Re: [10GMMF] TP3 3-impulse test proposal
From a practical implementation point of view, it makes a lot of sense
to limit the number of 'taps' to a minimum (three so far). As pointed
out by Lew, I'm also concerned that the limited time span of the
stressor (2xdT=200 ps) will not be a good emulation of real life channel
impulse responses even that the PIE are alike.
Many we should consider to have two tests with different dT(?). It
could, e.g., be implemented with 5 taps total.
B) Comments on the dynamic test: The biggest problem for real EDC
solutions may not be to handle the ~KHz variation (1 KHz has been
suggested) but more the problem of been able to handle the variety of
channel responses and make the transition from one response 'A' to
another channel response 'B' without causing errors.
The question is if we can imagine that any given Cambridge channel
response may be followed by any other arbitrary Cambridge Channel
response due to vibration of the fiber and/or connectors. If the answer
is 'no' then the suggested dynamic test of going from a state with
completely pre-cursor only to a new state with completely post-cursor
only may be a too stressful test that may lead to a non-desirable
trade-offs for the implementation.
[mailto:owner-stds-802-3-10gmmf@IEEE.ORG] On Behalf Of Lew Aronson
Sent: 17. september 2004 21:02
Subject: Re: [10GMMF] TP3 3-impulse test proposal
I think this is very close to where we want to be for the dynamic
penalty test, and I see no reason why the three impulse response
functions can't be aligned with those used for the static test. A
number of different comments:
1) We should do this calculation (and all similar calculations) for 300m
as well as 220m.
2) It is interesting to come up with 1.0 UI for the dT. In a way that
is good in that I know that many consider that a pretty demanding case,
so if it makes sense with the fiber model and EDC makers consider it
realistic to pass, than on that basis it might be a good choice.
3) I am concerned about deriving the details of the impulse response
parameters solely from PIE metrics. The reason is that PIE is for an
infinite ideal equalizer. As such, it's sensitivity to the total time
span of the impulse response is presumably a lot flatter (and more
non-monotonic) than is the performance of a real finite equalizer. I
think we see this in that the final choice in the values was not
terribly sensitivie to dT and was chosen to get a close fit on the PIE
value for the -D and -L cases simulataneously.
Thus, we risk coming up with a choice which is too short (compared to
actual fiber response) to adequately stress a real implementation, or
vice versa. Even worse, our choice is likely to drive the design
tradeoffs of the EDC's used and thus should really match the dT spands
from the fiber models.
I would suggest that, at least in the case of the dT choice, that the
fiber models and Petre's work be used to determine the approximate
choice, and that we live with a larger difference in PIE-L and PIE-D
4) With that said, I think the idea of having the same dT and the
pre-cursor and post cursor symmetric to each other is probably a great
goal for the static test as well and would then naturally align the
static and dynamic tests. [This would not change my thinking on the fact
that the static and dynamic tests should be seperate, I would still have
many reasons to want that]. Petre's work so far has emphasized the
goodness of fit to particular fiber examples. In the end, I think we
just want to have something of the same general shape chosen to
represent the 99% point by a combination of PIE value and represnetative
dT values, and to use that extra freedom to have equal dT's between
pulses and cases if at all possible.
Thus, I would say that we want to wind up with something of the form Ben
has given us, but with the dT and perhaps a value motivated more by
Petre's work than by the most exact fit to PIE numbers.
Lew Aronson (email@example.com)
1308 Moffett Park Drive
Sunnyvale, CA 94089-1133
From: Ben Willcocks [mailto:ben.willcocks@PHYWORKS-IC.COM]
Sent: Friday, September 17, 2004 8:13 AM
Subject: [10GMMF] TP3 3-impulse test proposal
Further to the preliminary data I presented in the phone conference of 7
September, I have attached slides showing our TP3 3-impulse test
proposal & simulation results. Many thanks to Sudeep Bhoja for his
assistance regarding the PIE calculations.
With Mike's permission, I would like to talk through this during next
Tuesday's TP3 conference call.