[SI-LIST] Re: Importance of Package Height
- From: "Istvan Novak" <istvan.novak@xxxxxxxxxxxxxxxx>
- To: <ARIAZI@xxxxxxxxxxx>, <si-list@xxxxxxxxxxxxx>
- Date: Mon, 1 Apr 2002 09:20:34 -0500
Dear Abe,
With respect to L_attached, I dont think I specifically mentioned either
series or parallel resonances. From the power-distribution planes, we have
several different series and parallel resonances. From the bypass
capacitors, we usually have only a series resonance; the parallel resonance
occurs at such high frequencies that we dont care. When one or more bypass
capacitors are mounted on planes, we get a mix of series and parallel
resonances.
For large-value capacitors, the capacitor's series resonance frequency is
typically in the MHz range, much lower than the lowest series modal
resonance frequency of a pair of not_too_big planes. For a pair of 10"x5"
planes with FR4 dielectrics, the first modal minimum is in the 110-160MHz
range . The minimum frequency depends on the location on the planes; it is
the lowest at the corners, highest in the middle, which is due to the
variation of equivalent plane inductance with location. The first modal
maximum for the same planes is around 288MHz: for laminates with non-zero
loss, this frequency shows a slight (~ +-1%) variation with location,
depending on how the resonance frequency is defined.
By mounting a capacitor on a pair of not_too_big planes, the series
resonance frequency of the capacitor will show up very distinctly in any
self or transfer impedances, because the series resonant frequency is so low
that at that low frequency the planes can be considered with their lumped
low-frequency equivalent circuit: static plane capacitance and
location-dependent plane inductance. If we first fully characterize the
pair of planes for its static capacitance and location-dependent
low-frequency inductance and loss, we can then easily deembed the attached
capacitor parameters (inductance and resistance) at the series resonance
frequency from any single self or transfer impedance measurement. Note that
not only L_attached, but also R_attached (which in this case is the ESR of
capacitor) should be deembedded: with one-ounce copper planes on the above 1
0x5" plane example, the actual ESR is 1.5 to 4 milliohms lower than the
value of measured impedance minimum. The difference depends on where the
capacitor is mounted on the planes and where the plane impedance is measured
for deembedding.
If we neglect the parallel resonance of the capacitor itself, parallel
resonances can occur as the modal resonances of the planes, and/or from
cross terms: the inductance associated with the capacitor mounting can
resonate with the plane capacitance.
As it was stated above, the lowest parallel modal resonance from a pair of
FR4 10x5" planes is around 288MHz. If the planes are ideal: perfectly
rectangular, no cutouts, plane separation does not change with location, and
the dielectric material is homogenious, this parallel modal resonance
frequency can be readily calculated as fres=1/(2*tpd), where tpd is the
propagation delay along the longer side of the rectangular. So as long as
these idealization assumptions are valid, calculating the modal resonance
frequencies is easy. In a validation process, however, the measurement
becomes tricky, because there is no unique way of measuring the parallel
resonance in a lossy circuit. If we measure the impedance at many different
locations on a pair of planes, we can, for instance, plot the extracted
resonance frequencies over the surface based on different criteria: we could
use either the frequency where the impedance magnitude is maximum, or take
the frequency where the phase is zero, or extract the frequency where the
phase derivative is zero. On a lossy pair of planes all three definitions
will yield a slightly different value. It was found that the zero phase
derivative condition yields values closest to the 1/(2*tpd) calculated
value.
The parallel resonance originated from the plane capacitance and mouting
inductance of bypass capacitor usually comes in the tenth of MHz range. If
we first characterize the planes in detail, again, from a single seldf or
transfer impedance measurement data L_attached (and R_attached) can be
deembedded at this parallel resonance frequency.
As it was pointed out in papers at ECTC2001, the capacitor parameters do
vary with frequency: L_attached (and also R_attached) is different at the
series and parallel resonances. The variation is bigger with lower ESR. As
there have been several threads on this list about the pros and cons of
low-ESR versus controlled-ESR (high ESR) bypass capacitors, it is
interesting to note that with higher ESR capacitors the frequency dependency
of parameters is minimal and negligible. Some recent publications indicate
that controlled-ESR (or high-ESR) bypass-capacitor solutions are becoming
commercially available.
Istvan Novak
SUN Microsystems
----- Original Message -----
From: "Abe Riazi" <ARIAZI@xxxxxxxxxxx>
To: <si-list@xxxxxxxxxxxxx>
Sent: Saturday, March 30, 2002 11:30 PM
Subject: [SI-LIST] Re: Importance of Package Height
>
> Istvan Novak Wrote:
>
> > We still should not expect the capacitor vendors to give us L_attached,
> but
> > it works the other way around: if we know the internal geometry of the
> > capacitor, for each connection geometry in our application, we could
> > determine L_attached by field-solver simulations. Or, if we have sample
> > parts, we dont need to know the internal geometry of the capacitor,
> > L_attached can be determined by measurements for each of our application
> > geometries.
> >
> Dear Istvan:
>
> From the above statements I have concluded that the preferred methods for
> determining L_attached consist of field solver simulations and series
> resonance measurements.
>
> This thread has contained discussion of series resonant frequencies
> (corresponding to zeros of board impedance), but no mention of
> parallel resonance (relating to poles in impedance expressions). It seems
to
> me that the parallel resonant frequencies can be ascertained via several
> different techniques including analytical calculations, simulations, and
> measurements.
>
> I am very interested to know what is your recommended approach for
> evaluating/treating parallel resonance when analyzing/designing
> a power distribution system?
>
> Thank you in advance for your reply.
>
> Abe Riazi
> ServerWorks
>
>
>
>
>
>
>
>
> ------------------------------------------------------------------
> To unsubscribe from si-list:
> si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
>
> or to administer your membership from a web page, go to:
> //www.freelists.org/webpage/si-list
>
> For help:
> si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
>
> List archives are viewable at:
> //www.freelists.org/archives/si-list
> or at our remote archives:
> http://groups.yahoo.com/group/si-list/messages
> Old (prior to June 6, 2001) list archives are viewable at:
> http://www.qsl.net/wb6tpu
>
>
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
