[ibis-macro] Re: EMD Pole-Zero Models (Why voltage controled sources?)
- From: "Sam Chitwood" <samchitwood@xxxxxxxxxxx>
- To: <ibis-macro@xxxxxxxxxxxxx>
- Date: Mon, 30 Jun 2008 22:01:11 -0500
Hi everyone,
I wanted to respond to a few specific points made in various emails.
- The original question was "Do we need controlled sources?". Multiple
emails have provided very valid reasons why we do (Scott's list of 5 was my
favorite). I think the question has been already answered, but I'll quickly
add a couple details to drive it home.
- Quoting one of Walter's emails: "It is a trivial exercise to convert any
one of these "Types" of EMD_Blocks to Hspice W, S, E and G elements. If all
of the above is correct, then there is no need for EMD_Blocks that are
specifically voltage controlled sources."
It is NOT a trivial exercise to convert between S-parameters to pole/zero
and/or controlled sources. As Brad mentioned in one of the conf calls,
there are tons of codes out there that do this - some are free and some cost
real money. Obviously there are big differences between them, and obviously
it is not well understood in the industry how to do this for any arbitrary
response.
- Pole/zero description of broadband package interconnect response is not
always the best choice. Most of you saw Sigrity's IBIS presentation at DAC
that reviewed various modeling techniques. Many packages are well described
with a highly compact, behavioral SPICE netlist, but controlled sources are
necessary to accomplish various aspects of the technique. (Using
S-parameters or pole/zero instead would seriously degrade efficiency with no
appreciable gain in accuracy.)
A few other responses.
- HSPICE has two types of pole/zero formats. (Thanks to Bob for showing one
of them.) I've attached an example of the other method. I believe you call
it just like a Touchstone file.
- Walter quote: "This is really simple. All we need to do is get agreement
from Agilent, Cadence, Mentor, SiSoft and Synopsis. I believe this is the
list of EDA/IBIS companies that have proprietary simulators." A couple of
points here: 1) Please add Sigrity to your mental list. :-) We have chosen
not to sell our SPICE engine as a stand-alone tool; it's integrated with EM
engines. 2) The list doesn't matter at all. Whoever pays IBIS dues gets a
Friday vote, regardless of whether they sell a tool or not. If enough users
vote against it, it doesn't matter what the "big 6" agree on. :-)
- And with regard to Arpad's question about adding new capabilities, I think
that's straightforward. All of these models will be used in nodal SPICE
engines, right? All you have to do is clearly define the voltage and
current relationships at the terminals of your new "XYZ" element. Every
SPICE simulator can easily implement support for the XYZ element in its next
release since the equations will have been clearly documented in the BIRD.
(The "SHORT" element that we recently discussed is a simple example. We
spec i1=i2 and v(1,2)=0. Whether you internally implement as a 0V DC source
or R=0 is up to you.)
Looking forward to a lively discussion tomorrow. :-)
Sam
Other related posts:
