Help with nanoamp meter

[Atomillo]

Hello:

First of all, this my first post on this forum, so if this is posted in the wrong spot, please tell and I will inmediately try and move it to the appropiate spot.

I tried to create a pico ammeter, similar to the ones already discussed in this forum, using a low leakage op amp and a high valued resistor in a TIA configuration. I used a LMC6082, a 1G 10% resistor and a 680pF polystyrine cap to avoid oscillations. I connected the power supplies and such in a protoboard, but lifted the input pin and soldered the resistor and capacitor in air. I then cleaned it all using 98 alcohol and dried it.

Because I dont have any of the specialized test equipment, I was unsure of how to test the thing. What I did at the end was to connect another 1G resistor to a variable power supply and to the input. To measure the current, I connected a Fluke 87V in the mV range in series. I then measured the output voltage, and estimated the gain in V/nA.

What surprised me the most is that the gain is not always the same! In fact, it always decreases with increasing current. I repeated the measurement several times and the result was always consistent. The left column is the input current, and the right one the measured gain in V/nA:

I(nA) G(V/nA)

3.5    0.9
4.4    0.83
5.6    0.8
6.6    0.78
8.7    0.76
10.8  0.73

What could cause this? The whole thing is also inside a cookie tin, and the noise floor is around 20mV).

Many thanks!

[RoGeorge]

You may want to attach a hand-drawn schematic and a picture of the physical setup.

[Atomillo]

This is the schematic, and as soon as I get home I will post pictures of the physical set up as well.
At first I suspected leakage, but you would expect the gain to go up with higher currents or remain constant, not actually go lower.

[exe]

Because I dont have any of the specialized test equipment, I was unsure of how to test the thing. What I did at the end was to connect another 1G resistor to a variable power supply and to the input. To measure the current, I connected a Fluke 87V in the mV range in series. I then measured the output voltage, and estimated the gain in V/nA.

I'm not confident this is an accurate measurement. I'd vary power supply voltage (in series with a resistor) and measured output voltage. This way we eliminate any potential non-linearity from Fluke.

[Gyro]

You may want to look at this thread regarding washing Polystyrene capacitors in IPA (alcohol) - It may take a few days for its leakage properties to return. The thread also contains a link to the the Picoammeter-design thread...

https://www.eevblog.com/forum/beginners/polystyrene-caps/

[Atomillo]

Exe:

I did another run measuring the voltages at the input and output. There is a still a trend towards lower gain at higher currents (calculated assuming that the input resistor was 1G) :

Input voltage   Output Voltage   Gain 1,3   1,075   0,826923077 2,09   1,643   0,786124402 3,7   2,675   0,722972973 4,66   3,3   0,708154506 5,87   4,04   0,688245315 6,85   4,7   0,686131387 7,34   5,037   0,686239782 9,3   6,5   0,698924731 11,72   8,13   0,693686007 12,43   8,69   0,699115044

[Atomillo]

Gyro:

That was the thread I was refering in my first post! The thing is that the circuit was soldered and cleaned yesterday, so I would expect the leakage to have gone down already.  I'm quite at a loss right now.

[exe]

Does the output drift? Esp. at higher input voltages.

[penfold]

Strange that its so non-linear, double check any voltage coefficient of resistance, I wouldn't typically imagine it to be significant at these voltages, but worth considering none the less

[Atomillo]

I think I've found the solution. In the attached photos I show my current set up. The 1G resistor connected with the yellow alligator clip is my input. Previously, I connected the other end using another alligator clip to the input pin of the op amp. I realized that whenever I moved this wire, the readings would change. Also, as exe suggests, sometimes at higher voltages the output would not be stable at all. Also, by letting it's insulation touch the black cable I use for ground the readings would change rapidly.

By connecting it directly this way, I'm getting much much better results:

Vin:   Vout:   Gain: 1,077   1,094   1,016 2,232   2,215   0,992 3,214   3,166   0,985 4,260   4,192   0,984 5,440   5,341   0,982 6,230   6,100   0,979 7,380   7,220   0,978 8,160   7,980   0,978 9,490   9,260   0,976 10,410   10,150   0,975

I feel like I start to start Dave "hold your tongue at the right angle". Next step is to get proper connectors, PCB and shielded box.

