Investigation of the Noise Immunity of Various Variants of the MMSE Algorithm when Implemented on a Fixed-Point Processor

The analysis of the noise immunity of two variants of the MMSE algorithm implementation was carried out on a fixed-point processor in this report. The first option is a matrix inversion using QR decomposition, and the second algorithm is implemented using a Kalman filter. Two variants of the algorithm behavior at the integer different lengths and fractional parts of the processor's number representation was considered. Noise immunity plots have been obtained for QAM-16 modulation method using a 4x4 antenna configuration. It is shown that for various integer lengths and fractional part of the representation of processor numbers, the MMSE algorithm, when implemented on a fixed-point processor and implemented using a Kalman filter, has higher noise immunity compared to the MMSE algorithm when implemented on a fixed-point processor and with matrix inversion using QR decomposition.