Improving MIMO detection performance in presence of phase noise using norm difference criterion

Practical MIMO communication systems suffer performance loss from oscillator phase noise. In particular, if maximum likelihood (ML) detection is performed naively without considering the phase noise, it results in an error floor in its symbol error probability. In this paper, we propose a method to detect the correctness of the naive ML solution in the presence of strong phase noise. A criteria based on the ML cost differences between the ML solution and the actually transmitted vector is used to determine a set of possible candidate solutions. Next we propose a novel algorithm for data detection using phase noise estimation techniques to obtain an modified ML cost for each of the candidate solutions. This approach results in symbol error rate performance improvement by reducing the error floor without incurring much additional complexity due to phase noise estimation. Theoretical arguments as well as simulation studies are presented to support the performance improvement achieved.