Beginner level: DIY an accurate picoammeter (measure picoamps!)

[Kleinstein]

The polarity change is a good idea as this would an offset at the current sink. The TIA type input can have an offset voltage, so that even when connecting a resistor from ground to the input one would read a significant current. This is one downside of the TIA type current measurement: zero is with a open input not with a short or resistor to ground.

A similar, but usually smaller offset can be there with the µcurrent: the OPs offset (is small) would give a non zero reading even with no current flowing.

With both polarities one can see the offset. Using the different from + 1nA to -1 nA would eliminate the offset.
So it would be 2 single pole 2 trough switches at the supplies to change the polarity with an off state (also isolates the battery).

The possible offset is a reason for favoring a large resistance. The OPs offset can be in the 1 mV range, possibly even larger if there is no adjustment of the offset. Even than the offset would be stable only to a few 10 µV. So with a 1 mV source voltage one can get avoidable errors.

[NeverDie]

OK, that makes sense.  It also explains why the uCurrent Gold (as compared to the picoammeter) seemed to have such a large offset in the earlier measurements.

[NeverDie]

Seems as though what's needed isn't a polarity reversal but instead a zeroing offset, perhaps driving the GND lower, such as what the picoammeter does in its circuitry.  Am I wrong?  I mean that way the test can continue as planned, whereas if the polarity is reversed, wouldn't you need some other calibration current to still test the accuracy of either the picoammeter or the uCurrent Gold?

Edit:  For instance, this may be crude, but it would seem to do just that:
 voltage_divider_v004.pdf (9.58 kB - downloaded 75 times.)

In this case, two AA's (the 3v source in the schematic) would zero the uCurrent Gold or the picoammeter by means of the 25-turn 100K trim pot, and then the regular testing could proceed as planned.

[Kleinstein]

The pA meter with the TIA should have some offset adjustment. Depending on the OP use this could be the adjustment pins at the OP, or just some offset added to the non inverting input.

Independent from that the polarity reversal still is useful for both the µCurrent and pA meter test. The low current meters should all read fine in both directions, so no other additions needed. The reversed polarity allows to test the scale factor even if there is some residual offset. So it can make the test of the scale factor more accurate. It is quite normal to test DMMs or similar with both polarities.

[NeverDie]

Here is the revision with the on-off switch and the polarity-change switch.  Also, the second stage is reduced to 100x intead of 1000x voltage division so that the output resistor can be 10M.

 voltage_divider_v005.pdf (12.08 kB - downloaded 86 times.)

[NeverDie]

A 10M ohm resistor should arrive in the mail today.  Once I build the device shown in the latest schematic (^^^above), what kind of reverse polarity measurements should I try to make?  Also, I'm still hazy as to what to do with reverse polarity data once it's collected.

Edit:  Here's the fancy 10-turn 1K potentiometer that arrived yesterday:

I thought that it would "click" with  each tick of the dial, but, alas, it doesn't.  I'll have to make due with visual sighting only.  At least it does have a turn-counter for keeping track of where it is in the 10-turns.  On the back it says it is a BOURNS 3590S-2-102L and on the bottom it says 1907M.  It says it has 5% tolerance and plus or minus 0.25% linearity, so it's probably the weak link in this setup.  Ah, well, that's all amazon had to offer in this category.  For anything better I'd have to look elsewhere and wait longer.  The next step up would be 3% tolerance and 0.1% linearity for a 3549S-1AA-102B.  I don't see anything better than 3% tolerance in the mouser catalog, so for anything better I'd probably have to go the resistor box route, which maybe isn't a bad idea anyway if the aim is to get single digit picoamp accuracy.  OK, then, I'll be switching over to that with something that will look approximately like:

[Kleinstein]

I see a limited need to fine trim the actual current. The point is more like measuring the voltage, either at the 10 mV level or the 1 V / 100 mV level (and assume the divider is accurate). Just to check the scale factor the 3 point of -1 nA , 0 and +1 nA are good enough. The 0 point is not even needed, more like a check for linearity / offset.

[SilverSolder]

The "trick" with these kinds of tests is,  you don't bother with super precise resistors...  you just measure the voltage and adjust the pot while looking at your DMM, until it is "just right".

Of course the resistors have to be "good enough" that they don't drift ridiculously, but that's unlikely to be an issue here.

