Packet loss recovery in the telemetry downlink via maximum-likelihood LDPC decoding

In this paper a novel framework for packet loss recovery in the Consultative Committee for space data systems (CCSDS) telemetry downlink is presented. The framework relies on packet-level LDPC erasure correcting codes and on low-complexity maximum likelihood (ML) decoding. A code design tailored to the ML decoder is presented, which is based on generalized irregular repeat-accumulate (GeIRA) codes. An optional concatenation with outer BCH codes is proposed as a mean for lowering the error floors. Numerical results show that for short block lengths the outer concatenation gives rise to very low error floors, while already for moderate-length codes the concatenation can be skipped without sacrificing the erasure recovery performance. Remarkably, all the proposed schemes under ML decoding tightly approach the performance of ideal maximum distance separable (MDS) codes down to low error rates.