system. MIMO has a wide variety of applications in wireless communications. MIMO is being used in many new technologies which include Wi-Fi, HSPA+, LTE to provide increased link capacity and spectral efficiency. The adaption of Massive MIMO for future generation 5G technology increases data rates and provides quality service to the rapidly increasing users. This paper consists of comparative analysis of different decoding techniques used at the receiver end for the detection of transmitted signal in the MIMO system. The decoding techniques are Zero Forcing (ZF), Minimum Mean Square Error (MMSE), Maximum Likelihood (ML), Ordered successive Interference Cancellation (OSIC), Sphere Decoding (SD), Lattice Reduction. This comparison is based on Bit Error Rate (BER), Signal to Noise Ratio (SNR), capacity, computational complexity, and computational time. This analysis is used to choose the specific decoding technique according to application. Keywords: MIMO, Zero forcing, MMSE, OSIC, ML, Sphere decoding, SNR, BER, computational complexity, computational time. 1 Introduction Mobile and wireless networks are developing in terms of higher data rates generation obtaining more processing power, more memory on board, and higher battery life for same type of applications. 5G is the latest developing technology and it has the potential to improve the access of mobile broad bands in rural areas also. MIMO is very useful in military communications because it has special feature i.e. anti-jamming capability. MIMO is an antenna technology for wireless communication. In which multiple antennas are used at both the source (TX) and destination (RX). These antennas provided in the MIMO systems are used to minimise the error which are caused by the noise while transmission and increase the data rate. There are several forms in antenna technology. They are SISO (Single Input Single Output), MISO (Multiple Input Single Output), SIMO (Single Input Multiple Output), MIMO (Multiple Input Multiple Output). MIMO is one of the smart antenna technology. It has many advantages compare to the other antenna technology. The rate at which data is transferred through the system cannot be changed under the conditions of interference and signal fading. The main theme behind the development of MIMO is it covers higher data rate over longer distance [1]. Figure 1 shows the normal MIMO system with K no.of users. In MIMO technology spatial multiplexing is a transmission technique in which high data rate signal is divided into several number of lower data rate streams. These streams are transmitted from separate transmit antennas through the same frequency channel. Spatial multiplexing 419
can also be used for transmission of data to multiple receivers known as Space Division Multiple Access (SDMA) on multi user MIMO. Fig. 1. MIMO system 2 Decoding Techniques Decoding is the reverse process of encoding MIMO has different decoding techniques, which are linear and non-linear [2]. There are used to get the original message signal at the receiver end of the signal. In MIMO the receiver end users an algorithm to detect the original signal from the multiple received signal. MIMO has different types of decoding techniques. These different decoding algorithms are as follows. 2.1 Zero Forcing ZF is one type of linear decoding technique [2]. In this technique, only one user can activate at a time and all other users nullifies at that time. The nulling all other users is done by using the pseudo inverse matrix [4]. It is used to take decision about one user. The general pseudo inverse m x n matrix [3] is given by w = (H H H) 1 H H (1) To find the signal or detect the signal we must satisfy H=1. ZF has some advantages that are, it has minimum computational time and complexity [4], [5]. It can be suitable in small loads. ZF has some disadvantages that are it losses the order of diversity when number of users increase. So, it can t handle over loads [5]. 2.2 Minimum Mean Square Error MMSE is a decoding algorithm which has good performance than the ZF under noisy conditions [2]. MMSE considers noise power in filter calculations instead of nulling all users as in ZF. In this approach, it reduces the mean square error (MSE) [4]. MMSE is given by MMSE=E{(x x) 2 } (2) Where x is unknown random variable, y is known random variable, x is estimate variable. It does not completely reduce ISI like ZF, but it can reduce the total noise power at the receiver. The weight matrix is given by W MMSE = (H H H + σ 2 Z I) 1 (3) 420
x MMSE = W MMSE =(H H H + σ 2 Z I) 1 H H y (4) =x + (H H H + σ 2 Z I) 1 H H z =x + z MMSE (5) z MMSE =(H H H + σ 2 Z I) 1 H H z (6) The noise enhancement effects due to minimum singular value for the MMSE linear detector is given below Nt σ 2 2 z σ min 2 + σ min E { Z MMSE 2 2 }= i=1 for MMSE (7) (σ z 2 ) 2 2.3 Ordered Successive Interference Cancellation This algorithm identifies the strongest user first, this strongest signal is subtracted from the received signal and then next identify the next strongest signal etc.., It can work in two manners, firstly it can subtract the soft information from received signal, secondly it can subtract the hard information from received signal [7]. Due to this there is a nonexistent error propagation is possible. This is the nonlinear decoding technique. This method can be used to avoid the co-antenna interference [6]. To avoid this there is a need for ordering when we are using the SIC. This ordering is done in this way first it identifies and detect the signals with lower error probabilities. This ordering process is done in the following ways a. Interference cancellation: To detect the symbol from the i th transmitter, suppose that there is a interference from i-1 transmitter, to cancel this interference it subtracted from the original received signal. y = y h i y i h i 1 y i 1 (8) b. Interference nulling: This is usually done with some form of linear decoding techniques. 2.4 Maximum Likelihood Detection Maximum likelihood detection is also known as optimal receiver. It is used to detect the transmitted symbol vectors i.e a set of all possible transmit symbol vectors. This technique calculates the distance (Euclidean) between the vectors which are received and the product of all transmitted signal vectors in given channel [8]. Here channel is denoted the product of all transmitted signal vectors in given channel [8]. Here channel is denote with H. It is used to find the vector which has the minimum Euclidean distance. Estimated signal of the transmitted vector x is given by 2 xˆ ML =argmin y h x (9) In this method complexity increases with increasing modulation order i.e if the no of transmit antennas increases then the complexity will increase. So, this method has computational complexity but its performance is much better than previous methods. It is the main advantage of this method. 421
2.5 Sphere Decoding Sphere decoding technique is used to detect the transmitted signal vector with less ML metric that means it is used to find the ML solution vector [10]. In this method, it considers a small set of vectors within the given sphere. This method adjusts the radius of the sphere until it reaches a single vector within the sphere. If there are no vectors within the sphere, then it increases the radius of the sphere and if there are more number of vectors within the sphere then it decreases the radius of the sphere to get single vector. It is the best technique among all previous techniques because it provides better BER and spatial complexity [11]. h 12R x 1R x 2R [ y 1R y ] = [ h 11R h 11I h 12I ] [ 2R h 21R h 22R h 21I h x 22I 1I ] +[ x 2I h 12I x 1R x 2R z 1R z 2R ] (10) [ y 1I y ] = [ h 11I h 11R h 12R ] [ 2I h 21I h 22I h 21R h x 22R 1I ] + [ z ] (11) 2I x 2I The above two Equations (10) and (11) yields the following expression: y 1R y 2R [ y 1I ] = [ y 2I h 11R h 12R h 11I h 12I h 21R h 22R h 21I h 11I h 12I h 11R h 21I h 22I h 21R h 22I ] [ h 12R h 22R x 1R xsss 2R x 1I z 1I z 1R z 2R ] + [ z 1I ] (12) x 2I z 2I Sphere decoding exploits the following relation: arg min y H x 2 = arg min (x x ) T H TH (X) (13) Fig. 2. Radius of sphere [6] Where R is obtained from QR decomposition of the real channel matrix H=QR. Fig. 3. Sphere decoding technique 422
In this method calculation of radius is the most important task. Figures 2 and 3 shows that the selection of radius. This will have done at the pre-processing level. So, we must take radius properly [6]. If radius is too small, then the solution vector can t satisfy the constraints of sphere. If radius is too large, then processing cycle will become large. 2.6 Lattice Reduction In general, the linear detection and OSIC methods may raise the noise component in linear filtering, thereby decrease the performance. Such noise enhancement problem becomes critical, especially when the condition number of channel matrix increases [12]. Lattice reduction method can be useful for reducing the condition numbers of channel matrices The basis vector set that is shown in Figure 4(a) has a larger condition number than that shown in Figure 4(b). A basis vector set with a small condition number reduces the noise enhancement in the linear detection and OSIC methods. Lattice reduction method can be used for reducing the no of channel matrices. A basic vector set with a small condition number [13] may largely reduce the noise in the linear detection and OSIC methods. (a) (b) Fig. 4 (a) A basis vector set with a large condition number (b) An orthogonal basis vector set There is no noise in the process of linear filtering when these basic vectors are orthogonal. The system equation can be represented as y Hx z QRx z (14) ~ ~ y QHy Rx ~ z (15) Where ~ z QHz This lattice detection method is simpler method with triangle matrix form compared to QR method. This conditioning number can be reduced by using LLL Algorithm. Lenstra-Lenstra-Lovasz (LLL) algorithm is used to reduce condition number of a triangular matrix. This method increases the sub-optimal detectors in MIMO point-topoint systems. This algorithm provides an output matrix that consists of transformations to be done over the initial channel matrix. In some applications, pre-coding that is the inverse of the transformation is also required. In this the important thing is to build an equivalent system equation which is better conditioned than Equation. whose condition 423
number depends on R. In general, it takes (N-1) steps in this algorithm for lattice reduction with an N N matrix. The resultant system equation is given by y = Hx + z = QRx + z = Q LLL R LLL T 1 LLLx + z (16) 3 Simulation and Results The comparative analysis of ZF, MMSE, OSIC, ML, Sphere decoding, Lattice reduction techniques are done by using MATLAB R2014a, and modulation technique is QAM. QAM is a combination of two amplitude modulated signals into a single channel, there by double the band width. Two carrier signals are used in QAM. Each having the same frequency but they are differing in a phase by 90 degrees. Fig. 5. BER Vs SNR for Nt=Nr=2 Fig. 6. BER Vs SNR for Nt=Nr=4 Fig. 7. Capacity Vs SNR for Nt=Nr=2 424
