COM for PAM4 Link Analysis – What You Need to Know/全体会议主旨演讲
(Room Ballroom H)
17 Oct 18
1:05 PM
-
1:35 PM
全体会议主旨演讲 COM (Channel Operating Margin) was adopted in the IEEE802.3bj standard. The intent is to qualify a channel in the context of a specification. The analysis is expected to be fast and efficient, utilizing transmitter and receiver specification parameters. This is in contrast to simulations requiring high quality device models. When the industry moved to 56G and now to 112G data rate range using PAM4 modulation, COM is continually widely adopted by IEEE and OIF standard bodies. Only minor changes and modifications were made when moving from NRZ to PAM4. For example, some of the modifications for 56G mainly affect the strength of equalization such as adding one more pre-cursor tap in TX FIR, and including the mid-band CTLE stage to take care the so-called long-tail effect. For 112G PAM4 the RX side FFE equalization scheme has been proposed. Besides the equalization change, there are also changes on the parameter side, such as tightening package model spec, relaxing TX SNDR and RX input noise spec, increasing equalizer setting ranges, narrowing model termination impedances with respect to the ideal, and so on. The handling of PAM4 in COM is very ideal. Essentially, COM treats PAM4 as a simplified NRZ case, with reduced signal amplitude. The consequence has been taken for granted. For PAM4 links FEC (forward error correction) is often indispensable, where not only the raw BER level matters, but also the error pattern (a.k.a. error signature) could matter even more. Thus, more detailed analysis is desired. This paper discusses selected aspects in COM with detailed analysis and specific examples which are supported by some time-domain silicon-correlated model simulations. The areas addressed in this paper include: (1) How jitter is handled in COM and its impact on the accuracy in predicting channel margin; (2) The sampling phase implications and its impact on error signature; (3) The DFE tap weight effect on link error propagation and hence on FEC coding gain; (4) The CTLE implementation feasibility and its effect on COM result; (5) How TX nonlinearity effect is treated in COM and its effect on COM result; (6) COM pass/fail criteria discussions to avoid both false negative and false positive errors. The goal of the paper is not to call for drastic changes in COM, rather to bring adequate attention to the COM community so that the result can be better interpreted and link margin assessment decisions can be more properly made.