Re: [SI-LIST] : A Question About Power Noise.

Ray Anderson ([email protected])
Wed, 17 Mar 1999 14:31:31 -0800 (PST)

> So yes, while I agree a 0.1uF cap will be an inductor @ 1GHz, a 100pF
> will still look largely capacitive.
> Ray brings up some points that perhaps explain the differences between
> what we are seeing. Typically my designs are 2 layer RF boards with
> coplanar lines. The "ground" is beside the signal/power line. That can
> yield a MUCH higher mounted SRF than if you have a 4 layer board with
> the vias to GND/VCC. I still see a reduction between quoted SRF's and
> board mounted SRF's, but the difference is not as severe as when vias
> are used. The difference in SRF's is only a couple X at most instead of
> a couple orders of magnitude. So how much does using an 0402 help when
> using > 2 layers since the PCB dominates the inductive parasitic?
> --

A short follow on to my earlier comments on capacitor

We've actually evolved our pad geometries and stackups far
enough along that we are at the point where the partial inductance
of the capacitor is about 50% or more of the total loop inductance.
To achieve this you need to be very aggressive with the board stackup
and pad escape and via design.

This means that very high performance layouts can now have
cap/spreading/pad inductance proportions of around 50%/5%/45%
or thereabouts. This makes it worthwhile to pursue a low inductance
cap if you are trying to achieve the highest SRF possible.

All 0603 and 0402 caps are not created equal as far as
parasitic inductance is concerned.

Our group has run a bunch of finite element simulations
on capacitor geometries and has come up with the following
#'s. So if you know the physical height of your capacitor you
estimate its inductance:

Thickness (mils) Inductance (pH)

20 200
30 300
40 350
50 400

So if you have multiple sources of 0603 caps and one vendor's
parts are 30 mils thick and the others is 40 mils thick,
for minimum inductance select the thinner part.

(those who attended the Power Distribution Workshop at the
EPEP conference last October should find these results somewhat

-Ray Anderson

Sun Microsystems Inc.

**** 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 ****