PIC18F2550 - Timer1 Interrupt causing error in Timer0

mfeinstein · Joined: 05 Jul 2012 Posts: 35

Hi everyone!
I am making a project using the PIC18F2550 and PCW 4.120. As you can see in the code below I have a Timer0 interrupt and a Timer1 Interrupt. The Timer1 Interrupt is currently disabled and the Timer0 Interrupt is expected to be called every 500 microseconds. Instead, the Timer0 Interrupt is being called every 250 milliseconds...after hours looking at the code, searching for a flaw, I finally found out that actually the delay_ms(75) in the Timer1 is triggering this flaw. I don't have any idea why it is happening, but it is! I commented all the delays in the Timer1 and it work as expected...but leaving them in the code makes the Timer0 go nuts....the weirdest thing is that Timer1 is disabled and every piece of code I place inside Timer1 is never executed....

Any ideas how to fix it? I don't know what else to do :/

Thank you very much!!

EDIT: Since many people were complaining, I posted a simpler code, just to show the real problem...so dont try to look at the code, figuring what it does functionally...just look at the Timers problem (Timer1 is disabled but its delay function makes the Timer0 interrupt be called late)

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

asmboy · Joined: 20 Nov 2007 Posts: 2128 Location: albany ny

1-learn to use the CODE button when posting( ah fixed -good)

2- where do you enable timer1 ints? ( answer=nowhere in this post)
the delays in the int won't hurt you if the int handler is not enabled

3-a more organized approach to SETUP of TRIS and timers would benefit you

just some EZ observations'

mfeinstein · Joined: 05 Jul 2012 Posts: 35

PCM programmer:
As you can see in the code, the Timer1 Interrupt is Disabled, so even if I had a delay_ms(1000) inside of it, this delay will never be executed, so this isnt the reason why I have this problem (as I already explained in my previous post)....as a matter of fact, even if I have just a delay_ms(75) and nothing else inside of the Timer1 Interruptor I still get this error...
again: Timer1 is disabled...no code is being run...but it still causes this error, if you have any doubts, you can try this code and simulate it yourself...

asmboy:

1 - Sorry, I was in a hurry

2 - That's the whole thing...I dont! It is disabled and it is hurting me! if I comment it, the Timer0 work as expected, but if the delay line is in the Timer1 code (even if that's and unreachable code) the Timer0 works flawlessly

3 - What do you mean? This is actually not my complete code, I made it simpler (since posting 870 lines didnt sound as a good idea :P )

4 - Do you have any ideas how to fix it?

Thank you all!

PS: Ttelmah this is the moment where you show up and save the day ; ) hahaha

asmboy · Joined: 20 Nov 2007 Posts: 2128 Location: albany ny

configuraSlewRate() is a functional mess.....
setup the ISR INT handler before setting up the timer hardware ???

you should 1- setup hardware 2- clear the int flag , then 3- THEN enable the ISR

AND
no return is required from an ISR or any callable function, unless you want an early exit

also

you might have a close look at ALL the fuses for that pic and make sure you are setting the part up, correctly for your intended use.

also PCM is 1000% correct in this advice:
never put a fixed delay of more than a few cycles or usecs in ANY ISR.
ONLY trouble flows from ignoring that

mfeinstein · Joined: 05 Jul 2012 Posts: 35

asmboy,
Thank you for your reply... as I said before, this is a 870 lines code, so I erased most of it to post in here, isolating just the minimum parts to show you guys the problem, and yes, it became a mess after the "edition"... (the return is indeed an early exit to debug the interruption...and if you look closer ALL the interrupts are disabled until EVERYTHING is already setup as it should...I just didnt clear the interrupt flags as you said)...

BUT, this has nothing to do with the actual problem...I just posted this code in here so people with more knowledge than me might figure out what is going on...so please dont focus on the code organization, just try to figure out why the Timer1 is messing with Timer0 even if it is inoperative...so far your advices were very productive in concern with "code organization" and "good programming techniques" but this is not the subject matter of this post...

Best Regards

Ttelmah · Joined: 11 Mar 2010 Posts: 19550

Key point is re-entrancy.

