How can reading a quadrature encoder fail?

Question

I am trying to the read the quadrature encoder on a ServoCity 624 RPM Premium Planetary Gear Motor w/Encoder with a SparkFun ESP32 Thing.

I am supposed to see 228 counts per revolution, but I see 230-232 instead.

Here is my Arduino code:

// Global variables modified by ISR
int state;
unsigned int count;

int error;

// ISR called on both edge of quadrature signals
void handleInterrupt()
{
  // Shift old state into higher bits
  state = ( state << 2 ) & 15;

  // Get current state
  if( digitalRead(34) ) state |= 2;
  if( digitalRead(35) ) state |= 1;

  // Check state change for forward or backward quadrature
  // Flag any state change errors
  switch( state )
  {
    case 1:
    case 7:
    case 14:
    case 8:
      count--;
      break;
    case 11:
    case 13:
    case 4:
    case 2:
      count++;
      break;
    default:
      error++;
  }
}

void setup()
{
  pinMode(33, OUTPUT); // PWM
  pinMode(32, OUTPUT); // DIR

  pinMode(34, INPUT_PULLUP); // QB
  pinMode(35, INPUT_PULLUP); // QA

  attachInterrupt( digitalPinToInterrupt(34), handleInterrupt, CHANGE );
  attachInterrupt( digitalPinToInterrupt(35), handleInterrupt, CHANGE );

  Serial.begin(74880);
}

void loop()
{
  int new_state;
  int old_state;

  // Start motor
  digitalWrite(32, HIGH); // clockwise
  digitalWrite(33, HIGH);

  // 57*4-17 found by trial and error to make a complete revolution
  for( int i = 0; i < (57*4-17); i++ )
  {
    // Busy wait for change
    do
    {
      // Get current state
      new_state = 0;
      if( digitalRead(34) ) new_state |= 2;
      if( digitalRead(35) ) new_state |= 1;
    }
    while( old_state == new_state );
    old_state = new_state;
  }

  // Stop motor
  digitalWrite(33, LOW);
  delay( 1000 );

  Serial.print( " state=" );
  Serial.print( state );
  Serial.print( " count=" );
  Serial.print( count % 228 );
  Serial.print( " error=" );
  Serial.println( error );
}

1. On my scope, quadrature signals are very clean
2. I instrumented the code to look at interrupts and they appear to be in the right places and not too close together.
3. I don't see mechanical slipping, and it would be obvious because over 100 revs, the count slips an entire revolution.

But the bottom line is: How can this be failing? If the CPU is missing transitions, I would get illegal transitions and errors. Noise would also cause errors. But I am not getting any errors at all (except a single error at startup).

Answer 1

ServoCity 624 RPM Premium Planetary Gear Motor w/Encoder has the encoder on the motor and a 19:1 gear reduction on the output shaft.

Mechanical tolerance in the gearbox accounts for the difference between 228 counts per revolution that I expected and 230.7 counts per revolution that I measured.