SRS Compression for Cloud RAN Massive MIMO

In order to satisfy capacity demands in urban environments, antenna arrays with a large number of digital ports, e.g., 32 to 64 ports, are being deployed using 5G massive multiple-input and multiple-output (MIMO) technology. The channel state information needed for MIMO operation in time division duplex (TDD) systems is acquired using sounding reference signals (SRS) transmitted periodically by a user equipment (UE) in the uplink. In cloud radio access network (C-RAN) deployments, physical layer processing is partly performed at the radio unit (RU) and partly at the distributed unit (DU) in a central location. Accordingly, we assume that SRS is received at the RU, converted to frequency domain, transported via a fronthaul network to the DU where SRS channel estimation occurs. In this case, a significant fronthaul bandwidth is required to convey the 32–64 port SRS received periodically from a large UE population. In this paper, we consider a beam-domain compression approach for SRS in order to reduce the fronthaul capacity requirements. Initial results show promising compression ratios can be achieved with some additional processing at the RU but without requiring SRS channel estimation at the RU.