Channel Capacity in the Finite Blocklength Regime for Massive MIMO with Selected Multi-Streams

In this paper, we study the capacity of massive multiple-input multiple-output (MIMO) systems in the finite blocklength (FBL) regime. Due to the impact of FBL and joint coding among MIMO multiplexing, the multi-stream transmission enabled by the MIMO technique does not perform in the same manner as the one in the infinite blocklength (IBL) regime where transmissions are arbitrarily reliable at Shannon's capacity. Having more streams in a MIMO transmission exploits a longer (equivalent) blocklength, which is more preferred in the FBL regime. On the other hand, with a given total power budget/limit, having more streams with poor channel gains sharing the budget actually reduce the contributions from the strong streams on the FBL performance. This tradeoff is addressed in this work. In particular, we characterize the FBL capacity of massive MIMO under selected multi-stream transmission (SMST) scheme and investigate the optimal multistream configuration maximizing the effective channel capacity with a given target decoding error probability. For the scenarios with no channel state information (CSI), the FBL capacity with equal power allocation policy among the selected streams is studied. In addition, when CSI is available an instantaneous power allocation policy is provided.