Performance analysis of a reduced complexity SCMA decoder exploiting a low-complexity maximum-likelihood approximation

This paper explores means of reducing the complexity of a Sparse Code Multiple Access (SCMA) decoder. SCMA was proposed to assist massive connectivity in 5G future wireless telecommunication standards. The existing SCMA decoding algorithm is based on the Message Passing Algorithm (MPA). It heavily relies on calculations of the exponential function to estimate the maximum likelihood decoded message. The exponential function typically requires a very wide dynamic range. MPA is reformulated by replacing exponentials with simpler functions. Implementation complexity of the proposed simplified SCMA decoder was characterized using Vivado HLS targeting FPGA implementations. Results reported in this paper show that this approximate algorithm utilizes 10 times fewer hardware resources than the original SCMA decoder and achieves an Area × Time complexity also reduced by a factor of 10. Moreover, when executed on a typical data-center processor, the run-time complexity is also reduced by a factor of 10. In terms of BER performance, up to 2.5 times improvement achieved for SNR less than 12 dB as well.