RE: [10GBASE-T] RE: EMI Discussion
Geoff and
Dan,
Yes, I was in the
middle of it. Apologies in advance because there were about 100 people
involved in 10BASE-T and I'm going to hit only a few of them here. We were doing
something similar to what some of you want to do - characterizing existing cable
so that it could be used beyond its intended and specified frequency
range.
Don Johnson of
NCR did a lot of work characterizing noise on cable in existing
installations (including the famous, to 10BASE-T at least, Dayton YMCA).
Hardware was left in place for days that would capture and count noise impulses
over a threshold. Traces of various sample noise were taken and analyzed for
frequency content.
Bob Conte,
AT&T, did a lot of testing and analysis of cable as did Bill Kind, HP,
and Bob Snyder, Wang. Crosstalk characteristics were studied on a wide sample of
cables. Efforts were made to identify the 99% cable characteristics and to find
cable for testing that was at the challenging edge of the range for
testing. Modeling and simulation was done of the crosstalk by those people and
others. Quite a lot of this was done and we had confidence in our results when
we had theoretical and test results that were matching from multiple sources.
For instance a lot of work went into finding the noise distribution and
demonstrating that the self-crosstalk was a truncated Gaussian (that truncated
way above our target BER). (At more than half the meetings we would have a
newcomer 10BASE-T wouldn't meet the required BER because it didn't have enough
SNR. It wouldn't if the noise was Gaussian but it wasn't - part of the price of
breaking new ground.)
Multiple engineers
from DEC and 3Com, including Ron Crane, put work into characterizing the noise
from a source we had identified early on as our worst disturber - analog phone
ringers.
There were also EMC
tests done for both emmissions and suceptability (with lots of
discussion of how you lay out cable in the chamber for a valid
test).
Fortunately, at
10 Mb/s the wavelength was big enough that the connectors were not much of
a contributor to the problem. At 100 Mb/s, the connectors became important and
required testing. I don't know how many of you remember, but the initial Cat 5
deployments were done with pretty much the same wall jacks, connectors, etc as
Cat 3 and some installations had to be redone to support 100 Mb/s after the
redesigned connectors became available. Connectors made a big difference in the
EMC testing for 100 Mb/s but not for
10BASE-T.
Everything was
distributed on paper back then. By the end the presentations took several feet
of file drawer space. I think the industry spent more per page developing the 60
page 10BASE-T standard then any other standard we have done. But our
investment per unit shipped is pretty good so we probably don't mind.
:^)
Even with all that
work, we had a nasty surprise near the end when we found that for some cable
insulation types the attenuation rose steeply above 40 degrees C. Once we
figured it out we had to add a caution that some cable types had a temperature
problem.
Regards,
Pat
Dan-
At 12:02 PM 8/13/2003
-0700, DOVE,DANIEL J (HP-Roseville,ex1) wrote:
Hi Geoff,
Good point of
clarification.
I was not directly involved in that effort. At the
time I was working on
802.4 (rf modems,CATV,etc) but I know a number of
the folks who worked on it
and actually have an original copy of the
massive binder full of work done
by Bob Conte et al. It was an impressive
effort and I think we are looking
at something similar
here.
I wasn't really involved in it. I was working on fiber a
the time. Pat was in the middle of it. The guy who did the major work in the
field was Don Johnson of NCR at the time. I have seen him recently in wireless
meetings so he might possibly be available to provide war stories.
It
was HARD!
Geoff
Regards,
Dan
>
-----Original Message-----
> From: Geoff Thompson [mailto:gthompso@nortelnetworks.com]
> Sent:
Wednesday, August 13, 2003 10:49 AM
> To: DOVE,DANIEL J
(HP-Roseville,ex1)
> Cc: stds-802-3-10gbt@ieee.org
> Subject:
RE: [10GBASE-T] RE: EMI Discussion
>
>
> Dan-
>
> At 12:03 PM 8/13/2003 -0400, DOVE,DANIEL J (HP-Roseville,ex1)
wrote:
> >In the case of 10BASE-T, where we were
>
>applying high speed (10MHz?) signals to CAT3 wiring which
> had
been installed
> >for phone support, a *huge* quantity of testing
was done to
> verify signal
> >integrity, EMI compatibility,
and noise immunity.
>
> Actually it was not CAT3, The installed
base was AT&T DIW (or worse).
> We considered DIW as the
baseline.
> The TIA CAT3 spec was not approved until after the
approval
> of 10BASE-T.
>
> Geoff
>