[sdiy] integrator / capacitor leakage

[Ian Fritz]

Hi Rene --

At 10:37 AM 12/3/05, René Schmitz wrote:

>>Yes, with care you can measure leakage that way.  But input bias
>>current of the amp can be tricky to sort out, especially if it is
>>highly temperature sensitive. Rather than use the intended opamp, I
>>would suggest using the *best* opamp you can get your hands on.  Also
>>the test setup must be carefully insulated from leakage to the
>>outside world.  I recommend building the circuit in air using teflon
>>standoffs.
>
>A good idea, if you wanted to characterise the leakage resistance of the
>cap. My suggestion to use the intended OP was made since I assumed it
>would need to be good enough for this anyway. Maybe a dead-bug
>construction with air wireing would be even better, you don't need
>standoffs then. But I guess you can already get a result on a
>breadboard, if its sufficiently clean.

Agreed.

Another method is just to bend the pin 2 up and solder the cap directly to 
it (in feedback mode).  That way you don't have to worry about leakage 
around your V source (switches or whatever), and you can build the test 
setup on perfboard or  whiteboard.


>One could also try to decouple the cap via a switch, and let it
>discharge on its own for some time, then the leakage from the OP doesn't
>play such a big role.

Ah, clever idea!


>Leakage per se isn't relevant here anyway, its the resulting time
>constants that are, because those will add to the integrating times.

Right.  But nonlinear errors will distort the waveform a bit.


>You'll find that with small caps the major contribution comes from the
>OP, but with large caps this more and more gets into the background,
>since the timeconstants will increase.

Oh yeah.  So parallel a number of the cap you want to characterize!


>But the question was if the leakage resistance is nonlinear, something
>which I don't know off hand (but doubt it for the +-15V we usually
>have). And this you can find out.

If the leakage mechanism is a microscopic material parameter, then it is 
the E field that matters rather than the voltage.  So maybe think about 
using an overrated cap, if nonlinearity is a problem?

   Ian