Re: [SI-LIST] : Medium range capacitors

About this list Date view Thread view Subject view Author view

From: Larry Smith ([email protected])
Date: Wed Mar 08 2000 - 08:59:46 PST


I have seen several references to this alleged "Larry Miller at Sun."
I think you probably mean Larry Smith. Ray Anderson, Tanmoy Roy and
myself have worked hard to develop a decoupling capacitor methodology.
It has been published in the IEEE Transactions on Advanced Packaging in
August 1999 and is available at:

        http://www.qsl.net/wb6tpu/si_documents/docs.html

(Sorry for the multiple plugs on this list..) The method involves the
use of many capacitor values with controlled inductance and ESR to meet
a target impedance over a broad frequency range. Steve has a good
summary of it below. We find that this is the most effective way to
meet the power needs of modern CMOS technology that demands low
impedance over a broad frequency range.

best regards,
Larry Smith
Sun Microsystems

> X-Sender: [email protected]
> Date: Tue, 07 Mar 2000 19:21:50 -0800
> To: [email protected]
> From: sweir <[email protected]>
> Subject: Re: [SI-LIST] : Medium range capacitors
> Mime-Version: 1.0
>
> Shayle,
>
> Sure, if you use enough of them, 100nF caps work ok to high freq, but they
> are acting as parallel inductors at high freq. The papers delivered by
> Larry Miller and co. of SUN show that for demanding applications, careful
> choice of a wide range of capacitor values provide a lower impedance at a
> lower component count than is possible either with many 100nF or similar,
> or capacitors on decade or century boundaries.
>
> The mid frequency performance is just a matter of impedance
> calculation. From 1KHz up, the 47uF capacitor starts life at approximately
> 3.5 ohms, and will bottom out at the combined ESR, ESL, which depending on
> the part will vary from the 10's to 100's of milliohms. RS-232 currents
> depend on the voltage swing. The typical today is +/-5V or less. That
> will yield about 100mA max on a mark-space excursion. If your noise budget
> is say 100mV, then one line requires a bus impedance of an ohm or
> less. How far down the spectra extends depends to some extent on how long
> the RS-232 lines are. Even if you have 100' of cable, the current will
> drop-off within one round trip at about 500nS, so even the low frequency
> stuff is still in the hundreds of KHz. After that, it is literally the
> difference in DC load, and RS-232 lightly loads, so this is very small.
>
> Over the space of a 3 inch board a wave will propagate across the whole
> board in less than 1 nS. Bulk capacitors respond on the order of 100's of
> nS, so distribution doesn't buy anything. The ceramic capacitors and the
> planes are doing all the work. If I were your customer, I would pay a lot
> more attention to the PLD than the 232 drivers.
>
>
> Regards,
>
>
> Steve.
>
>
> At 08:44 PM 3/7/00 +0000, you wrote:
> >Thanks, Steve, for your answers.
> >
> >I have used 0.1 uF successfully in speeds to 66 MHz with fast edge rate
> >logic (terminated when needed) and I suspect smaller values won't really
> >help with higher frequencies unless a smaller, or less inductive, package
> >is used instead of the 0805 of the 0.1 uF caps.
> >
> >I think you hit upon the crux of my client's concern, which is the voltage
> >regulator's output impedence at the mid range frequencies of the RS-232
> >transmissions. You mention that the bulk capacitor at the entry point of
> >the power input would help with this. I think my client would then ask
> >about that cap's influence elsewhere in the board, which is why he
> >mentioned "sprinkling" a few around.
> >
> >But my real question is, aren't we interested in edge rates (edge
> >transition times), covered by the 0.1 uF decoupling caps, or do the slower
> >(assume periodic) frequencies of the RS-232 transmissions have an affect
> >which is different than and in addition to edge rate EMI and which needs
> >mid-range filtering such as the 6.8 uF caps (and I assume at those
> >frequencies the extra inductance is not a factor)?
> >
> >I welcome more input from Steve and others on this issue of mid range
> >capacitors and how they are effective.
> >
> >Thanks,
> >Shayle
> >
> >
> >
> >
> >
> >At 05:42 PM 3/7/00 -0800, you wrote:
> > >If you are sadistic, they will drive your purchasing crazy, as tant's are
> > >hard to come by these days.
> > >
> > >The PLD edge rates need to be watched. You can optionally filter the
> > >outputs depending on the application, with anything from 10 ohm series
> > >resistors, to RC networks, should the edge rates pose a noise issue.
> > >
> > >The tantalum caps will filter only up to a couple of MHz. So the question
> > >that arises is, how much filtering do you need from at least a couple of
> > >KHz where your power-supply regulator impedance goes up to a couple of
> > >MHz? If you are really only running some RS-232 then the one 47 uF cap s/b
> > >plenty. The 100nF caps will go inductive around 8-15MHz ( mounted ). So,
> > >you will either need lots of them in parallel, or other smaller values to
> > >cover at least up to 80MHz. You can refer to recent postings on this
> > >thread for more details.
> > >
> > >Regards,
> > >
> > >
> > >Steve.
> > >At 06:20 PM 3/7/00 +0000, you wrote:
> > >>Hi,
> > >>
> > >>I have been asked to sprinkle 6.8 uF tantalum or electrolytic capacitors
> > >>evenly around the PC board of a new design.
> > >>
> > >>As usual, I have the 0.1 uF 805 ceramics at the power pins of chips and a
> > >>bulk storage cap. of 47 uF where power enters. The board is 3x4 inches,
has
> > >>a few chips, slow edge rates. Mainly RS-232 (up to 1 mb/s) and the
> > >>internals of a PLD clocked at 16 MHz.
> > >>
> > >>Can anyone tell me what the benefit of these 6.8 tantalums sprinkled
around
> > >>would be?
> > >>
> > >>Thanks,
> > >>Shayle
> > >>
> > >>
> > >>**** To unsubscribe from si-list or si-list-digest: send e-mail to
> > >>[email protected]. In the BODY of message put: UNSUBSCRIBE
> > >>si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
> > >>si-list archives are accessible at http://www.qsl.net/wb6tpu
> > >>****
> > >
> > >
> > >**** To unsubscribe from si-list or si-list-digest: send e-mail to
> >[email protected]. In the BODY of message put: UNSUBSCRIBE
> >si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
> > >si-list archives are accessible at http://www.qsl.net/wb6tpu
> > >****
> > >
> > >
> >
> >**** To unsubscribe from si-list or si-list-digest: send e-mail to
> >[email protected]. In the BODY of message put: UNSUBSCRIBE
> >si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
> >si-list archives are accessible at http://www.qsl.net/wb6tpu
> >****
>
>
> **** To unsubscribe from si-list or si-list-digest: send e-mail to
[email protected]. In the BODY of message put: UNSUBSCRIBE si-list or
UNSUBSCRIBE si-list-digest, for more help, put HELP.
> si-list archives are accessible at http://www.qsl.net/wb6tpu
> ****
>

**** To unsubscribe from si-list or si-list-digest: send e-mail to [email protected]. In the BODY of message put: UNSUBSCRIBE si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Thu Apr 20 2000 - 11:35:28 PDT