Accuracy of pulse in?

Forum 2005-2010 (read only) Hardware Troubleshooting
#1

Ok so I have an "arduino" and it is running @ 8mhz (i say it like that cause its not, its a tiny84 but for the point of this topic that is not need to know)

I am trying to cook up a ADB (apple desktop bus) interface, ADB is a 1 wire fun little serial protocol. Timing and flags are an important fact to get communications

When the machine first turns on I have a chance to sample the device 0 / reset signal which contains an Attention flag, a Sync Flag, and a command byte of 00001111, which is separated by a bit cell time period (lost yet?)

using the following test sketch I am reading the signal from the Macintosh and gathering the Attention flag, the Sync flag, and the Bit cell time,during void setup(), and during void loop I have the mcu output its findings on another pin

#define adbData 0
#define out 3
unsigned long atnSignal;
unsigned long syncSignal;
unsigned long bitCell;

void setup()
{
  pinMode(adbData, INPUT);
  pulseIn(adbData, HIGH); // waiting for bus to come online
  atnSignal = pulseIn(adbData, LOW);
  syncSignal = pulseIn(adbData, HIGH);
  bitCell = (pulseIn(adbData, LOW) + pulseIn(adbData, HIGH));
  pinMode(out, OUTPUT);
}

void loop()
{
  PORTA = 0x80; // pin 3 for attiny on
  delay(1);
  PORTA = 0x00; // pin 3 for attiny off
  delayMicroseconds(atnSignal);
  PORTA = 0x80;
  delayMicroseconds(syncSignal);
  PORTA = 0x00;
  delayMicroseconds(bitCell);
}

and here are the results

| * | MAC | AVR | | - | - | - | | ATTN | 648us | 640us | | SYNC | 56us | 48us | | Bcell | 80us | 80us |

now even though were dealing with pieces of 8's here I am a bit disappointed that most of the timings recorded by the arduino are short, so here are the questions

Would I get better timing data by constantly polling the pin and calculate time differences?

On the opposite, would I be able to do better by running the chip at 16mhz while saving code hassle and using pulseIn()?

How reasonable are my expectations and should I just move on

Thanks ;D

#2

Would I get better timing data by constantly polling the pin and calculate time differences?

What do you think "pulseIn" does?

Hint: The source is in "wiring_pulse.c"

#3

would I be able to do better by running the chip at 16mhz

yes running it at 16MHz would half the dead time, that is the time between the signal changing and you being able to detect that change.

#4