Hi Piers -
The text that caught my eye was line 54 on page 13 of D1.0, where it
explicitly defines "six" tests within this subclause. (And a single overload
test shows up again in the informative test). It does not use the words "shall
test", so I agree it does not make them all mandatory. But for our
document, I'm asking if we think six (plus one) is really the right number;
if not, we should write this differently. I'm hoping the number can be
reduced.
I agree that we may not know how to resolve this now without more study,
and it should not consume much time until we know more. I mainly wanted to raise
the issue so that is does get studied.
On your technical point, with no connector loss, would we not expect
both somewhat less modal noise and less dispersion?
Tom
----- Original Message -----
Sent: Friday, December 17, 2004 10:07
AM
Subject: Re: [10GMMF] TP3 call Meeting
Minutes Dec 14
Mike reported:
> Tom presented thinking that at
present we have to test for > overload conditions both in the
informative sensitivity test > and the stressed Rx sensitivity
test.
The draft says "The receiver under test shall satisfy the
comprehensive stressed receiver ... overload". Note "shall satisfy" not
"must be tested". We didn't hear any reason why not satisfying these
conditions would be acceptable.
Either these are readily satisfied, in
which case the burden is light, or they aren't, in which case an explicit
requirement is the more necessary. At present, the committee doesn't
know which. Mike suggested keeping just the "simple" overload
requirement: the disadvantage with this is that the "simple" requirement is
informative, and we need a normative overload requirement.
If the draft
is neither incomplete nor incorrect in this area, I suggest we spend our time,
for now, on the other areas where improvement IS needed. Later, when we
have some reports of overload testing in practice, we can refine
it.
One small wrinkle does occur to me: bad-frequency-response channels
are (nearly?) always ones with connector offsets. Connector offset with
power near the outside of the core is associated with loss. Modal noise
is strongly correlated with connector loss. So, we are not likely to
experience maximum power, bad frequency response AND high modal noise
together. But reducing the noise loading for the overload criterion may
not make much difference in practice.
Piers
> 5. Informative
Rx Sensitivity Testing > > Tom raised a point that we need to
check that we are > describing the end condition and not defining
implementation > choices in the test. This would lead to us defining a
nominal > rise time (around 129ps) rather than specifying a BT
filter > with a given bandwidth. Piers felt that it was ok as it
was > currently written. Piers and Tom agreed to look over it
again
Tom's right, my apologies. For the normative receiver test,
we even have a NOTE saying that other implementations are OK. The
informative one needs rewording - whether we want to describe the TP3 signal
by risetime, bandwidth or both.
Piers
> -----Original
Message----- > From: owner-stds-802-3-10gmmf@ieee.org >
[mailto:owner-stds-802-3-10gmmf@ieee.org]On Behalf Of Michael Lawton >
Sent: 16 December 2004 11:16 > To: STDS-802-3-10GMMF@listserv.ieee.org >
Subject: [10GMMF] TP3 call Meeting Minutes Dec 14 > ... > 4.
Overload proposal > > Tom presented thinking that at present we
have to test for > overload conditions both in the informative
sensitivity test > and the stressed Rx sensitivity test. There was
some > discussion around whether this could be reduced to a test >
that would be acceptable. Piers and Tom had views on this and > agreed
to keep working on this. > > 5. Informative Rx Sensitivity
Testing > > Tom raised a point that we need to check that we
are > describing the end condition and not defining
implementation > choices in the test. This would lead to us defining a
nominal > rise time (around 129ps) rather than specifying a BT
filter > with a given bandwidth. Piers felt that it was ok as it
was > currently written. Piers and Tom agreed to look over it
again >
|