[SI-LIST] Re: S11 or S21
- From: Thomas McGonigle <temcgonigle@xxxxxxxxx>
- To: weirsi@xxxxxxxxxx, zasio@xxxxxxxxxxxxxxxxxxx
- Date: Tue, 31 May 2005 12:39:41 -0700 (PDT)
touchy, touchy, touchy. Come on kids, play nice.
--- steve weir <weirsi@xxxxxxxxxx> wrote:
> John, since you assert that you do not believe my conclusions, then surely
> you must be able to articulate the exception that you take to my methods of
> reaching my conclusions, as I have of yours.
>
> Despite my best efforts to correlate: theory, field solver results, and
> careful experiment design, and to subject the entire body to expert peer
> review, you categorically discount my results relying on your own, very
> different experiment design. What I have yet to hear is any technical
> argument against my methods, or evaluations. Have you reproduced my
> experiments with different results? Do you see some flaw in my theory,
> experiment design, or data extraction? If so, I would greatly appreciate
> hearing about it. If anyone can show me wrong, and it has certainly been
> known to happen, I am happy for the education.
>
> If you believe that I have misinterpreted your experiments and data
> extraction and believe that my criticisms of them are not well founded,
> then everyone will be well served by you detailing my errors. I believe I
> have been very clear expressing the problems that I find with your methods
> and how they account for the differences in our results. By my estimation
> your experiments are under reporting the inductance of all three
> configurations tested ( 0603 2 via, 0603 4 via and AVX 0612 IDC ). Most
> troubling is that I believe you under report the 4 via 0603 by almost 2:1
> for the long via cases. This completely turns your conclusions upside
> down. So do you care to show why I am off-base? How do you account for
> any of the troubling issues in your reported data that I have
> outlined? How do you account for the sources of errors in your experiments
> that I have detailed?
>
> Steve.
> At 10:59 AM 5/31/2005 -0700, John Zasio wrote:
> >Steve,
> >I am not embarrassed at all. I have presented data measured on
> >small test boards and used this data to design systems. The
> >data from single capacitor measurement was used to design the
> >impedance profile for many large high power boards. It has been
> >verified with measurements of the impedance profile on these
> >boards and more importantly on the power plane noise on
> >functioning hardware. I have never had a single instance of a
> >system that did not function because of power supply noise or
> >signal integrity issues.
> >
> >This forum is to allow experienced engineers like you and I
> >to offer advice to those who have less experience. I feel like it
> >is pay back to all those that have helped me. My advice is
> >to not trust manufacturers datasheets for all the parameters
> >needed to build high speed systems. Build test structures to
> >create the models necessary to do your design.
> >
> >The manufacturers of capacitors generally present accurate data
> >on the equivalent series inductance. However they are giving you
> >the difference between a short across their test fixture and the
> >inductance with the capacitor replacing the short. This is of
> >little interest to me. What I care about is how the capacitor
> >functions on PC boards with the same stackup that I must use for
> >the product. The via inductance on typical boards is much higher
> >than the manufacturer's value for the capacitor and this board via
> >inductance dominates.
> >
> >I have chosen not to use the X2Y capacitors on my boards. For
> >me they do not meet my requirements as well as 0402 devices.
> >I admit that it takes 50% more devices but they are smaller than
> >the X2Y, cost less, and use same total number of vias. You
> >present an opinion that the X2Y is the right choice and only
> >choice. I present an opinion that there are other choices and I
> >think the people asking for advice should understand that there
> >is more than one solution.
> >
> >John Zasio
> >
> >steve weir wrote:
> >
> > > John, as we discussed, there are both problems with your measurement
> > > methods and your conclusions that are evident based on your writings.
> > >
> > > Let's take a look at your reported data points as reported in Lee's
> > > and your book:
> > >
> > > I have formatted the following table using spaces so just view in a
> > > monospace font like courier new to get the columns to line up:
> > >
> > > Table values reported from "Right the First Time" Ritchey w/ Zasio pp
> > > 143 Table 35.1:
> > > =====================================
> > > Vendor AVX AVX AVX |
> > > Case 0603 0603 0612 IDC |
> > > Cap 100nF 100nF 100nF |
> > > Vias 2 4 8 |
> > > Cesl 950pH 460pH 151pH |
> > > | Extracted
> > > Via length ESL ESL ESL | L/mil L/mil L/mil
> > > 13.5 1190 580 230 |
> > > 28.8 1460 720 310 | 17.6 9.2 5.2
> > > 62.5 2060 1010 500 | 17.8 8.6 5.6
> > > 77.5 2330 1150 580 | 18.0 9.3 5.3
> > > |Ave 17.8 9.0 5.4
> > > |*vias 35.6 36.0 43.2
> > > =================================================================
> > >
> >
> > >
> > >
> > > From this data, you offered the conclusion:
> > >
> > > "The Length Factor (Lf) of 35.5 pH/mil works for all 0603 size
> > > capacitors with either two or four vias."
