From: Ray Anderson ([email protected])
Date: Mon Feb 07 2000 - 16:58:17 PST
Larry Smith wrote:
>We must not forget the power plane capacitance. The parameters for the
>power planes vary greatly (depending on the pcb stackup) but try c=1nF
>r=.01 L=.1nH. You end up with a major impedance peak at a frequency
>where all decaps have gone inductive and the capacitance of the power
>planes is still there. True, you must do a distributed plane analysis
>to really understand this. But even with your calculator, you should
>be able to see the dangerous effects of this resonance. The high
>impedance resonance typically lands at several hundred MHz, right where
>we are trying to operate our modern processors... :(
I tooks the capacitors that Larry and DougB had used for their
single-node analysis and put them on our power plane model and ran a
multi-node Hspice simulation. Caps were 1, 1.5, 2.2, 3.3, and 4.7 nanoFarads.
The plane is 4X6 inches, 8 mil spacing between planes, Er = 4.
The plane was excited near the center and the 5 capacitors were randomly
sprinkled around the plane in no particular position or order. The plane
model is an 8 x 8 lossy transmission line grid.
There are two .pdf files attached at the end of this message.
File test1.pdf shows a frequency demain plot of the simulation with about
5 or 6 random plane nodes (out of a possible 64) displayed.
Note that up to about 215MHz all the displayed nodes exhibit
about the same impedance. Above that frequency their impedances at the
various positions on the plane diverge.
File test1plane.pdf is the result of post-processing the results
of the Hspice run in Matlab. It shows the impedance versus location on
the plane at a particular frequency. In this case I chose to look at about
524MHz as it showed quite a divergence in the frequency domain plot.
By proper choice of parts and location we can make the impedance
profile quite flat across the spectrum. In this case we did not attempt
to do that, but instead showed how the impedance on an inadequately decoupled
plane can vary wildly at various positions.
**** To unsubscribe from si-list: send e-mail to [email protected] 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
This archive was generated by hypermail 2b29 : Thu Apr 20 2000 - 11:35:01 PDT