| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
Second thread </Adee> From: Mike Dudek
mdudek@xxxxxxxxxxx I agree with Bill. These are different. Another contribution to the “D” term is any non-linear distortion. SNR_Tx is used in the simulations (e.g. COM) because it would complicate the models/simulations dramatically to include non-linear distortion or create Long reflections that create ISI outside
the Np range. Also if you tried to create these, there are an infinite number of different ways to create then and choosing one specific way (or asking for multiple different ways to be used) would further complicate the simulations. They are therefore
treated as “noise” and included in the SNR_Tx. Another difference as Adee has pointed out is that SNR_Tx is I think assumed in the simulations to remain constant at all points in the link. In a real link SNDR will vary with link/package
attenuation depending on the spectrum of the noise/distortion. (However if you assume a linear equalizer in the receiver this spectrum effect will disappear at the point that matters, as the linear equalizer will boost the high frequency noise that the link/package
has attenuated. A DFE rather than a linear equalizer will not restore the high frequency noise and will give better performance.
The values are set to be the same to close the budget with no margin. I have made proposals in the past on other projects to have more noise in the stressed Rx calibration to provide some link
margin for effects that are not fully captured. (e.g. reflections between the Tx and channel) but that was not accepted.
From: Bill Kirkland <wkirkland@xxxxxxxxxxx>
External Email
Adee,
When I brought this up in OIF XSR I was specifically told (verbally – during OIF meeting) that SNDR and SNR_Tx were not the same thing even though at the time they were both set to 32.5 dB. I let it be, because SNDR measures noise and distortion. There are
sn e and
sd terms in SNDR. SNDR contains the obvious “D” term which is not in SNR_Tx. I do see in XSR, under Receiver Interference Tolerance section that SNDR should be used for SNR_TX. Wish I had caught this back in the OIF meeting. (Shouldn’t that be “shall be used”, should implies it is a recommendation not required). I think folks testing XSR transmitters will have difficulties meeting SNDR requirements when they test
“over all Equalizer settings” in part because of what you have captured but also because of the Np,Dp settings. Again, I let it be after saying my piece.
To unsubscribe from the STDS-802-3-100GEL list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-100GEL&A=1 |