Re: [SI-LIST] : Decoupling between non-ground power rails, yes or no??

About this list Date view Thread view Subject view Author view

From: jrbarnes@lexmark.com
Date: Mon Jan 22 2001 - 10:25:37 PST


Michael,
I've used the technique of "stitching together" a partitioned power plane with
capacitors for the last three controller cards that I have designed. These
boards have 850 to 1100+ components on them, with up to six ball-grid arrays
(BGA's) and up to five supply voltages. For cost reasons we want to keep the
layer count as low as possible.

The 4-layer board was a S-G-P-S stack up:
* Topside active components, vertical signals, ground infill wherever possible.
* Ground plane.
* Power plane partitioned into:
    - 0.1-inch wide ground ring around the entire periphery.
    - +5V ring inside ground.
    - Patch for core voltage under processor and core-voltage regulator.
    - Remainder infilled with +3V.
* Bottomside passive components, horizontal signals, ground infill wherever
possible.

The 6-layer boards were a S-G-S-S-P-S stack up:
* Topside active components, vertical signals, ground infill wherever possible.
* Ground plane.
* Horizontal signals, ground infill wherever possible.
* Vertical signals, ground infill wherever possible.
* Power plane partitioned into:
    - 0.1-inch wide ground ring around the entire periphery.
    - +5V ring inside ground.
    - Patch(es) for core voltage(s) under processor(s) and core-voltage
regulator(s).
    - Remainder infilled with +3V.
* Bottomside passive components, horizontal signals, ground infill wherever
possible.

I "shotgun" the periphery of the boards with ground vias about every 1/2-inch to
form a Faraday Cage around the inner signal layers and the power layer.

I stitch the power-rings/patches in the power plane together with 0805 or 0603
100nF capacitors, spaced about 1-inch apart along the moat that separates two
voltages. That way, if a signal has to cross a moat its return current has to
make at most a 1/2-inch offset to follow the signal current. Since the
frequencies I'm worried about are well above the self-resonant frequency (SRF)
of even 100pF capacitors, the impedance seen by return currents jumping the
moats are due to the inductance of these "stitching capacitors" and their
mounting pads/ traces/vias to the power plane. This inductance is largely a
function of the length, width, and thickness (physical size) of the capacitors,
with their capacitance having a minor effect. So I use relatively large-value
capacitors to help with the decoupling of the supply voltages.

                                   John Barnes Advisory Engineer
                                   Lexmark International

**** 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 - 14:30:41 PDT