Sleep on pic24 with RTCC not working

[JMarysse]

I want to wakeup the pic every 10 seconds.
This is what i see on the terminal:
Program started
Friday 27/10/2017 09:02:00
Wakeup from sleep => after ±4 seconds
Friday 27/10/2017 09:02:00
Wakeup from sleep => after ±4 seconds
...
after 7 times this is as the program runs WITHOUT the sleep.

Please can someone advise ?
I have CCS PCWHD compiler V5.075

Thanks!

This is the code :

main.h :

[temtronic]

1st thought.
Have you actually got the RTC enabled? I don't use that PIC, but I'd like to see a small program that displays the RTC time say every second...

main()
{
display rtc time
delay_ms(1000)
}

a simple program, no interrupts, just read and display rtc data.

2nd thought..
get rid of WDT. It shouldn't be enabled until your product is ready to be shipped.

3rd thought..
if running through an ICD, several fuses and registers will NOT be correct.....you need to run in 'release' mode.

Jay

[JMarysse]

Hi temtronic,
Thanks for your answer.
1. if i don't use the sleep function, the RTC is running well.
2. if i disable the WDT, then the PIC does NOT wakeup anymore
3. i use the PicKit3.

[Ttelmah]

That it doesn't wake without the watchdog, tells you this is waking the chip.
You have 1mSec selected as the clock (32 prescaler), and 4096 selected as the divider. So the watchdog timeout is 4.096 seconds. Exactly what you are seeing.....

Obviously you need the watchdog to be set longer than the RTC timeout.

Then start by testing the RTC with the chip awake. Does the alarm trigger?.

I think you will find you have the RTC running much slower than you want. You are using a divider or 7837 * 256 = 2006272. You are meant to select the division to give 2Hz. You are running off 31KHz, so want 15500. So your 10 second alarm will trigger at about 1290 seconds.....

[JMarysse]

Hi Ttelmah,
Thank you for this info!
I will check/change this on Monday and keep you informed.
Have a nice weekend.

[JMarysse]

Hi Ttelmah,

I did 2 tests:
=> 1 with rtc_int routine
=> 1 with sleep.
The result is that in test 1 the RTC is running perfect, in test 2 it is indeed very slow...and is also it is the WD which is waking the device after 16 seconds.
Can not see why this happens, please help.
This is the code:
main.h

[JMarysse]

Hi Ttelmah,
The problem is solved.
I did not read your mail very well (monday morning...)

I changed the following line :

[jeremiah]

The key there is you need an oscillator that can run while the chip sleeps. The traditional internal oscillator does not. The LPRC will (on chips that have it) but usually it is very bad tolerance (Upwards of + or - 15% on some chips). I would normally recommend using an external oscillator tied to an asynchronous timer input. Normally for PIC24's this is timer 1, though I haven't checked your chip's datasheet.

If your LPRC has good enough tolerance then it isn't an issue though. Just make sure you check the datasheet to be sure.

[JMarysse]

New question : we need to communicate with a NINA-B112 module trough the UART.
Problem is that the modules uart is default set at 115200.
We use the INTERNAL 8Mhz osc and the compiler does NOT allow 115200 baud.

All comment/help is welcome.

[Ttelmah]

You need to be clocking faster.

From 8MHz, there just aren't enough clock cycles available to get reasonably serial accuracy at this rate.

If you look in the data sheet, they give how to calculate the error. In this case:
(remember the UART is clocked off FOSC/2)
U1BRG = ((4000000/115200)/4)-1 = 7
Baud = 4000000/(4*8) = 125000
Error = (125000-115200)/115200 = 0.085

8.5%

The maximum total error on a async link should be kept below perhaps 3%. The compiler will reject any setup that results in such a large error.

Using 16Mhz.
U1BRG = 16
Baud = 117647
Error = 0.0212

2.1%

Still not 'great', but close enough that it has a chance of working.

There is an option 'BRGH1OK', that on older chips allowed the /4 division (shown) to be used, but on PIC24's this is automatically selected, so this won't help (the calculation above is using this).

[temtronic]

The 'other' problem with using an internal clock even IF it is fast enough, has to do with temperature. You need to read the datasheet for your PIC and decide IF the clock will be within 'spec' for the temperature range your product will see. Often it's just OK, on the bench, in a warm building BUT put into an industrial/commercial site where the temperature drops, say due to setback thermostats or it gets really, really hot in the summer and the internal clock may easily go out of the 'OK' range.
Older PICs had a wide 'range' for say '4 MHz', some newer PICs are better and some 'families' are better yet so you MUST read the datasheet !

To get accurate, high speed (115K200) operation from most PICs you need an external clock, xtal and caps or canned xtal.
Jay

[JMarysse]

Hello,
Thanks for your replies!

So we will need to add an external xtal.
No way to do it with the 8 MHz internal one, not even if using the PLL?

[temtronic]

you need to download the datasheet and read the section on 'clock selection'.

I did a quik read and it may be possible to use the internal FRC with e PLL and get 64MHz operation.

Since I don't have that PIC I can't cut code/compile and test.

Try it...

Jay

[Ttelmah]

Key point to understand is that errors 'add up' in the worst case. Yes you can run at 16MHz off the internal oscillator. It'll then work, but the error will be 2.1% from the baud rate calculations, and potentially +/-1% from the oscillator. This may well give you problems. Go up to 32Mhz from the internal oscillator, and the baud error will drop to only a fraction of a percent, and it should work OK, but will be drawing a lot more power. You have been using sleep, which potentially implies low power is required?. If so, then look at an external crystal like 7.3728MHz. This will divide to give 115200bps _exactly_, and your baud accuracy will be limited just by the accuracy of the crystal.

[JMarysse]

Hi Ttelmah, temtronic,

I just made it work by indeed setting the fuses FRC_PLL and PLL2, and set the "use delay(clock=32M,internal,restart_wdt).

But you both are correct, we will need to redesign and use an external crystal.
Also "low power" will be needed because the system is to be used as a standalone data logger (4 MLX90614 sensors, 2 TSIC716 sensors, which are already implemented). The power of these sensor is controlled by the pic.

Thanks for your replies!