Hi Frank,
Thanks for your inputs.
The primary goal for my presentation was to show that although FDD is a full-duplex system, it does not necessarily mean there is a need for echo cancellation. The asymmetry in the signaling bandwidth is to the advantage of the US receiver for many reasons including the minimal power of the echo signal.
You're right about a 2.5G system that it would potentially result in higher echo. I also clearly mentioned that in when I was presenting my slides. The choice of 10G creates the higher asymmetry, therefore convey the message of my presentation more clearly. Having said that, you should also consider the following:
1- In 2.5G system, the DS transmit PSD is increased by 6 dB within the US band. This elevates the echo signal into the US receiver by the same amount. However, the echo signal still remains 10 dB below the tolerated noise floor, still a negligible source of interference with no need for cancellation.
2- My analysis was based on the assumption that the echo channel matches the limit line. We know that the limit for RL represents the envelop of the echo frequency response. The echo frequency response has a lot of peaks and valleys and with some of those peaks getting close to the limit line. Because of this, the overall power of the echo can be significantly lower than what is calculated from RL limit line. In typical cables, the total power of echo can be as much as 6 dB (or more) lower than what is calculated form the limit line. It would be good for this task force to consider a limit on total echo power besides the typical RL limit so that we have a predictable and guaranteed maximum echo power for this application.
3-Note that I didn't assume any filtering to separate the Rx and Tx bands. This is clearly an option for an FDD system which can further reduce the echo power significantly.
As for the issue related to LPF and hybrid, I agree that beyond the external link segment there is potentially an additional internal path of echo within the PHY. The choice of LPF or hybrid is part of the PHY architecture and there are multiple ways to deal with it. Note that while hybrid matching may not be a trivial task over a wideband as needed in symmetric system, for an asymmetric FDD scenario, the matching has to be done over a very small bandwidth (~50 MHz). Whether this simple hybrid matching is done completely in analog or with some digital assistance, it's the choice of PHY designer. In any case and without going into the detail, the complexity of this hybrid cancellation is no where close to the complexity of an echo cancellation filter that people typically imagine for a full-duplex system.
I think the conclusions of my presentation still stand valid. Namely, in an asymmetric FDD system (even with DS rate of 2.5GHz), the echo power is very low eliminating the need for a complex echo cancellation.
-Hossein
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