Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

[STDS-802-3-25G] What is COM?



In the November 19, 2014 25 Gb/s Ethernet Architectural ad hoc meeting it was suggested that we post some information about COM. COM is the normative channel specification used in IEEE 803.3bj and 'bm CAUI C2C.  Adee Ran and I put together this brief text.  A tutorial was also suggest but is likely outside of the ad hoc meeting IEEE scope. Suggestions on how to move forward on this would be  welcome. 

Thank You,

Richard Mellitz

________________________________________________________________

 

The channel specification metric channel operating margin (COM) is a essentially a signal to noise ratio . It is computed in the time domain as the voltage ratio of signal available in a reference signaling architecture (Tx and Rx) to noise at the reference receiver's sampler, where the noise is calculated at a quantile according to the specified error ratio.  For IEEE802.3bj and 'bm (C2C) the reference architecture is a 3 tap Tx FFE, Rx CTLE, a DFE whose number taps vary with clause, reference packages,  and filters.  For cables reference worst case host boards are included in the computation. In simple terms,  COM is a one dimensional  simulation of voltage at the cursor sample point , with the optimal equalization capability of the reference transmitter and receiver.

 

This channel inputs to the COM algorithm are a collection of 4 port s-parameters (s4p files). Each channel is a victim s4p file and all the relevant crosstalk s4p aggressor files.  For a cable configuration, the measurements are for the cable (using a well-defined test fixture) . Otherwise the channel is measured pin to pin. This would normally be from the BGA pad to BGA pad.  This includes and BGA escapes and DC blocking caps.

 

The COM end-to-end channel is computed from a reference transmitter pad to the sampler input in the receiver. It is not what may be classically called pin to pin or pad to pad.  Worst case packages plus pad termination and, in the case of cables, host boards are mathematically cascaded to measurement for the computation.  These package and board spec models are causal by construction and fit for time domain simulations.  All measured s parameters must also be fit for the time domain conversion.

 

Other inputs to the COM algorithm are:

The Transmitter specifications

 jitter

 distortion

 voltage swing

 noise

FFE tap ranges plus limits

The reference Rx

One-sided noise spectral density of 5.2e-8 V^2/GHz

Bandwidth limiting filter,

Reference CTLE,

number of ideal DFE taps plus limits,

the detector error ratio  related to FEC choice. 

 

COM method which be used to budget between loss, reflections, crosstalk, and noise supporting a wide range of platform configurations.  One dilemma was that the desired 35 dB loss between pins did not allow for  high volume variations and lower cost material and connector choices.  During the study group and working group phase for 'bj it was shown than budgeting needed to be in the context of reference architecture.  Power spectral (Gaussian) budgeting and frequency masks needed too much guard band to be effective for lower loss  higher impairment applications and enable the desired 35 dB budget. 

 

COM is used in receiver test as well. Using COM the amount of broadband noise that the receiver is expected to tolerate with any channel can be computed  and thus the test can be calibrated to get a desired stressed signal.  This can close the gap between test channel and products.

 

COM serves as a common reference for chip design and board design. The COM parameters represent the expected capability of a realizable PHY design. Channels that meet COM requirements are expected work with compliant PHYs with the specified BER or better.