Comparison of Two Suboptimal SISO Iterative Demodulators: LMMSE Estimation and Vector Gaussian Approximation

Abstract

The linear minimum mean-square error (LMMSE) estimator and the vector Gaussian approximation (GA) demodulator are two popular low-complexity suboptimal SISO demodulating methods for soft interference cancellation based iterative detection. In this paper, the equivalence of the two demodulation methods is established. In addition, the complexities of the two demodulators with arbitrary mapping constellations are reduced by reducing the number of matrix inverse operations in the preceding block processing. The comparison of the computational complexities demonstrates that the LMMSE demodulator is simpler than the one using the vector GA algorithm. Since the performances of the two methods are equivalent, the less complex LMMSE method has the advantage over the vector GA one for the SISO iterative detection.