Picoammeter Design

[David Hess]

Isn't the diode / diode connected jfet / transistor there to provide a wide range logarithmic scale?

Yes, but a bipolar transistor in the translinear configuration like the first half of the second circuit works over a wider range than a diode, which is why it is preferred.

[magic]

Not sure about 1/f, but I think wideband noise shouldn't be a problem in this circuit. In a bipolar transistor it is equivalent to the Johnson noise of a resistor conducting the same current at about 13mV, so with 13pA current wideband noise is the same as in 1GΩ resistor.

Thermal mismatch contributes some 7% output current error per °C. Error due to Q1B base voltage drift is similarly 7% per 2mV.

I would be more worried about leakage currents in the transistor (I suppose keeping base and collector voltages equal may help here) and base current errors, because β falls off at low currents in many transistors. Maybe it would make sense to diode-strap Q1A and Q1B, using a transimpedance amplifier to set the desired Q1B base voltage and simultaneously measure its current.

[pinout666]

Thanks for your comments, i search a little more on the net ^^ (maybe i buy 3 lmc662 and 3 opa 928 and test )

For very large resistor, mouser is a big place :

https://www.mouser.fr/c/passive-components/resistors/film-resistors/?resistance=1%20TOhms ~~50$ for 1Tohm

https://www.mouser.fr/ProductDetail/SEI-Stackpole/HVCB2512MDM500G?qs=IPgv5n7u5QZbu51hWHgoJw%3D%3D
 ~~15$ for 500Gohm

https://www.mouser.fr/ProductDetail/Vishay-Techno/TR10X100GKMEB?qs=sGAEpiMZZMtG0KNrPCHnjZX4yx0cXmo5eF2svPguICk%3D
~~ 18$ for 100Gohm


EDIT
maybe more idea here :
https://techlib.com/science/ionpage2.html
https://groups.google.com/g/sci.electronics.design/c/5nqKFRv_HnI?pli=1
https://www.electronicdesign.com/technologies/test-measurement/article/21766108/whats-all-this-femtoampere-stuff-anyhow

[David Hess]

I ended up going with the 1 GOhm, 1%, 25 PPM/C part from Mouser, so the output will be 1 millivolt per picoamp which is plenty.

https://www.mouser.com/ProductDetail/588-SM102031007FE

The OPA928 was interesting until I saw the price.  In the past I have used the LMC6081 which is a more precise version of the LMC662.  This time I might use the OPA140 simply because I have it on hand.

I need to test a bunch of parts for rebuilding old multimeters.

[David Hess]

I had another idea which I do not see covered in any of the discussion.   What about using a charge pump current-to-frequency converter?  Then the current in picoamps can be read out on a frequency counter.  Instead of a high value resistance, a low value of capacitance is used to pump charge.

[Kleinstein]

There is an alternative the resistor: one can integrate the input current and thus get the current as C*dU/dt. This works especially good for very low currents, as a small capacitor (e.g. 10 pF) is easy compared to a large resistor in the 100 Gohm range.
One still needs some form of reset or discharge. This could be from a charge pump, reset or maybe photodiode. For the charge pump one still needs the switches with very low leakage. It has some benefits, but is also more effort than the simple resistor in FB. 1 GOhm is still relatively easy to get and can be reasonable stable.

[David Hess]

There is an alternative the resistor: one can integrate the input current and thus get the current as C*dU/dt. This works especially good for very low currents, as a small capacitor (e.g. 10 pF) is easy compared to a large resistor in the 100 Gohm range.
One still needs some form of reset or discharge. This could be from a charge pump, reset or maybe photodiode. For the charge pump one still needs the switches with very low leakage. It has some benefits, but is also more effort than the simple resistor in FB. 1 GOhm is still relatively easy to get and can be reasonable stable.

I will have to do some math to check feasibility.  I am thinking in terms of integrator, followed by simple voltage-to-frequency converter, followed by charge pump.  Diodes can be used for the charge pump switching.

