Longer than expected delay with delayMicroseconds (Biphasic Square Wave)

[westfw]

I haven't seen the use of the avr-libc _delay_us() function yet. This produces a cycle-counted busy-loop and is more accurate than delayMicrosseconds() However, it can only be used with constant arguments; if you want a variable delay, you'd have to put it in a loop, which would complicate matters slightly (probably not significantly, at the 10us timescake,
So where are we?

1. The Arduino delayMicroseconds() function has a resolution of 4us, so it's not suitable for use in timing that need to be more accurate than that.
2. Moreover, Arduino functions like digitalWrite() can take several microseconds to execute (and soemtimes the timing is pin-dependent!)
3. Using the hardware timers could yield very accurate delays with little software involvement, but the desired biphasic output woud be "relatively complicated" to implement and may int6erfere with other Arduino functions (PWM, for instance.) There are three-phase motor controllers based on the same AVR family used in Arduinos, which I consider an "existance proof."
4. Using timers or _delay_us() will make code less portable to other architectures,

[ma7555]

@westfw

• The Arduino delayMicroseconds() function has a resolution of 4us, so it's not suitable for use in timing that need to be more accurate than that.

Honestly, I do not mind a 4us error on the long (800us) delay but on the scope it appears much worse than 4us. When I disable interrupt for the whole loop it looks pretty accurate.

@westfw

• Moreover, Arduino functions like digitalWrite() can take several microseconds to execute (and soemtimes the timing is pin-dependent!)
• Using the hardware timers could yield very accurate delays with little software involvement, but the desired biphasic output woud be "relatively complicated" to implement and may int6erfere with other Arduino functions (PWM, for instance.) There are three-phase motor controllers based on the same AVR family used in Arduinos, which I consider an "existance proof."

Correct. I will take a look over the three phase motor controllers.

@westfw

Using timers or _delay_us() will make code less portable to other architectures,

I don't really care about other architectures. Thanks for letting me know about the _delay_us function. I will take a look into it.

[oqibidipo]

@westfw

The Arduino delayMicroseconds() function has a resolution of 4us, so it's not suitable for use in timing that need to be more accurate than that.

Nope, micros() has 4 µs resolution, delayMicroseconds() is a tight busy loop, 4 iterations per µs at 16 MHz.

[westfw]

@oqibidipo

delayMicroseconds() is a tight busy loop

Huh. You are correct, and I was wrong.

[ua6em]

This is an example of how similar problems are solved, here is another one, a little more complicated than you need

void setup() {
pinMode(3, OUTPUT);
pinMode(4, OUTPUT);
TCCR2A = (1 << WGM21) ;
TCCR2B =   (1 << CS22 ) | (1 << CS21 );
OCR2A = 50;
TIMSK2 = 1 << OCIE2A;
}

ISR(TIMER2_COMPA_vect) {
static uint8_t tic;
tic=(tic&15)+1;
  switch(tic){
    case 1: PORTD&= ~((1<<PD3)|(1<<PD4));break;
    case 8:  PORTD= PORTD&0xe7|8; break;
   default: if (tic>8){ PORTD^= (1<<PD3)|(1<<PD4);} 
  }

}

void loop() {}

[dlloyd]

Yep, looks good ...

UNO-Timer-Scope.ino - Wokwi Arduino and ESP32 Simulator