Woohyuk Chang

Diversity-Multiplexing Tradeoff in Rank-Deficient and Spatially Correlated MIMO Channels

In this paper, we first generalize a recent work of Zheng and Tse on diversity-multiplexing tradeoff (DMT) for some rank-deficient channels (poor scattering). We show that rank deficiency lowers DMT curves from that of i.i.d. Rayleigh fading channels. We show an interesting observation that suggests a fractional diversity gain may be possible at integer multiplexing gains. As the scattering becomes rich, the DMT approaches that of the i.i.d. Rayleigh fading channels. We next focus on spatially correlated multiple-input multiple-output (MIMO) channels. We show that spatial correlation does not change the DMT but still degrades the outage performance and analyze such a degradation at high SNR

Compress-and-Forward Relaying Over Parallel Gaussian Channels

We propose a compress-and-forward (CF) strategy for relaying over parallel Gaussian channels. The proposed CF strategy compresses received signals at the relay under a constraint on the product of their allowable distortions and forwards them to the destination over all subchannels. Then, power allocation over all subchannels at both the source and the relay and bit allocation for compressing received signals at the relay are jointly optimized under a total power constraint on each node. The proposed CF strategy is shown to generalize the previous CF strategy which compresses the received signal from the i-th. incoming subchannel and forwards it only through the i-th outgoing subchannel. We show if for a good incoming link at the relay, its corresponding outgoing link is also good enough, both CF strategies have similar performances. Otherwise, the gap between their performances becomes significant.

Diversity selection combining to enhance the coding gain

Traditional diversity selection combining techniques are designed to minimize the probability of hard decision error (symbol-by-symbol) prior to channel decoding. When the diversity selection combining technique is employed in conjunction with the channel decoding, the traditional diversity selection combining technique does not necessarily minimize the post-decoding error probability because the decoder takes the soft decision of several different symbols in making a decision on a particular symbol. In this paper, we propose diversity combining techniques based on the log-likelihood ratio (LLR) that provide a higher coding gain than the traditional SNR- or LLR-based combining technique does. Simulation results are provided for the turbo and low-density parity-check (LDPC) codes in Rayleigh flat fading channels.

Low-density parity-check codes with power reallocation

Power reallocation among the code symbols in a codeword is investigated for regular low-density parity-check (LDPC) codes. The motivation is that when error-control coding is employed, the received SNR need not be equally maintained for all symbols in a codeword because a certain number of errors within the error-correction capability can be corrected regardless of the received SNR for those erroneous symbols. We propose allocating no power to symbols experiencing a less favorable channel and reallocating the saved power to the remaining symbols experiencing a more favorable channel in order to receive them with an even higher reliability. At the receiver, the truncated symbols are tagged with erasures and errors-and-erasures correction decoding is performed. The simulation results show that the proposed power reallocation provides a power gain of 1.7-1.8 dB over equal allocation of power resources among the code symbols in a Rayleigh fading channel.