En jämförande analys av Enkla och Dubbla FEC-kodare : ett projekt i samarbete med Ericsson AB

This thesis presents a comparative analysis of different Reed-Solomon (RS) Forward Error Correction (FEC) encoding strategies implemented using Siemens Catapult, a High-Level Synthesis (HLS) tool. FEC ensures reliable data transmission in high-speed communication systems by correcting errors without retransmission. Based on the properties of the SerDes-protocol JESD204D, which achieves higher bitrates using Pulse Amplitude Modulation 4-level (PAM4), this study explores encoding strategies capable of processing multiple symbols per clock cycle. Two approaches were evaluated: a single encoder processing eight symbols per cycle and dual encoders processing four symbols per cycle across two parallel streams, reducing the combinational path and improving slack while maintaining the same throughput. Four encoding models were developed, but only three met the required latency constraints and were analyzed further. Key metrics evaluated were slack, area, and power estimation. The study also investigated Symbol Interleaving, enabling encoders to run at half the clock speed by running two encoders and distributing the symbols in a round-robin fashion, maintaining throughput. Particular attention was given to optimizing Galois Field (GF) multipliers, the most computationally expensive component. Logic synthesis results showed that the single encoder approach using Look- up Table (LUT)-based GF multiplication provided the best overall results. While dual encoders offer shorter combinational paths, they did not show significant timing advantages for the tested clock periods. The final results highlight the potential of Symbol Interleaving to improve timing constraints, with one single encoder model achieving positive slack for a 2x clock period. Although the models could not achieve the desired clock period, the state-of- the-art encoding model delivered the best results, providing valuable insights into how optimized the current encoding model is.