[thingstodo]

Originally Posted by e*clipse

After more thought, I'd like to stay away from splitting the output busbars/pins/etc in order to add current sensing.

Not sure I have a good enough understanding to offer a specific comment. Splitting busbar is a bit tough for DIY with hand tools.. at least to do it accurately.

While the part P-hack suggested looks like a very good solution for this problem, it would still require adding a controlled impedance section... I suppose this could just be a thin part in one of the bus bars or connector pins... hmmm

Shunts are available that can be embedded. Not sure on the cost. I'd avoid narrowing sections of the bus or connections. The shunt designers put a lot of effort into temperature compensation and laser trimming of the shunts to get 5% or 1% (or 0.1% if you want to pay that much) accuracy.

The other option that would require this are the Allegro current sensors that are soldered to the bus bar.

Soldering to bus bar - do you do that with a torch? Copper dissipates the heat pretty well ...

In a general sense, I think non-contact sensors that use the magnetic field should be accurate enough for our purposes. Some of them require a flux concentrator, which is a gapped ferrite or laminate ring that goes around the sensed wire. The sensor fits in the gap. This would require some magnetic design to get right, but the ferrite might also serve the purpose of reducing stray EMI. A device from Allegro would be directly solderable to the main board with through hole pins, but would require an external field concentrator.

How do you guys feel about the issue of designing an external magnetic part (like a gapped torroid) for this?

Magnetic design is so far out of my competence that I am at a loss for a good comparison ... I'd likely be better at rocket science? If I can avoid magnetic design, I will.

Another option - would be the hall-effect sensors that don't require an external field concentrator. There is one made by Melexis that can do this and measure large nearby currents: - I think this is the one Paul made the daughter board for.
The MLX91206 is a contactless programmable current sensor

some advantages:
Isolated from high power circuit
Don't need external magnetics
Are spec'd for automotive use - - -50C to 150C storage and -40C to 150C operation.
Analog or PWM output

Issues:
requires calibration/programming
would require a separate "daughter board" that is perpendicular to the main circuit board.

I believe that I read about a hall effect sensor that uses AC excitation to cancel some of the drift and calibration requirements. EVTV sells a version, but it is CANbus only if I recall. Someone else should sell a competitive product by now - EVTV Motor Verks Store: LEM CAB300-C Flux Gate Current Sensor, Meters and Instruments, CAB300