Interrupt and Counter

[DanielPHuber 16]

A decoder from a motor has 2 square wave outputs that deliver 256 pulses/revolution (duty cycle 1:1). The two outputs are 90 degree out of phase for direction information.

I want to count pulse, in one direction up, in the other down. Towards this aim I attach to the first decoder output following routine:

 

void countStep(){

     if(digitalRead(_CIN2)){_pos++;} else {_pos--;}

}

 

"_CIN2" is the second decoder output and "pos" the counter variable.

This works well until up to approx. 16kHz.

But for higher frequencies it counts in the wrong direction.

I first guessed, that the phase relation between the 2 signals is not constant, but an independent measurement showed it to be rock stable.

Another guess is, that the interrupt is too slow (maybe because of the "digitalRead"). I am running a MSP430G2553 with 16MHz. Can somebody verify or falsify this guess?

 

[solipso 8]

I do not use Energia, but i doubt that digitalRead function is interrupt based. Just try to whack up a real ISR for both pins incrementing or decrementing your variable depending on previous state variables of respective pins.

[DanielPHuber 16]

In between I found the excellent post from this forum :

 

digitalReadFast and speed of digitalRead

Started by altineller, Aug 19 2014 05:41 PM

 

and you bet, the problem was the "digitalRead". I replaced

 

digitalRead(_CIN2)

by:

P1IN & 0b00100000)

 

and it works. It is astonishing that for a CPU running with 16MHz, digitalRead is so slow.

According to the above postdigitalRead takes approx. 18 uS what explains my problem.

[spirilis 1,265]

Yeah, digitalRead (and digitalWrite, pinMode, analogWrite, etc) is kind've a beast.  It's the cost of having an elaborate abstraction layer that takes a simple integer identifier and distills it into platform-native operations.