[SI-LIST] Re: Estimate ISI with S parameters -- Using TDR to do SDD21 correlation?
- From: Scott McMorrow <scott@xxxxxxxxxxxxx>
- To: John@xxxxxxxxxxxx
- Date: Tue, 17 May 2005 10:43:40 -0400
All,
Ye of little faith.
Let me clarify. To be more precise than Zhiping, no N-port s-parameter
model is fully accurate until all possible sources of structural and
material losses are accounted for in the matrices. So even 4-port
s-parameters are not enough to characterize a differential pair, unless
it is absolutely isolated from all other coupled or higher order mode
loss mechanisms. In the case of multiple loss mechanisms, even 4-port
s-parameters are not unique. We can take this to the absurd level if we
wish. However, I'm just a poor stupid engineer who likes to figure out
how to use the tools I have at my disposal as fully as possible.
Sometimes when you don't have the right tool, you just have to improvise.
But, as a real good engineering approximation, in lieu of all the
equipment necessary to do a multi-port frequency or time domain
extraction and correlation (how many of y'all can measure more than one
4-port differential pair at a time?), we use some simplifying
assumptions to get pretty darn close even with reflected waveforms off a
TDR. The assumptions are:
* Insignificant coupling to other structures.
* Good measurement launches.
* Insignificant mode conversion.
* Uniform transmission line structures
* Apriori knowledge of the transmission line topology. (Obviously a
single transmission line without discontinuities is best.)
* Access to the end of the line to make a resistance measurement.
* Two different line lengths to measure.
o The lines are significantly long enough that fringing
effects are negligible.
* Copper weight of the conductors is known
* We have a differential TDR and a very accurate microOhm meter.
Given these assumptions we can:
* Measure the round trip propagation time.
o From this, with two different length traces, it is possible
to compute the Er of the material with reasonable accuracy.
* Measure the risetime degradation of the reflected signal.
o From this, with two different length traces, it is possible
to compute the 3dB bandwidth of each trace.
* Measure the even and odd mode impedance of a differential pair
close to the launch point.
* Measure the resistance of two different length traces.
* Measure the risetime of the TDR at the end of the TDR cables.
* Using a good 2D field solver:
o If we know propagation delay, we then can compute Er
(stripline) or Er(eff) (microstrip).
+ If the trace is stripline, then Er can be entered into
the solver.
+ If the trace is microstrip, then you have to start
with your apriori known trace geometry and dial in Er
until Er(eff) matches the measurements.
o With a resistance measurement, we pretty much know what the
area of the trace cross section is.
+ With copper weight known, we use the lower end of the
potential thickness tolerance (since material vendors
always tend towards the low side in their process,
cause' it saves cost on copper) as a known parameter
to allow us to compute the etched trace width. For 1/2
oz copper use 0.6 mils and assume that the trace
profile is a rectangle. (Yes I know it's not, but
we're just trying to get within a reasonable error.)
o With two measurements of risetime degradation at two
different trace lengths, we can determine the 3dB bandwidth
of each trace.
o Using the apriori known structures
1. Dial in the individual trace geometry so that low
frequency resistance is a match.
2. Vary plane height and trace spacing until even and odd
mode impedance are a reasonable match to the measured
result within a few inches of the launch point.
(remember, you already have apriori knowledge of the
general trace geometry in the stackup, which is what
allows you to get quite close to the actual spacings.)
3. With 1 and 2 accomplished, we can now adjust tanDelta
in the solver to achieve the correct 3dB bandwidth for
each trace length measured
o At this point a w-element model of the traces that we
measured can be entered into Hspice. If your solver can
output frequency dependent table models, (like Apsim RLGC
and Ansoft 2D) then the fit will be much better, since
internal inductance change across frequency is better modeled.
+ The measured TDR response at the DUT can be imported
into Hspice as a pwl used to drive the simulation.
+ The measured TDR reflected responses can also be
captured as a pwl for comparison to the simulation.
+ A simple capacitive or inductive model for the launch
discontinuity can be entered as a simple LC circuit.
# Since we have a pwl of the actual TDR launch
waveform, we can adjust the LC circuit in
simulation to match the leading discontinuity in
the measured TDR profile.
* If you have multiple high and low
impedance discontinuities in the launch,
then add additiona L's and C's to cover
each and then optimize each for a
reasonable correlation to the measured
response.
+ We then compare the measured TDR reflected waveform,
from an open or shorted end of the line, to the
simulated reflected waveform.
# Tweek the transmission line parameters until you
get the best match for both line lengths.
regards,
Scott
Scott McMorrow
Teraspeed Consulting Group LLC
121 North River Drive
Narragansett, RI 02882
(401) 284-1827 Business
(401) 284-1840 Fax
http://www.teraspeed.com
Teraspeed® is the registered service mark of
Teraspeed Consulting Group LLC
John Lin (???) wrote:
>Hi Zhiping,
>Yes, I think you are right. It just like to build a RLGC models from
>S-parameter matrix. There are a lot of values for R,L,G,C.
>However, If not try to construct a complete S2P matrix, but only use S11
>measurement to prove correct S21 simulation result.
>It may be OK, right?
>
>Thanks,
>John
>-----Original Message-----
>From: Zhiping Yang (zhiping) [mailto:zhiping@xxxxxxxxx]
>Sent: Tuesday, May 17, 2005 1:14 PM
>To: scott@xxxxxxxxxxxxx; John Lin (ªL´Â·×)
>Cc: steve weir; si-list@xxxxxxxxxxxxx
>Subject: RE: [SI-LIST] Re: Estimate ISI with S parameters -- Using TDR to do
>SDD21 correlation?
>
>Hi Scott,
>
>I can not agree with your statement that S11 contains all the information that
>you need to characterize a channel. In order to completely characterize a
>two-port channel, the complete S2P matrix is required. In the real
>application, it is OK to use TDR (S11) to construct a channel model, but this
>model may not be unique and you may have different S21 for different models.
>It is not difficult to find a simple example to prove it. Thanks.
>
>Best regards,
>
>Zhiping
>
>
>------------------------------------------------------------------
>To unsubscribe from si-list:
>si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
>
>or to administer your membership from a web page, go to:
>//www.freelists.org/webpage/si-list
>
>For help:
>si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
>
>List FAQ wiki page is located at:
> http://si-list.org/wiki/wiki.pl?Si-List_FAQ
>
>List technical documents are available at:
> http://www.si-list.org
>
>List archives are viewable at:
> //www.freelists.org/archives/si-list
>or at our remote archives:
> http://groups.yahoo.com/group/si-list/messages
>Old (prior to June 6, 2001) list archives are viewable at:
> http://www.qsl.net/wb6tpu
>
>
>
>
>
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List FAQ wiki page is located at:
http://si-list.org/wiki/wiki.pl?Si-List_FAQ
List technical documents are available at:
http://www.si-list.org
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
