Detection of SCMA signal with channel estimation error

Fifth generation wireless communication systems should support, among other things, very large number of simultaneous connections. To address this requirement, various schemes of non-orthogonal multiple access (NOMA) were proposed that allow to increase the number of simultaneously active users. One of NOMA schemes is sparse code multiple access (SCMA), where sparse multidimensional codewords allow to use iterative detecting algorithms with reasonable complexity. In the paper, SCMA detection is investigated in the presence of channel estimation error. Uncoded and turbo coded SCMA is analyzed. Uplink channel with Rayleigh flat block fading is assumed. Simulation results show that required accuracy of the channel estimation depends on the turbo code block length. For full utilization of turbo code error-correction capability with short blocks (40 bits) normalized variance of channel estimation error should be less than 10-3, the same value applies to the case of uncoded SCMA. For turbo code with long blocks (1024 bits), estimation can be less accurate, with normalized variance up to 10-2. With such channel estimation accuracy, power loss is about 0.6-0.7 dB compared with the case of perfect estimation. Two different types of codebooks have shown the same performance for coded SCMA, that leads to conclusion that codebook with more simple structure that provides less complexity of detection algorithm is a good candidate for use in SCMA schemes. The comparison with traditional orthogonal multiple access scheme with the same overall spectral efficiency is provided for both uncoded and coded SCMA systems. In case of coded system, SCMA scheme is shown to have smaller BER in the range of bit error probabilities below 10-4-10-5 for long blocks. The power gain is 0.5-1 dB for long blocks and a few tenths of dB for short blocks.