Hi Joel,
All that’s OK as long as we don’t
double count the EIT voltage for channel self interference. I suspect well have
a good meeting today. J
… Rich
From: Joel Goergen [mailto:joel@FORCE10NETWORKS.COM]
Sent: Thursday, January 19, 2006
10:56 AM
To:
STDS-802-3-BLADE@listserv.ieee.org
Subject: Re: [BP] Interference
tolerance test channels
Rich,
Most of the materials cut out are the result of the high end changing, not the
low end. Your simulation assumes that Dk and Df are constant across freq,
length, temp, and humidity.
If you lower the last 8Ghz of the model, you will gain back the ability to use
fr-4 in some cases across production volumes.
My personal opinion is that we are making the EIT channel concept tougher then
it needs to be. It seems to me that us guys designing back plane channels
could bear some of the burden rather then taking it out on the receiver and
transmitter.
I would suggest going back to the old channel equation constants, raising 4 to
5Ghz by about 1.5dB, and then assign a small group to develop a real channel as
the minimum floor in design practice, material construction, etc .... using
this channel as the minimum test channel. Sort of what we were about to
do 2 years ago, but not quite that deep. Then there are no arguments as
to what the minimum EIT should be and how to get it there ... simply because it
'is' there.
-joel
Mellitz, Richard wrote:
Actually that’s backwards of what I did. I adjusted parameters to move the line
up 3dB with out affecting the 1GHz characteristic very much. I though
that’s what we were talking about at the dinner meeting. Then I created a
channel to fit that line that had little return loss.
The intent
was to create an EIT test channel where we could control interference and not a
real 1 meter backplane channel.
The channel
has no connectors (step dielectric launch) and has targets of:
Trace
wide=3.6 mil spacing =6.5mil
Length 37.69
in
Plane space =
9 mil
Symmetric
Stripline ~ ½ oz
Er 4.4 tand
0.0117
Surface
roughness= 0.915 micron
17degree
routing
Target mean
that it wound need tuning but initial values would be close enough get you
started.
Also If we
pay $$$ for BERT why wouldn’t we expect to pay $$ for a test channel?
I’m also wrestling with the something else too. It seems that since we
add interference to the EIT channel which effectively adds in return loss and
crosstalk, why should AF max be the approximately the same since it does have
return loss and via effects. Maybe we need a different line for EIT
channel and AF max.
… Rich
Rich,
I looked at your data and it looks like you just changed the channel equation
constants to fit the limit line to the data.
If we adopt this, in general, we eliminate the use of several glass
constructions and pretty much any feature set less then 5mils.
Is this what you meant to do?
-joel
Mellitz, Richard wrote:
Oops…
missed one.
Sorry,L
Rich
Please be advised that your presentation
title page contains a confidentiality disclaimer. I will
disable access to this file via the reflector archive. I would recommend
that you re-post an appropriately scrubbed version.
-----Original
Message-----
From: Mellitz,
Richard [mailto:richard.mellitz@INTEL.COM]
Sent: Wednesday, January 18, 2006
4:09 PM
To: STDS-802-3-BLADE@listserv.ieee.org
Subject: Re: [BP] Interference
tolerance test channels
Did it!
I built a channel that has 3db less loss at 5GHz with out impacting the
loss at 1GHz. Also I created the AF coefficients to match this channel. See
attached zip file. I also checked the causality of the s4p as well. J
…Rich
Steve,
But also realize that the general line
also has built in margin for it to account for temperature, environmental, and
material variation as well. The line as currently proposed has to
be examined to look at from several aspects. For example, the Molex
channels are hugging the new proposed 23 dB line. 5” are on the
daughtercard and 35” are on the backplane, which uses a typical 7 mil
line. So we are saying that to meet the skin effect at the lower
frequencies we need a 7 mil wide line? I think that is too far.
Look at the attached figure – 7 mil wide traces hug that line. I
think we have moved it too far upward.
I don’t see any efforts yet on
reducing the problem via the crosstalk aspect of the problem. Has that been
abandoned? I don’t think all of the burden at this time should be
shifted to the channel, but should also be shared with the total allowable
crosstalk. Many of the channels did have margin. We should look to
striking a balance between the two.
John
-----Original
Message-----
From: Steve Anderson [mailto:steve.anderson@xilinx.com]
Sent: Monday, January 16, 2006
2:52 PM
To: DAmbrosia, John F; STDS-802-3-BLADE@listserv.ieee.org
Subject: RE: [BP] Interference
tolerance test channels
John, all:
But does a line made with the squared and cubed terms create a physically
realizable channel?
In the real channel I think there may be only two variables to play with:
skin effect and dielectric
absorption. If we
base simulations on something other than this, then I think bad things can
happen like
non-causal effects.
Steve A.
Guys,
Goergen asked the magic
question. Is it possible? Yes it is. We have a squared and
cubed term to play with. I am hoping Joel has some suggestions as well.
I just had a chance to do a quick scan and saw this. I will be working on
this stuff tonight
John
-----Original
Message-----
From: Oganessyan, Gourgen [mailto:Gourgen.Oganessyan@MOLEX.COM]
Sent: Monday, January 16, 2006
2:24 PM
To: STDS-802-3-BLADE@listserv.ieee.org
Subject: Re: [BP] Interference
tolerance test channels
I see
now what you refer too. I am not sure how you physically relaize a channel you
are suggesting, keep low freq the same and come up at 5 GHz? Any physical
channel should result in a tilted line?
-----Original Message-----
From: Mellitz,
Richard [mailto:richard.mellitz@INTEL.COM]
Sent: Monday, January 16, 2006
1:05 PM
To: STDS-802-3-BLADE@listserv.ieee.org
Subject: Re: [BP] Interference
tolerance test channels
The line didn’t only tilt. It also shifted. John D looked
at a few channels as I attached. If we shift, it’s got an impact for KX
and KX4.
…Rich
The new channel appears to be inline with what we stated - a tilt of the
line from DC to about a 3dB drop at 5GHz. Why do you feel the lower
frequencies need to stay fixed?
Thanks, Joe
Joe Abler
abler@us.ibm.com
IBM Systems & Technology Group
919-254-0573
High Speed Serial Link Solutions
919-254-9616 (fax)
3039 Cornwallis Road
Research Triangle Park, NC 27709
Hi Charles,
DC and low freq's went down way to much! Can you create a model with the
same losses at say 1GHz or so and 2 dB less at 5GHz? I thought that's
what we agreed.
... Rich
-----Original Message-----
From: Charles Moore [mailto:charles.moore@avagotech.com]
Sent: Friday, January 13, 2006 8:42 PM
To: STDS-802-3-BLADE@listserv.ieee.org
Subject: [BP] Interference tolerance test channels
guys,
I propose that we use the model ITTC_23.s4p for reduced
attenuation
EIT modeling. If it really will not work, ITTC_20.s4p is available as a
bail out channel. If it looks too easy, let me know and i will step it
up just
a tad.
The numbers refer to the fitted attenuation at 5.15...GHz.
charles
--
|--------------------------------------------------------------------|
| Charles Moore
| Avago Technologies
| Image Solutions Division
| charles.moore@avagotech.com
| (970) 288-4561
|--------------------------------------------------------------------|