generating PWM from timers --> debugging generated freque

foxOnTheRun · Joined: 17 Apr 2010 Posts: 43

Ok, I was reclutant to post this, but untill now I haven't found an explanation.

I'm using timer0 (or timer1), every overflow I just toggle the value of a pin, so I can generate a pwm signal, nothing fancy (PIC16F690).

I'm expecting to see:

Starting value: set_timer1(65436);
100 counts to overflow; 8mhz, 1 increments every 4 clock.
So, one effective timer increment every 4/8000000 --> 0.5us

pwm is set at 50%, so 100 ticks UP, 100 ticks DOWN ==> 200 ticks_period * 0.5us = 100us duration

F = 1/100us = 10kHz square wave - correct assumptions up to here?

But I'm getting 7.03 kHz using timer 1, and 6,98kHz using timer0 (measured by a multimeter with freq mode). As a check, if I setup the internal pwm generator to whatever freq, I can get a matching reading.

I know, writing a post about pwm+timer+settings is as dangerous as getting hit by a bullet train carrying a backpack full of knives.. but.. Clearly, I'm missing something :\ - but what? :roll:

This is the minimum, compilable code which reproduce the issue:

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

Look at the #int_timer1 routine in the following link, and notice how it
compensates for interrupt latency by reading the current timer value
and adding it to the preload value.
http://www.ccsinfo.com/forum/viewtopic.php?t=45554&start=1

Remember, before the program gets to the isr, it has to go through the
first half of the CCS interrupt handler, and this may take 25 to 30
instructions (at 1/2 usec each for an 8 MHz oscillator). So about 12 to 15
usec passes before you can reload the timer in the isr. So by reading
the timer in the isr, you find the amount of time that has already passed
after the interrupt fired, and in effect, subtract it from the time before the
next interrupt fires.

Also notice how this is done at the very beginning of the isr. Your code
loads Timer1 within the isr in two places. There is no need for this.
You can move it to the top of the isr as shown in the example.

foxOnTheRun · Joined: 17 Apr 2010 Posts: 43

Oh my!

You were.. ARE right!

To tell the truth, before posting I tried a slower PWM, like having an int after a full count of timer0, and it was interrupting/overflowing correctly every 32ms.. but I didn't persisted :\

To get a nice 10kHz pwm I use: