Superimposed Training in Cell-Free Massive MIMO: Is It Really Worth It?

This paper derives an achievable rate for centralized uplink cell-free massive MIMO (CF-mMIMO) networks when using superimposed pilots in combination with advanced combining schemes. This rate relies on a novel channel estimation technique that uncorrelates the combiner applied to any given information symbol from the channel estimate and its corresponding channel estimation error. The performance of powerful combining strategies such as minimum mean square error (MMSE) or zero-forcing (ZF) can then be assessed in the context of superimposed pilot transmission. Results show that despite superimposed training might be suitable when using the simple maximum ratio combining (MRC) technique, its performance when compared with regular pilot transmission falls short when advanced combining strategies are used.