[10GBT] Questions Related to Power Back Off

[Bijit Halder <bijit@PLATONETWORKS.COM>]

Hi fellow 802.3an members,



Being a layman, it was hard for me to keep up with the giddy pace of
progress at the last meeting in San Antonio. As a result, I decided to air
my ignorance in the relatively slow pace of email reflector. The use of
email reflector is also motivated by my wish to resolve my confusions long
before the next meeting. Given the urgency for progress, I do not with to
take up valuable meeting time.



One of the questions I attempted to raise on the floor is the system models
and methods that were used to calculate the power back off numbers. I was
assured by the experts that my concerns were completely irrelevant to the
discussion, or it was vender depended and beyond the scope of the
standardization process. However, that failed to dispel my confusion. And
hence, I am here to seek some clarification, without contradicting any of
the experts.



The confusion I have with the power back off numbers is as follows. I have
done a simple calculation for best achievable SNR for class E cables using
the models defined in draft D1.1. The system model I used includes
background noise (-150 dBm/Hz), flat transmit (-81 dBm/Hz), and ANEXT using
the PSANEXT model from D1.1. AFEXT, and residue Echo, NEXT, and FEXT are
ignored for simplicity. Using the model, I calculated the best achievable
SNR as a function of length of the class E cable. The insertion loss and the
ANEXT are changed according to the length. Please see the attached file
(power point) for details. According to my calculation, the optimal SNR does
not increase significantly with the decrease in cable length. In fact, my
calculation shows the opposite: the SNR is less for lower cable length.



Given the numbers there are two possibilities:



1. I have made serious mistakes in my calculations, and the numbers are
useless. In that case my apology to all.



2. My calculations have some merit, and in that case I would like to know
how on Class E cable we are going to have any power back off, let alone 10
to 16 dB? I would appreciate any engineering feedback on this, from power
back off experts and non-experts alike.



I would like to add that in this mail my comments are limited to Class E
cable. Also the addition of AFEXT, and residue Echo, NEXT, and FEXT will
change the achievable SNR, but would not improve it.



Thank you.



-bijit



P.S. Please see the attached file (ppt) for simulation details.

Powerbackoff_Comments_Bijit_11_19_2004.ppt