Fig. 8. Capacity Vs SNR for Nt=Nr=4 Capacity, BER/SNR graphs of different decoding techniques are shown in fig (5-8). These are used for comparative analysis of ZF, MMSE, OSIC, ML, Sphere decoding, Lattice reduction. From the graphs ZF has the less complexity and computational time. ZF and MMSE has similar values. Computational time is more in SIC method. Complexity is more in ML method. For higher values of SNR sphere and lattice reduction methods has very less BER. Lattice reduction method has less computational time. Finally, it is the best and has improved performance among all techniques. 4 Conclusion This paper represents the comparative analysis of the different decoding techniques of MIMO technology. Zero forcing is the simplest technique among all decoding techniques. It has minimum complexity and computational time, but signal losses will be increased with increasing number of users. MMSE is advantageous method than ZF but it has the same order of diversity as ZF. ML is the best technique among the previous technique but computational complexity is more than ZF. Sphere Decoding has the better performance than previous techniques in terms of BER and SNR. By this analysis, we conclude that lattice reduction has the better performance among all decoding techniques and it is best technique in terms of computational time, complexity, capacity, SNR, BER parameters. References 1. G. J. Foschini and M. J. Gans, On limits of wireless communications in a fading environment when using multiple antennas, Wireless Personal Communication., vol. 6, no. 3, pp. 311-335, Mar. 1998. 2. Er. Navjot Singh1, Er. Vinod Kumar Linear and Non-Linear decodingtechniques in Multi-User MIMO Systems A Review Vol. 4, Issue 5, May2015. 3. Xiantao Sun, Leonard J. Cimini, Larry J. Greenstein, Douglas S. Chant, and Jan Kruys Coordinated zero-forcing beam forming in multipoint MIMO networks for backhaul applications, University of Delaware, Cisco Systemsxsun@udel.edu 4. Shivendra singh, sumith raghuvanshi, comparative analysis of various optimization techniques with coded OFDM-MIMO transmission 2011 International Conference on Computational Intelligence and Communication Systems 5. J. G. Proakis, Digital Communications, 3rd ed. New York, McGraw-Hill, 1995. 425
6. MIMO-OFDM Wireless Communications with MATLAB_ Yong Soo Cho, Jaekwon Kim, Won Young Yang and Chung G. Kang _ 2010 John Wiley & Sons (Asia) Pte Ltd. 7. J. Wang and S. Li. Soft versus hard interference cancellation in MMSEOSIC MIMO detector: a comparative study, in The Fourth International Symposium on Wireless Communication Systems, Norway, 2007. 8. Liu Peng Department of Electronic Information and Automation, Tian Jin, University of Science and Technology Tianjin, The Maximum Likelihood detection for MIMO Communications Systems China diarylp@sina.com. 978-1-4244-7161-4/2010 IEEE. 9. Shinobunagayama and Takeshi hattori Department of Electrical Engineering Sophia University Tokyo, A Proposal of QRM-MLD for Reduced Complexity of MLD to detect MIMO signals in Fading Environment Japan shinobu@mmc.ee.sophia.ac.jp, thattori@mmc.ee.sophia.ac.jp 1-4244-0063-5/06/ 2006 IEEE. 10. InversionYi Wang and Harry Leib Sphere Decoding for MIMO Systems withnewton Iterative Matrix 1089-7798/13$31.00 _c 2013 IEEE. 11. H. Yao and G. W. Wornell, Lattice-reduction-aided detectors for MIMO communication systems, in Proc. IEEE Global Tele. Conf., Nov. 2002,pp. 424 428. 12. Purnima K Sharma, Dinesh Sharma, R.K.Singh Evolution of Mobile Wireless Communication Networks (0G-8G) International Journal of Applied Engineering Research ISSN 0973-4562 Volume 10, Number 6 (2015) pp. 14765-14778 Research India Publications http://www.ripublication.com 13. Purnima K. Sharma and Dinesh Sharma Mobile Wireless Technology is Boon or Curse Indian Journal of Science and Technology, Vol 9(40), DOI: 10.17485/ijst/2016/v9i40/84293, October 2016 Authors Ms. I. Sri Devi Mounika, Mr. Sk. Mansoor Ali, Mr. K. Nagalingeswara Rao, Ms. U. Venkata Lakshmi, Ms. M. Yogitha Prabandha have completed our B. Tech degree in Electronics and Communication Engineering from Velagapudi Ramakrishna Siddhartha Engineering college, Vijayawada, Andhra Pradesh. Our team consists of 5 members and we have done a project on Analysis of Different Detection Techniques of MIMO In Future Generation of Wireless Communication. Dr. Dinesh Sharma was born on 5th Dec 1982 in Narnaul District Mohindergarh of Haryana (India). He received his Ph.D. from UTU, Dehradun (India) in 2013 and M.Tech. Degree in Communication Engineering from Shobhit University, Meerut, India in 2009. He is an Associate Member of the IETE. He has published several Research papers in national and international journals/conferences. His present research interest is in Signal Processing and Wireless Communication. Mrs Purnima K Sharma was born on 2nd June 1983 in Eluru, Andhra Pradesh (India). She received her M.Tech. Degree in Communication and Signal Processing Engineering from Nagarjuna University (A.P.), India. She is an Associate Member of the IETE. She has published several Research papers in national and international journals/conferences. She received her Ph.D. from UTU, Dehradun (India) in 2015. 426
427
428