Thanks for all the help!!

[Kleinstein]

Many OPs don't like capacitive loading to the output. Directly driving an output from an OP is usually not that great - just a DMM or 2 m of coax max be enough capacitance to make some OPs oscillate.  So it may be a good idea to have something like 100 Ohms in series with the output - this usually solved the capacitive loading problem.

Edit:
One may have superimposed mains hum, and this may drive the OP into saturation at some times. This can cause a nonlinear relation.

[penfold]

Many OPs don't like capacitive loading to the output. Directly driving an output from an OP is usually not that great - just a DMM or 2 m of coax max be enough capacitance to make some OPs oscillate.  So it may be a good idea to have something like 100 Ohms in series with the output - this usually solved the capacitive loading problem.

+1 to that, also also noticing no supply decoupling, so that's another route for PSU noise and stability issues

[Atomillo]

Hi:

Adding a 100 ohm didn't change anything. Scoping the output didn't reveal any oscillation besides down in the 20mV range due to my poor shielding.

The results In showed before are reapetable to a good degree of accuracy. I'm starting to suspect that the only reliable way to test circuits in this current range is to solver everything in a shielded PCB.

EDIT: I said the measuremts showed before were not respetable. They ARE respetable. Sorry for any confusion cause by my poor use of the english language

[Atomillo]

+1 to that, also also noticing no supply decoupling, so that's another route for PSU noise and stability issues

Added a 10pF decoupling cap. No difference, because the supply I'm using is quite clean (noise in the 10mV range)

[penfold]

+1 to that, also also noticing no supply decoupling, so that's another route for PSU noise and stability issues

Added a 10pF decoupling cap. No difference, because the supply I'm using is quite clean (noise in the 10mV range)

Is 10pF a typo? Whilst its probably a little overkill I'd put 10uF or so where the power comes into the breadboard and 100nF as close as possible to the opamp's power pins.

Have you tried checking the output with an oscilloscope? as Kleinstein suggested, oscillation is possible... also, it would be wise to terminate the inputs of the spare amp in the package, it could be doing something unexpected and coupling to the amp under test via the PSU line

One note about the breadboard... you've clearly gone to some efforts to minimise its problems, kudos!

[Atomillo]

Oops I'm used to using small capacitors for high frequency decoupling sorry. Used a 10uF and 100nF as you described and grounded the unused inputs. It worked the same.

Probing the output, it is a clean DC signal. The only noise is down into the 20mV region, and is electromagnetic noise (I suspect coming from one of my lights) caused by my poor and improvised shielding.

Regarding the breaboard: before embarking in this project, I read a bit about low level measurements (Keithley handbook is a must) and drooled quite a bit at the sights of input stages of electrometers (I specially liked the 617). In my next order to Mouser I will buy some Teflon standoffs to make the final version tidier (along with 1G resistor with better precision than 10%...)

[exe]

Well, lets assume the noise is 20-30mV p-p. Looking at your table, in lower range I see offset at around ~20mV or so. Which might be because of noise rectification in IC. Of course, this is very vague calculation, but it seems that overall gain around ~0,975. It is higher at low voltages because of the noise.

Breadboard circuits don't always behave as expected due to parasitic inductances, noise pickup, leakage, etc. The clue is how wiggling wires affect output.

[penfold]

Regarding the breaboard: before embarking in this project, I read a bit about low level measurements (Keithley handbook is a must) and drooled quite a bit at the sights of input stages of electrometers (I specially liked the 617). In my next order to Mouser I will buy some Teflon standoffs to make the final version tidier (along with 1G resistor with better precision than 10%...)

Whilst breadboard is a terrible choice for anything more than a 555 timer, there shouldn't be any major problems with what you've done. That being said, it is far easier to debug something from pictures when its "solidly" built, there isn't anything wrong with the circuit per say, but there's a lot of unknown (remember there's also a black rubber adhesive strip connecting all those breadboard wire clip thingies that will have some resistance, its just very hard to judge from here how much that is and whether its affecting your result).

I would recommend building on fibreglass perf. board, its a little more expensive than the SRBP but it can be more "predictable". TE Connectivity HBA series resistors are good for high values.

