RE: [SI-LIST] : Re: delay lines with PCB traces

Mellitz, Richard (
Thu, 25 Sep 1997 17:47:41 -0400

Hey y'all.
I used Dimitri's W models loaded with field solver data was able to
predicted this effect quite nicely. Good work Dimitri!

... Richard Mellitz, NCR
>From: Michael Chin[]
>Sent: Wednesday, September 24, 1997 5:59 PM
>To: si-list@silab.Eng.Sun.COM; smithh@VNET.IBM.COM
>Subject: [SI-LIST] : Re: delay lines with PCB traces
>I would like to echo that this zip-zag trace pattern has been known
>and oberserved in the lab to reduce the total expected delay. I
>came across a situation when I was using PCB trace to add some "skew"
>into the clock nets. The actual clock skew that was induced thru this
>kind of topology was measured to be less than the min. expected delay
>by 5% to 10%.
>I ended up running the serpentine in a wider gap (24 mil to 30 mil)
>to allow an accurate clock skew distribution. But, this costed us
>more space on the PCB to delay the clocks.
>Michael Chin
>Cisco Systems, Inc
>> From owner-si-list@silab.Eng.Sun.COM Wed Sep 24 14:35:44 1997
>> The 2nd order effect that Andy alludes to is what I call current jumping
>> in a zig-zag pattern. Primarily due to inductive coupling, the active
>> signal induces a return current on itself which due to the zig-zag pattern
>> propagates in the same direction as the active. The outcome of this effect
>> is a net delay decrease.
>> So when you force wire in a zig-zag pattern to obtain a certain net delay
>> (i.e. for clock balancing, etc.) and expect a To*L result, be careful.
>> You may have just added an unwanted skew in your clock distribution or
>> pop up early mode surprise in your design.
>> > ...
>> > What about second-order effects; for example, does the little zig-zag
>> > approach allow a weaker wave mode to zip right along as if the
>> > zig-zags weren't there?