manual hardware PWM setup on 16F1824

[comwarrior]

ok, here is what i want...
OSC = 32MHz
PWM bits = 8
PWM frequency: as low as possible but grater than 200Hz
PWM to use TMR6

ok, so, most of it i have done and figured out by scouring the datasheet for the last week.

However, the fun bit is the PWM bits... nowhere does it tell you how to set it to 8 bits not 10...
How is this done?

Also, TMR2 or 4 or 6 are capable of a 1:64 prescaler, but the PWM section referes to the maximum prescaler of 1:16.... why can't i use 1:64?

Also, be aware i'm using all 4 PWMs including the two enhanced mode PWM's in standard mode with default port assignments.

Also, i can't quite understand the 'formulas' in the datasheet for calculating PWM values.

Any help appreciated.
Thanks

[mister_e]

Did you already used PicMultiCalc for PWM on previous PIC before?

I haven't open the datasheet for this specific one, but I think the PWM formulas shouldn't be that different from others. There's quite a few thread about that in here (try http://www.picbasic.co.uk/forum/show...ht=duty&page=2 ). Fortunately you have other PIC on hand for testing purpose?

I would start with a single output and then, play with related registers to add some outputs.

[comwarrior]

that doesn't help...
I looked through that thread, first off your website seems to be on strike, so couldn't get that calculator application thing...
Secondly... i compared the PWM section of the 16F1824 datasheet witht the 18F4550 PWM section of the datasheet. The differences were number of PWM's and clock selection...

Other than that the sections on duty cycle and resolutions were almost the same, so it's not chip specific...

How do i make the PWM's 8 bit not 10 bit?

[ScaleRobotics]

Here is the location for Steve's Multi-Calc. It is quite helpful.

http://www.picbasic.co.uk/forum/cont....-PICMultiCalc

If you wanted 8 bit PWM using the hardware, you could shift your 8 bits to change it to 10, matching the hardware.

something like this:

DUTY1 var word
 
DUTY1 = 255                    'put in some 8 bit value
DUTY1 = DUTY1 <<2          'shift your 8 bits to 10 bits
CCP1CON.5=DUTY1.1       ' load CCP1 duty value            
 CCP1CON.4=DUTY1.0       '      with Duty1 
CCPR1L=DUTY1>>2

[cncmachineguy]

This is not a perfect solution, but you could just leave the 2 LSB's set to 0, then just use 8 bits for the MSB's. This will allow you to set PWm to zero, but not 100%

[comwarrior]

or... i could just times the 8 bit duty by 4...

but still, why does the datasheet, even the one for the 18F4550 'indicates' a variable bit resolution...

TABLE 24-5: EXAMPLE PWM FREQUENCIES AND RESOLUTIONS (FOSC = 32 MHz)

PWM Frequency                1.95 kHz 7.81 kHz 31.25 kHz 125 kHz 250 kHz 333.3 kHz
Timer Prescale (1, 4, 16)   16          4           1              1          1           1
PRx Value                         0xFF      0xFF       0xFF         0x3F     0x1F      0x17
Maximum Resolution (bits) 10         10           10            8          7           6.6

Thats a copy of the table from the datasheet... as best i can get it over
So, saying their is maximum resolution that is less than what the physical hardware can do indicates that you can change the resolution...

So, where do you change it?
HOW do you then only use 8 bits as indicated?
I've now read through 3 'guides' but they have not told me what i need to know...
However...
One of them indicated that PR was the period register used for setting the time lenght for each cycle...
However this can't possibly be true since that would mean the chip was built back to front... It would be more logical to set the frequency and then set the number of bits....

???????

[ScaleRobotics]

Good luck operating it at 6.58 bits



That's its maximum resolution at that PWM frequency. It's still a 10 bit pwm.

[cncmachineguy]

You've never used 6.58 bits? The trick is what to do with the little bit (1.42) left over all the time

[ScaleRobotics]

You've never used 6.58 bits? The trick is what to do with the little bit (1.42) left over all the time

That's easy, I just sweep them under the rug.

[cncmachineguy]

I would try that but my 2.3 kids will eat them

[HenrikOlsson]

but still, why does the datasheet, even the one for the 18F4550 'indicates' a variable bit resolution...

It's not variable in the sense that you can change it, it changes as function, or relation, between the clock frequency and PWM frequency.

If you had a 1MHz clock (Fosc/4=1Mhz) and wanted a 1Mhz PWM frequency you'd get two different dutycycles, 0% and 100% -1 bit resolution (not very useful). If you lower the PWM frequency to 100kHz you'd have 10 different dutycycles available. How many bits does the value 10 occupy? Well, you can't fit it in 3 bits and 4 is more than you actually need. Using the formula in the datasheet tells us that it's 3.32 bits. Obviosuly there are no such thing as partial bits and you'd need 4 bits to actually be able to use all 10 dutycycles but writing anything "above" 10 to these four bits will result in a dutycycle of 100%.

Now, if the PWM module gives you a higher resolution than you want you can obviously "scale up" your low resolution value to fit whatever number of bits of resolution there is, as have been shown earlier.

/Henrik.

[comwarrior]

right...

So, if TMR6 is setup to feed the PWM's... and 'overflows' @ 80KHz...
If the PWM's were 8 bit then i'd expect a cycle frequency of 312.5Hz...
If the PWM's were 10bit then i'd expect a cycle frequency of 78.125Hz...

Where on earth are these stupid decimals coming from?
Who on earth made a PWM system so complicated?
and which idiot approved their design?

