Multiuser Precoding in Massive MIMO Systems: From a Precoding Granularity Perspective

We discuss the impact of downlink (DL) multiuser precoding granularities (in the frequency domain) on massive multiple-input-multiple-output (MIMO) systems. In practice, multiuser precoding is performed on a per block basis, i.e., all subcarriers in one block use the same precoding matrix calculated from the average channel state information (CSI) of this block. Per block precoding enables per block estimation of the (post precoding) effective channel using DL reference signals (RSs) inserted in each block. A large block size reduces the overhead of DL RSs, but also increases the mismatch between the precoding and channel matrices, causing higher multiuser interference (MUI). In MIMO systems with a small number of antennas, the advantage of a large block size outweighs the disadvantage. While with the increase of antenna number and user number, multiuser precoding becomes more sensitive to MUI, so small precoding blocks or even pre-subcarrier precoding is desired. A small precoding block makes the insertion of DL RSs too costly or even impossible. To solve this problem, we propose a new technique for DL effective channel estimation that does not need DL RSs. Both performance analysis and simulation are used to compare different block sizes under different antenna and user numbers. It is shown the in massive MIMO systems with a large number of users, per subcarrier precoding without DL RSs outperforms the traditional per sub-band precoding with RSs.