Larry
----- Begin Included Message -----
>From ldsmith@lisboa Fri Aug 1 10:46:43 1997
Date: Fri, 1 Aug 1997 10:41:22 -0700
From: ldsmith@lisboa (Larry Smith)
To: si-list@silab.eng.sun.com, chen@sigrity.com
Subject: Re: [SI-LIST] : Power/ground plane models
> Some examples that I tested have been placed in SIGRITY's web site
> (www.sigrity.com). These examples show the input impedances of the power
> and ground plane structure from DC to 1 GHz, and the effective capacitance
> and inductance from DC to 50MHz. You can also download these examples
> and a demo SPEED97 program from the web site so you can run these examples
> on your own machine.
Cool Web site! Thanks for modeling the 10 inch square planes for us!
For anybody on the list that has an interest in power plane analysis, I
would strongly suggest that you visit the site.
> We have been doing chip carrier and printed circuit board analyses with
> our newly developed software tool SPEED97 so I did some simulation on your
> "10 inch square board and 2 mil FR4 between power and ground plane".
>
> Just as what you said, we found package resonance at about 590 MHz,
> 830MHz, and so on. The power and ground plane structure is capacitive at
> the low frequency and its capacitance is about 45 nF in the low frequency
> limit.
Here are a couple of comments for you:
- It would help to make the vertical axis (impedance) a log scale.
You are clearly seeing the resonant peaks in impedance (multiples
of half waves), but not the resonant minimums that are associated
with multiples of 1/4 wavelengths.
- The position of the source on the plane will make a big difference
in the results. Yours appears to be about in the center. Because
of symmetry, you will not pick up resonances where the edges of
the board have high voltage, but the center of the board has
a lot of current (relatively low voltage). Place the source
asymmetrically off center and you will see more stuff, at least
different stuff.
- The impedance maximums are more associated with board size,
but the impedance minimums are completely related to source
position.
> However, the power and ground planes in a PCB are somewhere connected to a
> power supply that has a very small internal impedance. In that case, the
> power and ground plane structure is inductive instead of capacitive at the
> low frequency. The value of this inductance depends on any factors, such
> as where the power supply is connected to the power and ground planes,
> how many connections are made between the power supply and the power and
> ground planes, and the radius of via(s) connected to the power and ground
> planes, etc. The inductance value does not seem to be easy to estimate.
> Moreover, we found package resonance at frequencies much lower than 600
> MHz.
>
> Regards,
> Raymond Chen
> SIGRITY
- Yes, there is a very low impedance power supply connected
somewhere, but it is only low impedance at low frequency.
A switching power supply will hold down the impedance to
about 1/5 of it's pulse rate (probably 10's of kHz) but
goes high impedance after that. It has no effect on
power plane impedance after 100 kHz or so, far below the
frequency of interest of digital designers. A linear power
supply is a little better but not much. Decoupling
capacitors are the only answer...
- The low frequency resonance that you have found has to do
with the inductance of the conductors used to hook up
the 'ideal' power supply. Put in a real power supply
and the resonance will go away. The impedance curve
should trace out the impedance curve of the switching
power supply at low frequencies and then rise to the
point where it is intercepted by the plane capacitance
curve (or curve of whatever decaps are on the board).
regards,
Larry D Smith
Sun Microsystems
----- End Included Message -----