SPWM_INT - Multiple Software PWM Question

[RossWaddell]

(I'm starting a new thread as none of the ones I searched which referenced 'SPWM_INT' seemed appropriate)

I've implemented Darrel's excellent SPWM_INT include file together with ADCIN (reading a trim pot) to adjust the duty cycle for one pin (see my thread here). The problem is that (a) using 8-bit resolution (0-255) makes the LED brightness adjustment a little coarse, and (b) there's a delay in changing the duty cycle no matter how low I adjust the ADC_SAMPLEUS value. I'm willing to live with (b) but I'd like to move to 10-bit resolution to address (a).

Is it just a matter of changing

DEFINE SPWM_RES 256

to

DEFINE SPWM_RES 1024

and then altering

DEFINE ADC_BITS 8

to

DEFINE ADC_BITS 10

and

ADCON1.7 = 1

for right-justified values? Will that give me the finer control of LED brightness that I'm after?

BTW, I'm using a PIC16F1825 with 16Mhz internal oscillator setting.

Darrel's page (below) doesn't have the frequency range definitions like his earlier Multi-SPWM page, but I think this won't work with DC motors as the frequency needs to be greater than 16,000 Hz. Is that right?

DT_INTS-14 (SPWM_INT - Multiple Software PWM)
http://www.darreltaylor.com/DT_INTS-14/SPWM.html

[RossWaddell]

(I guess I really have 2 issues here: finer LED brightness control and doing PWM for a motor (since I also want to 'fade in/out' the motor RPM by changing the duty cycle). If this should be a separate thread, please let me know)..

For the motor application, I think I can configure the registers for the duty cycle (CCP1CON.4 & .5, CCPR1L) but I can't wrap my head around the frequency calculation (with 10-bit resolution at FOSC 32Mhz, I think I need PR2 = 255 with a timer prescale of 1 to get 31.25kHz judging from Table 24-5 in the PIC16F1825 datasheet). But what if I want to aim for 16-20kHz?

[Darrel Taylor]

SPWM_INT has a maximum resolution of 8-bits (0-255), as detailed on the page you linked to.

For using PWM with the CCP module, mister-e's PicMultiCalc is very useful for figuring out what frequencies and dutycycles you can get.
http://www.picbasic.co.uk/forum/cont....-PICMultiCalc

[RossWaddell]

Thank Darrel. I mis-read "PWM resolution is no longer limited to 8-bit (256)" on your web page to mean higher resolutions were possible - I guess you mean lower resolutions are now possible.

I just grabbed Mr E's MultiCalc, but it doesn't seem to output results for a 1:1 prescaler (which I think would give me 10-bit resolution for 32Mhz & 16kHz). Also, when you say it will give me available duty cycles does that mean I can't just use 0-1023 (10-bit) or 0-255 (8-bit) when incrementing the CCPRIL & CCP1CON.4 & 5 registers?

[Darrel Taylor]

No, it's not very often that you get a full 0-1023 resolution with hardware PWM.

And you can't get 16Khz from Timer2 with a 1:1 prescaler, which is why PicMultiCalc didn't show it.
With a 32Mhz OSC, the number of timer ticks required is ...

32,000,000 / 4 / 16,000 = 500

But Timer2 is only an 8-bit counter, so the max is 256 ticks till it overflows.
The next available prescaler is 1:4. So divide 500 / 4 = 125
Timer2 overflows at PR2 + 1, so subtract 1 to make it 124. This is the PR2 value for 16Khz.

The maximum dutycycle value is (PR2 + 1) * 4 because of the CCPxCON.4,5 bits.
(124 + 1) * 4 = 500
So 500 is the "Always On" point (100%), and anything above that is 100% too.

If your control value is 0-1023, it must be scaled to 0-500.

500 is slightly less than a 9-bit value, which is 1-bit better than the 8-bit resolution of SPWM_INT.

At 32kHz, you can get 0-1000 dutycycle, because a 1:1 Timer2 prescaler will give 32Khz when PR2 = 249.

[RossWaddell]

If I switch to 16Mhz OSC I can get 10-bit resolution at 16kHz:



So here's the thing: if I set up my ADCIN to use 10-bit then presumably I can get 0-1023 values from my trim pot but since 1000 is the max duty cycle anything above that changes nothing. Will it cause a problem if I pass that in to the duty cycle registers (e.g. 1012) or do I need to set a programmatic limit at 1000? Same applies to 8-bit - max duty cycle is 250 but 8-bit resolution on ADCIN would give me 0-255, so anything greater than 250 does nothing.

[RossWaddell]

The maximum dutycycle value is (PR2 + 1) * 4 because of the CCPxCON.4,5 bits.
(124 + 1) * 4 = 500
So 500 is the "Always On" point (100%), and anything above that is 100% too.

Is that calculation dependent somehow on the prescaler? I'd like to programmatically determine the max duty cycle available to me with PR2=249 and prescaler of 1:1 (16Mhz OSC, 16,000Hz freq). With Mister E's MultiCalc it shows me '1000' is the max duty cycle for these paramters, which also works with (PR2 + 1)*4 where PR2 = 249 but is that always the rule?