ESP32S3 : Fifth or sixth stepper getting stuck moving.

[dronus]

On my ESP32S3 Controller, using pioarduino Framework, I had some issues with steppers stuck in motion:

• the problem occurs rarely (eg. sometimes only after days)
• the stepper runs at constant slow speed beyond the target position of stepper->moveTo(target)
• stepper->getCurrentPosition() is actually reflecting the wrong position - it is counting.
• further stepper->moveTo(target) do not change anything
• I am issuing stepper->moveTo(target) commands at about 100 Hz for all six channels.
• there is no other IO code I am using despite USB Serial and WiFi (that may interfere with the PCNT or RMT peripherials) and some digitalWrite for enable signals.

It seems only to occur on the fifth or sixth channel, but hard to say because of it occurs that rarely.

I can not give a minimal example, because it happens so rarely and no good experiments are possible at that pace. This is very annoying to debug of course.

[gin66]

using espidf5? then 4x RMT and 2xI2S. I think, I2S are in this setup stepper 5 and 6.
If the steppers are assigned with FasDriver, will the stopping motors change accordingly?

[gin66]

If the position is still counting, this means the whole pulse generation machinery is working. Pulses are prepared in the rmt and i2s buffers, drivers are processing the buffers and commands are processed.

This leaves these possibilities:

• internal connection from rmt/i2s module to the output pad is lost
• gpio is reverted to input
• pulses on the wire are not recognized by the driver
• driver is going into thermal shutdown
• gnd shift

Agree. hard to debug. you could connect pulse counter (via FastAccelStepper). If the pulse counter is too counting, then the first idea can be excluded and the second idea most likely too.

[dronus]

The last issues are external issues - (pulse on wire not recognized, driver problems, gnd problems) - this is not the case, as it would still allow FastAccelStepper to drive the ESP pin as intended.
Basically this bug still trigger without anything attached to the ESP, it is happening in the ESP SoC itself.

About I2S: is this prefered now over PNCNT ? So motors 5+6 will use I2S by default?

I wonder because I experienced this issue earlier than the I2S implementation was done as far as I know.

This gives hope, that it is already fixed! To be sure however takes a lot of time, as it triggers rarely.

[dronus]

"If the position is still counting, this means the whole pulse generation machinery is working." - I would say partially yes, it feels like the pulse generation is running on itself, without control. So the counter is still updating the current position, but no new commands are accepted in some way.

I wonder if the same would happen if the generation was modified by other means? Could this be possible? Eg. something activated some PWM on the pin, would FastAccelStepper then count this pulses and update the motor position, even if it is not in control of the pin anymore?