I'm starting to make a very simple brushless DC motor. It's just a single coil and a couple of magnets glued to a needle.
Mechanically it's simple, but the programming will be fun. As expected it won't run stable with just a fixed pulse, so a microcontroller will be working out it's speed and position from the back EMF from the motor so that it can send the pulse at the correct frequency and duty cycle.

Here I managed to get it to run for long enough to see what was happening. It's spinning at about 3800rpm (4 times the speed of the pulses), which is not the aim, but it's enough that I can see it is possible to make it work.

 

That isn't how we did it in the 1950's!

 

That isn't how we did it in the 1950's!

Valves are expensive these days, I think I'll just stick to the microcontroller.

 

Nice essay. Do you want me to move it in another forum? The Projects one seems more suitable.

 

You can if you feel it's more appropriate, I'm not really asking for advice, but I don't mind if any is offered.

 

I've had another play with the motor and upgraded to an H-bridge for now. There's a 4K7 resistor divider connected between the rails to one end of the coil and the oscilloscope is connected to the other end.
It's still running open-loop at a random period and duty cycle, but I've put a switch so that it will hold when I see it's running stable. Open-loop it's much more stable than the one transistor version.
I think that for closed-loop control I need to be aiming to get the peak of the back EMF into the driven part of the cycle.
In these shots it's running at about 800 RPM and about 100mA average on 5V.

 

Does it self start? Bet you have to give it a spin to start rotating. Since there is not a full "rotating" field.

 

Yes, it needs a little flick to get going. Apparently if you put a little magnet off-axis it's possible to get them to self start, but with back EMF sensing you can't tell which direction it's rotating in until it gets up to a certain speed. If it goes the wrong way you have to try again.

 

It seems I was wrong about the power being applied at the peaks of the back EMF. It looks like it needs to be at the crossover point, although I'm not sure why. I've now got it running closed loop, although something is not quite right with the timing, but it increases the duty cycle when there is more load (second trace) and will also run with just one transistor (third trace).
Still a bit of work to do, but it's looking promising.
<ed>After some more experimenting I'm now sure that I was right the first time. The maximum torque is possible at the peaks of back EMF.<ed>