Subject: Output_ECL: ibis vs dml, hyperlynx vs spectraquest
From: iwatsuki, david (iwatsuki_david@emc.com)
Date: Fri Feb 22 2002 - 13:09:49 PST
I am working with an ibis model of an ICS pecl part and
am getting different results in Hyperlynx Linesim vs.
Spectraquest SI (in both cases, I have a 3.3 v differential
driver into a 56 ohm tline terminated with a 56 ohm pullup to 1.25v).
In the Hyperlynx sim, the output never swings below 1.2 v,
while SQ-SI swings low to 0.8 v.
The Hyperlynx sims correlate to lab and hspice results, which makes
me believe there is a model/simulator issue with ECL and SQ-SI since it
uses .dml format files that are translated by the program ibis2signoise.
The 2.1 IBIS spec says that for Model_type Output_ECL, the data in the
pullup and pulldown tables are BOTH referenced to VCC.
My questions are then:
- should the [Pulldown reference] be explicitly stated for
Output_ECL types within the ibis model? The ICS model does not have this.
The 2.1 IBIS spec says that if there is no [Pulldown reference], then it
defaults to 0.0 v which seems to contradict the earlier ECL spec
statement.
- should the Output_ECL type override the reference voltage in the
simulator?
I am assuming this is sometimes the case since Hyperlynx reads in ibis
directly and
produces a "good" simulation, while ibisi2signoise translates to dml with
an ECL-type keyword, and sets the pulldown reference voltage to 0.0v, and
then
produces a "bad" simulation under SQ-SI. (setting the pulldown reference
to
0 or 3.3 doesn't seem to affect Hyperlynx, so another reason I think
there's
an override in that simulator)
I don't see ibischk or dmlcheck programs flagging any errors around this
model
syntax. I also looked at similar PECL parts from ON and Motorola to compare
their
models. ON explicitly sets the [Pulldown Reference] to VCC=3.3V for the PECL
outputs;
Motorola avoids the issue by using a Model Type of Output, not Output_ECL.
The fix of adding the pulldown reference to the model seems to resolve my
low voltage
swing, but rise and fall times in SQ-SI still don't match either Hyperlynx,
Lab, or HSPICE.
(I have asked the same question on the Cadence SQ forum)
- david i
This archive was generated by hypermail 2b28 : Sun Feb 24 2002 - 08:04:28 PST