On 08:28 AM 5/06/2003, John Sheahan said:
> <snip..>
> Value   Size  Impedance
> 103      0603    ~1 ohm
> 103     0805    ~0.5 ohm
> 103     1206    ~0.3 ohm
> 104     0805    ~1 ohm
> 104     1206    ~1 ohm
> So if you spec a 10nF 0603 you have a resistor, not a decoupler, at 100
> MHz.

probably only a NPO lower than 1n and the interplane capacitance is
working here.  The 100n/10n argument is null as we are essentially
at/past series resonance.   Ovviously here we have an L or an R , not
a C. But we do have a low value L or R so we are still decoupling,
- the point of the dcoupler is to attenuate noise. C is just the way
we often tend to do this.
A sprinkling of 10-100pF value caps appeal to some in the >100Mhz range.

I beg to disagree a little - the interplane cap maybe, but insofar as the bulk C's ... I am not looking at theoretical impedance (1/jwC) rather the published impedance curves from the manufacturer. Operating past resonance is nothing special, and not necessarily a problem. All you need to worry about is minimising the AC impedance between the power nets over a suitably broad range of frequencies and to an adequately low level - both application dependent. Operating past resonance simply means the impedance is inductive and rising with freq - so what, as long as it is low enough, that is generally all you need to know.

(One big issue here is production spread of course - self resonance varies widely from device to device. If your circuit relies on self resonance to null a specific noise spike, and you are using high-Q caps to do this, you may have problems in production. I can imagine what the production team would think about trimmable decoupling caps. :-)

An NPO is only available in small cap values and so will be operating as a pretty good cap at low freqs and hence fairly ineffective. At 30 MHz the Kemet 1nF 1206 NPO is speced at typically 2 to 3 ohms. Even if you had 1nF of high Q interplane capacitance (which you won't - see next para) you would still only have an AC impedance of 1.6 ohms. 0.2A switching current -> 0.3 V ripple. I would therefore look at the published impedance curves and find the most appropriate set of caps that supplement the plane capacitance at the dominant freqs of interest. Forget the cap value - look at the impedance curve - for high speed decoupling situations this is all that matters.

1nF of *high-Q* interplane capacitance! Not likely, even if the parallel plate capacitance equation gave that much in practice the resistance and inductance of the plane copper, and very uneven current distribution will mean that much of this theoretical capacitance is out-of-circuit for a particular component supply pin. At higher frequencies you also have to consider the effects of propagation delay - some of the capacitance is too far away for the electrons to whizz over and smooth the voltage ripple due to the current spike - 1ns is about 180 cm assuming a 0.6 velocity factor (conservative as the Er is likely to be such that the velocity factor is lower (1/sqrt(Er)). Even if I had superconducting power planes I would still have an effective area that is dependent on the speed of the transistions.

> Note also that the 100n 0805 has roughly *twice* the impedance @ 100 MHz
> than the cap *one tenth* the value in the same pkg!  In this case, if you
> are operating above 30 MHz the 1206 10n wins, followed closely by the 0805
> 10n. 100n in any pkg and 10n in 0603 have about twice the impedance.

here you would do better with 1n NPO however..

In a 1206 though, then you potentially get into the issue of longer tracks to the component pad so more inductance, and the 1206 1n NPO only performs better at 100MHz and a small range either side - due to the high Q the resonance is sharp. So you have poor decoupling at 30 MHz - hence the need to parallel a big and sloppy with a sharp and quick. Due to the series elements (look at the cap model) the interactions between caps is not all that much, when one is looking like a high impedance the other is becoming a low impedance - a high impedance in parallel with a low is a low.. Take the published cap models and do some SPICE sims - it is easy to see the results. Better still use real caps and a spec-an/tracking generator or VNA.

The above is my, possibly flawed analysis...I suspect that is enough from me on this,

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
* To leave this list visit:
* http://www.techservinc.com/protelusers/leave.html
* Contact the list manager:
* Forum Guidelines Rules:
* http://www.techservinc.com/protelusers/forumrules.html
* Browse or Search previous postings:
* http://www.mail-archive.com/[EMAIL PROTECTED]
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Reply via email to