Inconsistent Pulse Output on ESP32

[Samthecoolman]

Howdy,

While playing around with the fastAccel library I found that I was getting inconsistent step outputs from my ESP32 dev board. Switching to different dev boards results in the same issue, have not tested on AVR.

Using a scope to probe the step output I can verify this issue is not my servo motor (Yellow is step output, blue is enable):

I only have a moderate understanding of Cpp so I am sure there could be an issue with my code, however after spending countless hours trying to debug I have no clue where to proceed from here.

Here is the snippit of my code (Modified example sketch);

#include "FastAccelStepper.h"

// As in StepperDemo for Motor 1 on AVR
//#define dirPinStepper    5
//#define enablePinStepper 6
//#define stepPinStepper   9  // OC1A in case of AVR

// As in StepperDemo for Motor 1 on ESP32
#define dirPinStepper 18
#define enablePinStepper 26
#define stepPinStepper 19

FastAccelStepperEngine engine = FastAccelStepperEngine();
FastAccelStepper *stepper = NULL;

void setup() {
  Serial.begin(115200);
  engine.init();
  stepper = engine.stepperConnectToPin(stepPinStepper);
  if (stepper) {
    stepper->setDirectionPin(dirPinStepper);
    stepper->setEnablePin(enablePinStepper);
    stepper->setAutoEnable(true);

    // If auto enable/disable need delays, just add (one or both):
    stepper->setDelayToEnable(50);
    stepper->setDelayToDisable(50);

    stepper->setSpeedInHz(5000);  
    stepper->setAcceleration(100000);

  }
}

void loop() {

  stepper -> moveTo(1600);
  while(stepper -> isRunning()) {Serial.println(stepper->getCurrentPosition());}
  delay(50);

  stepper -> moveTo(0);
  while(stepper -> isRunning()) {Serial.println(stepper->getCurrentPosition());}
  delay(50);

}

Any help would be greatly appriciated!

• Sam

[gin66]

Your code looks good and the result is surprising.

Questions:

• Which FastAccelStepper version do you use ? Arduino library manager is stuck at 0.30.0
• From the picture it seems, that the stepper switches between few and many pulses depending on the direction. Can you confirm this ?
• Any different behavior, if stepper->setAutoEnable(true); is changed to false?
• what is the output at the serial console ? Is the position deviating, too ?
• plain esp32 or esp32s2/s3/… ?

[Samthecoolman]

Thanks for the quick reply!

• Version is 0.30.0, I am using arduino IDE, looking at the issue link you provided this sounds like the cause
• Direction does not appear to affect step output, it does look like a pattern in the picture I provided though it is not consistant
• Same behavior is observed regardless of autoEnable state
• Position does not deviate, serial output counts up and down to desired setpoint
• Tested on plain esp32 as well as esp32s3

Something else that is interesting is that after running for ~30 minutes the steps become consistant?

[gin66]

could you please upgrade to 0.30.10 ? Even the Changelog does not list a related issue, it is better to use the latest version.

[Samthecoolman]

Upgraded my version and the inconsistancy appears to have been resolved, however there is slight drift in setpoints over time. I am waiting on some level shifters to arrive, so missed steps here could be chalked up to my servo needing 5 instead of 3.3 volts.

[gin66]

sounds good and thanks for support.

With an esp32 you could use a free pulse counter using stepper-> attachToPulseCounter((7). This allows to verify, if pulses are lost on uC side or not. expectation is: no pulses lost

[Samthecoolman]

Had some time to continue working on this project and the issue is still present; attatching a pulse counter verifies that steps were lost during movement, inconsistency is present in both directions. Code is here:

#include "FastAccelStepper.h"

// As in StepperDemo for Motor 1 on AVR
//#define dirPinStepper    5
//#define enablePinStepper 6
//#define stepPinStepper   9  // OC1A in case of AVR

// As in StepperDemo for Motor 1 on ESP32
#define dirPinStepper 18
#define enablePinStepper 26
#define stepPinStepper 19

FastAccelStepperEngine engine = FastAccelStepperEngine();
FastAccelStepper *stepper = NULL;

void setup() {
  Serial.begin(115200);
  engine.init();
  stepper = engine.stepperConnectToPin(stepPinStepper);
  if (stepper) {
    stepper->setDirectionPin(dirPinStepper);
    stepper->setEnablePin(enablePinStepper);
    stepper->setAutoEnable(true);

    // If auto enable/disable need delays, just add (one or both):
    stepper->setDelayToEnable(50);
    stepper->setDelayToDisable(50);

    stepper->setSpeedInHz(5000);  
    stepper->setAcceleration(10000);
    stepper-> attachToPulseCounter(7);

  }
}

void loop() {

  stepper -> moveTo(1600);
  while(stepper -> isRunning()) {Serial.println(stepper->getCurrentPosition());}
  Serial.println(stepper -> readPulseCounter());
  delay(500);

  stepper -> moveTo(0);
  while(stepper -> isRunning()) {Serial.println(stepper->getCurrentPosition());}
  Serial.println(stepper -> readPulseCounter());
  delay(500);

}

[gin66]

Sorry to hear that. I have set up an esp32 with a digital scope connected and I cannot reproduce your issue with either example code.:

• The first code outputs as expected, with enable inputs being operated on a regular pattern. The scope screenshot, you have provided, is not regular.
• The second code yields result as expected

FastAccelStepper uses mcpwm/pcnt module for the first six allocated steppers and then rmt module. The mcpwm/pcnt combo uses a feedback loop: mcpwm generates steps and pcnt counts those. The theory is, that the pcnt-module gets input from the actual pin. So any noise on the pin could lead to wrong counts. To verify this, please try:

• Use rmt-module using engine.stepperConnectToPin(stepPinStepper,DRIVER_RMT);. rmt generates pulses without looking at the actual pin
• Disconnect your hardware from the step pin, so induced noise should vanish

If this does not help, then it will get tricky.

[Samthecoolman]

Using rmt module seems to fix it, I will do some more testing to verify stability.

Thanks for the help!