Another note.
I think the switch in my design was driven by something that was effectively a D or J-K flip flop (though there were two of them inside an integrated dual H bridge chip), clocked at the peak of the sawtooth clock waveform, with the D or J and reset driven by the current comparator. So each clock period, if the comparator showed current below reference, the clock edge would start a power pulse, and when the comparator showed the current above the reference, the reset would terminate the pulse, and that decision was irrevocable till the next clock edge began another pulse. The clock circuit consisted of an external RC, in parallel to ground with a fast charge and an RC time constant discharge. I added a bit of that waveform to the current reference input to make the sawtooth version of the current reference.
I wasn't micro stepping the driver, but operating in half step mode. I also varied the average reference level between two different values, depending on whether one or two windings were being powered to produce a uniform break away torque, regardless. This took a lot of rotational torsional vibration out of the rotation.