Efficient tree search-based detectors for multiple input multiple output wireless systems

In Multiple Input Multiple Output (MIMO) systems, multiple antennas are deployed at both ends of the link to introduce both transmit and receive diversity. The MIMO principle has significant advantages to combat channel fading, increase reliability and data rates without increasing bandwidth or transmitted power compared to Single Input, Single Output (SISO) counterparts. However, MIMO detectors are more computationally demanding. The tree search based Maximum Likelihood (ML) optimum MIMO sphere detectors (SD) minimize Bit Error Rate (BER) and can collect the diversity offered by the MIMO channels. This comes at a high complexity that grows with the size of MIMO system and the modulation constellation used that makes them difficult to implement. Linear detectors are of a much lower complexity but with a much inferior performance. Therefore, a great research effort has been devoted to reduce the complexity of optimum detector and to devise sub optimum solutions to satisfy needs of real applications. In this thesis we design methods to speed up the operation of the SD to reduce its complexity in terms of the average number of the visited nodes per the detection of one received vector. This reduction is achieved by searching smaller spaces utilizing the Complementary Distribution Function (CDF) of the noise encountered by the MIMO receiver. While maintaining the diversity order of the known SD, our methods are very effective in reducing its complexity especially when it is high in large MIMO system dimensions and at low to medium Signal to Noise Ratio (SNR) values. Our first method is efficient, yet simple and tolerable to fit with different implementations of the SD and adopt to make use of their particulars. It is based on speeding up the reduction of the tree search radius under certain condition using the noise CDF. Testing shows that the proposed SD is capable to offer an average complexity reduction of 20% for a 8X8 MIMO system with 16 QAM, compared to an efficient implementation of the known SD. Our second design is the First Point Found (FPF) based SD. We show that this FPF solution, despite the fact that it is sub optimum, can still preserve the diversity order of the SD with a typical SNR gap of 0.2 dB at the BER of 10 -2 for the 4 X 4 MIMO system with 16 QAM. We elaborate further on the FPF idea and define a single probability parameter, the tuning of which can continuously adjust the BER performance-complexity tradeoff between the bounds of the FPF and the SD solutions. We show that this mainly affects the SNR range where the SD complexity is high while maintaining the BER performance and the diversity order, simultaneously. Testing shows that the FPF based SD is especially suitable for large number of antennas where it can offer 60-80% complexity saving for the system in the SNR range of 6-24 dB with 16 QAM compared to another efficient reference SD. Our third design is the statistical FPF SD (SFPF). Based on the FPF idea, the SFPF SD offers another form of performance-complexity tradeoff while maintaining the diversity order of the SD but with some SNR gap. Testing shows that an additional 10% of nodes saving is possible by using the SFPF compared to the FPF SD. In the coded case, the thesis considers Bit Interleaved Coded Modulation (BICM) MIMO systems where the tree search is used to find a group of transmitted vector candidates used to compute soft bit values to be fed to the soft input decoder. Based on noise characteristics, a method is proposed to define the group of such candidates without doing a full tree search which can reduce complexity by about 30% for the 4 X 4 system with 16 QAM compared to the known List SD (LSD).

Preview PDF FK 2012 44R.pdf Download (989kB) | Preview

Actions (login required)

View Item