[Kleinstein]

For a slow circuit a breadboard is not that bad for a first test. A main problem is quite some parasitic capacitance.
For shielding a circuit on a bread board, I like an upside down cooky box (metal). One has good access and when done putting the circuit together just cover with the shied (main part of the box) - thin cables can sitll go out to the side.

Decoupling is not so much about reducing the supply ripple, but to provide a low supply impedance at higher frequency. With a circuit that is essentially DC one would not see too high a supply impedance as ripple, but it can still make the circuit more prone to oscillation and RF pic-up. In a linear circuit it is more the GBW of the OPs that matters not the actual frequency used in the circuit.

It is only with a circuit that is actually switching and this has significant AC current to the supply that one would see high impedance as supply ripple.

[Terry Bites]

The dark forces of dielectric adsorption creeping up on you. See https://www.edn.com/wp-content/uploads/2016/11/1982-10-13_EDN_Pease-capacitor-soakage.pdf from Mr Pease. The input wiring may be charging and leaking in uncontrolled way.

Guarding will help kill stray input capacitance and leakage. Make sure you're breadboard is bone dry.

Give the op-amp some input protection before it gets zapped.

You can reconfigure the feedback (at the expense of some increased noise) With a tee network. This allows you to use less exotic resistors and lower the charging voltage across the feedback capacitor.

[Kleinstein]

Extra diode for input protection are not that practical for an pA meter. The more practical way is to increase the resistor just before the OPs input to some 100K to 1 M, so that the OP internal diodes can stand the current from a more usual ESD event.

2N4393 does not make a good low leakage diode - of at all a 2N4117 or a purpose made low leakage diode like BAV199.

[RoGeorge]

Could be the capacitor, but I would rather suspect the breadboard.

Before debugging, I would remove the breadboard completely.  Use a copper clad PCB instead, with islands to solder and the remaining copper grounded, all well cleaned before tests, preferably in an ultrasonic bath.

Put everything in a shielded box, or at least put the board inside a grounded metal food-cane while measuring.

[Atomillo]

Hello all:

I have great news!!! I did what you all told me and soldered the circuit in a small piece of perfboard (sadly I don't have copper clad) and shielded it using an upside down metal cookie box. After cleaning and drying the results are great! The output follows closely the input voltage with a small difference (as stated previously, the 1G resistors have quite an error margin) that seems to always be the same. I will wait until tomorrow to make sure the measurements are repeteable before reporting them, but so far (three measurements with around 1h in between) everything is looking golden.

Terry and Kleinstein:

This is just a prototype, in the final version I intended to use input protection. In the X Chapters, image 1x.124 there is a very interesting plot of leakage current of FETs and BJTs used as diodes (is it ok for me to post a photo of the graph? Am I breaking any copyright or rule?). What I didn't know was that the IC had protection diodes incorporated.... That certainly makes it easier to work with.

I also will look at the tee network. Also in the Art (that book sure is a gold mine!) they refer to it as the Resistance Multiplier (in page 259 of the third edition). What compelled me to not use it is the increased offset voltage. If I switch to an op amp that has external compensation pins available or as in the already linked pico ammeter thread add an offset correction, then I would consider it as a pretty good option for pA/nA level measurements. I would imagine that the increased noise make it unsuitable for fA measurements, although that level of precision is well beyond me! 

[RoGeorge]

Great!   

No copyright problem with posting just a photo of a page, and state the source and say all the copyrighted materials are not yours, and belonging to their legal owners.

[Atomillo]

I have good and bad news.

The bad news is that the circuit worked previously because I had not connected the capacitor. Thus, scoping the output showed oscillation at around 255Khz and 800mVpkpk.

Getting rather demoralized at the thought of complex things I have no idea about like dielectric absorption and such, I realized that my capacitor was a lot bigger than that used by Alan Yates and other (680pF).

 Connecting another capacitor in series (thus a total capacitance of around 340pF) immediately solved the issue, reducing the aforementioned oscillation to less than 50mV pk pk, and also gets me a perfect dc response. It seems like I had reduced the bandwidth way too much and made a poor integrator without realizing!

Oops! I will let this rest for the night and tomorrow I will make sure everything is right before closing this case.  In any case, many thanks to everyone!!!!