Re: [SI-LIST] : Routing signals between ground and carved-up

sweir (weirsp@a.crl.com)
Wed, 22 Sep 1999 23:51:44 -0500

Neil,

You are likely to get a lot of reaction to your posting.

What happens to layers 1, 3, 8, and 10 depends on how you elect to pair for
routing, and how you space your stack-up. As long as you keep signals on
those layers off of layers 5 and 6, then the return currents will travel in
layers 2 and 9. It is easy enough to insure that these layers will not
cause you grief.

The signals on layers 5 / 6 are in real trouble if they cross the cuts in
layers 4 / 7. The return current will track the signal up to the cut and
then must divert along the path of least inductance around the cut. This
may mean jumping up to layer 2 / down to layer 9, through decoupling
capacitors and then back down to the original layer past the cut, or
wiggling around the cut through the isthmus'. In either case, the current
goes blasting right into the islands, but only after diverting long enough to:

Corrupt the primary signal integrity due to big impedance bumps.
Corrupt nearby signals due to coupling from the handy inductance loop created,
Increase the susceptibility of the primary signal to nearby signals by the
same mechanism.
Make contact with alien worlds via the EMI transmitted through the
efficient antenna this creates.

If you want to direct the return current, then control the routing of the
signal traces. One goes with the other.

Regards,

Steve.
At 05:54 PM 9/22/99 -0400, you wrote:
>Hello,
>This is my first message to this list. I hope I can describe our problem
>coherently.
>
>We are designing some high-density digital boards with clock frequencies
>of 40 and 50 MHz. Our general policy (virtues of which are for another
>discussion) is to create a separate filtered power supply area for each
>block of logic, using 3-terminal EMI power filters. As a result, each
>power plane is carved up into a bunch of local islands, fed by either
>a common "trunk" on that plane, or from a redundant, solid plane (the
>latter is a possibility we haven't explored too deeply yet.)
>
>The most immediate consequence of this is that the power plane is no longer
>a return path for signals that cross the cuts, which leads to the question:
>What is the effect of having a signal sandwiched between a continuous
>ground plane and a carved-up power plane? How do we model the impedance?
>It would seem that signals which don't cross the cuts are standard
>striplines, whereas for those which do cross the cuts, the plane is not
>considered a plane anymore.
>
>Clearly, we would not use these oddball layers for controlled-impedance
>critical signals, but how bad would they be for high-speed non-criticals?
>
>A sample 10 layer stackup would look like this:
>
>1 S
>2 Ground (continuous)
>3 S
>4 Power (carved)
>5 S
>6 S
>7 Power (carved)
>8 S
>9 Ground (continuous)
>10 S
>
>
>Layers 5 and 6, sandwiched between two carved up power planes, are
>restricted to routes which do not cross any plane cuts.
>
>Layers 3 and 8 are the more interesting ones in question.
>
>Any useful references would also be appreciated.
>
>Thanks,
> - Neil
>
>------------------------------------------------
>Neil Weinstock ViaGate Technologies
>weinston@viagate.com
> - Neil
>
>------------------------------------------------
>Neil Weinstock ViaGate Technologies
>weinston@viagate.com (908) 595-6400 x4525
>
>**** To unsubscribe from si-list: send e-mail to
>majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE
>si-list, for more help, put HELP. si-list archives are accessible at
>http://www.qsl.net/wb6tpu/si-list ****

**** To unsubscribe from si-list: send e-mail to majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****