Floating digital pot query.

You don't really care what voltage the pot is at other than the DC imposed on the two caps, right? Maybe just add a single high value resistor across one of the caps??? That will put all the DC on the other cap. Add a resi stor across each cap and the DC voltage will be split between them.

What sort of cap are you using? Do you care if there is much DC on them?

Rick C.

I think having both ends fully floating is unwise as you then have no way of keeping it from drifting towards one or other of the supply rails

- the chip esd protection diodes will then start conducting and distorting.

piglet

Yes, high value Rs across each C is a better idea, thanks.

The original circuit used a single cap in series with the digipot, but it's in a differential amplifier circuit and noise from the pot's supply was thus coupling unevenly into the two halves of the diff amp and being amplified, hence two Cs of twice the value to balance things up, which reduces the noise to about a tenth. That's quite good enough.

Cheers

That's what worried me, though it does work well. Rick C has come up with a better solution than mine.

Cheers

If the dpot is not bipolar voltage rated (most aren't) and you AC couple into both ends, the DC voltage is undefined. As you get AC signal, you can touch the ESD diodes, positive or negative rail sort of randomly, and the DC voltage will sort of self-define. There could be some transient distortion there, and a slim chance of a latchup hazard from big signal transients.

It would be prudent to apply some mid-rail voltage through a high-value resistor, but with small AC signals, probably not necessary.

tages.

s the

igh

ing.

Isn't the current causing this bias very, very small? Would something like a 10 MOhm resistor to ground on one side of the pot be a problem? Sure, t wo resistors makes it more balanced, but wouldn't an adequately large value of a single resistor do the job? How much current are you expecting?

Rick C.

Yes a single resistor would almost certainly do, but they're cheap and there's something satisfying about symmetry.

Cheers

I sort of feel that the leakage from the few hundred FET switches will keep the resistor array somewhere between the digipot rails, the signal is small so it doesn't have to be perfectly central. The capacitors are

I just measured the voltage on the digipot resistor array, and it is about half supply. I think I'll go with Rick C's suggestion for the next board spin (there are a few other minor issues too), but stick with letting the array float for any retrofitting, as the extra resistors would be a bit hard to fit neatly.

Cheers

No, a c*ck up. It measures roughly zero volts between array and either

Cheers

So there is just some low-leakage ESD diodes. You could just let things float.

Yeah, it should work. One possible issue is that the switch resistance will depend on the bias level, so it won't be as stable in the rheostat configuration if you let it float around like that. (Of course cheap dpots aren't very stable used that way in any case.)

Cheers

Phil Hobbs

The on-resistance of the FET switches likely depends on the bias voltage

- like any analog switch. If he cares much about the exact value of the "wiper" resistance, (i.e. using it as a variable resistor rather than as a potentiometer with no wiper current), then maybe letting it float to a different voltage after it has been adjusted might not be ideal.