Timer 3 IT does not trip

Franck26 · Joined: 29 Dec 2007 Posts: 122 Location: Ireland

Hello,

I've got a bug on a 18F2520 that I'm fighting with for few months now.

I've already ask for some help on this post (no need to read it):
http://www.ccsinfo.com/forum/viewtopic.php?t=42279&highlight=franck
At the time I thought that my problem was linked to the GIE bit, but it looks like it was not that.

I use the timer 3 overflow to generate an interrupt.
It use the internal clock (FOSC/4), the oscillator is the internal RC at 8MHZ.

Everything works fine for a while. But after 1 day, 1 week or sometimes 1 month, the interrupt stopped being triggered and I have to reset the system to get it to work again. Everything else seems to work.

I've implemented a trap in my software to check the registers that I thought would be the problem.
I'm monitoring: INTCON, RCON, T3CON, and PIE2.

Below are the bits that I think are linked with the timer 3 interrupt.
INTCON:
GIE/GIEH = 1 => Enables all unmasked interrupts
PEIE/GIEL = 1 => Enables all unmasked peripheral interrupts

RCON:
IPEN = 0 => Disable priority levels on interrupts

T3CON:
RD16 = 1 => Enables register read/write of Timer3 in one 16-bit operation
T3CCP<2:1> = 00 => Timer1 is the capture/compare clock source for the CCP modules
T3CKPS<1:0> = 00 => 1:1 Prescale value
T3SYNC = 0 => This bit is ignored. Timer3 uses the internal clock when TMR3CS = 0.
TMR3CS = 0 => Internal clock (FOSC/4)
TMR3ON = 1 => Enables Timer3

PIE2:
TMR3IE = 1 => TMR3 overflow interrupt enabled.

All those register bits seems fine, in fact they don't change from when it works and when the bug is acting.

Is there any other registers that I should check?

Thanks for any help,
Franck.

Note: I've posted the same message on the microchip website.

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

You know what I'm going to say. Post a very small test program
that is compilable, that show the problem. Something like 10-20 lines.
How do you know the test program is failing ? Explain.

Also post the following:

1. Compiler version.

2. What is the Vdd voltage ? Also describe the voltage regulator circuit.

3. Do you have bypass capacitors on the Vdd pins? What type & value ?

3. Describe the external MCLR circuit, or are you using the NOMCLR fuse ?

4. Post anything unusual about the test environment, such as the
temperature, electrical noise, static discharge, etc.

Franck26 · Joined: 29 Dec 2007 Posts: 122 Location: Ireland

Hi PCM,

Thanks for your answers.

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

Franck26 · Joined: 29 Dec 2007 Posts: 122 Location: Ireland

Hi,

1. Yes it fail in the lab, same way than on the field.

2. The time before failure is not constant (it would be too good Crying or Very sad

). The system runs 24/7 and sometimes it fails after several hours, sometimes several weeks and the bug stay until manual reset. So you can imagine how frustrating it is to run a test on this system. The worst is when someone unplug it before the weekend Shocked

!!!

3. That's a very good point, I haven't tested the bug without RS485 comm. I'll set the test on Monday.

4. I change the timer overflow periode during RS485 comm to leave some dead time between 2 RS485 trams.
The clock divisor is always the same: div=1.
I reload the timer 3 preset value at each IT. I load directly the register, I don't use set_timer3().
I don't have the code in front of me, but from what I remember TMRH is loaded first followed by TMRL which should load the 16 bits at ones.
I can post this part of the code if you think that it can help.
INT_TIMER3 is never disabled. It is enable during init and I don't touch it anymore.
I often disable INT_GLOBAL to protect some variable used both during IT and normal operation. I use the "while loop" that you suggested me to use a while ago to disable and restore GIE. At the time I thought that it was the bug...

Solution:
The watchdog is already used as normal.
Using the watchdog on timer 3 IT means reseting the system when the bug happens. This is not acceptable for this application (camera inspection). This would mean stopping the production. Now when the bug happens the system keep running, the only problem is that there is no communication and if an operator wants to change a setting he has to stop the production and reset.

I'm going to order some 18F2620 and run some test.

Thanks for your help, this is much apreciated Razz

!!!

Franck.

pad · Joined: 29 Nov 2007 Posts: 15

hey Franck26,
Restarting your old topic. Currently I am facing same problem with timer3. It has to update on every 1 ms. But after 1 day, 2 day or 30 days its value changes and starts giving interrupt of 400 to 600 us. My controller is PIC 18F85j90.
Do you find any solution regarding this.

regards
Pad

Ttelmah · Joined: 11 Mar 2010 Posts: 19496

I'm wondering if an _occasional_ mistiming, could result from the way that the timer3 value is updated.
Franc26, was doing the load himself, now when you load the upper byte, this is effectively 'latched', waiting for the lower byte to be loaded. which then triggers the transfer of both bytes into the register. He talks as if he is loading the registers both in the RS485 code, and in the interrupt. Now, what would happen if he loaded the first byte in the RS485 code, and then an interrupt occurred?. The upper byte is latched, but will then overwritten in the interrupt code when it loads the upper byte. The lower byte is then written, and the interrupt exits, and the code then loads the lower byte waiting from the first transaction. Result 'wrong value'....
However I can only see this giving 'one' wrong time, very rarely.
I would suggest though disabling interrupts round the write in code outside the interrupt.

Best Wishes

Douglas Kennedy · Joined: 07 Sep 2003 Posts: 755 Location: Florida

I had similar issues with the 2620. It would hang on the timer even with the watchdog. It rarely ran 24 hours and often froze within 5 to 6 hours. The code sequenced a water purification system and the main code was heavily interrupted by all three timers plus rda and tba and button presses.
Trouble shooting this was very unpleasant.
Protecting the 16 bit timer sets and resets from interrupts was the work around.

Ttelmah · Joined: 11 Mar 2010 Posts: 19496

Interesting.
I wonder if there is something 'deeper' in the timer logic when performing 16bit updates, so that an interrupted update, actually causes a problem in the timer circuitry.
I must admit, I'd 'automatically' disable interrupts around anything like this, just to avoid the timing problem if an interrupt does occur, but it does sound as if something worse is actually happening....

Best Wishes