I am currently working on a modular power supply/front end that will allow me to get use it as an adjustable constant current source/constant voltage supply with current limiting general purpose lab supply. I settled for the ATX supply not because I really need oodles of power (proper technical term here) but I want to hook a couple of those front end boards to a single supply.
Those front ends will be connected to one front panel control board so I don't need two expensive ten turn pots per output channel.
In order to set my current limit/output current and output voltage I would prefer a DAC over simple PWM control. I'm currently looking at the STM32F100 value line devices as those are pretty cheap and they include a 12-bit ADC and DAC. They will communicate with the front panel controller over I2C.
Throwing a 32-bit ARM at this task might look ridiculous at first but the truth is it's cheaper than having a PIC and an external DAC, plus I get 12-bit resolution also for the ADC.
Unfortunately the low pin count LQFP-48 package of the STM32F100s don't have separate voltage reference inputs, so I would like to use the controller's supply voltage as the reference voltage for ADC and DAC.
The devices can't operate down to 2.048V so I chose 3.072V as the next base 2 divisible reference voltage which should give me a resolution of 750µV/bit.
And now for the actual question: how should I go about supplying 3.072V to the microcontroller? I have thought about three options:
1. Using an adjustable voltage reference (output current might be too low to run the microcontroller
2. Whipping up my own regulator with a voltage reference, an op amp, a trim pot and an NPN transistor and adjust the pot until the output reaches 3.072V
3. Using a bog standard LM317 two resistors, a trim pot and some caps to generate the reference voltage
I don't know if option 1 is viable so I currently tend towards using option 3 because based on my experience with building Dave's constant current source it might be hard to get the loop stability for option 2 right.
The only reason I think of why the LM317 might be problematic is that the LM317's internal voltage reference might be complete crap, but on the other hand the temp coefficient of the trim pot probably would be much more of a concern.
Any advice that some more experienced people might have?
Cheers,
Elia