[SI-LIST] Re: Importance of Package Height
- From: Larry Smith <ldsmith@xxxxxxxxxxxxxxxxxx>
- To: ldsmith@xxxxxxxxxxxxxxxxxx, zhiping@xxxxxxxxx
- Date: Mon, 18 Mar 2002 13:51:56 -0800 (PST)
Zhiping - First, I completely agree with your last statement. There is
an inductance associated with the capacitor itself and there is an
inductance associated with the mounting structure including the power
planes, vias and pads. The mounting structure is under the control of
the PCB designer (and possibly the SI engineer..). The inductance of
the mounting structure may be found by electromagnetic extraction of
geometries and materials or by some kind of measurement that does not
involve a discrete capacitor. The MLC capacitor inductance has already
been determined by the capacitor manufacturer when the capacitor was
designed. For the library of capacitors used in a PDS simultion tool,
we only want the inductance of the capacitor itself to be associated
with the cap model. The rest of the loop inductance comes from the
mounting structure. The whole question here is "how do we obtain the
best value of inductance to associate with a capacitor to be used
in a PDS simulation?"
I usually do not make inductance measurements directly, they are too
difficult at high frequency and low impedance. I usually obtain
inductance numbers from system resonances. Resonant frequencies are
very easy to measure. A parallel RLC circuit gives a peak and a series
RLC circuit gives a dip. The frequency of the dip or peak is usually
pretty obvious on a VNA and is given by the formula f0 =
1/(2pi*sqrt(ESL*Cap)). Capacitance and frequency are easy to measure
and ESL is inferred from the measurement.
Calibration issues are very difficult for power distribution
measurements whenever S11 is involved. VNA's are 50 Ohm systems. The
impedance of power planes on our products are likely to be 10 to 100
mOhms at the frequency of interest, a factor of 500 to 5000 away from
our 50 Ohm VNA. I like to use S21 measurements and then just do a
simple through calibration using 50 Ohm coax cables. Then, solder the
cables to the device under test (power planes, capacitor, whatever).
The frequency measured for the peak or dip is extremely accurate even
though a complete VNA calibration was not performed. It turns out that
the magnitude of the low impedance measurement (dip) is also pretty
accurate.
So, I have not directly measured the inductance for your cases 1, 2 and 3
below. I measure the resonant frequency of a resonant system and the
capacitance in the system, then back calculate the ESL.
It turns out that the resonant frequency of a capacitor mounted on low
impedance (50 mOhm or so) power planes is higher than the same
capacitor mounted on a 50 Ohm transmission line. Therefor, the ESL is
lower for the cap on power planes than it is for a cap on 50 Ohm
traces. The physical reason has a lot to do with the resonant
structure for each measurement. The most meaningful measurments of
capacitors intended for decoupling will be made on low impedance (low
inductance) mounting structures.
regards,
Larry Smith,
Sun Microsystems
> Date: Mon, 18 Mar 2002 12:44:35 -0800
> From: Zhiping Yang <zhiping@xxxxxxxxx>
> X-Accept-Language: en
> MIME-Version: 1.0
> To: ldsmith@xxxxxxxxxxxxxxxxxx
> CC: si-list@xxxxxxxxxxxxx, ARIAZI@xxxxxxxxxxx
> Subject: Re: [SI-LIST] Re: Importance of Package Height
> Content-Transfer-Encoding: 7bit
>
> Larry,
>
> Thanks for sharing your ideas.
>
> Can you explain why the measurement from direct VNA with 50Ohm fixture
> is greater than on PCB?
>
> Here are my understandings:
> 1. VNA with 50Ohm fixtures (Calibrate VNA to the cap pads, can we do this
> in lab? I am not positive.)
> L_measure=L_cap_self_inductance
> 2. VNA with 50Ohm fixtures (VNA is not calibrated)
> L_meausre=L_cap_self_indutnace+L_fixtures+M_mutual
> 3. Measurement with cap on PCB structures
> L_meaure=L_cap_self_inductance+L_vias+L_plane+M_mutuals
>
> For my understanding, case 1 gives the smallest inductance. Maybe you
> are refering the VNA measurement to case 2.
>
> For the case 3, the results are also heavily depended on cap layout
> and PCB stackup. How can you build the universal cap component
> library for different PCBs?
>
> Here is my thought. When you build the cap libraries, the ESL should
> only be self inductance of the cap. The mounting inductance (vias, plane
> and traces) should be compensated in simulation tools. In this way, the
> cap library can be applied to any PCBs. Anyone disagree?
>
> Thanks.
>
> Zhiping
>
>
>
> --
> Zhiping Yang, Ph. D.
> Hardware Engineer
> Cisco Systems
> 270 West Tasman Drive
> Mail Stop:SJCG/2/2
> San Jose, CA 95134 | |
> email: zhiping@xxxxxxxxx :|: :|:
> Tel : 408 525 5690 :|||: :|||:
> Fax : 408 526 5504 .:|||||||:..:|||||||:.
> *****************************************************
>
> Larry Smith wrote:
>
> > Abe - As mentioned earlier, the inductance you get depends greatly on
> > the assumptions you make, either in measurement or software
> > extraction.
> >
> > The thing we really want to know is "what is the series resonant
> > frequency of a decoupling capacitor mounted on the pads, vias and power
> > planes of our products?" From the series resonant frequency, we back
> > calculate the ESL. My definition of ESL is the equivalent inductance
> > that causes the capacitor to have the low impedance dip at the
> > frequency that I measure.
