RTC on PIC 24FJ64GA306 drifting

JamesW · Joined: 23 Apr 2007 Posts: 91 Location: Rochester, England

Hi folks,

Very quick question - I'm running the internal RTC on a PIC24FJ64GA306 and it's drifting badly. I set the time at 1pm and it is now running 50 minutes slow (@ 7pm)

I'm running it on a bog standard watch crystal (uk.farnell.com part 1652573 )

With a couple of 15pF caps across it - it stops completely, but by removing them it does run - albeit slowly.

Can anyone offer any pointers - the datasheet doesn't give any concrete values to use, and I'm not sure if I'm able to tune/trim it in code.

Thanks in advance.

James

temtronic · Posted: Mon Dec 19, 2016 1:00 pm

slow... I'd say too much cap.... also check PCB. It MUST be CLEAN, use qtip and alcohol to clean ANY solder residue.
I'd have thought Microchip would specify xtal/cap specs 'somewhere' in teh 500-600 pages though.....

Jay

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

Your crystal has an ESR of 35K. Microchip appnote AN1798 says the
it should be 50K.

It also says the recommeded capacitors are 22 pf:

newguy · Joined: 24 Jun 2004 Posts: 1907

Layout and the crystal caps are big issues. The one time I had to implement a RTC I managed to get a clock accuracy of about +/- 3 seconds over a weekend, or about 1 second per day with nearby circuitry active/running. With that circuitry off and the RTC operating from a backup battery I got about +/- 2 seconds per week. ....At room temperature.

It was one of those instances of ehhh, good enough.

JamesW · Joined: 23 Apr 2007 Posts: 91 Location: Rochester, England

Thanks guys,

The problem appears to be a lot better - albeit I wouldn't say it is solved.

I've fitted a different crystal, and ended up with 33pF rather than 22pF caps. With 22pF the clock still wouldn't run.

It's been on now for 3 hours and has lost about 4 seconds. (Which is a definite improvement over the 30 minutes it would have been previously.

I will order in a crystal with and ESR of 50K, and see if this makes things even better.

James

temtronic · Posted: Tue Dec 20, 2016 7:03 am

I went through the same 'exercise',got very frustrated and ended up using a discrete RTC chip/battery combination ( DS1307). While it's a few pennies more than xtal/caps it works very,very well. Also has a bit of backedup RAM and 1HZ interrupt.
Perhaps an option, at least look into the 'specs and space' for it.

Jay

gaugeguy · Joined: 05 Apr 2011 Posts: 303

Are you using the low power or high power mode? High power mode would be preferred for better stability and accuracy.
Are you using the I/O pin adjacent to the SOSC pins? Any switching on this pin can inject noise into the clock.
It can also at times help to add a series resistor into the oscillator circuit.

JamesW · Joined: 23 Apr 2007 Posts: 91 Location: Rochester, England

That is a very interesting question. How do I know what mode the oscillator is running in?

I currently have

gaugeguy · Joined: 05 Apr 2011 Posts: 303

Section 9.5.2 of DS39996F discussed CW3<9:8> determining operation and power mode.

Section 9.5.2 of DS39996G has the low power section omitted and only defines one bit CW3<8>

It appears I was looking at an older version data sheet...

JamesW · Joined: 23 Apr 2007 Posts: 91 Location: Rochester, England

Ordered one of these 2064057 from Farnell.

Allegedly it's 50K ohm - so will try it tomorrow and see what happens.

After 8 hours of settings - its about 30 seconds behind.

Thanks for all the help

James