A novel conflict-free memory and processor architecture for DVB-T2 LDPC decoding

In this paper, we present a flexible architecture for an LDPC decoder that fully exploits the structure of the codes defined in the DVB-T2 standard (Digital Video Broadcasting - Second Generation Terrestrial). We propose a processor and memory architecture which uses the flooding schedule and has no memory access conflicts, which are encountered in serial schedule decoders proposed in the literature. Thus, unlike previous works, we do not require any extra logic or ad hoc designs to resolve memory conflicts. Despite the typically slower convergence of flooding schedule compared to serial schedule decoders, our architecture meets the throughput and BER requirements specified in the DVB-T2 standard. Our design allows a trade-off between memory size and performance by the selection of the number of bits per message without affecting the general memory arrangement. Besides, our architecture is not algorithm specific: any check-node message processing algorithm can be used (Sum-Product, Min-Sum, etc.) without modifying the basic architecture. Furthermore, by simply adding relevant small ROM tables, we get a decoder that is fully compatible with all three second generation DVB standards (DVB-T2, DVB-S2 and DVB-C2). We present simulation results to demonstrate the viability of our solution both functionally and in terms of the bit-error rate performance. We also discuss the memory requirements and the throughput of the architecture, and present preliminary synthesis results in CMOS 130nm technology.