ML co-channel interference estimation from SINR measurements for multicell ofdm downlink : Bounds and performance analysis

We consider the downlink of a multicell OFDM system where the frequency reuse factor is 1. Base-stations cooperate and precode the symbols intended for the mobile stations (terminals) located in different cells and operating on the same frequency, in order to precompensate the co-channel interferences (CCIs). To design the precoders, the CCI coefficients have to be estimated. We assume that the mobile stations are able to estimate their SINR that they feedback to their base-station. By properly designing a perturbation signal, it is possible for the base-station to estimate the complex valued CCI coefficients from the SINRs fed back. We investigate the maximum likelihood (ML) estimation of co-channel coefficients from SINR measurements. We obtain the associated Cramér-Rao bound and the mean square error of a practical estimator. Finally a two-cells system is considered and we investigate the impact of such a co-channel estimation on the sum rate of the system when precoding is operated based on these estimates.