The PIC does not support re-entrant code. Hence if _any_ complex function, is used inside an interrupt, and also outside the interrupt, then interrupts have to be disabled in the external copy. This will affect everything else that uses interrupts. Since the compiler _must_ ensure this is done, it will disable interrupts this way, even if the interrupt is never called.

Hence, as soon as you have a delay function inside an ISR, and the same function in the main code, interrupts _will_ be disabled in the external copy, _even if the interrupt is not enabled_. So if you have a routine for timer0, using interrupts, this will have to wait till the delay_ms(1000) in the external code completes, before it'll be called.

Seriously, your code, is a 'bad example' of how to debug a problem. 90% of it is not needed to show what you are worrying about, and what is posted, won't actually display the problem, since it lacks the timer0 initialisation...
As another comment, is ADC_CLOCK_INTERNAL, legal for a chip running at this speed?.

Best Wishes

Mike Walne · Joined: 19 Feb 2004 Posts: 1785 Location: Boston Spa UK

mfeinstein · Joined: 05 Jul 2012 Posts: 35

Mike Walne,
1 - I am not familiar with PIC assembly code, and the code generated will be huge in any case (its more than 870 lines of C).
2 - I cant use other pins, the board is already full and functional.
3 - The code is well organized...but it is in Portuguese, so the function names dont make much sense for you guys, and they look confuse, but there is a reason for it...I am sorry :P I tried to delete most of the parts, but I can see the example was still a bad one, sorry again!

Ttelmah,
As usual, thank you my friend! Could you give me some links to study more about reentrancy in the PIC?
I do have the Timer0 initialization...I dont understand why you said I dont... its the

FvM · Joined: 27 Aug 2008 Posts: 2337 Location: Germany

I can't see the point where Timer1 get's enabled in the present code. I even don't see PIE1 referenced in the compiled code of V4.120. And there are no known silicon erratas related to a similar problem.

So for the time being, I'll assume that the problem can't be reproduced with the posted code.

To find out how timer1 ISR can be called though, hardware debugging is the most promising way.

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

Errors are often related to probability. The more code the higher the probability. This board would ideally like all issues reduced to one line of code. Most accept this as impracticable so a few lines is next best. Dozens of lines does violence to the concept of paring down the issue. Statistically 99% of issues are of the coders own making. Breaking the code down into simple blocks and testing them one by one is the analytical approach. This board offers the opportunity of learning from the mistakes of others but more importantly learning from your own mistakes by the process of simplifying your code so you can spot your own mistake.When TV's had tubes sometimes they could be made to work by just thumping them. No electrical knowledge was needed. Posting dozens of lines of code culled from a larger body of code without even bothering to test them as exhibiting the issue might be considered by some as the thumping the TV approach to resolving issues.

mfeinstein · Joined: 05 Jul 2012 Posts: 35

FvM...please read through the post...I already said it before, many times indeed, that Timer1 ISNT enabled at all AND it's causing an error (and that's the whole question.."how may it be possible?")

asmboy · Joined: 20 Nov 2007 Posts: 2128 Location: albany ny

here is the way to converge on a solution then:

keep on reducing the code to:
the FEWEST lines that WILL compile - AND still .
show the problem.

get rid of EVERYTHING you can - and still see the problem.

!! START WITH the RETURN in your TIMER0 ISR !!

I C 2 much forest in your code to let ME see the rotten tree i guess.

besides - i have my own project to debug this afternoon Very Happy

mfeinstein · Joined: 05 Jul 2012 Posts: 35

asmboy,
I edited the code so you all could see it in a more simple way.

Ttelmah,
I agree with you that this is a reentrancy problem, but still I dont undestand why the PIC interrupts at 250ms instead of 1000ms, as the delay_ms should be executing and then the Timer0 will be called...
Thank you!

jeremiah · Joined: 20 Jul 2010 Posts: 1357

You're not understanding what they are saying. It absolutely does NOT matter that you disabled timer1. The fact you have the delay statement in the timer code at all (regardless of disabled/enabled) tells the compiler that you could enable it at some point, so it automatically disables all interrupts while any delay statement, which will mess up timer0 isr timing. That's why the behavior changes when you comment the line out. Even if you never ever ever enable the timer1 interrupt, the code there will cause the compiler to disable all interrupts while in delay_ms().