From the app note:
STEP input MIN pulse width (common in H/L): 500ns (MAX input frequency: 1MHz)
Even $1=3 (ms) ought to be 6 times longer than enough by that.
However, constant current control is performed by PWM during chopping period, which is set by the capacitor connected between OSC1 and GND. You need to perform chopping more than once per step.
For this reason, for the actual STEP frequency, you need to take chopping frequency and chopping count into consideration.
For example, if chopping frequency is 50kHz (20μs) and chopping is performed twice per step, the maximum STEP frequency is obtained as follows: f＝1/(20μs×2) = 25kHz.
I didn’t even try to probe with an oscilloscope to find the chopping frequency of the driver boards I got, so 10µs is just a guess for
I measured max current with a good (fluke) multimeter and increased current on all four drivers to just under 2.8 A. It didn’t improve anything.
Data sheet indicates 3.3V should be fine:
But I don’t know whether there are optoisolators under the heat sink that I’m really driving. If so, then the classic MOSFET-based bidirectional logic level shifter might not source enough current, and I should instead use active level shifters to switch my 5V source. I ordered both TXS0108E and TXB0104 devices to try for this purpose, since I’m sure I’ll need more as I do more work with ESP32s and having some of them immediately available is likely to come in handy.