Nice project and thanks for the link. If you need help with calibration of your 34401A, feel free to contact me.

— xDevs.com on 06.04.2020 at 07:44 (UTC)

Nice project. Have you considered making and selling more of them - I'd be interested in having one.

If not, perhaps I'll build one myself. I often sell (more copies of) things I've built on Tindie. Perhaps there'd be a market for more of these there...

— Paul "LeoNerd" Evans on 16.04.2020 at 14:56 (UTC)

@Paul Currently I'm not planning to make these for sale. If you want to, feel free to use & sell the design!

— Petteri Aimonen on 16.04.2020 at 15:36 (UTC)

Constructive criticism follows.  This is nice work and I can tell you value precision- The current sink circuit as designed will not be very accurate.  You must have gotten lucky with your tolerance stackup because I would expect an error of about .5% if you made a few dozen.  Don't forget the collector current is not the same as the emitter current in a bipolar (even for a follower driven by a MOSFET follower).  Luckily the bipolar transistor you chose has a very high beta 200-500, this makes the error on the collector contribution 1 part in 200-500, about .25%.  In order for the FET base driver to develop .7V across the Vbe of bipolar, it will take 150 uA.  The transistor will take 40 uA (at best beta).  These these currents get summed again at the emitter of the bipolar back to 190 uA which the FET leg will draw from the output.  The collector however is generating beta  x Ib and the emitter is sensing beta+1  x Ib. The FET isn't really doing you and good and is just creating an extra 150 uA of common mode error which gets mostly cancelled out but is subject to CMRR of the amp.  Just use the FET as the sinking element and you should get what you expect.  At 20 mA, a small signal mosfet won't have a problem.  The Wison current mirror solves a similar problem in a very clever way- take a look at it.  Good luck- enjoyed your Owon DMM post.

— John M Wettroth on 15.07.2021 at 00:08 (UTC)

On second thought- the emitter current is accounted for after all and Vbe variation will also get cancelled.  The circuit is akin to a Wilson mirror. You'd still be better off with a FET for output transistor- the common mode is still there.

— John M Wettroth on 15.07.2021 at 00:17 (UTC)

@John M Wettroth

I agree that the NPN is quite unnecessary and just a MOSFET should be enough. In Hackaday comments it was pointed out that too much amplification there could lead to oscillation, though I didn't see any.

However I can't quite follow your concerns. All of the DC accuracy should come from the feedback loop through U2 and the R3 shunt resistor. With the 140 dB open-loop gain of MCP6V51, the 0.7V extra voltage required to turn on Q2 gives just 70 nV error, which is insignificant compared to the input offset of the opamp.

— Petteri Aimonen on 15.07.2021 at 05:43 (UTC)