From: Istvan Novak - Board Design Technology (firstname.lastname@example.org)
Date: Mon Nov 08 1999 - 11:41:55 PST
I beleive at the dimensions that we are now using around bypass capacitors we cannot use
the 'incremental' counting of total inductance. As you stated accurately, it is the
mounted inductance that matters. The mounted inductance (except the plane inductance) is
really an inductance of a loop, where the inductance of a stand alone capacitor is its
partial inductance. If we move the capacitors and planes closer, it will increase the
mutual inductance, and it eventually will reduce the loop inductance as well. So for a
500pH mounted inductance of a 805 capacitor, the self inductance of the capacitor may be
higher than 500pH to start with, and it is reduced by the mutual inductance.
Date: Mon, 8 Nov 1999 10:14:19 -0800 (PST)
From: Larry Smith <Larry.Smith@eng.sun.com>
Subject: [SI-LIST] : Capacitor Inductance (was: Comment on Johnson's article)
I wanted to address several comments that have come back on the
subject of inductance of decoupling capacitors and vias. Rather
than further increasing the traffic on SI-list, I'll do it all
in one email. I have changed the thread title to better reflect
the subject at hand.
> From: "Rick Brooks" <email@example.com>
> Has anybody ever tried 0508 caps (not 0805) with multiple trace/vias to
> the planes?
> It seems like a better idea for low inductance, but I don't know about
> the cost or availability.
> - Rick Brooks
The 0508 capacitors have a little less inductance than 0805 capacitors
because the current path is wider and shorter. But this only effects
the 'partial inductance' due to the capacitor, which is usually small
compared to the total inductance. To understand the total inductance,
we must examine the entire loop associated with the capacitor
current. That includes not only the capacitor but the mounting pads,
traces (if any) vias and power planes underneath the capacitor.
For this discussion, let's define inductance in terms of energy stored
in the B field caused by the current flowing in this loop. Energy =
(L/2)*I^2, or one-half L I-squared. The energy is proportional to
inductance and the square of the current.
To reduce the inductance, we have to reduce the energy stored in the B
field. The best way to do that is to reduce the 'loop area' traversed
by the current because the B field is associated with this loop. The
loop may be marginally reduced by using shorter, wider capacitors, but
there are other things that are much more important. For example, the
distance between the vias to the power planes and the height of the
vias between the power planes and the solder pads. These are the
factors that greatly effect the inductance. As I mentioned in a
previous note, the partial inductance due to an 0805 capacitor is 0.2
to 0.4 nH, depending on the height. The partial inductance contributed
by the height and separation of vias is 0.5 to 2nH. The mounting
inductance must be carefully controlled before it makes sense to use
'low inductance' capacitors.
> From: Keith Howard <firstname.lastname@example.org>
> The low inductance capacitors from AVX have 8 or more leads, so require
> multiple vias into the planes.
> Unless you're talking about multiple vias in parallel with the required
Some configurations of the AVX capacitor have multiple leads in
a vdd/gnd checkerboard pattern, which nearly forces you to make a
checkerboard pattern of vias (possibly 8 of them) under the capacitor.
This is VERY good for low inductance. We have now broken up the loop
into at least 8 parallel loops, greatly reducing the inductance.
But in many cases, you could have just put the multiple vias in parallel
(not in a checkerboard pattern) and mounted a dumb old 0805 capacitor
on them. There would be almost as much improvement in the inductance
and you can skip the expensive part. That is because the inductance
is in the design of the vias, mounting pads and distance to the power
planes, with only a minor amount attributed to the capacitor.
> Date: Fri, 05 Nov 1999 17:53:09 -0800
> From: Ron Miller <rmiller@Brocade.COM>
> Also note that ATC recommends that their microwave capacitors be
> mounted on their side so that the resonances from the top layers and
> bottom layers of the cap are all the same length from the input and output
> trace, which shifts the SRF up somewhat.
> For those caps that have very thin height it probably doesn't matter, but
> for the cube shaped ones it does.
> Ron Miller
After significantly reducing the pad and via inductance, the biggest
inductance contribution is from the height of the capacitor. For large
'cubic' shaped capacitors, current must go a long distance up the post
of the capacitor before it can cross over through the plates of the
capacitor and go back down the other post. This essentially increases
the loop area of the capacitor. Note that power planes in the PCB may
be 10 mils or less below the surface of the PCB, but capacitors may be
40 mils high. With carefully chosen PCB stackups (power planes near
the surface) and carefully designed vias and solder pads, it is
possible for the dominant inductance to be in the height of the
As Ron has mentioned above, you can see the effect of the capacitor
height inductance on an impedance analyzer just by turning the cubic
capacitor 90 degrees on it's side. It is difficult to know which
direction the capacitor posts go just by looking at it. But if you
mount it on an impedance analyzer with the posts vertical, two
impedance dips can sometimes be seen. The low frequency resonance is
from the mounted loop inductance. The higher frequency minimum is due
to inductance internal to the capacitance. Current is sloshing around
between the top and bottom plates. There is inductance associated with
the current-loop path inside of the capacitor. Turn the capacitor on
it's side so that the capacitor posts are perpendicular to the mounted
inductance, and the 'anti-resonant' peak will go away. This
anit-resonance is due to a parallel resonant circuit and behaves just
like the parallel resonant circuit formed by inductive capacitors
mounted on PCB power planes, but all inside of a single component.
**** To unsubscribe from si-list: send e-mail to email@example.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 ****
Istvan Novak Sun Microsystems, Inc.
Istvan.Novak@Sun.Com Workgroup Servers, BDT Group,
One Network Drive, Burlington, MA 01803
Phone: (781) 442 0340
**** To unsubscribe from si-list: send e-mail to firstname.lastname@example.org. 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 ****
This archive was generated by hypermail 2b29 : Tue Feb 29 2000 - 11:39:42 PST