| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
Hi Claude, Just to chime in. I also found the TDD presentation using external clock generator experiment unconvincing.
In that system there is still a clock reference albeit drifting. The experiment shows tracking the frequency
but in my opinion did not demonstrate it can hold frequency in absence of any reference clock source.
Having implemented EEE systems where the follower has to track leader’s frequency in the absence of a signal from the leader during long quiet periods I learned of a lot of nuances in reliably tracking frequency.
Not having a reliable crystal reference at the follower during quiet periods, it is not clear to me how a PHY can use the
information it gleaned during the active periods to hold frequency at the follower during quiet periods.
Perhaps you or someone can shed some light on how this can be done in TDD. It is not obvious to me.
Thanks, William From: George Zimmerman <george@xxxxxxxxxxxxxxxxxxxx>
Claude – glad to see you’re issues with the reflector seem resolved, and thank you for correcting that obvious error. If you have such problems in the future, I’m sure others would be happy to forward emails
from you. I think it is only fair that you apply the same scrutiny to the TDD tests that you are asking for ACT. We will move forward by looking at both dispassionately and with equal scrutiny. Crystalless HAS been demonstrated for ACT. Is it a thorough test of all possible corner cases, no. But it is a feasibility demonstration with functional silicon. It deals with more real-world effects than
most simulations. Implementations are what they are – generally I have NOT seen implementers of a silicon demonstration go into detail of the implementation (on either side) – nor would I demand they do. It is not imcumbent
on presenters to teach others how they are implementing systems. There is no “magic” being proposed by either side. Both systems have multiple implementation options. But it is important to be fair and evenhanded in considering evidence. You have listed TDD as “proven” crystal less. I don’t know of a presentation showing results from a working crystal-less TDD link, let
alone in an EMC environment. If not, please point to it. The Ng presentation isn’t a working crystal-less TDD link (it’s not even a TDD link as proposed – it’s ASA-ML), it doesn’t even claim to have an operating crystal less link. It claims an ‘experiment’,
with an external clock generator. It is, in my opinion, far less credible evidence, yet you are willing to call it ‘proven’. Additionally, you give credit to “analyses” which are little more than simulations with eye diagrams (and considering DME as a double-rate PAM signal as well, something that we know is suboptimal) – far less
than a debugged and implemented timing recovery loop necessary to make silicon work (at any reach or temperature). No temperature or length effect was claimed – only that the jitter was too high. It is pretty easy to make a broken simulation (even easier
if you avoid parts of signal processing and just show plots) – but harder to make a working solution – yet you give these criticisms credence over and above a demonstrated working device? That just doesn’t seem even-handed. I would suggest that you apply an even threshold for proof in any comparison. Simulations from individuals not invested in making a working solution with a technique have never been a reliable statement that
“something doesn’t work” in my experience. Likewise, working silicon doesn’t have to prove how it achieves functionality, but – I will say that the test parameters should be considered, and should be even on all sides. George Zimmerman, Ph.D. President & Principal CME Consulting, Inc. Experts in Advanced PHYsical Communications 310-920-3860 From: Claude Gauthier <claude.gauthier@xxxxxxx>
Unfortunately, I’ve had issues w/ the reflector – mostly on inbound emails getting filtered (per user, not per alias.) But George’s email did get through, I did see this, and made the change, removing instances
of the “Not Demonstrated,” text. There is some additional text on the ACT crystal topic that could be resolved with a dedicated contribution. For example, does the demonstrated implementation use an equalizer to clean up the data-dependent
jitter? Can tests be run at worst-case cable lengths (maybe 15m, maybe higher as proposed by TJ), lower voltages, higher temps (basically whatever you conceive as a possible mission profile). In my mind, those kind of experiments put the issue to rest. Best regards, Claude From: George Zimmerman <george@xxxxxxxxxxxxxxxxxxxx>
The statement was made that ACT hadn’t demonstrated crystalless. Please see slide 3 of
https://www.ieee802.org/3/dm/public/0725/wu_3dm_01a_072925.pdf , presented to this group. There may be others. This statement should be corrected. George Zimmerman, Ph.D. President & Principal CME Consulting, Inc. Experts in Advanced PHYsical Communications 310-920-3860 To unsubscribe from the STDS-802-3-ISAAC list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-ISAAC&A=1 To unsubscribe from the STDS-802-3-ISAAC list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-ISAAC&A=1 To unsubscribe from the STDS-802-3-ISAAC list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-ISAAC&A=1 |