Following my well proved example code:

#include <reg51.h>

sbit    PHASE_A = P1^0;
sbit    PHASE_B = P1^1;

int     enc_delta;              // -32768 ... 32767

void t0_interrupt( void ) interrupt 1
{
  static char enc_last;
  char i;

  i = 0;
  if( PHASE_A )
    i = 1;
  if( PHASE_B )
    i ^= 3;                     // convert gray to binary
  i -= enc_last;                // difference new - last
  if( i & 1 ){                  // bit 0 = value (1)
    enc_last += i;              // store new as next last
    if( i & 2 )                 // bit 1 = direction (+/-)
      enc_delta--;
    else
      enc_delta++;
 }
}                               // 45 cycle


void main( void )
{
  TMOD = 0x02;                  // Mode 2: reload
  TH0 = -111;                   // 111 cylce (30kHz at XTAL = 20MHz)
  TR0 = 1;

  for(;;){
    P2 = enc_delta;
  }
}

Signal oscillations only critical on using external interrupts, but not on this solution with the timer interrupt !

Since there is absolutely no way to trigger the interrupt faster as expected and thus no way for false counting.

False counting can only occur, if the rotation speed was faster as the sample rate.
E.g. with an AT89C51RD2 a timer interrupt of about 30kHz is possible very easy, which can count up to 15000 pulse per second.

Thus only for very high speed external hardware was needed.


Peter