I am experiencing some difficulty getting the current sense working as desired.

Originally, I had it after the regulator (the theory being that it would be more accurate that way, since all the current which was going to the load would have to go through the current sense resistor). It worked, but there was voltage drop at high currents due to the resistor. When pulling 5A, there was a 0.5V drop, which is unacceptable for me. (I am not too worried voltage drops on load, but 0.5V is too much. 1% (so a 0.1V drop at 10V) would be fine by me. (For the record, the LM317 has nominal 0.5% load regulation).

I then decided to move it before the regulator. Since the voltage is regulated after the current sense, the regulator should be able to handle all sorts of loads (as mentioned above, the 317 has decent load regulation). After wiring it all up, I tried it.... and it didn't work.

It turns out that the problem is due to the common mode voltage on my op amps. When the current sense is before the regulator, it is running at about 0.7V lower than my rail voltages (+/- 16V, so the voltage reading on the high side would be about 15.3V). My op amps are not rail to rail, and so cannot output this much. (I am using a three-op-amp instrumentation amplifier configuration, so the first amp buffers the input voltage, but the most it can output is about 14.5V).

If I switch to the simpler single op-amp differential configuration with a gain of 10x, I can just barely squeak by (there are implied voltage dividers in this configuration, so it drops the voltage down a bit before the op amp inputs). Perhaps if I switched to a 5x gain it may be better, albeit a bit less precise. (There are other benefits to doing it this way, including reducing the number of op amp ICs I need for each channel).

Of course, another option is to use a dedicated current sense IC... for instance I already have some INA169s lying around. While tempting, they do not work on negative voltages, so would pose a restriction on one of the major goals of this project (i.e. having the same board used for positive and negative channels). Or, if not a current sense IC, a differential amp with high common mode voltage. I am not sure that I want to go that route, though... I am trying to use generic components as much as possible here, at least for each channel (the control board is a bit more specialized, but such is life).

So at this point I am a bit lost. From what I can see, the only reasonable option is to use a single op amp as a differential amplifier, with a reduced 5x gain. I suppose that using a three-amp instrumentation amp would not really buy me much, since the values I am measuring are very low impedance anyway (it's a freakin' power supply!) I'm still a bit annoyed that I can't get it to work, even though I know exactly what is wrong and why it is failing.

Oh well. Onward and upward...