Sleep and Interrupts

[plinder13]

I have had a difficult time getting sleep to work with interrupts as expected on various PICs

I want to put a PIC to sleep using the sleep() function. But, before I put it to sleep, I want to enable global interrupts. As soon as it wakes up, I want to execute the ISR for any interrupts that occured and then immediately disable global interrupts so that background tasks can run. I have much experience with the MSP430 for which the following code works great:

while(1)
{
if (flags.sleep)
{
LPM0_EINT(); //(enter low power mode and enable global interrupts)
DINT(); //(disable low power mode and disable global interrupts
}
// Do background tasks and check flags
}

The reason I need global interrupts enabled immediately before sleeping is that I set various flags in ISRs that determine whether the MCU should sleep. I don't want those ISRs to run between the time I check the flags and the time I place the MCU into sleep (if an interrupt occured between the checking of the flags and the sleep instruction, the MCU could get "stuck" in sleep mode).

Here is the code I have tried which has not worked for me:

if (flags.sleep)
{
enable_interrupts(GLOBAL);
sleep();
delay_cycles(1);
disable_interrupts(GLOBAL);
}

In my specific project, the comparator does not seem to wake up the MCU. However, if I comment out the sleep() instruction the application works fine. But, I have had a hard time understanding how to do this in other applications as well. I also have tried inserting another delay_cycles(1); before the sleep instruction.

Thanks for any help.

[PCM programmer]

[Guest]

PIC16F883
CCS Version 4.044

Specifically, I am trying to use the comparator to wake up from sleep mode.

[Bill Boucher]

Try copying the ISR controlled flags variable into a local variable in your main(). The value of this local copy will not be changed if an interrupt happens. Check the value of the copy and do your sleep thing based on that.

[PCM programmer]

Before you execute the Sleep() function, you also need to do
these three things:

1. Clear the "mis-match" condition that causes the interrupt
by reading the CMxCON0 register for the comparator that
you're using. Use the #byte directive to declare the address
of the register, and then read it into a variable with a line of code.

2. Clear the comparator interrupt flag by calling the CCS
clear_interrupt() function, with INT_COMP as the parameter.

Doing items #1 and #2 will clear any pre-existing comparator
interrupt condition, prior to going to sleep.

3. Enable comparator interrupts by using enable_interrupts()
with INT_COMP as the parameter.

Then you can enable Global interrupts and go to sleep.
That's three additional lines of code that you need to add.


Also, inside the comparator isr, you should do item #1. This is in
the data sheet. In addition to that, if you don't want to get any more
comparator interrupts (for the time being), you should call
disable_interrupts(INT_COMP) inside the isr.

[SET]

If you want to execute the code after the sleep() when an interrupt occurs then you DONT want to set the GIE bit - just set the interrupt enable flag(s) for the interrupts that you want to happen. When the interrupt occurs the PIC will continue execution after the sleep instruction - if you have the GIE flag set it will jump to the interrupt handler.

[plinder13]

Let me ask my question a different way.

How do I force my ISRs to execute only at the sleep instruction? I have more than one interrupt that may occur (comparator, port b change, etc.).

[SET]

Let me see if I understand - you want the PIC to be brought out of sleep by an interrupt (which could be from a number of sources), execute some code depending on the interrupt that happened and then go back to sleep? If so, then that is what my code fragment will do - just enable each relevant interrupt, and in the code following the sleep() test the flags to see which interrupt it was.

[Guest]

[SET]

Sorry for delay in replying, back from weeks holiday..

In theory that might happen - the PIC really should have a single instruction that enables global interrupts and puts it to sleep atomically. In reality it probably wouldnt happen, the time difference between setting the GIE bit and then executing sleep would be 4 clocks, surely not enough for an interrupt to be recognised? Maybe..

[RLScott]

[Ttelmah]

[SET]

Thanks for that Tlemah - I assumed that Microchip would have taken it into account but hadn't looked at the exact mechanism.

Yes RLScott, I agree - this is exactly what we do in a number of low-power designs and it works well.