Would mounting a hall effect sensor to the outside of the starter work?
Might. I now have some evidence that the residual magnetization is pretty weak. I am trying to detect rotation after the battery is disconnected when residual magnetization is all there is.
Here is what I learned: I borrowed two different aircraft starters (Delco-Remy and Prestolite) that have a Bendix drive. I also borrowed one that has a permanent magnet field. I spun them with a 12V battery, and recorded the voltage produced by the starters as they spin down and coast to a stop. The wound-field starters both exhibited similar voltage generation, as shown in the attached plot. Y-axis is V; X-axis is time in sec.
View attachment 68141
Battery voltage was applied at ~0.25s, and the starter spins-up for about 2 sec. Note that as it speeds up (~1s), the brushes begin skipping on the segmented commutator, causing the noise on the trace. The Delco starter was an old dog, and exhibited this much more than the Prestolite, which is near new.
By the time the battery is disconnected at ~2s, the starter is spinning fast, much faster than if the Bendix was engaged. As the battery is disconnected, the voltage drops from ~10V to <1V, with that short duration what looks like an RC decay with a time-constant of about 0.1s. Both of the wound-field starters exhibited that RC decay.
The important thing is what happens between 2.2 and about 6.5s. The starter is spinning, but slowing down. You can see the voltage go from about 400mV to zero as the starter coasts to a stop. Both of the wound-field starters did this, at nearly identical amplitudes.
I tried "loading" the starter with various resistors connected to ground to see if drawing some current (presumably it would flow through the field coils) would enhance the output, but all I saw was the same 400mV linearly decaying to zero as the starter wound down.
I tried "backfeeding" the starter with a bit of current by using a resistor connected between the starter terminal and +12V. This had an effect, although not the one I was looking for. I was hoping that the current would excite a larger output, but what I saw was due to the skipping brushes. The resistor acts as a pull-up during the short durations when the brushes loose contact with the commutator, putting a high-amplitude fuzz on top of the 400mv baseline. The old starter did this much more than the new one, so probably not good to exploit this in a practical circuit.
So, it looks like I need a voltage level detector circuit that lights a LED indicator when the voltage is at ~100mV or above. This would light the LED while cranking normally (starter relay closed, starter running, Bendix engaged. Assuming that the start relay releases normally, and the Bendix kicks-out because the engine caught and is now running, the LED might stay on for a second or two as starter coasts to a stop. However, if the Bendix hangs up, and the engine is backdriving the starter motor, it would continue to spin, generating >100mV, keeping the LED lit. If the starter relay hung-up, keeping power applied to the starter after the engine fires, the LED would indicate that too, warning the pilot to stop the engine and to investigate.
Looking for ideas for a simple 50mV to 100mV level voltage detector/LED driver. It should have as few parts as possible, be capable of operating on dirty 12V power, work over a temperature range of -40 to +45deg C. It should drive a red LED with about 20mA. It should draw less than 1mA when the LED is not on.
ps. The permanent magnet field starter initially generates almost 8V as the battery is disconnected, which decays linearly to zero as the starter spins down. The voltage output is adequate to light the LED without any active circuitry.