So, here is the new concept for current sensing. The regular Allegro Hall-effect sensors only go as high as 200A, which doesn't look like it will provide enough range for high-power setups. There are ways to extend that by setting up a current divider with an external shunt, but that involves other complications. I'd like a simpler solution.
What I plan to do is to use the standard linear Hall-effect sensors to measure the current without being in the electrical path. These are similar to the sensors used in motors, but with a linear analog output instead of digital. I can use these to create a similar setup to the packaged current sensors, but with the range and scaling I need. Here's the configuration I've devised in my FEMM model:
The three rectangles are the Hall sensors, which measure about 4 x 1.5 mm. The three smaller circles that form the corners of a triangle are the conductors. Here I've used wire with a similar diameter to 8 AWG. Surrounding the conductors and sensors is a steel ring which serves to concentrate and direct the flux. There's also a steel piece in the center of the sensors. This drawing shows it more clearly:
The configuration of the conductors and sensors is important. The flux is concentrated between the two active conductors with the help of the steel. Because the flux must loop back around, each of the other two sensors also see the flux, but of the opposite polarity and half the amplitude (ideally the sum of the three fluxes will sum to zero). With a little signal processing, I can combine the signal from all three sensors to get the final measurement. One really useful thing about combining the signals is that it'll reduce the noise and the tolerance of the individual sensors should average out somewhat for a more accurate measurement. It also lets the system operate in a sort of less-linear mode for really high currents. The sensor between the active conductors will saturate first, but the other two sensors will remain linear until twice the current, so the combined signal is still useable although it won't be quite as accurate.
My FEMM model of this configuration shows that the primary sensor will see about 4.4 G/A. The lower-sensitivity sensors are 2.5 mV/G with a usable range of +/- 2V. All of that works out to give me a fully-linear range of up to 180A and a "less-linear" range to 360A. If I need to go even higher, I can reduce the amount of steel slightly to lessen the flux coupling. For smaller units that don't need a super-high range, the sensors come in versions with 3.125 mV/G and 5.0 mV/G which would give linear ranges of 145A and 90A respectively. The averaging will hopefully make the signal useful at lower ranges, too.
I have some of the Hall sensors and an assortment of other parts on the way to do some testing. A 1" x 1/8" bar is just about the right size to make the flux ring out of, so I can pick one up at the local big-box store and gin up a quick-and-dirty prototype. I can work on the signal processing circuits to make this work in the meantime.
.
