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

Re: [802.3_100GEL] Crosstalk Cancelation

Rich –

I’d consider crosstalk cancellation, and it should improve robustness, but it isn’t great for extending the reach – because reach always comes with more insertion loss.  When you’re overdesigned on the equalization and on all the other noise budgets, (or cancelling everything else out like in BASE-T) and only crosstalk is limiting you, it then helps tremendously.  However, we’re still in the early days of adding signal processing to PHYs at this rate and pretty limited in complexity.  From where we are right now, if I were limited only by crosstalk,  I’d probably go beyond what is in COMS and use some additional signal processing to improve noise robustness by a few dB or more, at lower complexity cost.  With the single-duplex, you’d need fext cancellation, or fext precoding, which has its own challenges.


However, in most of the channels I’ve seen, you’re not limited only by crosstalk.  Insertion loss at high frequencies is an issue, driving a bunch of media, routing and physical design issues.  It’s a battle against the really bad behavior, which has been pushed above 30 GHz at great effort, but still  starts to pull the whole curve down at frequencies of 20 GHz or higher.  For example, we see this in the ‘group 2’ models in Zhang ( see particularly slide 10).  As such, I think the signal processing complexity is much better spent on using a dual-duplex approach, which halves the frequency you’re occupying, getting you plenty of margin out of the insertion loss against all impairments.

And even then, you can still mitigate the crosstalk on top of that if needed.

Changing the frequency range needed opens all kinds of doors for improvement and design flexibility.



From: GMAIL Richard Mellit [mailto:richard.mellitz@xxxxxxxxx]
Sent: Thursday, March 01, 2018 5:44 PM
To: STDS-802-3-100GEL@xxxxxxxxxxxxxxxxx
Subject: [802.3_100GEL] Crosstalk Cancelation


Any thought to crosstalk cancellation to help make the longer 100G reaches?

... Richard Mellitz

To unsubscribe from the STDS-802-3-100GEL list, click the following link:

To unsubscribe from the STDS-802-3-100GEL list, click the following link: