GigaOhm measurement 1

[bubulindo]

Hello,

i'm building a GigaOhm measuring device.

I'm making a voltage divider with the GigaOhm resistance and a 68MOhm resistance and i apply that signal to a non-inverting amplifier with a gain of 40.

My problem is that my signal takes too long to settle. How can i make it faster??



Best regards, Carlos Cardoso

 

[IRstuff]

What do you consider too long? Your circuit is completely dominated by your high resistance and input/node capacitance. It still shouldn't be much more than milliseconds.

If you want lower than that, you'll need fF capacitance and VERY careful board and wiring layout. AND you might need to precharge, although that'll add capacitance as well



TTFN

 

[bubulindo]

about 12 seconds.

I'm doing my prototype in Vero Board, i've already lowered my voltage divider resistance to 10 MOhm's but it is still too slow.

What did you ment by fF capacitance?? Does it mean add a capacitor in my circuit??

Precharge is out of the question because my microcontroller will only work for 3 maybe 4 seconds and then will go to sleep. So i cannot precharge it from time to time.



Best regards, Carlos Cardoso

 

[IRstuff]

but you still have a "gigaohm" resistance in your voltage divider.

But, there still seems to be a disconnect somewhere. The implied load capacitance of the divider is something like 3 nF, which is absurdly high. What are you trying to drive with your amplifier? Some amplifier configurations will reflect their load capacitance into their inputs, so if your amp is driving 120 nF, that might account for it.

Your voltage divider needs to see not more than maybe 10pF. This would require minimizing the node capacitance using a JFET input op amp, short traces on the board, etc. You might look at reconfiguring your amplifier configuration to isolate its load capacitance from the input.

TTFN

 

[zeitghost]

Testing PAD5 (5pA diodes) used to take about 30 seconds to settle, so 12s sounds reasonable to me.

Can't remember what the node capacitance was in the tester.

rgds
Zeit.

 

[bubulindo]

At the exit of my amp-op (i'm using OPA602) i have a zener diode, an RC filter ( 10kOhm and 0,47 uF) and an Analog to digital converter!

Best regards, Carlos Cardoso

 

[logbook]

I bet your gigohm resistance isn’t a resistor, but is insulation from something you are testing. Dielectrics can take a long time to settle when you try to measure their resistance. I have tested relay coil-to-contact isolations above 1E9 ohms and they take at least as long as 12 seconds to settle, and probably longer.

 

[dspDad]

The time constant of 30 seconds would point to something other than the op-amp. As noted by IRStuff, the time constant for a 68 mOhm resistance with a 10 pF input capacitance would be about a millisecond.

Even if you waited 5 or 10 time constants to settle, this is pretty short.

Does the 68 mOhm terminate on ground or on some floating point. Especially if this is battery operated, you might be seeing a power supply feedback component.

Are switching the power to the op-amp on and off? There may be some other compensation as simple as bypass caps that don't settle. And if the 68 mOhm terminates on anything other than ground then switching the power would be a problem for more than one reason.

What is the reference to the A/D. Are you measuring this drift with the A/D or with a scope?

Does the circuit have a single r/c approach to its final value, or does it simply take a long time to get to within the required tolerance? The 0.47 cap on the output, if electrolytic, could have enough memory to keep the output from settling for quite a while.

Consider using an inverting amplifier referenced to ground. This eliminates voltage changes at the input and so the input capacitance can be ignored. If you need a reference voltage, use a voltage reference chip that settles in milliseconds after power up.

These are just some ideas. Usually you will find what the real problem is in the process of eliminating potential problems.

DspDad

 

[BrianR]

At g-ohm impedance levels veroboard will have unacceptable leakage and this will vary with humidity and contamination. Consider mounting the hi-Z nodes on teflon insulated standoff pins.

Regarding the long stab time, you might get some useful ideas by looking at the technique of 'shield driving' used on hi-Z instrumentation amplifiers to effectively cancel shield capacitance.

 

[gfzoom]

Measuring GOhms is not easy! I'm an engineer at Keithley Instruments, where we make instruments that measure up to TOhms, among other things. To do this be prepared for the heavy use of Teflon, guard traces, plastic gloves, MOSFET input opamps. The OPA602 is a JFET input with up to 10pA of bias current; if you want any kind of accuracy, this won't do. There are also issues with dielectric absorption (which I'm sure the vector board has pleanty of).

Try connecting the measurement node to the op-amp input in the air, with none of the connections touching the vector board.