Thursday, February 09, 2006

Doff Your Cap

I write only because some random Explorer tab was open to Blogspot and I was really tired of combing through Digikey to find the absolute cheapest caps for a given value. I'm working on finding a cheap, minimal decoupling network for our chip to go on our customer application board (imagine it as an eval board wandering around the lab in nothing but its socks). This thing has nothing on it; we're using the background capacitance of the board for power decoupling. It has to be small, cheap and easy... and it has to work right the first time.

I'm excited. I get to fill my engineering notebook with TABLES upon TABLES of hand calculations and notes about package geometries. I will, hell or high water, find a way to estimate the ESR of these Panasonic-ECG caps, despite the fact that they don't provide curves or values in their datasheet.

5 comments:

Anonymous said...

You know...

I have some ECGs. You can measure 'em if you'd like.

"Hey baby, wanna measure my cap?"

-rsw

w0z said...

Panasonic-ECG seems to have the cheapest caps on Digikey. So, I poked around the ins0rw3b and it tells me that Chinese foundries are churning out more 0603s than anything else (although market share for 0402s is on the rise). Rationale achieved, I just spec'd the cheapest 0603s on Digikey for our "widely accepted" range of power decoupling caps (all ECGs, unsurprising) and then pulled the equivalent cap curves out of AVX's modelling software (spicap3).

Why doesn't anyone give approximate equivalent series inductance for their various cap packages; isn't it primarily dependent upon the package aspect ratio?

All this is back-of-the-hand in any case. Our part works (with, admittedly, a solid power supply) on a board with nothing but 30nF of background sandwich cap. Our logic? LVDS -- the only transient current draw is if one side of the output diff pair is faster than the other (in which case, intra-pair skew on the signal line is going to hurt you more than supply compression). Our digital block? A mere 75MHz. A frequency at which any set of our decoupling caps will have an effective impedance of less than one ohm.

Anonymous said...

Hey, if it's 0603s that you want, I measured about ten metric assloads of those for my thesis.

...but don't take it from me. The Fonseca 10 year Port has got me rolling tonight.

Oh yeah, also my share of the bottle of sake I split earlier tonight. That helped.

-rsw

w0z said...

I went to the MFA yesterday. There was a wry comment (on the placard for one of the Japanese screens) that the artist was known to be an individualist and it was rumored that he would work under the influence of sake.

Some day, someone's going to say that about your circuit designs, Wahby.

And, dude. I soldered 0402s and SOT-363s on New Years Eve. Admittedly, before I started drinking. You've got my empathy.

If you still have plots for those 0603s kicking around, send them my way. I'd love to have real data. Otherwise I'm going to just estimate the package ESL (around .6nH, eh?) and assume the minimum ESR for NP0 and X7R is going to be less than .7Ohm. It's ballpark. Good enough for Taiwan.

w0z said...

oh, I lied that panasonic ECGs were the cheapest caps around. mark found some from this random manufacturer (i have it in email) who's low-low prices can only be explained if they use horsemeat for their dielectric material.

Choice!