Finite precision implementation of LDPC coded M-ary modulation over wireless channels

This paper studies the finite precision effects on the LDPC decoding performance with high order modulation over Rayleigh fading channels. We compare the performance of different quantization schemes and decoding algorithms using discretized density evolution over code ensembles of various code rates. We find that the min-sum algorithm with a simple correction term using 5 bits for quantization offers a very good performance in all code rate and modulation combinations. The loss is only 0.1 dB compared to floating point implementation. Direct implementation of the sum product algorithm with 6-bit quantization suffers a much larger performance loss when high rate LDPC codes are used. Both software simulations and test results of a hardware decoder implemented on FPGA verify these conclusions.