From: Scott McMorrow (firstname.lastname@example.org)
Date: Mon May 21 2001 - 14:07:25 PDT
I cannot speak for the accuracy of skew prediction at the chip level.
However, especially with source synchronous busses, timing (skew)
is still based upon the most accurate error analysis that we can do.
In your example, lets assume +/- 50 ps skew at the driver inputs
and +/- 50 ps skew at the receiver outputs. Total worst case skew
without other effects is +/- 100 ps.
Skew due to package, and SSO effects is probably +/- 100 ps
Skew due to board crosstalk, routing and ISI effects is probably +/- 100 ps.
Total skew = +/- 300ps.
If +/- 25 ps of board effects are unaccounted for then this
would constitute an error of 8%. Not a significant error percentage,
but certainly a noticable one. My guess is that in many cases there
is much more than +/- 25 ps of effects that are missing from many
end to end system simulations.
If we know that there are board effects not account for by simulation,
we can certainly guardband for them. For our hypothetical system the
total system skew would then be +/- 325 ps.
However, one must be quite careful with these "simulation modeling errors" and
be sure that there are not cumulative errors that can creep up. Such
cumulative errors can occur due to cascaded models, each with errors
that occur either in their fundamental modeling aspects, or due to simulator
algorithm errors. (The old HSPICE w-element cascaded section problem
is an example of this. Hopefully fixed by now.) Fortunately, in source
synchronous systems, cascaded modeling errors tend to drop out of the
equation, since for skew we are dealing with relative skew, not absolute
timing numbers. In that case, if all signals which require low skew are
"nearly identical", and also nearly identical in characteristics with the strobe,
( this is Scott's same animal rule), then most of the small modeling errors
drop out because of symmetry.
-- Scott McMorrow Principal Engineer SiQual, Signal Quality Engineering 18735 SW Boones Ferry Road Tualatin, OR 97062-3090 (503) 885-1231 http://www.siqual.com
Todd Westerhoff wrote:
> Scott, > > Very good points - but they beg the question: > > If I'm doing my own ASIC, and even if I characterize the behavior through > the I/O buffers to the Nth degree with HSpice (or whatever) - > > - how good, on average, is the internal skew information I can get for the > path delays inside the ASIC?. This is the same question as before: if I can > only predict the signal arrival times at the inputs of the output buffers to > +/- 50 pS - how important are 25 pS effects at the board level? And more > importantly, what level of timing resolution for internal ASIC delays should > be considered "good" for a modern device? > > Todd. > > Todd Westerhoff > SI Engineer > Hammerhead Networks > 5 Federal Street > Billerica, MA 01821 > email@example.com > ph: 978-671-5084 >
**** To unsubscribe from si-list or si-list-digest: send e-mail to firstname.lastname@example.org. In the BODY of message put: UNSUBSCRIBE si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu ****
This archive was generated by hypermail 2b29 : Thu Jun 21 2001 - 10:12:03 PDT