an IC we need

The Maxim thing? No, it's two pins. To set the capacitance to N (N=0 to 31) you...

Raise the EN pin

Pulse the DAT pin N times

Drop EN.

Good thing it's only 31 pulses max ; 65535 would get tiring.

John

But that's nonlinear, has limited range, has lots of distortion versus swing, is useless for real analog signal handling (say, +-5 volt signal swings) and has a really rotten TC.

Imagine all the slick things you could do with a good programmable capacitor!

John

I assume the tics from DAT drive a binary counter inside the beast. The block diagram shows 5 binary-weighted caps. SPI would make more sense for a part with more steps; I guess they just ran out of pins.

John

Monolithic Equal-value caps sounds like a good idea. Of course, then there's the issue of switch capacitance. These are pretty low-Q caps. Hmm, for a 13pF cap, a Q of 30 at 100MHz implies 4 ohms of FET series resistance. But that implies a rather large FET switch, which should lead to more than 6pF of parasitic capacitance... Hmm...

I've done that sort of thing to tune on-board-chip filters and oscillators.

As for the interface... isn't that your typical digital-dork SPI serial interface ?:-)

...Jim Thompson

All of those things are possible now: D/A driving a Vari-Cap diode. Two whole parts.

Aha! So you don't load a serial word like in SPI?

That's totally weird. That would require decoding counts into binary for the switches... unless there's actually 32 equal value caps ??

...Jim Thompson

Drive two identical varicaps with the same voltage, and measure one for the feedback... measuring the capacitance is probably not going to be fast enough for what you want, though.

I'd like to see a two part circuit with a D/A and a varicap that works up to

Since it's load-only and not-reporting, you should need only two pins for SPI ??

...Jim Thompson

I used a reference frequency plus a phase detector.

Started at one extreme and swept the tuning.

When phase flipped I stopped.

...Jim Thompson

Hello Jim,

It doesn't have to be weird. You could, for example, have a slow oscillator with one R, one C and a 1/6th of a Schmitt chip such as the HC14, then run it until something is "in tune", then stop the oscillator with another of the Schmitt gates plus a diode from its output to the oscillator gate input. In terms of simplicity that would be hard to beat.

I'd still use a varicap ;-)

Regards, Joerg

I'm just saying that there are plenty of real-world problems where your proposed architecture doesn't work yet other architectures do...

Yea, so would I. But lots of applications are in the millivolt level and even less. I rather think about just what *can* be done with the available technology - not just what *could* be done with hypothetical components. Hey, but thats just me :>

MOS capacitors are some of the best caps there are. Very low DA, low TC, very linear, superb on-chip ratio matching (which is why most fast ADCs now use capacitive ladders.) Their main disadvantage is low available capacitance and fairly low Q.

Vishay makes some surface-mount MOS caps that are fabulous; too bad they're impossible to solder.

John

Well! They are GLASS capacitors ;-)

The main use for a chip of this sort is auto-tuning.

...Jim Thompson

I was referring to these...

a few other people, Metelics and such, make MOS caps.

John