Precision(ish) 1A current source

[max_torque]

If i wanted to build a programmable current source, that could provide up to 1 amp of current, what practical level of precision could be achieved without going too mad with specs and design?

Assume i use a 16b DAC to provide the voltage reference to the closed loop current controller, that's a theoretical maximum resolution of just about 15uA, although in reality, i doubt all 16b could be sensible realised?

Then, the control loop would need to amplify the voltage across a sense resistor, which i guess could be a relatively large value ( i don't need a particularly low output impedance, as the current is just being run through a shunt to calibrate that shunt).  Obviously, there will be some offset and gain errors here too

Finally, the unit will need to operate at a sensible range of room temperatures, although i can include temperature measurement and hence correction in the firmware


What have people achieved with similar designs?

[nctnico]

With high stability resistors and a good reference you might achieve 100ppm (0.01% / 100uA). I'd go for a good >18bit DAC though from TI or AD.

[cellularmitosis]

This is exactly the project I'm working on right now!  (Actually, it will be two sources, used to characterize a transistor: it will try a bunch of different combos of base current and collector current and spit out a table).

I made a programmable LM317 a few years ago, so I'll be reusing the digital side of that (ATTiny85 + MCP4801)

I'll keep you posted!

[cellularmitosis]

Forgot the link: https://github.com/pepaslabs/ProgrammableLM317/blob/master/README.md

Schematic: https://github.com/pepaslabs/ProgrammableLM317/blob/master/releases/v2/ProgrammableLM317_schematic.pdf

[David Hess]

The thing that always limits precision on these is the self heating and temperature coefficient of the current shunt resistor.  So power derate the current shunt resistor and use the lowest value of resistance consistent with the offset error and noise of the error amplifier.

[tszaboo]

I was working with similar circuits. The design goal was 0.03%. For that, you need bulk metal foil resistors (or similar), or temperature control, reference voltage which is very stable, auto-zero amps, and 24 bit ADCs, closed loop, auto calibrated. It is not trivial, but all parts are off the shelf. And you need a better multimeter, because a  simple 6.5 digit one is just not good enough.

[max_torque]

The thing that always limits precision on these is the self heating and temperature coefficient of the current shunt resistor.  So power derate the current shunt resistor and use the lowest value of resistance consistent with the offset error and noise of the error amplifier.

I was considering putting the shunt resistor on/in a temp controlled section of pcb, with a heater, and closed loop controlling that temp to say 50degC, for consistency?  In either case i will at the very least measure the temp of the shunt resistor (as closely as possible) and correct the DAC output for this offset.

[Hensingler]

as the current is just being run through a shunt to calibrate that shunt).

Then you don't need a precision current source just a stable enough one which most bench power supplies will be. Measure the current with a reference shunt or a meter.

[tszaboo]

The thing that always limits precision on these is the self heating and temperature coefficient of the current shunt resistor.  So power derate the current shunt resistor and use the lowest value of resistance consistent with the offset error and noise of the error amplifier.

I was considering putting the shunt resistor on/in a temp controlled section of pcb, with a heater, and closed loop controlling that temp to say 50degC, for consistency?  In either case i will at the very least measure the temp of the shunt resistor (as closely as possible) and correct the DAC output for this offset.

These shunts have significant self heating, so that is not really working like that.
Also, do a search for these resistors first. It is not trivial. You need at least 100mV to get any reasonable accuracy, so the shunt has to be in the 0.1 Ohm-0.5Ohm range, so you end up with 0.1-0.5W self heating.  That is about 15 degrees for a 5W resistor.