> >
> > We find that the ESL obtained from measurement depends greatly on the
> > fixture used to make the measurement. Most people are now using a VNA
> > S21 measurement to characterize capacitors. The purists will make a
> > fixture that has a 'perfect' 50 ohm through impedance and mount the
> > capacitor on that fixture. That S21 measurement gives a clear
> > capacitance, resonance and inductance portion of the curve. Most
> > people will argue that the "ESL" determined by the resonant frequency
> > of that measurement is the "ESL" of the capacitor. Valid argument, but
> > it does not apply to our products.
> >
> > Ok, now go mount that same capacitor on a set of low inductance pads
> > and vias that are connected to thin power planes near the surface of a
> > PCB (my product..). Use the same VNA to make an S21 measurement of the
> > power planes. You will find that the resonant dip is as much as 25%
> > higher in frequency than the cap measured on the 50 Ohm fixture. Back
> > calculate the inductance and you find that the ESL for the capacitor is
> > 50% higher on the 50 Ohm fixture than it is on the power planes.
> > Hmmmm. Once again, the ESL of the capacitor depends greatly on the
> > assumptions and fixture that you use when you measure it.
> >
> > The inductances given on both the Kemet and AVX web sites are too high
> > for my purposes. They are not useful for calculating or simulating the
> > series resonant frequency when the capacitor is used for power supply
> > decoupling in our products. I would encourage any SI'ers that are
> > truly interested in power distribution to go repeat some of these
> > measurements on your own products. Find the ESL of a capacitor by
> > measuring the resonant frequency when mounted on low inductance
> > products and fixtures.
> >
> > Sun and Cadence sponsored a capacitor consortium meeting last October
> > which was attended by about half a dozen major capacitor suppliers. We
> > described the software tool techniques that we use to design power
> > distribution systems (Specctra Quest Power Integrity Tool). Very
> > quickly it becomes obvious that software tools for PDS analysis rely
> > heavily on the accuracy of capacitor models. Unfortunately, the
> > standards for measuring capacitor parameters (capacitance, ESR and ESL)
> > do not apply very well for capacitors used for PDS decoupling above
> > about 10 MHz. We found that the capacitor suppliers are very receptive
> > to these ideas and would like to make measurements that are meaningful
> > for their customers. The problem is that there is not a strong
> > consensus (much less a standard) in our industry on what ESL is and how
> > it should be measured.
> >
> > As you know, Sun has been working in this area for some time. We are
> > motivated to share this information with the rest of the industry in
> > hopes that meaningful measurements can be made by the cap vendors, which
> > lead to accurate simulation models for our PDS tools. But, there must
> > be agreement on what ESL is and how to measure it. That's why you get
> > all these loooonng emails from me.. We also publish our data in IEEE
> > conferences to get peer review on these concepts. Eventually, meaningful
> > measurement standards and simulation techniques will come out of this.
> >
> > regards,
> > Larry Smith,
> > Sun Microsystems
> >
> > > From: "Abe Riazi" <ARIAZI@xxxxxxxxxxx>
> > > To: <si-list@xxxxxxxxxxxxx>
> > > Cc: <ldsmith@xxxxxxxxxxxxxxxxxx>, <jrbarnes@xxxxxxxxx>
> > > Subject: Re: [SI-LIST] Re: Importance of Package Height
> > > Date: Sun, 17 Mar 2002 22:52:29 -0800
> > > MIME-Version: 1.0
> > > Content-Transfer-Encoding: 7bit
> > > X-Priority: 3
> > > X-MSMail-Priority: Normal
> > > X-Mimeole: Produced By Microsoft MimeOLE V5.50.4133.2400
> > >
> > > Larry Smith Wrote:
> > >
> > > > >
> > > > I'm not sure what assumptions Kemet is making for
> > > > inductance in their calculator. Their inductance seems way high to
> > > > me. We have measured a lot of caps from a half dozen vendors and Kemet
> > > > is no different than any of the others. When mounted on a low
> > > > inductance structure, the Kemet cap will behave as if it is
> > > > significantly less than 1 nH.
> > > >
> > > >
> > >
> > > Larry,
> > >
> > > Thank you for an instructive reply.
> > >
> > > The AVX SpiCap program may output considerably
> > > smaller ESL for similar capacitor than the kemet Spice Simulator.
> > > As an example, for the ceramic capacitor we recently discussed
> > > the AVX answer is 1.0 nH compared to kemet's 1.94 nH.
> > > Sometimes, use of Kemet software is preferable because it can show
> > > variations of ESL with capacitance values (for fixed package
> > > size) while the AVX calculator can not.
> > >
> > > I think (please correct me if I am wrong) the 1.0 nH by AVX and
> > > the Kemet's 1.94 nH (and all other ESL results by these two
> > > programs) apply to un-mounted capacitors. Therefore, based
> > > on your expected 600 pH for that 100 nF 0805 X7R cap
> > > both of these manufacturer calculators are providing ESL values
> > > which are too large/inaccurate in this case.
> > >
> > > Best Regards,
> > >
> > > Abe Riazi
> > > ServerWorks
> > >
> > > P.S. I also compared the AVX 1.0 nH value to the following measurement
> > > result earlier reported by John Barnes:
> > >
> > >
> > > Nominal
> > > Case Capacitance WV ESL
> > > Size Dielectric (pF) (V) Manufacturer Partnum (pH)
> > > ------ ---------- ----------- ---- ---------------------------- --------
> > >
> > > 0805 X7R 100000 25 AVX 08053C104JAT2A 900
> > >
> > > and found them reasonably close.
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> >
> > ------------------------------------------------------------------
> > 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:
