PWM issues

[Lord Grezington]

Hello All

I am trying to use the PWM outputs form the dsPIC33EP32MC504 to gernerate various waves from data stored in a LUT. In general it works very well, however I seem to be getting glitches on the falling edges (aways seems to be on falling, and never rising).

As you can see in the triangle wave example, as the PWM duty is reducing, for some reason the duty glitches in the high state so stays on longer than it should (never in the low). The glitch also seems to move accross the peroid of the waves on a sort of pattern that repeats every few periods.

Things I have tried:
I have tried chaning the clock sourse from the PWM component
I have tried changing the PWM frequencies.
I have tried changing the number of values within the LUT per period of the sine wave
I have tied using both pull and and pull down reistors.
I have tried changing all the values between bytes and ints (for some reason when the addresses are ints, this resets the whole program every few seconds).

I have added two examples. One where you can see both the PWM generated and the trianlge wave (the same signal but run via a filter). The other one is a sample of two different PWM modules filtered running together and showing a similar patten.

Any ideas will be useful, I have attached my test program.

Thanks

glitch Same signal.jpg (122.73 KiB) Viewed 2275 times

Two diffenet PWM's with pattern.jpg (150.68 KiB) Viewed 2275 times

[QMESAR]

Hi.
In my humble opinion the problem is a PWM Resolution problem
In basics the PWM has a specific resolution for a given frequency and if there is a small mismatch between the set period and set DC you will see this in any signal generation probably not when driving for example a DC bridge for a DC motor but if you generate a signal (which has level and phase )
you will see this clearly .
If I would do this I will use a C Icon and use the correct vales for Period and duty cycle which correspondence to the PWM Resolution for that frequency.
I know this sounds weird but the PWM is a tricky module on the dsPIC or any higher level micro

my 2 cents on this without investigating the issue deeper.

[Lord Grezington]

Thank you QMesar

I have done what I can with attempting to vary the perieods and frequencies with no luck. I think what I will try next is use the high speed PWM peripherals on the dsPIC33EP32MC504 with C (rather than the PWM component) to see if the issue persists.

Thanks

[QMESAR]

That would be a good thing and make sure that you have the correct values for Period and DC as calculated by the Resolution ,check the PWM section in the datasheet very carefully

example (just numbers for explanation)
100KHz Period 4099 and then 50% Dc will be 2049 and so on obviously 100% Dc will be 4099 value written to the DC registers
20KHZ Period 3500 then 50% Dc will be 1750 and 100% will be 1750

Goodluck

[Benj]

Hello,

I've seen these glitches before. It basically happens when you write a lower duty but the timer has already couted past the value you set. And so instead of immediatley outputting low the output remains high for the rest of the period.

When writing to the PWM duty try resetting the timer count register to 0 using a c icon and see if that makes any difference. I might make this part of the component CAL code if that is what's happening to save this problem appearing again.

[Benj]

I've pushed a change to the update system that should solve the problem automatically for you.

This is the fix I have implemented when changing the duty.

Code: Select all

	#if (MX_PWM_TMR_X == 1)
	  if (duty <= TMR1)
	    TMR1 = 0;
	#endif
	#if (MX_PWM_TMR_X == 2)
	  if (duty <= TMR2)
	    TMR2 = 0;
	#endif
	#if (MX_PWM_TMR_X == 3)
	  if (duty <= TMR3)
	    TMR3 = 0;
	#endif
	#if (MX_PWM_TMR_X == 4)
	  if (duty <= TMR4)
	    TMR4 = 0;
	#endif
	#if (MX_PWM_TMR_X == 5)
	  if (duty <= TMR5)
	    TMR5 = 0;
	#endif

Let us know how you get on.