[NeverDie]

Yeah, that's what I realized after my first schematic, where I tried to come up with precise resistance values.  It might have worked well enough, except that the parallel resistance threw in a wrench and created oddball equivalent resistance values.  With the latest schematic, though, which uses an abundance of 25 turn-trim pots and divides voltages by using fairly high resistance values, I expect the voltage division can be dialed in much more precisely than the original approach ever could have, and with less expense too.  Ironic, isn't it?  I'm actually looking forward to dialing in those resistor ratios and seeing just how close I can make it.  Kleinstein is probably right that I don't need this many trim pots.  However, I figure having one trim-pot for every digit will make the dial-in as accurate as I reasonably can without going to too extreme.  I'll know one way or the other after I attempt the dial-in.

[SilverSolder]

If you like tweaking and tuning, knock yourself out!  -  it is an enjoyable art...

[NeverDie]

And when things don't go perfectly, it's an irritating art:  when set to zero, the resistor box has 340 milliohm resistance that will creep into every measurement. 

Not really that big a deal, but I see no way to subtract it, except by post processing after measurement.  Maybe a negative offset voltage would do it?  Bah, it's not worth the bother, as we don't yet know for sure whether the noise can be suppressed enough that accurate single digit picoamp measurements are even possible.  I'll just run with it.

Edit: Aha!  I'll diminish that significance of the 340 milliohm by making the first stage voltage divider use higher resistances.  That oughta fix it.  Well, at least a lot of it.

[SilverSolder]

And when things don't go perfectly, it's an irritating art:  when set to zero, the resistor box has 340 milliohm resistance that will creep into every measurement. 
(Attachment Link)
Not really that big a deal, but I see no way to subtract it, except by post processing after measurement.  Maybe a negative offset voltage would do it?  Bah, it's not worth the bother, as we don't yet know for sure whether the noise can be suppressed enough that accurate single digit picoamp measurements are even possible.  I'll just run with it.

Edit: Aha!  I'll diminish that significance of the 340 milliohm by making the first stage voltage divider use higher resistances.  That oughta fix it.  Well, at least a lot of it.

Don't forget, you actually just care about the voltage coming out (feeding into the big resistor), you don't care what the nominal resistance is...

If you put 5V across your resistor divider, the output voltage at the "tap" will probably be reasonably accurate.

[NeverDie]

And when things don't go perfectly, it's an irritating art:  when set to zero, the resistor box has 340 milliohm resistance that will creep into every measurement. 
(Attachment Link)
Not really that big a deal, but I see no way to subtract it, except by post processing after measurement.  Maybe a negative offset voltage would do it?  Bah, it's not worth the bother, as we don't yet know for sure whether the noise can be suppressed enough that accurate single digit picoamp measurements are even possible.  I'll just run with it.

Edit: Aha!  I'll diminish that significance of the 340 milliohm by making the first stage voltage divider use higher resistances.  That oughta fix it.  Well, at least a lot of it.

Don't forget, you actually just care about the voltage coming out (feeding into the big resistor), you don't care what the nominal resistance is...

If you put 5V across your resistor divider, the output voltage at the "tap" will probably be reasonably accurate.

Nah, I've already setup for just a single AA battery and drilled the holes for all the other modules accordingly, so for now ixnay on the 5v.  I'lll just stick to the build plan.  I can always circle back and change it later if we collectively decide it would be worthwhile to do so, after the next data collection.  Anticipating that such things might happen, I'm making the build modular.

[NeverDie]

I like your idea though!

[SilverSolder]

1.5V will work just as well -  it was just a figure I plucked out of thin air! 

It will be interesting to see what turns up in the wash.

[NeverDie]

Doh!     These 7 decade resistor boards aren't designed to work like potentiometers.  They have no "wiper."  It would require two of them to do the job.  There do exist decade potentiometers, but in blinding hindsight this little resistor board is not one of those.  Such a board would be a handy thing to have, though.  I wonder if there are published projects which do exactly that, complete with downloadable gerber files.....

So.... back to the 10-turn potentiometer, which I expect will be "good enough" to discover whether we have a shot at single-digit pa measurements or not.  i.e. if it turns out there's too much noise even with the lunchbox test enclosure, then it won't have mattered anyway.  On the other hand, if it turns out we can either suppress, eliminate, or compensate for noise enough to discriminate single digit picoamps, then I'll circle back to improve the accuracy.  Until we know for sure one way or the other, though, pursuing it further now might be just wasted effort.  Hopefully this next set of measurements will reveal whether we've already reached the practical limit or whether this is just the start of the next level.

[SilverSolder]

The 10 turn pot is great.  You can always put a resistor in series with it (on "top" of it) if the voltage going in is too high to divide down accurately.

I believe Conrad Hoffmann of this parish published a Kelvin-Varley divider that looked like the board you got (same construction principles with the jumpers) in his Mini Metrology Lab project.

[NeverDie]

The 10 turn pot is great.  You can always put a resistor in series with it (on "top" of it) if the voltage going in is too high to divide down accurately.

I believe Conrad Hoffmann of this parish published a Kelvin-Varley divider that looked like the board you got (same construction principles with the jumpers) in his Mini Metrology Lab project.

Great find!  You're right: on its face, it looks like a Kelvin-Varley circuit might be the missing puzzle piece for the voltage divider that we're building.  The way Dave Jones describes it at time index 26:54:



the Kelvin-Varley circuit is notionally similar to the decade resistor board, but it also has virtual "wipers" that function nearly equivalent to the wiper of a potentiometer.

Bobby Dazzler!

[SilverSolder]

You may want to take a look at this article as well, very interesting stuff for DIY accurate voltage division:

http://conradhoffman.com/HamonResistor.html

[Gyro]

You may want to take a look at this article as well, very interesting stuff for DIY accurate voltage division:

http://conradhoffman.com/HamonResistor.html

Here's a thread on one I built...  https://www.eevblog.com/forum/metrology/anyone-else-built-a-hamon-divider/

Dr Frank has built a very nice one (linked in the thread). There are several other hits if you do a forum search for Hamon divider too.

[NeverDie]

What do you typically use it for (I mean aside from the obvious of using it to divide voltages).  i.e. For what applications do you need a standalone precision divider like that?

[SilverSolder]

If you are aiming for super high precision, adding more pots and resistors to improve resolution doesn't really help - things will never get more accurate than the DMM you are using to measure the output.

With a Hamon divider, you can make something that is better than your DMM (almost no matter how good it is!)

Typically you might use a high precision divider to calibrate and verify the operation of other instruments.  Or, in this case, to accurately divide down a higher voltage (that your DMM can measure reasonably accurately) to a very low voltage (that is difficult for a DMM to measure accurately).

[Gyro]

What do you typically use it for (I mean aside from the obvious of using it to divide voltages).  i.e. For what applications do you need a standalone precision divider like that?

I use it for meter calibration. The key thing about a Hamon divider (unlike other dividers) is that it is effectively self-calibrating. When you set the jumpers for divide by 2, you simply compare the voltages between Input Hi - Output Hi and Output Hi - Output Lo and make sure they are identical (which is easy, to whatever resolution instrument you have).

When you switch to Divide by 10 mode, you have confidence that it is an accurate divide ratio and can then use it with a voltage reference to calibrate two decades of your voltmeter. If you only have an accurate 10V reference for example (as my situation) you can calibrate the 10V range and then use the divider to calibrate the 1V range, based on the 10V reading. With a 2 decade divider, you can get even better certainty and range, by being able to calibrate the 100V range while cross checking it with your calibrated 10V range and freshly calibrated 1V range. In this way, you can leapfrog ranges and verify / calibrate them from a single known reference.

The limitation is the input resistance of the meter, it cannot significantly load the output of the divider. In my case, my Datron meters have 10G input resistance up to 20V so there is no significant loading of the 900R effective output resistance of my 10k divider. I can calibrate the 100V range with reference to the 10V one - because I can use an ordinary 100V PSU to drive both divider and meter - the meter's 10M input resistance on the 100V range has no consequence. With the 2 decade divider that I have, I can verify the 100V range and the 1V range against the 10V reference... or the 1V range and the 100mV range. Once I have confidence in the 1V range and the 100mV ranges, I can then step down again with an arbitrary PSU (eg 1.5V battery) to calibrate the 10mV range against the 1V and 100mV ranges.

There was recently an in-depth discussion on the effect of the divide by 2 / divide by 10 switch contact resistance on the certainty of the divide ratio (talking in very low ppm terms here). I used a terminal strip and hard screw-down links to minimise contact resistance on my 10k decade (see photo in the thread I linked).

Here's the contact resistance discussion:  https://www.eevblog.com/forum/metrology/absolute-divider-concept/msg2924250/

Edit: Err, sorry, no this one:  https://www.eevblog.com/forum/metrology/influence-of-switch-resistance-in-hamon-dividers/

[magic]

The limitation is the input resistance of the meter, it cannot significantly load the output of the divider.

IIRC there is no precision penalty besides the obvious "input current of the meter times output impedance of the divider". Both are easy to calculate and the effect can be accounted for.

But I also built mine with 1k5 resistors to avoid hassle

[SilverSolder]

The limitation is the input resistance of the meter, it cannot significantly load the output of the divider.

IIRC there is no precision penalty besides the obvious "input current of the meter times output impedance of the divider". Both are easy to calculate and the effect can be accounted for.

But I also built mine with 1k5 resistors to avoid hassle

Doesn't that just move the hassle somewhere else?  (I.e. now your reference has to supply a higher current and may lose precision instead?)