Wireless Detections in MIMO System

Transcription

1 Wireless Detections in MIMO System Sudarshan Adhikari Masters in computer Engineering (MECE), Nepal College of Information Technology (NCIT) Balkumari, Lalitpur Dr. Sanjeeb Prasad Panday Asst. Professor, Department of Electronics and Computer Engineering Institute of Engineering, Pulchowk Campus,TU Abstract Multiple-Input Multiple-Output (MIMO) communication technology has received a significant attention in the area of wireless communication systems. MIMO provide reliable transmission and increasing data rate than traditional single input single output system. Increasing the number of transmit or receive antennas increases channel throughput. In this paper linear and non-linear detection techniques are analyzed. The linear detection techniques such as Zero Forcing (ZF) Minimum Mean Square Error (MMSE) are tested and results are analyzed. In linear MIMO wireless detection schemes, pre-coding at the transmitter is provided. It is observed that this detection scheme is sensitive to the impact of spatial correlation caused by mutual coupling between antenna elements and the spatial propagation conditions of the wireless channel. Also, the non-linear MIMO wireless detection techniques which include Maximum Likelihood (ML), Minimum Mean Square Error-Successive Interference Cancellation (MMSE-SIC) are tested and results are analyzed. The results depicts how to apply non-linear detection, which can help mitigating the impact of correlated signals, in a closed-loop MIMO system and compare the average rates achieved with different detection schemes. Keywords: MIMO, Wireless detections, MMSE, Zero- Forcing, Maximum Likelihood, MMSE-SIC, Wireless communication. I. Introduction Wireless communication system uses antennas on transmitter and receiver side for transmission of radio waves. The frequency spectrum is the scarce resource for wireless communication systems and the rapid increase of wireless applications has demanded the new techniques to achieve higher spectral efficiency. The availability of spectrum has been an influential factor impeding a rapid and efficient convergence of technologies. The multiple-input multiple-output (MIMO) system utilizes the spatial diversity to increase the data rate and spectral efficiency. MIMO is considered to be the potential solution in combating the challenges of bandwidth constraints and high data rate demands. MIMO systems are regarded as one of the most promising research areas of wireless communications. This is due to the fact that a MIMO channel can offer a significant capacity gain over a traditional Single Input Single Output (SISO) channel. This provides a fundamental limit on data throughput in MIMO systems. MIMO systems improve data transmission reliability without increasing transmits power and bandwidth. The spatial diversity obtained from transmit and receive antennas can be combined with channel coding. This combined process leads to space-time coding in a coded system [1]. II. Related Works MIMO Detectors Figure 1: MIMO System Model MIMO detectors consist of different linear and non-linear detections techniques which include: a) Zero Forcing (ZF) Zero Forcing (ZF) is a linear detection method which treats all the transmitted signals as interferences except for the desired stream from the target transmit antenna. Therefore, interference signals from other transmit antennas are minimized or nullified in the course of detecting the desired signal from the target transmit antenna. ZF receiver works 1

