Software PWM 40KHz with variable duty cycle

[reddyysnr]

Hi all,

I'm new to PIC and have worked with the basics on PIC. I am trying to explore features like Timers, software PWM etc.

I'm using 16F677 running at 8MHz internal clock. I am trying to generate a software PWM of frequency 40KHz. I have seen methods which use

[Ttelmah]

Not possible.
Just too fast.
It takes _time_ to setup timers, read timers, detect timers have expired. 40KHz, is a total cycle of just 25uSec. Just 50 processor cycles.
Even using the delays approach, you could only do a very limited range of cycle durations.
Seriously, the whole point of hardware PWM's, is to allow frequencies above what software can handle and allow reliable operation while the processor is doing other things.

Best Wishes

[Mike Walne]

Sit down and do the maths.
(With the data sheet if you need it)

You want 40kHz frequency with an 8MHz clock.
That defines your range of duty cycles using the PIC's built in hardware PWM.

With your posted code you've totally ignored the time taken to perform the output_xxx(pin) functions.

Mike

[PCM programmer]

I suspect you want to make an IR TV remote transmitter. If that's what
you want to do, you should tell us.

[FvM]

PIC16F677 can do hardware PWM for a single half- or full bridge. Modulation schemes are limited to basic 2-level, however.

[reddyysnr]

@Ttelmah,@Mike Walne

Correct me if I'm wrong.

At 8MHz, every instruction cycle takes 0.5uS. (I read on some page that 1/(clock/4) gives the time period of one instruction cycle)
If we need to generate something which has a duty cycle of 25uS, why wouldn't that be enough?

Even with the approach of

[Mike Walne]

You said this in your first post.

[Ttelmah]

First understand that a single C instruction is lots of PIC instructions. Something like a multiply, _hundreds_. However even something simple like reading a timer register, for a 16bit timer, has to select the bank used, read an eight bit byte, select another bank, and write this to a RAM location, then do the same for the second byte. Just loading a timer, with a value from a variable, will take something like ten machine instructions. 0.5uSec/instruction, is only 50 instructions in 25uSec. Yes, you could generate a 40KHz pulse, but using two variable widths, no. Using a timer, and setting/reading it, will just take too long. You could have two delay based routines, giving two different pulse widths, and switch physically between these.

delay_us(variable), has a minimum time it can generate (just the time needed to access the variable and test), and a minimum 'step size'. Minimum time, for your processor will be 5uSec, and 1.5uSec 'step'.

Then simply outputting a level will take typically four instructions. So, you could have a couple of routines to develop say a 35% duty cycle, and a 65% duty cycle, adjust all the timings to work, and switch between them, but how are you then going to find the time to do this switch?.
Seriously, play with the hardware PWM, that is what it is _for_.
With this, the hardware generates all the timings once told, so you can do the jobs like making a decision to change pulse width....

Now, there are ways of being 'ingenious', and getting close, Primarily (funnily), using Timer2. Alone of the 'standard' timers, it can handle generating short timings, with automatic resetting. It's designed this way, to drive a PWM.....

[asmboy]

WHAT are you actually trying to DO !!!

unless your goal is to simply get a good grade,
with knowledge of your specific design task,
much more help could come your way.

but with the opaque presentation so far- not so likely

[reddyysnr]

@Ttelmah: That was pretty helpful. Though I didn't understand a few things, I'll use them as pointers to try and explore. I use Proteus to simulate; I'll try out MPLab..

@asmboy: There's nothing at all to be hidden.. I'm not working on any secret project. Like I said, I'm just trying to experiment with software PWM. 40KHz was a number I found interesting because I see it's used in IR Protocols, Ultrasound Transceivers and also some AC lamp dimmers.. If I succeed, I MIGHT take up and any of those in the future.. and if 40KHz isn't possible, my goal is to achieve the maximum possible frequency..

[Mike Walne]

asmboy is not suggesting you are purposely hiding anything.
Your questions are so vague we can't figure out what you are trying to do.
You learn by having a play with REAL hardware.
FORGET PROTEUS. It will cause more problems than it solves.
If you must simulate, use MPLAB.

Like I said before the answers to your PWM questions are all in the data sheet.
Ttelmah has explained the software PWM limitations.

If you want to work at 40kHz with an 8MHz clock the hardware PWM is the sensible way to go.
That's a 25us period.
The duty period resolution will be 125ns.
So you will be able to go from duty period of zero, through 125ns, 250ns in 125ns steps.
All the way to 24.750us, 24.875us and 25us.
That's 201 steps or ~0.5% each.
If you want more resolution:-
(1) you use a faster clock for 40kHz PWM or
(2) accept lower frequency PWM.

Mike

[reddyysnr]

Thanks Mike.. I'm on it.. Sorry for being vague.. Next time I have a doubt, I'll post clearly..