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

[10GBT] Results (?) from channel diags teleconference


Unfortunately one hour was not enough time to discuss the matter fully &
we ran into the hard cutoff after 60 minutes.

Some resolutions, but first a note:

This was an informal, ad-hoc call. There was no attempt made to list
participants, record motions or anything like that. My intention is
that, based on what I learn from the unnamed voices on my telephone, I
will prepare a new baseline for channel diagnostics that should be more
acceptable to the Task Force.

The first group of items discussed was labeled as TDR but the biggest
stumbling block turned out to be the label! It is not reasonable to
assume that TDR techniques (or techniques usually implied by the term
TDR) will be used to get single-ended, unterminated measurements.

I think that most people would be comfortable with the idea of defining
access to the results of optional measurements of this type but we will
need to come up with a new name to refer to the subject (perhaps
single-ended, unterminated - SEUT - would be best). It seems reasonable
to expect that reflection time, to within 5nS, should be achievable;
reflection amplitude may be possible but some of the PHY developers will
consider what accuracy is reasonable. It may be that the best reasonable
accuracy is too coarse to be useful, in which case we would define the
reading as: positive, negative or zero.

It is recognized that a properly terminated line might not return a
measurable echo. We should calculate what the smallest echo from a
non-compliant termination would be. Finally, there was agreement that we
would not expect the PHY to detect mid-span echoes from connectors
unless there was a gross failure.

Next we discussed the equalizer and what information would be available
for a MIB. Unfortunately, many (most) of the questions simply raised
more questions. I think that the main conclusion ws that we need to
resolve the equalizer startup/operation/conformance issues before we
consider what diagnostics information will be available.

One thing that was clear - the equalizer will naturally be defined in
terms of SNR vs frequency not attenuation vs frequency. It may take a
significant amount of processing to get from the SNR to the attenuation
definition. My opinion is that the SNR vs frequency is as useful as the
attenuation vs frequency so that this issue is not a problem.

Another complexity to consider for the equalization is the difference
between the feed forward and feed back paths. Typically the DFE
information will be transferred from the receiver to the transmitter but
the FFE information will not. Also, an implementation that has
significant receiver equalization (that would be legal) may not send as
much detailed feed back information for the precoder. I conclude that
the receiver should be responsible for providing the required diagnostic
information as a combination of its own equalizer settings plus the
information that it is sending to the remote transmitter.

I think that the TF will benefit from some detailed discussions
regarding the THP setup, startup sequences, adaptive (or static)
equalization and conformance requirements in the near future. The
diagnostics discussion will depend heavily on the results of the
equalization decisions.

Alas we were cut off before we could discuss canceler information. It is
my intent to organize a similar con call either in the week of Aug 9th
or Aug 16th. Watch this space...