Online compressive diagnosis of massive MIMO calibration state

Abstract

With hundreds of antennas at one single base station (BS), it is critical to take full advantage of time-division duplexing (TDD) channel reciprocity to learn the downlink (DL) channel state information (CSI) from the uplink (UL) channel measurements at the BS. However, due to different radio frequency branches for transmitting and receiving, calibrating the antenna array at the BS is a necessary and prerequisite task to effect this reciprocity between the end-to-end DL and UL channels. This calibration procedure becomes extremely challenging in massive multiple-input multiple-output (MIMO) since all the existing calibration approaches do not scale well with the size of the antenna array. To release all the potentials promised by massive MIMO, we rely on compressive sensing theory to develop an efficient and robust way to diagnose the massive antenna array and further restore the full calibration at the BS without service interruption. The diagnosis only requires CSI feedback in the amount of O(log N) instead of O(N) as required by those conventional calibration schemes. Extensive simulations demonstrate our approach can maintain the whole large scale antenna system in a calibrated state with only limited amount of feedback overhead. The proposed novel scheme thus allows us to make full use of the channel reciprocity in massive MIMO under TDD operation.