BER- and COM-way of channel-compliance evaluation: What are the sources of differences?
What is Channel Operating Margin (COM) analysis?
COM analysis is a simplified version of the statistical simulation of SERDES links. Since the IEEE 802.3bj 100 Gb/s Backplane Ethernet standard has been officially approved, Channel Operating Margin (COM) will likely become an important quality-evaluation method for SERDES links. Though there is an IBIS standard (IBIS 5.0 to 6.2) that describes the use of Algorithmic Model Interface (IBIS-AMI) models, it mostly defines the rules describing how a simulation platform interacts with the model libraries, calls their interface functions, forms inputs, and reads the output parameters.
The main sources of difference between BER and COM
There are number of reasons causing the differences between results of BER and COM that a few are listed below but more are discussed extensively in the white paper:
Since COM doesn’t use device models, it doesn’t know the exact package parameters. “Packages” in COM are “template approximations” of those that exist in devices, containing shunt capacitors and transmission lines.
COM doesn’t know the actual termination conditions that exist on both sides of the channel.
COM considers a receiver noise filter, with a flat bandwidth of 75% of the data rate. Most eye/BER analyzers don’t have the characteristics of this filter and don’t apply it.
Signal-to-noise ratio of the transmitter is rarely considered by BER analyzers. The same is true for receive noise.
In COM, the contribution from aggressors is taken with the worst possible phase combinations, which may cause crosstalk overestimation.
Comparison between COM and BER
The two simplifications common for COM and BER include using the bit response in jitter evaluation and taking the worst possible phase combination for crosstalk components. The white paper has shown that the two factors, on average, modify COM in different directions. COM most likely underestimates jitter but overestimates the effect of crosstalk. Overoptimistic prediction may lead to design failure. Still, it has been shown that the method can be easily improved with minor (if any) computational cost.
This paper also demonstrates an improved COM method that perfectly matches the results given by an accurate statistical eye and BER analysis, provided that both are based on the same channel response and identical equalization settings.
To learn more about SERDES, please visit.