Error-resistant differential feedback of long-term covariance matrix in closed-loop MIMO

One method of obtaining channel information for closed-loop SU-MIMO or MU-MIMO is for the mobile to feed back a covariance matrix based on the measured downlink channel. However for MU-MIMO to be able to operate properly (i.e., have low crosstalk), the covariance matrix must be fed back with fairly high precision which would result in a very high level of overhead. This paper describes two new methods for feeding back accurate covariance matrix estimates from a mobile to a base station with low overhead, while being very robust to transmission errors. System-level results show that one of the proposed feedback methods has performance very close to an unquantized covariance matrix when the feedback bit error rate is less than 10%. When the feedback bit error rate is 20% one of the proposed feedback methods has better performance than an existing single-shot covariance matrix quantization method operating with no feedback errors while additionally requiring less feedback overhead than the existing method (16 bits versus 28 bits).