Asymptotic analysis of space-time codes with Mahalonobis distance decoding in non-gaussian noise and interference

In this paper, we derive simple and asymptotically tight expressions for the pairwise error probability (PEP) of coherent space-time codes (STCs) which are valid for any type of noise and interference with finite moments and detection with general Mahalonobis distance (MD) metrics including Euclidean distance (ED) and noise decorrelating (ND) metric. We show that while the diversity gain of an STC is independent of the type of noise and the type of MD metric used, the coding gain depends on both the noise distribution and the MD metric. We show that in the case of correlated noise, significant performance gains can be achieved with the ND metric compared to the ED metric. While noise correlations are beneficial at high signal-to-noise ratios if they can be exploited by the metric, they are harmful if this is not the case and the simple ED metric is employed.