Counter PWM with no jitter?

Dave Matthews

I need to output a signal that will span from 1.0 to 1kHz that will have a duty cycle span of 1 to 99%. I thought I could do that with the Counters, but I had less than good results. Perhaps it is my bad coding or perhaps it is because "for values of FRQA which are not a power of 2, there will be jitter present on the output signal since the most significant bit of PHSA will toggle at an inconstant rate". (Propeller application note AN001, page 17.)

I tried a few examples in the OBEX for counter use to provide such PWM with duty cycle control, but they all had jitter. The spin examples work fine.

Am I correct in my conclusion that using counters exclusively will not give me a jitter free signal for my needs?

Dave

tonyp12

Use PLL Mode, The fixed multiply by 16 and then a user selectable divide by 16 should smooth things out

jmg

If you want to control Frequency and duty cycle, then you effectively need to alternate ON time, then OFF time.
That's probably easier with just WAITCNT.

There will be a SW limit on how short one can be, but 1kHz 1% is 10us, which will be fine in PASM.

With On/Off time loading there should be no jitter between cycles, and the granularity will be small at 1Hz~1kHz
(I think ~12 MilliHertz)

Phil Pilgrim (PhiPi)

tonyp12 wrote:

Use PLL Mode, The fixed multiply by 16 and then a user selectable divide by 16 should smooth things out

You can't get there with PLL. The NCO clock has to be between 4 MHz and 8 MHz for the PLL to lock. 4 MHz / 16 = 250 kHz. He needs 1 Hz to 1 kHz.

I'm with jmg: do it in software with PASM.

-Phil

Duane C. Johnson

While the software solution should have no jitter when running there may be a big phase jump when changing frequency or duty cycle.

Duane J

Peter Jakacki

Phil Pilgrim (PhiPi) wrote: »

You can't get there with PLL. The NCO clock has to be between 4 MHz and 8 MHz for the PLL to lock. 4 MHz / 16 = 250 kHz. He needs 1 Hz to 1 kHz.

I'm with jmg: do it in software with PASM.

-Phil

I think I've tested the 8 channel 8-bit PWM module in my Tachyon to 5kHz, It's table based so a 256 byte table holds the pattern for 8 channels.

Ariba

I think the best approach is to use a software loop for the frequency and a counter to generate the pulslength.
At such low frequencies you can do it all in Spin:

CON
  _clkmode  = xtal1 + pll16x
  _xinfreq  = 5_000_000

   OUT_PIN = 16

VAR
  long  freq, width
  long  stack[16]
    
PUB Main
  freq := 1000   'Hz
  width := 33    '%
  cognew(pwm(OUT_PIN),@stack)

  repeat
    'you can change freq and width here

PUB pwm(pin) : time | period, puls
  ctra := %00100<<26 + pin
  frqa := 1
  dira[pin] := 1
  time := cnt
  repeat
    period := clkfreq/freq
    puls := period * width / 100
    time += period
    waitcnt(time)
    phsa := -puls
'

Andy

jmg

Ariba wrote: »

I think the best approach is to use a software loop for the frequency and a counter to generate the pulslength.

A benefit of this approach, is a second (same freq) PWM comes almost for free by using the 2nd counter.

I can't see a way to avoid a small phase shift on the two though ?

kuroneko

Ariba wrote: »

I think the best approach is to use a software loop for the frequency and a counter to generate the pulslength.

This may not be an issue for the OP but

waitcnt(time)
    phsa := -puls

will suffer from hub aliasing. IOW +1 for a pure PASM solution.

Ariba

Hmm... if it is a problem then this modification may help:

period := (clkfreq/freq) & $FFFFFFF0

It will add a little frequency error of max 0.02% but should remove any jitter.

Andy

Dave Matthews

Thank you for the example code, I will learn from it!

Dave Matthews

Thanks everyone for the excellent and prompt help and advice. And extra thanks to Ariba for the code, I have done some initial testing and it works fine. This forum is a great resource!

Dave

JonnyMac

I've attached a simple dual-motor object that uses a code loop like Andy showed you. I use a variation of this (with more resolution) in a friend's commercial pan/tilt controller for movie cameras.

Dave Matthews

Thanks Jon, very much appreciate this example, and -all- of your examples!
Dave