2 best with high SNR level. Zero forcing method is based on the calculation of pseudo inverse of channel matrix H i.e. facilitate the detection of desired signals from each antenna, the effect of the channel is inverted by a weight matrix W [3] [8]. In simple form, Y= HX+N..(1) To solve the X the weight matrix W is to solved out which satisfies WH=I. The Zero Forcing (ZF) linear detector for meeting this constraint is given by, W = (H H H) -1 H H..(2) Where,H H is the Hermitian transpose. b) MMSE Signal Detection Minimum Mean Square Error (MMSE) is linear detection method in which mean squared error (MSE) is minimized between the transmitted signals. MMSE equalizer does not usually eliminate ISI completely but, minimizes the total power of the noise and ISI components in the output [3]. In order to obtain the unknown transmitted signal MMSE use a weight matrix given by W = (H H H + N 0 I) -1 H H..(3) where I is the identity matrix. In the classical approach, the Minimum Variance Unbiased Estimator (MVUE) is derived by first considering minimization of the mean square error i.e., θ1 = arg minθ mse(θ) where: mse(θ1)=e[(θ1-θ) 2 ] = ʃ (θ1- θ) 2 p(x; θ) dx..(4) and p(x; θ) is the PDF of x parameterized by θ. c) MMSE-SIC Signal Detection Minimum Mean Square Error Successive Interference Cancellation (MMSE-SIC) is signal detection method where weight matrix is used same as linear detector MMSE. The receiver can obtained an estimate of the two transmitted symbol x1, x2 as: xˆ1 xˆ2 =(H H H+N O I) -1 H H..(5) This method is similar to that of ZF-SIC except the weight matrix used to identify the unknown transmitted symbols. In classical SIC an approach, the receiver arbitrarily takes one of the estimated symbols and subtracts its effect from the received symbols. Instead of choosing random estimated symbols, another intelligent way is optimal ordering in which received power at the transmitted symbols are determined [3] [8]. d) ML Signal Detection Maximum Likelihood (ML) is optimal decoding method that compares between the received signal and possible transmitted signal. As the transmitted signal is modified by channel matrix, we need to estimate transmit symbol by using maximum likelihood detection algorithm. xˆ = argx k {x1, x2..x N } min r-h xk 2.(6) Where r-h xk 2 is ML metric which achieve maximum performance when transmitted vectors are equally likely [1],[4].r be the received signal and h be the channel matrix. Since the receiver has to consider M ntx possible symbols so it has complexity issues as the number of transmitter increases. Here M defines the modulation constellation and ntx defines the no of transmitted antenna system. For example 2 2 MIMO system and QPSK system, the total possible symbol is 16. II. Related Work Several research works have already been done and many research papers have been published regarding detection algorithms. Since, each of the papers has focused on different detection techniques being implemented in MIMO with their resulted output in simulation tools as well. However, the comparative analysis is very rare and proposed research is crucial in today s time in order get the de-facto standard for efficient detection techniques implementation. Some of the related works that are closely related to proposed work are highlighted below along with their scope of research. The research work done by Xiaoqing Peng, Weimin Wu, Jun Sun, and Yingzhuang Liu [1] put forward about the detection of symbols via compressive sensing algorithms, to reduce the original MIMO system to a new one, whose input dimension is much less than the output dimension. Also, the research done by Max Scharrenbroich, Michael Zatman and et.al [2] explain about non-linear detection and requires a prior knowledge of the target SNR and it depicts more result for the stationary case only. Similarly, the research completed by Gurpreet Singh, Rahul Vij and Priyanka Mishra[3] put an approach for spatial multiplexing technique with various decoding techniques got more optimal result specifically for 1x4 antennas. Besides, the result obtained by Shreedhar. A. Joshi Dr. Rukmini T S, Dr. Mahesh H M[4] on V-BLAST technique with MIMO exhibits better bit error rates. 2

3 III. Methodology This chapter deals with the explanation of observed system model. Here, the binary data generator generates the random binary data with equal probability of being 1 and 0. The binary data are created as they were easy to analyze, and they are matched to the real digital communication world. The encoder encodes the data using the Convolution Encoder. If the encoder takes k input bit streams (that is, it can receive 2 k possible input symbols), the block input vector length is L*k for some positive integer L. To solve the problem, systematic steps are needed which is better to explain in the flow chart form: Start Input Data Streams Convolutional Encoder QPSK Modulation Rayleigh+ Rician Channel+ AWGN Noise Define SNR Transmitter [2x2] Receive Transmitted Data Demodulate Decoder Part Detection Techniques: ZF, CMMSE, MMSE, ML, MMSE-SIC, ZF-SIC Plot BER Vs SNR Figure 2: Flowchart of the system Similarly, if the encoder produces n output bit streams (that is, it can produce 2 n possible output symbols), the block output vector length is L*n. Additive White Gaussian Noise (AWGN) is present in every electronic system. It cannot be removed by any means. It has to be modeled in an electronic system if the system is to be made noise tolerant. AWGN is modeled in this dissertation. The randomly generated AWGN is added in the sample points after the modulation is performed in the randomly generated data sample points. The mean value of the noise is 0 where as the variance is 1. The decoded bits from the different decoding techniques at the decoder are compared with the originally generated random data bits. The difference gives the number of errors encountered by the receiver. Bit error rate defines how good the communication receiver is. IV. Simulation Results and Discussions To verify the proposed method, MATLAB simulation is done. Parameters which are necessary for the simulation are taken from the standards in the MIMO communication system. The simulation results are obtained, after the MATLAB programming by using the parameters as shown below: Table 1: Simulation Parameters Simulation Parameters Values No. of bits 10 6 Encoder Convolutional Encoder MIMO configuration 2x2 Modulation type QPSK No. of transmitter/ 2 receiver Channel Rayleigh, Rician fading channels Noise type AWGN Detection Scheme ZF, MMSE,ML and MMSE- SIC The fig. 3 shows that the capacity of channel increases as the value of SNR increases i.e. lower most line has lowest capacity compared to other. Although the capacity increases as the value of SNR increases (here from 2 to 20), it is least compared to MIMO system which has 2x2, 3x3 or 4x4 antennas. SNR value 18dB corresponds to the capacity of 6 bit/sec/hz for SISO, but for the same SNR value the capacity for antennas are 10.5, 15.5 and 21 respectively. It can be analyzed that for the fixed SNR the capacity of system increases as the number of antennas increases. From this it can be said that any system can increase its channel capacity by increasing the number of antennas at both transmitter and receiver side without increasing SNR value. 3

