[BP] my plan 4 February
--
|--------------------------------------------------------------------|
| Charles Moore
| Avago Technologies
| Image Solutions Division
| charles.moore@avagotech.com
| (970) 288-4561
|--------------------------------------------------------------------|
I said that i would present my own personal proposal at next weeks meeting.
I have not settled all issues in my mind, i a still waiting for more inputs
but here is a snapshot of where i am now. Comments welcome.
a) Reduce channel loss: I will propose 3dB reduction at Nyquist, and i am
looking for guidance from people who know boards
better for possible shape change. I think that
less than 3dB will spell grief for the Rx and more
will but an intolerable burden on meeting 1m.
I will support a channel return loss spec as long
as it has some sort of smoothing of the return
loss. If someone wants my help in writing such a
spec let me know.
I will support a distinction between the KR, KX4,
and KX channel specs.
I will support some disclaimer about not making all
channel impairments worst case at the same time,
especially if it applies mainly to KR. If someone
wants help writing this let me know.
b) Reduce crosstalk: I will propose an integral measure of crosstalk such as
the power gain ratio described in brown_c1_0106. This
will more accurately predict the problems caused at
the Rx and allow the maximum improvement in performance
while rejecting the minimum number of boards. The
33dB power gain ratio as calculated in moore_c1_0106
or the equivalent will be my target.
John D'Ambrosia tells me that ICR provides useful
guidance to the channel designer on how to improve
crosstalk and it should be retained for that purpose
but without a specific limit.
c) Increase EIT baseline: If i look at abler_01_0106 and abler_c1_0106, and
do some eyeball interpolation (the data is too
jumpy to justify anything more precise.)
Source of Interference EIT equivalent sum as
Crosstalk 10mV RSS
Self Interference 10mV RSS
50mU p-p DCD 7mV Linear
______________________________________________________________________
total 21mV
So i will propose this value.
This assumes that the interference tolerance channel
has low self interference and that Agilent ITTC1016
is near worst case, and that my crosstalk
recommendation is followed.
To help validate this assumption and improve
repeatability of EIT, I will support a return loss
spec on the interference tolerance test channel which
is tighter than any interconnect channel return loss
spec. If you want help writing such a spec let me
know.
d) Convert some of jitter in EIT test to DCD: Our test department has not
found any equipment which allow
this so i will not support it.
e) Reduced and/or modify Tx jitter: Joe's simulations (see abler_c1_0106)
does not show any substantial benefit
from a reasonable level of reduction.
The burden of proof that it is cost
effective will lie with any proposer.