Problem with Timer 1 and mega88

Christian Olsson · Joined: 08 Mar 2005 Posts: 9 Location: Sweden, Gothenburg

Hi all,
I have a problem when I changed the AT90S4433 to a Mega88
The Timer 1 did not work any more.
I am a beginner whith timers.
I have read the datasheet, but i am not getting any cleverer.
And i wonder how to change the led exactly one time per second.
Thanks in advance Christian.

DToolan · Posted: Thu Mar 31, 2005 12:18 am Post subject:

Christian,

I don't know why your code used to work and doesn't now... but timer / counters can be fairly simple. Timers and counters both do the same thing (count). The only variables are how fast and how high. How high is obviously determined by the use of 8 or 16-bit counters. How fast is the more tricky part. This is determined by 1. the system clock (crystal) and 2. the prescale value. Configuring timers as counters gives a lot more options (count up / down / trailing edge, etc) which only confuse the simplicity of timer / counters. In the simplest of configurations, a counter always counts up, overflows to zero after reaching it's max count and can trigger an overflow interrupt (this is considered "normal mode"). We'll do that first.

Since you have said that you want a 1 second interval and you are using a 16-bit timer, selecting a prescale value is determined by the calculation...
PScale = Crystal / (DesiredTimerOverflow * MaxCountOfTimer)
PScale = 1,000,000 / (1 sec * 2^16)
PScale = 1,000,000 / 65,536
PScale = 15.25 (select the next available Pscale value... not one that is lower)
PScale = 64

Now that we have determined a prescale value, we can see how many times per second (the frequency) that this timer will overflow (TOV).
TOV = (Crystal / PScale) / MaxCountOfTimer
TOV = (1,000,000 / 64) / 2^16
TOV = 15,625 / 65,536
TOV = .238

Since this is roughly .25, we can quickly see the timer will take approximately 4 seconds to overflow... or the exact time is determined by...
Time = 1 / frequency (TOV)
Time = 1 / .238
Time = 4.19

Obviously, this more time then you want. To create a 1 second overflow, you'll have to pre-load the timer with some value so it overflows faster than 4.19 seconds. Since it will take 4.19 seconds to count the full range to overflow (65,536), we can figure out how much time (at what time interval) each count takes place.
Time4EachCount = Time / MaxCountOfTimer
Time4EachCount = 4.19 / 65,536
Time4EachCount = .000064 or 64. us

You want a one second overflow so (1 sec / Time4EachCount) tells us that it will take 15,625 counts to equal 1 second. Let's subtract 15,625 from the MaxCount to tell us what our pre-load value should be.
PLoad = MaxCountOfTimer - 15,625
PLoad = 2^16 - 15,625
PLoad = 65,536 - 15,625
PLoad = 49,911 (Hex C2F7)

Ok, so now if you pre-load your timer with 49,911... there will be 15,625 counts left until overflow which will equate to the one second you desire. This probably seems like a whole lot to do in order to figure these numbers out (or at least I think so). Fortunately, there is a handy utility that does all of these calculations for you. It looks like this and it can be downloaded here.

Christian Olsson · Joined: 08 Mar 2005 Posts: 9 Location: Sweden, Gothenburg

First and foremost thank you very much, I really learned a lot about timers.

I have read the datasheet and trying to write to the Power Reduction Register - PRR with (Prr.3 = 0) but it doesn't seem to work.
The portd.1 is constant low.
I don’t know if this is the right way to write to the register.

DToolan · Posted: Tue Apr 05, 2005 10:15 pm Post subject:

Christian,

Yes, that is the right way to write the register. I've found though that the default state for the PRR register is all zeros (no power reductions enabled), so you probably won't need to change it. What you have been fighting is a Bascom bug and not some mis-coding or lack of knowledge on your part.

BASCOM is supposed to take the line, "Enable Compare1a" and with that it should set TIMSK1.1 (OCIE1A). Instead, it is mistakingly setting TIMSK0.1 (OCIE0A for the 8-bit timer/counter 0). It is also making the same mistake with "Enable Timer1" (setting TIMSK0.0 instead of the correct TIMSK1.0). The following should fix your problem until Mark corrects it (you might want to email him to let him know of the problem).

Christian Olsson · Joined: 08 Mar 2005 Posts: 9 Location: Sweden, Gothenburg

With a thing like that I could have fought for ever...
Now it works just fine, Thank you.
/Christian

RickB · Joined: 27 Jan 2005 Posts: 29 Location: Eugene, Oregon USA

Which version(s) of Bascom is this bug in?

Rick

Christian Olsson · Joined: 08 Mar 2005 Posts: 9 Location: Sweden, Gothenburg

I am using BASCOM-AVR Compiler: Version 1.11.7.7
I don’t know if there is a bug in older versions.

The Mega88 is only a year old I think,
So for that reason BASCOM versions older then a year might not have support for Mega88.
/Christian

DToolan · Posted: Fri Apr 08, 2005 11:04 pm Post subject:

Because of the similarities between the mega48 and mega88, I figured this bug would also appear when $regfile = "m48def.dat" (and it does). Just a heads-up for anyone working with the mega48 as well.