LDPC code-based bandwidth efficient coding schemes for wireless communications

摘要

This dissertation deals with the design of bandwidth-efficient coding schemeswith Low-Density Parity-Check (LDPC) for reliable wireless communications. Codedesign for wireless channels roughly falls into three categories: (1) when channel stateinformation (CSI) is known only to the receiver (2) more practical case of partial CSIat the receiver when the channel has to be estimated (3) when CSI is known to thereceiver as well as the transmitter. We consider coding schemes for all the abovecategories.For the first scenario, we describe a bandwidth efficient scheme which uses highorderconstellations such as QAM over both AWGN as well as fading channels. Wepropose a simple design with LDPC codes which combines the good properties ofMulti-level Coding (MLC) and bit-interleaved coded-modulation (BICM) schemes.Through simulations, we show that the proposed scheme performs better than MLCfor short-medium lengths on AWGN and block-fading channels. For the first case,we also characterize the rate-diversity tradeoff of MIMO-OFDM and SISO-OFDMsystems. We design optimal coding schemes which achieve this tradeoff when transmissionis from a constrained constellation. Through simulations, we show that witha sub-optimal iterative decoder, the performance of this coding scheme is very closeto the optimal limit for MIMO (flat quasi-static fading), MIMO-OFDM and SISO OFDM systems.For the second case, we design non-systematic Irregular Repeat Accumulate(IRA) codes, which are a special class of LDPC codes, for Inter-Symbol Interference(ISI) fading channels when CSI is estimated at the receiver. We use Orthogonal FrequencyDivision Multiplexing (OFDM) to convert the ISI fading channel into parallelflat fading subchannels. We use a simple receiver structure that performs iterativechannel estimation and decoding and use non-systematic IRA codes that are optimizedfor this receiver. This combination is shown to perform very close to a receiverwith perfect CSI and is also shown to be robust to change in the number of channeltaps and Doppler.For the third case, we look at bandwidth efficient schemes for fading channelsthat perform close to capacity when the channel state information is known at thetransmitter as well as the receiver. Schemes that achieve capacity with a Gaussiancodebook for the above system are already known but not for constrained constellations.We derive the near-optimum scheme to achieve capacity with constrained constellationsand then propose coding schemes which perform close to capacity. Throughlinear transformations, a MIMO system can be converted into non-interfering parallelsubchannels and we further extend the proposed coding schemes to the MIMO casetoo.