4 Capacity (bit/s/hz) MIMO Capacity According to Number of Antenna Shannon Capacity(NT=NR=1) MIMO, NT=NR=2 MIMO, NT=NR=3 MIMO, NT=NR=4 log10(ber) BER for QPSK modulation with 2x2 MIMO and ZF equalizer(rician Channel) ZF 2 TX 2 RX SNR(dB) Figure 3: SNR vs. Capacity plot for different transmit and receive configurations SNR in db Figure 5: SNR vs. BER plot for 2x2 MIMO case with ZF equalizer in Rician Channel BER for QPSK modulation with 2x2 MIMO and ZF equalizer (Rayleigh channel) theory (ntx=2,nrx=2, ZF) sim (ntx=2, nrx=2, ZF) BER for QPSK modulation with 2x2 MIMO and MMSE equalizer (Rayleigh channel) theory (ntx=2,nrx=2, ZF) theory (ntx=1,nrx=2, MMSE) sim (ntx=2, nrx=2, MMSE) Average Eb/No,dB Figure 4: BER vs. SNR plot of the ZF receiving technique over the Rayleigh Channel ZF detection algorithm for MIMO is the most simple and basic algorithm, and the basic idea of ZF algorithm is kept of MIMO-channel interference by multiplying received signal and the inverse matrix of channel matrix. Here, over Rayleigh channel for BER, there exist 12 db of SNR value of ZF detection technique as in fig. 4. The fig. 5 depicts the MIMO case of 2x2 in the Rician fading channel using ZF as a linear detecting technique. The simulated results with a 2 2 MIMO system using QPSK modulation in Rayleigh/Rician fading channel shows matching results as obtained in for a 1 1 system for same modulation in Rayleigh/Rician channel. The ZF equalizer helps us to achieve the data rate gain, but not take advantage of diversity gain (as there is two receiving antennas). It might not be able to achieve the two fold data rate improvement here. A MMSE estimator is a method in which it minimizes the mean square error (MSE), which is a universal measure of estimator quality. The most important characteristic of MMSE equalizer is that it does not usually eliminate ISI totally but instead it minimizes the total power of the noise and ISI components in the output. BER Average Eb/No,dB Figure 6: BER Vs. SNR plot of MMSE equalizer for Rayleigh Channel MMSE 2 TX 2 RX SNR in db Figure 7: BER Vs SNR in case of MMSE equalizer for 2x2 MIMO case in Rician fading 4

5 The fig. 8 is the case of 2x2 MIMO, which uses the MMSE equalizer at the receiver. This depicts the analysis of various cases, performance, and comparison of the simulated and theoretical bit error rate of the MMSE with that of the ZF. linearly with the number of antennas and link range without additional bandwidth and power requirements. MMSE-SIC Rayleigh for 2*2 MIMO with QPSK Modulation 2TX 2RX Bit error probability curve for QPSK with ZF and MMSE equalizer for 2x2 Case ZF MMSE BER Eb/No, db Eb/No (db) Figure 10: BER vs. SNR plot for MMSE-SIC Rayleigh of 2X2 MIMO with QPSK Modulation Figure 8: Comparative analysis of ZF and MMSE over 2x2 MIMO case BER for QPSK modulation with 2x2 MIMO and ML equalizer (Rayleigh channel) theory (ntx=1,nrx=1) sim (ntx=2, nrx=2, ML) Average Eb/No,dB Figure 9: BER for QPSK modulation with 2x2 MIMO and ML equalizer for Rayleigh Channel This shows that the performance of the MMSE is highly impressive compared to that of the ZF. Equalization techniques can combat for ISI even in mobile fading channel with high efficiency. MMSE equalizer uses LMS to compensate ISI. The MMSE equalizer results in around 3dB of improvement when compared with Zero Forcing equalizer.from the simulation result as shown in fig. 8, it can be summarized that, ZF equalization in addition of noise gets boosted up and thus spoils the overall signal to noise ratio. Hence it is considered good to a receiver under noise free conditions. The multiple antennas are used to increase data rates through multiple antennas to improve performance through diversity. This technique offers higher capacity to wireless systems and the capacity increases 5 In fig. 9, for BER there is 5.8 db of SNR for ML equalizer for Rayleigh Channel. In fig. 10 for BER, there is 7.5dB of SNR. Comparing the results of ML and MMSE-SIC of Rayleigh fading channel for same BER i.e., SNR of MMSE-SIC gets improved than that of ML signal detection. The result of fig.10 is obtained after average of 10 times of MATAB simulation for MMSE-SIC. In fig. 11, for BER there is 8 db of SNR for MMSE- SIC Rician Channel. In fig. 10 for BER, there is 7.5dB of SNR. Comparing the results of MMSE-SIC of Rayleigh fading channel and MMSE-SIC of Rician fading channel, for same BER i.e., SNR of MMSE-SIC Rician channel gets improved than that of MMSE-SIC Rayleigh channel. BER 2x2 MIMO MMSE SIC Rician fading with QPSK MMSE-SIC Eb/No (db) Figure 11: BER vs. SNR plot for MMSE-SIC Rician of 2X2 MIMO with QPSK Modulation The SNR is 0.5 db more in MMSE-SIC of Rician Fading. The result of fig. 11 is obtained after average of 8 times of MATAB simulation.

6 V. Conclusion and Future Work Conclusion: The Rician channel has the good SNR value than the Rayleigh in all detections and in varying configuration of the MIMO system. MMSE-SIC is much better than MMSE. It can be concluded that wireless MIMO receiver with ML detection has the least BER for a given SNR and ZF detector has a greater BER. Also, the simulation results obtained shows that by combining SIC with MMSE provide better BER performance characteristics than normal receiver consisting simple MMSE or ZF respectively. And, MMSE-SIC provides better overall system performance than MMSE, ZF and ML the increasing diversity order. Future Work The different modulation schemes and antenna configurations can be used to analyze the performance of the MIMO system. Detections techniques can be used for designing of adaptive channel estimation with beam forming antenna and MIMO-OFDM. References [1] XiaoqingPeng, Weimin Wu, Jun Sun, and Yingzhuang Liu, Sparsity-Boosted Detection for Large MIMO Systems, IEEE COMMUNICATIONS LETTERS, VOL. 19, NO. 2, FEBRUARY 2014 [2] Max Scharrenbroich, Michael Zatman and et.al, Performance of a Practical Two-Step Detector for Non-Fluctuating Targets, IEEE seventh sensor Array and Multichannel signal processing workshop(sam), 2012 [3] Gurpreet Singh, Rahul Vij and Priyanka Mishra, Performance Evaluation of ML-VBLAST MIMO decoder using different antenna configuration using Rayleigh and Ricean Chhanel,2013 [4] Shreedhar. A. Joshi, Dr. Rukmini T S, Dr. Mahesh H M, Performance Analysis of MIMO Technology using V-BLAST Technique for Different Linear Detectors in a Slow Fading Channel 2010 IEEE. [5] Analysis of MIMO System Through ZF & MMSE Detection Scheme 1SwetamadhabMahanta, 2AnkitRajauria Dept. of ECE, RIET, Jaipur, Rajasthan, India, IJECT Vol. 4, Issue Spl - 4, April - June 2013 [6] Sugandha Aggarwal Characteristic Parameters Evaluation for MIMO based Communication Systems, ICCCSIM VOLUME- 8, Issue No. 2, page , July 2012 [7] Kuldeep Kumar, A comparison of different detection algorithm in a MIMO system, (IJAEST) Vol No. 7, Issue No. 2, , Punjab, [8] Wei Liu, Kwonhue Choi, et.al., Complexity-Reduced Channel Matrix Inversion for MIMO Systems in Time-Varying Channels, Vehicular Technology Conference (VTC 2010-Spring), May 2010 [9] Cheng-Yu Hung and Wei-Ho Chung, An Improved MMSE-Based MIMO Detection using Low- Complexity Constellation Search, IEEE Globecom workshop on Broadband wireless Access, [10] SajidBasir, Symbol Detection In Mimo Systems, a Phd. thesis submitted at the University of Engineering and Technology Taxila, July [11] Mingxi Wang, Weiliang Zeng, and Chengshan Xiao, Linear Precoding for MIMO Multiple Access Channels with Finite Discrete Inputs, IEEETRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 10, NO. 11, NOVEMBER 2011 Sudarshan Adhikari, from Likhu VDC- 6, Nuwakot, Nepal, did B. E. in Electronics and Communication from nec, Changunarayan in 2010, and M.E. in Computer Engineering from NCIT in He served as a Lecturer in various Colleges for 1 year. Since October 2012, he has been with the Dept. of Electronics Engg. and currently he is working as YC of Computer II Year in Dept. of Computer Engg. KEC, Kalimati. His research interest includes C-Programming, Signal Analysis, Communication Systems and AI. Dr. Sanjeeb Prasad Panday from Kathmandu, Nepal, passed his Doctor of Engineering (Ph.D.) in Information Systems Engineering from Osaka Sangyo University, Japan in He passed his Masters of Science in Information and Communication Engineering from Tribhuvan University, Nepal in He has been working as Asst. Professor in the Dept. of Electronics and Computer Engineering at IOE, Pulchowk campus, TU, Nepal since 2002.His research interests include image processing, digital holography, optimization algorithms and their applications to flow field measurements. Dr. Panday is a member of IEEE. 6