[Smokey]

Just noticed the Keithley 6485 has a normal BNC as the input connector and is still specced to 10fA resolution on a 2nA scale.  I figured that was triax territory.  How do they do that with just a normal BNC and no guard?

http://uhv.cheme.cmu.edu/manuals/picoammeter-6485.pdf

[David Hess]

Just noticed the Keithley 6485 has a normal BNC as the input connector and is still specced to 10fA resolution on a 2nA scale.  I figured that was triax territory.  How do they do that with just a normal BNC and no guard?

http://uhv.cheme.cmu.edu/manuals/picoammeter-6485.pdf

I would assume that the difference is that Keithley models which use a triaxial connector, like the Keithley 617, have 100 times the resolution extending to 100 attoamps.  I doubt the leakage of the connector itself using a Teflon insulator is limiting performance.

My hacked together picoammeter achieves a resolution of 10 femptoamps limited only by external noise because of lack of a fully shielded enclosure.  Shown below is a BC337 with 50 femtoamps of leakage at 20 volts.

[Hydron]

Feedback ammeter should have ~0 potential between the inner and outer co-ax connections - essentially creates it's own guard hence being able to get by with just a BNC. On the triax instruments the outer connection is just PE for noise shielding (which is probably needed for getting into atto-ampere territory tbf).

[Kleinstein]

The triax is not so much needed for a current measurement. The main point is with voltage measurements that the electrometers also offer.

For the simple test system a closed case would be a good idea. I would consider a case not only for the pA meter, but also for parts to test. So a bit like the pA meter with an attached test chamber.

[Terry Bites]

Who can read through all the replies?

Its not easy even with electrometer grade opamps and full pcb voodoo. Strays will kill you. That's why the real thing costs an arm and a leg.
Guard traces are very necessary. Seriously pricey ultra low leakage expensive relays are needed for range switching.
Guarding and (active) shielding will be the make or break of any design.
That also applies to the measument set-up. Triax is a must.
Input protection circuits leak to much unless you apply some bootstapping arrangement.
Even cleaning the PCB is a major challenge!

See this missive from Bob Pease. https://www.electronicdesign.com/technologies/test-measurement/article/21766108/whats-all-this-femtoampere-stuff-anyhow

My advice, dont try to build on unless you absolutely have to!

[BD1QMP]

some X-ray dosemeter used this methode. such as Solidose 300.   see below picture
AD744 with 220pf capacitor and LM393 to generate V-F. then MCU to read the frequency.

and PTW electrometer K1s  (for ion chamber dose meter)
LMC6082+ capacitor, LM311   

I had another idea which I do not see covered in any of the discussion.   What about using a charge pump current-to-frequency converter?  Then the current in picoamps can be read out on a frequency counter.  Instead of a high value resistance, a low value of capacitance is used to pump charge.

[BD1QMP]

Today, the cheapest design for picoAmper meter is DDC112+Arduino.   DDC112 was used in old CT das board. The DDC112 is an integrating current input ADC. You can use it for acquiring small currents (of the range of femto to picoamps (or uA)。 second hand market has a lot of DDC112, there is total new chip with only $20 cents.

below is a DIY project of DDC112 picoAmper meter (there are some bugs in the code)
https://41j.com/blog/2018/08/ddc112-board-design-bring-up/   

[exe]

There is an alternative the resistor: one can integrate the input current and thus get the current as C*dU/dt. This works especially good for very low currents, as a small capacitor (e.g. 10 pF) is easy compared to a large resistor in the 100 Gohm range.
One still needs some form of reset or discharge. This could be from a charge pump, reset or maybe photodiode.

I think AoE or elsewhere another approach was mentioned: to reverse polarity of opamp's supply pins (I presume with some current limit). This way the capacitor will be drained through ESD diodes. This sounds fun, so I might try this...

[David Hess]

There is an alternative the resistor: one can integrate the input current and thus get the current as C*dU/dt. This works especially good for very low currents, as a small capacitor (e.g. 10 pF) is easy compared to a large resistor in the 100 Gohm range.
One still needs some form of reset or discharge. This could be from a charge pump, reset or maybe photodiode.

I think AoE or elsewhere another approach was mentioned: to reverse polarity of opamp's supply pins (I presume with some current limit). This way the capacitor will be drained through ESD diodes. This sounds fun, so I might try this...

Robert Pease wrote some articles about how difficult it is to reset a low capacitance integrator.  I think he ended up using a custom made switch with small pieces of copper tape and Teflon.  Too much Teflon makes things worse.  Electronic switching or charge balancing is difficult because of excess charge injection.

