sleep mode in 12f683

[mahdihir]

Hi
I’m working on project with 12f683, the purpose is to get sample from ADC every 10 sec and if the sample is bigger than X, do something.
The problem is I need to save energy as much as possible for battery life,
Is it possible to put MCU in sleep mode for 10 sec and after that wake up MCU for 10ms to get sample from ADC and go back to sleep?
thanks

[40inD]

Set WDT to closest round value and do sleep() so much times to get 10 s delay then reset cpu:

Example from my program:

[temtronic]

just be aware that the WDT is run from the internal 31KHz osc and is NOT precise !! You'll have to experiment to get reasonable,repeatable operation.
VDD, temperature, initial calibration all affect the timing AND each PIC is different.

For a 'one-off' project , not a big deal , but for a 'commercial' product, you're best to use a proper xtal/caps.
Jay

[mahdihir]

thanks a lot
but my problem is the sleep mode is in the middle of program, once WDT reset the MCU and goes to the first line of program and i want to continue after sleep instruction, the rest of program.

[40inD]

In SLEEP mode WDT does not reset cpu, just continue program. Read dataheet.
that's why I add reset_cpu() in my program above.

[Ttelmah]

First of all, don't reset the cpu.
It's bad practice. A reset should only happen when there is good cause.

What happens on a watchdog, changes between chips. Some of the very early PIC12's, will always physically restart on a watchdog. However most modern ones won't. Old line _read the data sheet_.....

Section 12.7 on the data sheet for your PIC, says:

"The device can wake-up from Sleep through one of the
following events:
1. External Reset input on MCLR pin.
2. Watchdog Timer wake-up (if WDT was
enabled).
3. Interrupt from GP2/INT pin, GPIO change or a
peripheral interrupt.
The first event will cause a device Reset. The two latter
events are considered a continuation of program
execution. The TO and PD bits in the STATUS register
can be used to determine the cause of a device Reset.
The PD bit, which is set on power-up, is cleared when
Sleep is invoked. TO bit is cleared if WDT wake-up
occurred."

Now, see how an external MCLR causes a reset, but "the two latter events are considered a continuation of program execution". Exactly what you need to know.

So your code structure should be something like:

[mahdihir]

thanks,
its working properly.

my another question is, i setup a ADC to get sample every 2.3 min, and if the value is bigger than MIN, do something,
is my code correct?

[Ttelmah]

1) Move the setup_adc_ports outside the loop.
There must always be Tacq, between using the setup_adc_ports, and reading the adc. This is what the delay_us at the start of my loop, is there for.
2) Forget float. Unless you absolutely 'have to', FP should not be used when an integer will do. FP, is bulky, slow, and inaccurate. For things like financial calculations they are never used because of the errors. Remember the value from the ADC, is just an integer count. int16 is what you should be using. If you want to work with a 'number' that has some direct relevance to volts, then consider multiplying, so that the integer, gets scaled to represent mV.
3) You have the watchdog enabled in hardware. Again look at how I do it. The watchdog is _not_ enabled, till it is to be used. You have a very high risk now of a watchdog occurring while you are awake. Particularly because of 4 below.
4) You are now using a very slow clock. Slightly 'pointless', since all it does is makes the time spent awake, longer. The effect on consumption, will actually be to make it worse!. At 3v, the chip draws typically 500uA at 4MHz, and 16uA at 31KHz. 31.25* as much, _but_ it executes code 129*faster. So you end up drawing more power running at the slower speed (4* as much to execute the same code....). Use the slower speed, if you are wanting to keep the chip awake. If you are sleeping, look at which processor speed gives the best performance per uA. The slow clock will make things take an age. The FP comparison, will take nearly 1/10th second at 31KHz. If your routine reached after the comparison does anything more than sneeze, you will end up having a watchdog restart in this - since you are leaving the watchdog running. Because you have it enabled in the fuses _it cannot be turned off_. Read the data sheet. Look at what the WDTE bit does.
5) Don't 'goto'. Using goto, in some computer courses is treated as a 'capital' offence. You should really never need to use it (there are a few situations in embedded programming, primarily related to interrupts/recovery, where it may be used), but otherwise treat this instruction as if it doesn't exist. It is _not_ a primary part of C, but is only there for those odd occasions. As it stands, depending on whether you want to get back into the loop, you may well end up with the processor stack becoming unbalanced. It is a very easy way to shoot yourself in the foot.
6) At your slow clock, the ADC will take over 6mSec to perform the conversion. A very long way outside it's recommended operating range (57* slower than the recommended slowest operation). This will degrade it's accuracy appallingly.

[guy]

PIC12F683 also supports Ultra Low Power WakeUp (ULPWU). If you connect in series a 130 ohm resistor from GP0 - to a 10nF capacitor - to GND you can have the PIC charge and discharge the cap to keep track of time while in sleep mode. Power consumption is less than 1uA (much less than the WDT).
GP0 >-----/\/\/\/\----||--- GND
Another thing to look out for - disconnect the cap when programming in-circuit (GP0 is also ICSPDAT).

Let me know if you want me to post the code - I don't think the feature is supported by the compiler. It also requires calibration using the internal oscillator - the code takes care of everything.

[mahdihir]

My friend
thanks for your help.
I have done what you said from No 1 to 4.
But for No5 and 6, can you explain more? did you mean I should change [spam] from 31k to 4M?
The battery life is the most important factor in my project.
Thanks

[temtronic]

re: #5 xtal value.
yes, you should use the 4 MHz xtal instead of the 32KHz xtal. Although the PIC will consume more power at the higher frequency, it gets the program done and back to sleep a LOT faster than the 'slow'(32KHz) PIC will. Overall the 'fast' PIC will SAVE energy.
Microchip has an application note(AN606) you should read. It deals with 'low power design' and is well worth reading.

Jay

[Ttelmah]

I'll elucidate just a bit.

Consumption, is power*time. Power is volts*current. Assume 3v.

Assume the 'awake' code is 1000 instructions long (the number doesn't matter). At 31KHz, this will take (4428ips), 0.226 seconds. At 4MHz (1mips), it'll take 1mSec. When awake at 4MHz, it'll draw 500uA. When awake at 31Khz, it'll draw 16uA. Total consumption to run the 'awake' code at 31KHz = 3*0.226*16uA = 10.8uW. At 4MHz, = 3*0.001*500uA = 1.5uW. Total battery consumption 7.2* _less_ running the code at 4MHz....

The chip runs less efficiently at the low clock rate. Where the low clock would be used, is if you were keeping the chip awake all the time, so that it could perhaps keep time. Given that the low clock also takes your ADC out of spec, it's a 'no brainer', to use the faster clock.