As some of you may know, I and my colleagues Mike Baxter, Jinhua Chen, Tom
Savarino and one of our client researchers, Dick Elco have been addressing
the questions that Chris Simon posed in his recent message.
The issue of time domain skin effects is at once one of the oldest questions
in signal transfer going back to signalling in Morse code across the
Atlantic, leading directly to the use of inductive loading coils and the
theory of the distortionless line. At the same time, it is one of the
newest and most important due to the clear emergence of differential
signalling and Gigabit transfers desired in the HIPPI-64 and copper Fibre
Channel contexts.
Over the past year, we have been asking ourselves exactly how does one
address the problems of skin effects in the time domain with arbitrary
risetimes and non-linear sources and receivers. The frequency domain
expanations for finite lines are not easily analytically specified from
first principles, and a variety of conformal mapping and numerical methods
have been applied successfully. The most readily studied is the round wire
where anlaytic solutions are possible. Many authors, this one included, have
developed concentric ring analogs for skin effects. We at NESA have
extended these to finite elements for transmission line segmentation.
(Design SuperCon'97)
Recently, in investigations involving cabling and 1.0625 Gbps Fibre Channel
signalling, we have had the opportunity to compare the .W model in
Avanti/HSPICE to the NESA segmented model in both the time and frequency
domain for long cables. To date, calculations have been made for both the
NESA and .W model. Both the .W model and NESA models have proven to allow
excellent modeling of real cable attenuation vs. frequency response (defined
in the frequency domain from measurements). The issue for time domain
simulations is the phase response, for it is the phase vs. frequency
response which determines the dispersive characteristics of the transmission
system. We have performed numerous time domain (.TRAN) simulations using
both matched lab generators and realistic semiconductors, ECL and CMOS,
driving these cable models. Variations in wire gauge are being considered
as well as dielectric losses. As soon as we have experimental verifications
of the simulation models, we will be publishing the results. Suffice to say
so far that to date, the .W model seems to be holding up pretty well. [We
refer the readers to Dr. Kuznetzov's notes on the .W model, avaliable from
Avanti, for more details concerning the frequency modeling of Rs and G]. The
trick is learning to relate the .W parameters to the actual interconnect you
are working with. For best use of the .W model, (that to avoid a SWAG), you
have to have frequenxy domain results for a sample of the interconnect.
(The NESA model comes from first principles, but is limited in shunt loss
considerations.)
Lastly, I wish to note that NESA has no affiliation with any simulation tool
company and respects all trademarks, and other proprietary marking for
product manes and features. We will keep the community informed as results
develop.
Ed Sayre
PS: If you are planning to come to the IBIS meeting we aresponsoring on
September 18th, please let us know ASAP by e-mail so we can plan properly.
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~+
| NORTH EAST SYSTEMS ASSOCIATES, INC. |
| ------------------------------------- |
| "High Performance Engineering & Design" |
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~+
| Dr. Ed Sayre e-mail: [email protected] |
| NESA, Inc. http://www.nesa.com/ |
| 636 Great Road Tel +1.508.897-8787 |
| Stow, MA 01775 USA Fax +1.508.897-5359 |
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~+