lockups when using delays in the main loop w/ I2C bus

DaveKush · Guest

I'm experiencing consistent lockups when I use any type of delay or printf command inside the main while loop. The I2C interrupt routine is active, and this seems to be the source of the problems. It is looking like the program never returns correctly after servicing the I2C interrupt . The code model looks like this:

while (1) {

//An I2C transaction happens here

if (I2C_byte == 1) {
delay_ms(500);
//lockup happens here like the I2C never interrupt never returned correctly
}

}

Any ideas guys?

Dave

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

Are you using any CCS delay function inside your isr ?

Dave Kush · Guest

No delays in the I2C interrupt handler, just the main loop. I configured I2C manually because the CCS routines didn't seem to work with the PIC18f6585. Everything works great with the I2C both directions unless I have a pause in the main loop. Weird, yes.

Dave

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

The implication is that the delay routine is disabling global interrupts.

Do you have a CCS delay routine inside any isr (not just the i2c isr) ?

DaveKush · Joined: 10 Jan 2006 Posts: 6

Yes, I do have a delay in another external interrupt to handle a switch debounce, but the problem persists even if I never touch the switch.

Dave

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

If you have compiler warnings enabled (in MPLAB Project/Build Options),
you will see a message like this:

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

quote]I do have a delay in another external interrupt to handle a switch debounce.
[/quote]
One additional comment:

There are no nested interrupts. So if you have a delay in your INT_EXT
isr, you will lock out all of your i2c interrupts for the duration of that
delay and until you exit the isr.

Suppose you're running i2c at 100 KHz. Then at 10 bits/char, you could
get interrupted at roughly a 10 KHz rate. That's once every 100 us.
If your debounce delay in the isr lasts 1 ms, or 10 ms, you'll miss
large numbers of i2c interrupts. So I think you have to do your
debounce in a different way.

One way might be to start a timer in the INT_EXT isr. (and disable further
INT_EXT interrupts). Setup the timer so that it will roll over and cause a
timer interrupt in 10 ms. This is your debounce delay. Then handle the
key press in the timer isr. Then clear the INT_EXT interrupt flag and
and re-enable INT_EXT interrupts. Also disable the timer.

DaveKush · Joined: 10 Jan 2006 Posts: 6

Thanks for the suggestions PCM Programmer. I can't seem to get those warnings turned on in the compiler though. The switches I'm using are: +DF +LN +T -A +M +Z +Y=9 +EA +EW +FH but I still see no warnings like you mentioned.

On a side note, is there any docs that show which subroutine calls cause interrupts to be disabled with the CCS compiler? My code is getting complicated enough where this would now be helpful.

Dave

DaveKush · Joined: 10 Jan 2006 Posts: 6

Looks like I figured it out. Apparently the SSP interrupt wasn't being cleared correctly right before the call to the printf or delay. I inserted a clear_interrupt(INT_SSP) and this fixed the problem. I'm guessing that my interrupt handler isn't working correctly though. Time to dig into that.

Dave