Initially he used an air capacitor for low leakage, but the physical volume was so large that it was detecting cosmic ray strikes and they got better results with a physically smaller Teflon capacitor.

I do not know what he did in the production design for testing LMC6001 operational amplifiers at 25 femtoamps.

[exe]

After reading this thread I decided to replace resistor in my picoammeter. It was 100M, and now it's 10G. I confirm that noise went down by a factor of ~71x, it's now 13.2fA p-p (leads disconnected), as measured by ADC.

[exe]

Has anyone designed an integrating ammeter? I'd like to make one, but not sure how to properly reset the capacitor.

I want to try to short opamp pins to ground (slightly below that compensate for voltage drop of ESD diodes) to reset the capacitor, but I don't know if it's going to disturb the load. There are very few sources on the internet discussing this topic. They say that the load needs to be disconnected, but this would increase complexity and potentially increase leakage.

[Kleinstein]

I have seen a schematics from a kind of integrating pA meter as part of a mass spectrometer. They used diodes in a glass case as low current photodiodes to allow adding charge pulses (both signs) via LEDs. With likely fixed carge pules this would make it some kind of AD converter for pA range currents. I don't remember if there was and extra ADC to read the integrator, or just a comparator. For best resolution an ADC (like µC internal or even better) would probably be a good idea.

I don't see that one would need to disconnect the input. So keep the capacitor at the intgrator and add a switchable extra feedback path (essentially in parallel to the capacitor). For better isolation a simple, non critical CMOS SPDT switch can act as an extra series switch so that the low leakage switch sees essentially zero voltage.
For switches I see a few options: a mechanical relay, a 2N4117 JFET, the "photodiode" + LED option (pulsed or maybe linear) or low leakage diodes (e.g. BAV199) that are switched to essentially zero voltage for off. The diodes would need an extra amplifier to really get zero or just do a partial reset.
One may well get an extra charge pulse from switching off. So the reset would not be perfect, but still get a reasonable fixed starting charge close to zero (like 1 pC range).

[David Hess]

I have done it at the picoamp level.  I used the base-collector junction of a 2N3904 as a better than 1 picoamp leakage diode for the reset circuit.  I have since found BC327s which have a leakage of 80 femtoamps, but try a bunch of different small signal transistors.  Bootstrap the low leakage diode to reduce leakage even further.

Bob Pease had some success with glass reed relays, but needed electrostatic shielding and needed to minimize the voltage change across the coil to reduce charge pumping.

I suggest air wiring the sensitive node at that level of leakage.

[iMo]

Cannot the contacts of the reed relay be used as the capacitor itself??
It might be, say, a fraction of a pF?? Or perhaps 1-2pF??
Say 0.5pF capacitance and 10fA will charge the contacts in 50secs to 1V..
Then apply the mag field (with some elstatic shielding like a brass tube) and reset it..

[David Hess]

Cannot the contacts of the reed relay be used as the capacitor itself??
It might be, say, a fraction of a pF?? Or perhaps 1-2pF??
Say 0.5pF capacitance and 10fA will charge the contacts in 50secs to 1V..
Then apply the mag field (with some elstatic shielding like a brass tube) and reset it..

For a capacitor I would use a length of Teflon coaxial cable, which has the advantage of being self shielding when the shield is driven by the output of the transimpedance amplifier.

Bob Pease started out using an air capacitor made from sheets of copper clad board, but this will not work because the large volume will pick up too many cosmic ray strikes.

[exe]

About "photodiodes", do you think this can work? The idea is that the voltage difference between opamp's inputs is well below 1mV, so LEDs, in theory, wouldn't leak. There are two diodes, to charge capacitor in different polarities.

They do increase input capacitance, not sure how much this is a problem, considering low differential voltage.

[Kleinstein]

In principle the idea with the LEDs as photodiodes can work. The question is finding a LED (or other low leakage diode in glass case) with low capacitance. Even though operated at essentially zero voltage the leakage current of the diodes still matters. Related to the leakage current there is genration / recombination noise that causes some current noise  (a bit like the shot noise that would come with the leakage current).

[David Hess]

Also, not all diodes have low forward leakage.  At least some LEDs are suppose to be better about this however.

Trimming the input offset voltage of the operational amplifier should help, although not as much as you might expect since leakage is non-linear and can occur at low voltages across a diode.