OK, tried disabling interrupts just before the bit-bang calls and reenabling right after as suggested and the behavior is unchanged. I’m surprised by this. As mentioned previously, I have an S3 on order, though Seeed’s fulfillment time leaves something to be desired… In the interim I may try a timer interrupt to see if that drives the motor without the glitching, but I’ll need to look at how to achieve step counting in the ISR and what the overhead would be for speed changes…