If the hardware is static 10bit... why the hell would the datasheet even suggest any other resolution? I don't even understand how you couldn't use 10bit resolution on hardware PWM...

I really don't understand how these PWMs work... They are backward and alien to me...

So, can someone please explain in English how to setup the PWMs (ie how the hell you get a different bit resolution)?

Thanks

[mister_e]

Look at the related registers... it's timer based, then look which feature are in Timer... YUP you got it... prescaller OSC/2,4,8,16... hence the decimals.

Use the PICMulticalc, set a low PWM frequency see the resolution... set it way higher... see the resolution.

Conclusion?

[mackrackit]

comwarrior,
Henrik and others are trying to help you understand. There is no need to get upset.

I am asking you to keep an eye on your language. This is a "family" type forum.

[ScaleRobotics]

I really don't understand how these PWMs work... They are backward and alien to me...
So, can someone please explain in English how to setup the PWMs

I found this helpful, maybe you will as well. It is mostly about the hardware, so it is a good example to understand it better. It was done by a company that makes a C compiler.

http://www.mikroe.com/eng/chapters/v...ntroller/#c3v7

[comwarrior]

No one is getting my point... I'm experienced with hardware PWM chips... and I've never seen a setup like what pic's seem to have...
I'm very aware that the pwm cycle frequency comes from a timer... I believe i have demonstrated this knowledge...

It's one thing to say 'yes this is what will happen'... what i want to know is why?
WHY would the resolution change with different input frequencies?
If the hardware is 10bit resolution then it should always be 10 bit regardless of input frequency...

[cncmachineguy]

You might get more joy about this over at microchip forum. Seems you are looking for a legimate answer to why did you make it this way question.

[rmteo]

....It's one thing to say 'yes this is what will happen'... what i want to know is why?
WHY would the resolution change with different input frequencies?
If the hardware is 10bit resolution then it should always be 10 bit regardless of input frequency...

Henrik answered your questions very succinctly in post #11.

[comwarrior]

rmteo,
No, post 11 doesn't answer the question... it again does not tell me why...

If i wanted 1 MHz pwm frequency and the pwm hardware is 10 bit then i'd have to give the PWM 'chip' a 1024MHz clock frequency...

If i give the PWM 'chip' a 1MHz clock and the resolution is 10bit then i would expect the pwm frequency to be (1,000,000/1024) = 976.5625Hz

Changing the bits based on clock input frequency (i believe your saying) is stupid...
Thats like building a car that will only turn left when the engine RPM is above 6,000 RPM...

If this 'variable' bit resolution is because the PWM hardware in the micro-controller can't keep up above certain speeds then it should be specified as rated upto a certain frequency...
This is how all but one PWM chips that I've previously programmed have worked... the one exception was a royal pain in the ass!

As it is, the frequency that i want is supported at 10 bits (apparently)... while my duty is 8 bit, it's just a few cpu cycles to times it by 4 and then do the PWM 'update'...

scalerobotics

I found this helpful, maybe you will as well. It is mostly about the hardware, so it is a good example to understand it better. It was done by a company that makes a C compiler.

http://www.mikroe.com/eng/chapters/v...ntroller/#c3v7

oky, so two lots of PR registers... one for 'pules duration' aka duty... the other for 'pulse period' apparently thats frequency... the PWM output frequency should be defined by the input clock divided by the resolution...

[ScaleRobotics]

Changing the bits based on clock input frequency (i believe your saying) is stupid...
Thats like building a car that will only turn left when the engine RPM is above 6,000 RPM...

Why are you so worried about the resolution it is going to have at that high frequency? What is your application? I thought you didn't need more than 8 bit resolution anyway!!!

[rmteo]

No where is it stated that the PWM hardware is 10-bit (it doesn't make sense anyway). The duty cycle is set by loading a 10-bit register (comprising CCPxCON5:4 and CCPRxL).

[ScaleRobotics]

The key is, it produces UP TO 10-bit resolution output.

15.4 PWM Mode
In Pulse-Width Modulation (PWM) mode, the CCPx pin
produces up to a 10-bit resolution PWM output.

[Bruce]

PWM resolution refers to the duty cycle. Not the PWM frequency. The value you load into PR2 sets the period for the frequency you want.

Say you want a 125kHz frequency. The period of 125kHz is 1/125,000 = 8uS. Now say you're using the 32MHz internal oscillator.

With a prescaler of 1 the timer will increment every 125nS or 1/8MHz.

With the timer incrementing every 125nS, it takes 64 instruction cycles before the timer is = PR2.

When it's = the value in PR2, the timer is automatically reset, and the PWM output pin is toggled high.

You can see this in Figure 24-3 in the data sheet.

The duty cycle is set by loading CCPR1L and CCP1CON bits 5 and 4. CCPR1L is an 8-bit register and the 2-bits in CCP1CON are the 2 least significant bits of the up to 10-bit duty cycle.

When the timer = the value in CCPR1L, and the lower 2-bits = the 2-bit value in CCP1CON bits 5 and 4, the PWM output pin is forced low.

Figure 24-4 and note 1 explain how the timer is concatenated to form a 10-bit counter used to compare the timer count to the 10-bit duty cycles registers.

The higher the frequency, the lower the resolution of the duty cycle, since there's less time in the period for more steps to adjust the duty cycle.

It looks confusing at first, but if you spend a few minutes reading the data sheet and have a calculator handy to plug-in the values shown in tables 24-5, 6 and 7, to see how they work out, it gets a lot clearer.

[mister_e]

<IMG SRC="http://gobnf.org/i/jg/gwb_approved.gif">