problem with sleep mode.

[hemnath]

[Ttelmah]

There are several things 'problematical' about this.

Buttons rarely give simple on/off edges. When pressed, you normally get several short make/break pulses over a few uSec. You need to design your hardware, to give clean make/break pulses to have any hope of the current approach working. Now you have a long delay in the interrupt (too long), but the problem is that flags will have been set on the first edge, even if this is not the actual edge the button ends up giving, and worse, this results in interrupts being disabled in the main code during the led on/off delays....

Step back.
Either design the hardware to give clean edges (hardware filtering on the buttons), or rethink, and have the button 'event', trigger an interrupt based delay, and _then_ re-sample. The code needs to only be triggering the sleep, once it _has_ verified that the button is cleanly made.

Then put a 'delay_cycles(1)' after the sleep. The instruction after the sleep is 'pre-fetched' when you go to sleep. ideally it should be a nop (which is what a single cycle delay generates).

Best Wishes

[jgschmidt]

I'm assuming you are trying to achieve low power sleep. I've had some experience and good results with that and I have several suggestions:

1) To verify that you are truly sleeping, you might just try straight-line code that doesn't require a button press. Do a delay_ms(10000) after main() so you can get a steady reading and then call sleep().

2) You will probably see a reduction in current but less than the spec sheet claims, even if you don't have any other hardware connected. See some suggestions for that below.

3) For detecting button presses I generally use a polled I/O check in my main processing loop. For Inputs held high, I check for an input held low for several consecutive loops. Once that is detected I go and process that button.

4) Once you get your button working you'll need to do several things before calling sleep(). For systems like this I have two functions - HWSetup() to setup the environment for running and HWSleep() to arrange for minimal current sleeping. Depending on what else you have connected to your processor, you need to turn off UARTs, timers, voltage references and ADCs, set the I/O ports so current doesn't flow in or out of the processor, etc. Turn off power to peripherals via relays or pFETs. Set up your conditions for waking up, usually an interrupt and then call sleep().

For my power on/off applications that is followed by interrupt resets and a reset_cpu(), still within the HWSleep() function. The first call in main() is to HWSetup and my application is off and running. You can check to see if you started from a reset or a true cold start with restart_cause().

Microchip publishes some guidance for low power systems and there are posts here and on the Microchip forums related to sleep() which will help you achieve low current sleeping.
_________________
Jürgen
www.jgscraft.com