Multiple PWM and port C

[HenrikOlsson]

Hi,
The reason it runs slow is due to the interrupt driven PWM routines. You have a PWM frequency of 200Hz and a resolution of 256 steps, this results in an interrupt frequency of 51200Hz.
According to Darrels page (if I'm on the right one) each "PWM-channel" needs 20 instructions - in your case 51200 * 20 * 2 = 2,048,000 instructions per second, just for running the 2 PWM channels. At 20MHz there's 5,000,000 instructions per second to play with which means that 40% of the processors time is spent running the PWM.

You tell PBP that you're running at 20MHz but then the assembly interrupts silently eats away 40%.

In this case each and every Hz you decrease the PWM frequency will give more time to the actual PBP program, if 100Hz is "all" you need then don't try running it at 200Hz. But again, it doesn't seem to be the timing per se that makes it NOT work since it using different ports for the PWM and I2C works.

/Henrik.

[Scampy]

Thanks for the explanation...

I think I'll go back to using port D for the outputs and revert back to the mulit_swpm file. Not ideal, as I'll have to modify the PCB, but at least it works without the delays or screwing up the clock etc. If I get chance I'll try this on the old 16F877A chip rather than an 18F to see if this issue is chip related...

[Scampy]

I've recorded the wave form when using the hardware PWM modules on the chip to explain the issue I'm getting.



If you compare this to using the multi_swpm file and port D you notice there are loads of spikes in the HWPM signal and some harmonics ? - the amplitude is also a lot less (I had to increase the volume for the HWPM to get a similar trace)

[HenrikOlsson]

As I've said a couple of times now:
You can not run the hardware PWM at a frequency below 1221Hz - it's not a bug in PBP and not a hardware issue with the PIC or the circuit, it's just the way it is.

The reason is because TMR2, which is used for the PWM timebase is 8 bits wide and has a prescaler with a maximum ratio of 1:16. At 20MHz the timer ticks at 5MHz, so 5,000,000/256/16=1220.7Hz. If you try anything lower than that it will not work properly and I'm pretty sure that's what you're seeing when trying to run it at 200Hz.

To prove or disprove, try it again at 1500Hz and see if the waveform looks better. I understand that it won't work in your particular application due to the LED drivers but it'll show that actual problem is that you're trying to run the PWM at a frequency (200Hz) which the module is incapable of producing.

/Henrik.