[SI-LIST] Re: Flight time measurements and post-route analysis
- From: "Baranauskas, Dave" <Dave.Baranauskas@xxxxxxxxxxxx>
- To: <Michael_Greim@xxxxxxxx>, <twester@xxxxxxxxxxx>,<si-list@xxxxxxxxxxxxx>
- Date: Wed, 30 Jan 2002 17:27:49 -0600
Hi All,
I agree with Michael, when he says he prefers to use input thresholds.
The fundamental issue I have with taking flight time measurements at a
single point on the receiver is that the part's Vil and Vih can be used
for binning at the foundry. Selecting a typical CMOS crossing level of
say 1.65v for a 3.3v IO does not mean every component from a given
manufacturer will switch at exactly that voltage. They will guarantee
input switching at the specified thresholds. This is especially
important if you are trying to design an open standards interface such
as PC133, RDRAM, AGP, etc., and trying to broaden your vendor supply
base as much as possible (think high volume products). In both cases,
using thresholds or a fixed voltage point, you will still wind up with
two flight time numbers for min and max cases. If the differences are
large enough to matter you may want to take a look at other areas of
improvement and go down the safe path. This is indeed a conservative
practice; however, just because there is slight negative margin on paper
does not mean you are going to have a problem -- just a place to start
looking if you do.
Regards,
Dave
-----Original Message-----
From: Michael_Greim@xxxxxxxx [mailto:Michael_Greim@xxxxxxxx]=20
Sent: Wednesday, January 30, 2002 4:44 PM
To: twester@xxxxxxxxxxx; si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: Flight time measurements and post-route analysis
Hi Todd,
Well, I will make these comments with my Specctraquest hat on.
In the interest of processing efficiency, perhaps the other tools
work the same. Specctraquest specifies a drivers delay driving=20
into a test load at Vmeas. Inputs can be measured at the thresholds.
The basic idea of this is to remove the clock to out component of=20
the test load so that you aren't double adding delays and being=20
incredibly conservative.
The problem with using the driver vmeas point and the receiver=20
threshold point is that your flight time numbers become conservative
to the tune of the driver time from vmeas to each threshold. From my
understanding,
the vmeas number is used so that a table of values does not need to=20
be preserved, as in theory receivers hooked to the driver could have=20
a range of threshold values. In any event, subtracting a constant
from the time crossing threshold at the receiver simplifies the
operation.
One school of thought goes that this should be good enough, if you
violate this extra pad that you are pushing the design too hard. The
other school says that tweaking a chip to meet this artificial extra
pad is a waste of time and effort. What it all comes down to is what=20
one's general philosophy is. Should you design beyond worst case
because
it gets you additional design margin. When you pass your worst case=20
corner, you are done. I have worked for some companies that did one=20
and others, the other. I am a personal fan of when you make worst case
you are done. However, you need to make sure that you have considered=20
all the things that you need to in order to define your worst case=20
corner.
One thing that I did want to point out is that if a signal shelves but=20
does not turn the other way it could still pass a tools monotonic check
but offer you a delay that you didn't expect.
To answer your question. I like to measure at the thresholds that the=20
receiver recognizes as valid highs and lows. Doing it the other way you
may be right most of the time but I believe you open a window where you
could get surprised and violate setup or hold if things are tight.
Just my 0.02=20
MG
-----Original Message-----
From: Todd Westerhoff [mailto:twester@xxxxxxxxxxx]
Sent: Wednesday, January 30, 2002 4:23 PM
To: Si-List@Freelists. Org
Subject: [SI-LIST] Flight time measurements and post-route analysis
Question for the group:
Most commercial SI tools seem to measure flight times to the receiver's
input thresholds. Thus, for a given transition, they report a "min" and
"max" flight time based on the threshold settings.
However, there is a school of thought in SI that says measuring flight
times
to Vil/Vih is too conservative. It's based on the observation that
setup/hold specs for a part are measured the same way Tco is - with
reference to a specific threshold. Therefore, the reasoning goes, you
can
measure flight times to the point where the receiver crosses the
reference
threshold, and, as long as the signal is within certain quality (i.e.
non-monotonic) and slew-rate limits, the flight time measurement is
valid.
The required slew-rate is (or should be) part of the receiving device's
input spec. If the input signal does NOT meet the quality/slew rate
requirements, well, then you have to use Vil/Vih method, or in some
cases, a
derating formula.
My question is: what techniques are people predominantly using: the
Vil/Vih
method, the reference voltage, or something in-between?
For those who use the reference voltage method, I'm curious to know how
post-route analysis is performed. You can adapt any post-route SI tool
to
measure at a single voltage by changing the models (if necessary), but -
do
any of the tools report slew rate at the receiver inputs during post
route
analysis?
Your comments, both on and off the list, are greatly appreciated.
Todd.
Todd Westerhoff
SI Engineer - Hammerhead Networks
5 Federal Street - Billerica, MA - 01821
email:twester@xxxxxxxxxxx - ph: 978-671-5084
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
"Oh, but ain't that America, for you and me
Ain't that America, we're something to see
Ain't that America, Home of the Free
Little pink houses, for you and me"
- John Mellencamp
------------------------------------------------------------------
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 archives are viewable at: =20
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages=20
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
=20
------------------------------------------------------------------
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 archives are viewable at: =20
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages=20
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
=20
------------------------------------------------------------------
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 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:
