The Simpson 260 VOM requires a 50uA constant current source for calibration. So, I've been using that as a target for my circuits output. Here are a couple of JFET circuits I've been working with for the last few days. Adding the 2nd JFET "flattens" the output "slope" across varying loads but also adds instability to the output. I do realize that I am talking about being down in the nanoamp realm. Any tricks as to adding stability to these circuits? The circuits slowly (hours) drift ±1uA and rapidly (minutes) drift ±1-10nA. Touching/pinching the JFETs had no effect on drift so I don't think it is thermal drift in the transistors. I am using a decade resistance box as the "potentiometer" to set the current so I suspect the several resistors in the box may be the culprit. Or, is it simply "noise" from outside effects?

 

Don’t see how the first circuit works at all. There is no current flowing in the adjustable resistor, so Vgs is always zero.

 

Works fine, except for drift...

 

Your first circuit is incorrect.
Below is the sim of a single N-JFET circuit with a transistor model I have:
Don't know what the cause of you instability is other than thermal drift.



For a more stable current I suggest you use a stable voltage reference such as the TL431 in a constant current circuit (below)

 

I have a 1kΩ resistor as the load and the source pin tied directly to ground with the resistance decade box clipped to the source and gate legs of the transistor. Having photo problems ATM or I'd post a pic. Watching it, it has drifted from 50.001uA to 49.919uA currently. I would suspect thermal drift as I had problems with it using a current mirror circuit that uses three 2N3906 transistors and a 6.2Vz. Just touching any of the 3906s sent it flying. Not so with the single JFET here. Next up is back to the LM317L and I do have some TLV431 chips that I'll try as well. The three 3906s mirror circuit actually calls for a thermally compensated 6.4Vz which just came in the other day so will revisit it as well. Ammeter now back up to 49.956uA. Is it even possible to control this thing to within ~10nA span or just wishful thinking?

Edit: Just out of curiosity, I clipped the Decade Box directly into my bench DMM. It is solid as a rock on 4.4979kΩ. Hasn't budged even 1/10 of an ohm! And that is with the ceiling fan on...

Which led me to wonder just how solid the voltage is. The PSU says 6.00V but the DMM says it is 5.9932 or 3. So, flitting 100uV or so. Definitely need to look at a precision regulator!

 

Definitely need to look at a precision regulator!

Don't need that if you have a stiff constant-current circuit such as the TL431 design.
Notice that once the supply voltage gets above 5V, there's no appreciable change in the current from 5V to 10V (the sim shows <50nA or about 10nA/V)

 

Is it even possible to control this thing to within ~10nA span

That's 0.02% of full scale which is rather overkill for calibrating an analog meter with a 1.5% or worse accuracy.

 

No, not for calibration. When I did use the three 3906s mirror circuit for calibration it was get it on spec and use it before it drifted. Just experimenting with constant current using the 50uA as a target. It makes things "interesting". Here is the mirror circuit but the ground got clipped off. All I had at the time was a 6.2Vz 1N4735A without temp compensation but it worked fine except for the drift.

And yeah, the ole 260s specs at best are from 1.5%-4% depending on range being used. The resistors inside are all 1%.

 

Isn’t your two-JFET circuit what is called a “lambda diode”?
EDIT: no it is not. The LD requires n-chan and p-chan JFETS

 

Studied and worked a bit with the TL431A. Put @crutschow 's circuit together and waiting for it to fully settle but it's pretty rock solid. I used a multiturn 50k trimmer for the 49.9k resistor as I couldn't cherry pick anything close to that and still wanted to trim it exactly on 50uA. While composing this it has drifted only 5nA. Looks like it's quit drifting and gone up a nA. Best results so far and good enough to work with. Thanks!

I had these in stock but had never worked with them before. They are certainly going into my "toolbox". Neat component!

Yup, it's as I suspected. The 2N3904 is causing the drift but the TL431A must be slowing the drift to compensate?

 

The 2N3904 is causing the drift but the TL431A must be slowing the drift to compensate?

Basically.
It's the magic of negative feedback.
(Understanding negative feedback is key to the operation of many analog circuits).

The internal gain of the TL431 is high enough so that it adjusts its shunt current (and thus the current/voltage at the transistor base) to keep the emitter voltage (the TL431's reference node input) across the emitter resistor essentially constant at the TL431's 2.5V reference voltage, thus changes in the transistor gain or Vbe voltage have no significant effect on the constant-current.

 

I swapped out the 50k trimmer (too hard to trim) for a 5k trimmer+47k resistor and gave it a while to get to stable temp. It's now staying within a 1nA span. I can definitely settle for that. Now back to seeing what a LM317L will do but I'm not putting any bets down on it against this TL431A. Might be interesting to see what real difference the extra 0.5% tolerance of the TL431B makes but I'm satisfied with these results.

I added 5 TL431Bs to my Mouser cart for my next shipment.

 

Now back to seeing what a LM317L will do

It may have slightly more drift due to a poorer temperature coefficient, but should still be pretty good.

 

You might look at CLD's, "Current Limiting Diode". Its a two terminal device and works using the same principle as the JFET limiter.
In fact, I think you can buy a 50uA CLD.

 

Actually, it didn't work out very well. After doing the calculations and building the circuit I gave it 10Vs and got ~0.716mA out. Turned the voltage down and at 1.5Vs it was close enough to trim it to 50uA. Not at all stable. Constant drop and touching the LM317L made it climb rapidly. I even put a heat sink "hat" on the TO-92 to see if it would help stabilize and it didn't help much. Not at all stable. It may work better in the 10-100mA range but not sub 1mA. I ended up using R1=12K, R2=10K (20K was too high).

 

Current Limiting Diode

Sounds interesting and hadn't heard of them before. Looking at the PDFs for the 1N52xx and 1N53xx they run from nominal 0.22mA to 4.7mA if used simply as a current limiting device. Unless there is some other adjustable current ciruit method? No typicals seen in the PDF or found googling.

 

R1=12K, R2=10K (20K was too high)

Oops!
I forgot that the LM317 needs a minimum, worst-case load current of 10mA to properly operate (below), so can't be used for such a low constant current.

 

It may work better in the 10-100mA range but not sub 1mA.

LM317L aren't guaranteed to work at less than 10mA.

 

Yeah I saw that on the pdf and wasn't even sure it would actually get down in the uA range but it does. This is the TO-92 cased LM317L version that only can handle 100mA max. Just not very stable and with only 1.5V driving it. But I tried and did learn something from it... I did put a 47k resistor on the output as a load and the Vs had to be 3.8V to drive 50uA but still unstable.

 

It may work better in the 10-100mA range but not sub 1mA

As I mentioned before...