Re: [SI-LIST] : effect of trace width on the performance

About this list Date view Thread view Subject view Author view

From: George Shaw (gshaw@ptsc.com)
Date: Wed Nov 01 2000 - 12:21:43 PST


Muhammad:
Presuming ground bounce or power may be a problem, here are some things to
check.

* Are the design rules for the connection lengths/arrangements for the ICs
and decoupling caps the same on both boards? If the power traces connecting
the ICs on the misbehaving board are longer than the on the good board, or
the loop area is greater, that could be the culprit. Longer traces/loop
areas will have more inductance (other things equal) and cause worse ground
bounce.
* Are there any breaks in the power planes caused by the clumps/rows of
vias and the associated clearance rings? With that many signal layers the
planes will look like swiss cheeze and it is possible that a power/ground
pin is unconnected or seriously isolated.

Alternatively:
* Do you trust the PCB shop? We recently had a batch of boards built where
the manufacturer used dull drills that caused FR4/copper to be
shredded/punched rather than cut, resulting in contaminated/faulty vias.
Just about every board was dead/flakey. We had to have the boards
microsectioned to determine the problem.

George Shaw

At 09:41 AM 11/1/00 -0800, Muhammad S. Sagarwala wrote:
>Sorry...I guess I should have been a bit more precise...
>
>My response is given below:
>
>
>
> >...
> >> The problems is the board with 9 mil wide traces and 1oz. copper is
> >> performing very good and the other board with 7.5 mil wide
> >> traces and 0.5 oz. copper is behaving very very bad.
> >...
> >
> >To make this a fair comparison you need to be more precise and scientific.
> >Any number of things could cause the difference you are seeing. It is
>quite
> >possible that there is something else wrong that has nothing whatsoever to
> >do with the trace widths.
> >
> >How many boards of each kind were built and tested, and found to be "very
> >good" versus "very very bad"?
>Two of each kind and the behaviour is very very consistent.
>
> >
> >Is the stackup "pretty much the same" or exactly the same? Is the
> >decoupling "pretty much the same" or exactly? How about the component
> >placement and trace routing patterns?
>
>The stackup is exactly the same...the decoupling is exactly the same...the
>component placement is exactly the same...
>the routing is different but the routing the rules are exactly the same...
> >
> >What exactly do you mean by "performing very good" and "behaving very very
> >bad"? Did you measure and compare the ground bounce, or is this an
> >observation of the end result, perhaps BER? Did you use the same
> >conditions? Might you be real close to a threshold, such that a very small
> >difference in (say) ground bounce could cause this big difference in
> >performance?
>I did not measure and compare the ground bounce...this is just an
>observation of the end result..
>The conditions were exactly the same.... There is a possibility that we are
>pretty close to the threshold and that is the
>reason I posed this question...I wanted to find out whether ground bounce on
>the board would significantly improve
>if I change the weight of copper and whether you people think (based on
>experience) change in trace width contribute
>significantly to improve the quality of the signal...

George William Shaw
Director of Microprocessor Systems Engineering
Patriot Scientific Corp., 10989 Via Frontera, San Diego, CA 92127
Direct line: 925-820-0781, FAX: 909-257-9779, email: gshaw@ptsc.com
Main Office: 858-674-5000

"It takes less time to do a thing right than it does to explain why
you did it wrong." Henry Wadsworth Longfellow

**** To unsubscribe from si-list or si-list-digest: send e-mail to
majordomo@silab.eng.sun.com. 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
****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Tue May 08 2001 - 14:29:57 PDT