Simulation of a Multiuser MIMO System

 

Manas Deb

EE 360 Project Proposal

Winter Quarter 2010, Stanford University
 
Prof. Andrea Goldsmith

 

The 802.11n wireless LAN (WLAN) standard, which uses MIMO as the underlying technology, was recently ratified by the IEEE. Multiple antenna systems, which transfer data packets between one access point (AP) and one user station (STA) at a time, based on the 802.11n standard, are steadily gaining in popularity.  Even as this standard is being adopted by the industry, an IEEE Task Group known as TGac has already begun work on the next generation of the WLAN standard. For the next generation of WLAN systems, TGac is considering multiuser MIMO (MU-MIMO) systems. These systems can increase the total data throughput by effectively exploiting the spatial degrees of freedom, to simultaneously transfer data between a MIMO AP and multiple STAs.

My project will simulate the broadcast channel of such a multiuser MIMO system, in MATLAB. It will be based on the existing 802.11n standard, but with multiuser MIMO enhancements. TGac has already proposed some enhancements to the 802.11n Channel Models to support multiuser MIMO channels. I will leverage the knowledge that I gained from my EE359 project to incorporate these changes into the existing 802.11n channel models. The MATLAB model will implement a MU-MIMO OFDM transmitter, a channel and a receiver.
 
The transmitter will take an input stream of bytes and pass it  through the scrambler to ensure a uniform distribution of 1s and 0s. The scrambled data will be encoded by a convolutional encoder and then passed through a puncturer block to obtain the desired code rate. The resulting bit stream will be re-arranged into Nss groups of bit strings where Nss is the number of spatial streams per user. For this project, I will use 2 spatial streams per user and a total of 4 users. If time permits, I will incorporate 4 spatial streams per user, in my simulation. The spatial bit streams will be interleaved to spread the codeword across the subcarriers. The resulting bits will be mapped to constellation points based on the desired modulation (i.e. BPSK, QPSK, 16-QAM or 64-QAM). The complex numbers generated by the modulator will be mapped to the data subcarriers. Then the pilot subcarriers will be set with the appropriate values based on the 802.11n specifications. The resulting group of complex numbers will be input to a spatial mapper. In this project, I will compare the performance of TDMA, beamforming and linear and non-linear precoding techniques  of transmission. In the TDMA case, which is a single user MIMO system, the spatial mapper will cyclically shift the spatial streams to avoid unintentional beamforming. In the beamforming case each spatial stream will be multiplied by complex weights obtained by performing a Singular Value Decomposition (SVD) on a perfect estimate of the channel matrix. In the precoding case, I will use the Tomlinson-Harashima precoding technique to mitigate the interference between users at the receiving STAs. If time permits, I will try to come up with a low complexity linear precoding technique which will be suitable for implementation in a WLAN system. The output of the spatial mapper will be sent to the IFFT block which will output the time domain signal samples for each spatial stream. A cyclic prefix will be added to each OFDM symbol. The samples of all the OFDM symbols will be concatanated together and the 802.11n preamble comprising of a short training field, two long training fields and the SIGNAL field will be prepended to the signal samples. The signal will then be upsampled and output to the channel.  
 
The channel impulse response coefficients will be obtained from the MATLAB implementation of the 802.11n channel model written by Laurent Schumacher of Namur University and available for download from the Internet[7]. This is the same program that I used for my EE359 project. I will incorporate the changes proposed by TGac to this channel model  program, in order to support multiuser MIMO.
 
The receiver will be implemented assuming perfect carrier detection and gain control, perfect timing synchronization and no carrier or sampling frequency offset, since the main purpose of this project is to compare the performance of the various multiuser MIMO transmission techniques. The receiver at each user's STA will first remove the cyclic prefix and then perform an FFT on the signal samples received on each antenna. It will estimate the channel based on the Long Training fields. After the FFT, the signal samples will be sent to a Maximum Likelihood (ML) demodulator which will generate soft-decision bits. After de-interleaving, the soft-decision bits will be sent to the Viterbi decoder.
 
The output of the decoder will be used to match the input to the transmitter in order to generate packet error rate (PER) versus received SNR graphs for each user. These graphs will be used to evaluate the performance of the various multiuser MIMO transmission schemes for the broadcast channel. If a transmission technique is effective in reducing the interference between the users, then it will have a relatively low PER compared to the other techniques. The Error Vector Magnitude (EVM) versus subcarrier for each of the transmission techniques will also be plotted to analyze the degradation of EVM between single user MIMO systems and multiuser MIMO systems using precoding techniques.
 
Approximate Timeline:
 
 Start Date  End Date

 Task

 2-1-10 2-4-10 Incorporate the TGac multiuser MIMO channel model changes into the existing MATLAB program. Generate channel matrix coefficients corresponding to several independent instantiations of the various 802.11n channels. Verify the channel matrix correlation between users.
 2-5-10  2-12-10 Implement the scrambler, convolutional encoder, puncturer, interleaver, stream parser, OFDM modulator of the transmitter.
 2-13-10      2-20-10 Implement the transmitter spatial mapper for CDD, beamforming and Tomlinson-Harashima precoding. Submit project progress report on 2-19-10.
 2-21-10  2-28-10 Implement the channel estimator, ML demodulator for the OFDM symbols, de-interleaver, Viterbi decoder of the receiver.
 3-1-10  3-5-10 Run simulations. Graph and analyze the results. Start project report.
 3-6-10  3-10-10 Incoporate 4 streams per user in the model and also a new low-complexity linear transmit precoding scheme. 
 3-11-10  3-14-10 Complete final project report
 
 
References
 
[1] IEEE (29 October, 2009), IEEE 802.11n-2009—Amendment 5: Enhancements for Higher Throughput
 
[2]Costa, M., Writing on dirty paper (corresp.), IEEE Transactions on Information Theory, Volume 29, Issue 3, May 1983, pp.439-441
 
[3]Hiroshi Harashima and Hiroshi Miyakawa, Matched-Transmission Technique for Channels with Intersymbol Interference, IEEE Transactions on Communications, vol, COM-20, no.4, August 1972
 
[4]Sharif, M. and Hassibi, B., A Comparison of Time-Sharing, DPC and Beamforming for MIMO Broadcast Channels With Many Users, IEEE Transactions on Communications, vol. 55, no. 1, January 2007
 
[5]N. Jindal and A. Goldsmith, Dirty Paper Coding vs. TDMA for MIMO broadcast channel, IEEE Transaction on Information Theory, vol. 51, no. 5, pp.1783-1794, May 2005
 
[6] V. Erceg, et al, Indoor MIMO WLAN Channel Models, IEEE 802.11-03/940r4, May 2004.
 
[7] L. Schumacher, WLAN MIMO Channnel  Matlab program at http://www.info.fundp.ac.be/~lsc/