Channel-Optimized Information Bottleneck Design for Signal Forwarding and Discrete Decoding in Cloud-RAN

Inhalt:
We consider the single user uplink of a Cloud Radio Access Network where a single radio access point forwards quantized received signals to the central unit. The focus of our investigation is on the quantization step in the radio access point and the decoding in the central unit. First, we investigate the impact of different quantizer approaches on the performance of the belief propagation decoder applied for low-density party check codes. Second, we investigate the performance of discrete message passing decoders which are optimized via the Information Bottleneck method in order to process quantized samples. The resulting decoder has a low bit representation for each variable and all internal decoder functions are determined by lookup tables. For the investigated scenario the discrete message passing decoder performs close to the floating point implementation of the belief propagation decoder processing real valued log-likelihood ratios.