Semi-Blind Estimation of FDD Massive MIMO Channels using Correlative Coded Analog Feedback

Abstract:
Knowledge of accurate channel state information (CSI) is crucial for Massive MIMO systems in order to unfold the full potential of spatial diversity. In TDD systems, uplink (UL) and downlink (DL) channels are estimated by transmitting pilots in the uplink and exploiting channel reciprocity. FDD systems are more complicated because they require a dedicated downlink training and corresponding CSI feedback in the UL. For the latter task, linear analog modulation has been proposed which avoids digitizing and coding the CSI (see e.g., Echo-MIMO). We investigate the possibility to utilize these feedback signals for the (blind) estimation of UL channels, making a separate UL training obsolete. The method requires the correlative coding of the feedback signals prior to transmission. As we will show, the blind channel estimation benefits from the large amount of DL CSI feedback that arises in Massive MIMO systems. Consequently, for sufficiently many base station antennas M (i.e., M > 32 with K = 4 terminals) and low SNR, our method exhibits larger coherent processing gains than the Echo-MIMO scheme which requires K additional uplink training symbols. On the other hand, employing Echo-MIMO with M/2 uplink training symbols is always superior to SBCE method.