> > >
> > > In reference to the equation:
> > >
> > > Cesl = ESLcap + Via_length * Length_Factor / Number_of_vias
> > >
> > > In other words, vias do not interact.
> > >
> > > Your first clue that something is wrong should have been the notion
> > > that four vias mounted on the corners of a 50mil / side box should
> > > exhibit 1/2 the inductance of just one pair on one side ( pg 142,
> > > fig.35.5 ). Solenoids still work even if loosely wound.
> > >
> > > Your second clue that there is a problem with your beliefs, data or
> > > both should have been that inductance of the interdigitated vias on
> > > AVX IDC went up instead of down compared to conventional capacitors,
> > > without any further explanation.
> > >
> > > Your third clue should have been that your extracted inductance per
> > > unit length does not match published and verified derivations,
> > > including those readily available as published by Dr. Johnson, which
> > > are themselves readily verifiable using a physics text such as
> > > Halliday and Resnick to apply the method of Biot and Savart. You
> > > report 35.5pH/mil for 10mil via pairs on 50 mil centers. The correct
> > > value should be about 46.8pH, over 30% higher. How do you reconcile
> > > that your vias should be better than theory?
> > >
> > > Your fourth clue should have been that your extraction for the IDC raw
> > > component inductance came out 150pH, when those parts as reported by
> > > both AVX and independent measurement show up as very close to 60pH.
> > >
> > > So how did your experiments go bad?
> > >
> > > First, your fixture, an approx. 4" x 10" board sets the RF source and
> > > the receiver close together at the far end of a narrow aspect ratio
> > > board, and the target capacitor positions essentially along the length
> > > of that board. This minimizes the incremental transfer impedance
> > > changes for the various capacitor positions that you used on
> > > approximate 2.5 inch spacings.
> > >
> > > Second, the tests that you did use only a single capacitor at a time
> > > on that platform with 3mil dielectric plane pairs. The impedance of
> > > the mounted capacitors is so high compared to the capacitors that you
> > > erroneously concluded that they contribute a negligible amount of
> > > impedance. This is true for the tests, but is false for any system
> > > requiring anything but an late 1980's early 1990s impedance profile.
> > >
> > > Third, the test fixture attempts to derive inductance values for
> > > attachments to planes at various depths, but where apparently, all the
> > > planes have been joined together at the RF generator / spectrum
> > > analyzer connections in the upper left hand corner of the board.
> > > Since there is no pictorial evidence, nor any comment about the issue,
> > > I doubt that you did anything to either insure that the instrument
> > > connections to each plane formed contiguous transmission lines, nor
> > > found a way to deembed the parasitics of those attachments. This
> > > means that your measurements were colored in two important ways:
> > >
> > > 1. The parasitic inductance of the attachments increased the apparent
> > > insertion loss above the actual values, and tended to equalize all of
> > > the measurements.
> > > 2. The measurements reflected the distance from the uppermost plane in
> > > the entire board to the uppermost plane that any given capacitor
> > > attached to.
> > >
> > > Your experiments measured lots of combined effects that you failed to
> > > account for, leading you to very faulty conclusions. For unknown
> > > reasons you discounted, or ignored the prominent warnings readily
> > > visible in your data. I hope for your sake and the sake of customers
> > > of you or Lee that your faulty beliefs and conclusions have not been
> > > relied upon.
> > >
> > > I am sorry if all this is embarrassing to you in a public forum, but
> > > as you know, I have offered multiple times to show both you and Lee
> > > how you have gone wrong privately.
> > >
> > > Steve.
> > >
> > > At 09:11 AM 5/28/2005 -0700, John Zasio wrote:
> > >
> > >> Steve,
> > >>
> > >> I have come to the opposite conclusion than you in regard to the
> > >> benefit of
> > >> IDC or X2Y capacitors for decoupling. In order to obtain the data for my
> > >> conclusions I build small multi-layer test boards and measure several
> > >> caps
> > >> with varying via length to the power planes. What is important to me
> > >> is the
> > >> total effective inductance of the mounted capacitor.
> > >>
> > >> The most recent test board has 26 layers and is 100 mils thick. I
> mounted
> > >> 0402 capacitors with 4 vias and X2Y capacitors with 6 vias. Power plane
> > >> one (PP1) is closest to the surface of the board and PP6 is near the
> > >> bottom.
> > >> Measured ESL is as follows:
> > >>
>
=== message truncated ===
Thomas E. McGonigle
8305 48th Avenue
College Park, MD 20740-2401
tmcgonigle@xxxxxxxx
301-441-1175
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
------------------------------------------------------------------
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 FAQ wiki page is located at:
http://si-list.org/wiki/wiki.pl?Si-List_FAQ
List technical documents are available at:
http://www.si-list.org
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:
