Microcontroller issue

Hi John,

You seem to be under the impression that you will be doing this in software.

Previous postings have stated that hardware PWM is the only way to go.

A 300Khz to 500Khz PWM is not PWM, its a range of frequencies.

If you want a 500Khz PWM, it would be 500Khz from two raising edges with a variable high to low time.

--| |__________| ^ ^ | 2 uSec |

This is also 50% duty cycle PWM.

_ _ | | | | __| |___________________| |__________ ^ ^ ^ | | 1.8uSec | .2uSec .2uSec + 1.8uSec= 2 uSec or 500Khz PWM with a 10% duty cycle

or ___________________ ____________ | | | __| |_| ^ ^ ^ | 1.8 uSec | | .2uSec 1.8 Usec + .2 uSec = 2 uSec is still 500Khz with a 90% duty cycle

Your original post asked for 10bit resolution, thats 1024 steps, so each step would be 2usec/1024 ~ 1.95nanoseconds per step !!

Is this what you really want ??

This is not going to be done is software.

hamilton

Hi,

The hardware is available on the microcontroller chip. The dedicated PWM generator. Am I right? or did you mean that I have to design it byself. For example the C8051F411 has a on board 8 bit PWM generator that can can go up to 93 KHz of different duty cycles. Am I on the right track?

John

"Jim Thompson" wrote = in message news: snipped-for-privacy@4ax.com...

Look up fractional convergents and continued fractions.

I wrote a program which accepts a rational number and converts it to = whatever convergent you prefer. By ending the process at a certain = limit, for instance a divisor less than 256, you get a best case = resolution much better than crude 8 bit PWM. Worse-case isn't any = worse, since you can't count less than 1/256 in either case. With = proper selection of convergents, you can keep frequency similar as well = (e.g., only pick convergents in the 224-255 range).

It's too bad I'm in your killfile, it seems the futility of your = arrogance has been proven ;-)

Tim

--=20 Deep Friar: a very philosophical monk. Website:

Yes, if you are asking for 93KHz at 8 bits, but you are asking for

You _did_ read mine:

This works if you're following the PWM stage with a _really_ low-pass filter. It is often the case in a control loop that you need eight bits or so of resolution for efficiency, and the plant forms a _really_ low-pass filter. You also generally need more than eight bits of output resolution for good control -- but the "good control" bits can use the plant to average out a dithered 8-bit signal, so you can use a sigma-delta technique either at the loop's update rate or (better) at the PWM's update rate.

"Jim Thompson" wrote in message news: snipped-for-privacy@4ax.com...

Isn't that called Dithering? I may have a paper tucked away here somewhere....

Cheers

Where can I get the whole enchilada ?:-) ...Jim Thompson

How can I get that as a PDF? ...Jim Thompson

Try:

Jon

Found it, Check your Fromwebsite mail addy.

Cheers

Thanks! ...Jim Thompson

sorry the gifs seem a little hard to get

This one with GIF's

I sent a copy of my original to your website email. Note that it's long on explanation and short on a description of the technique -- that's not because I was trying to whitewash anything, that's because the technique is dead easy.

One of these days I need to find a good book on sigma-delta stuff: the first-order sigma delta is almost trivial to analyze, but I've never been able to do so from first principals for the higher order ones, and the descriptions I've found on the web all seem to lack detail. Usually that's a sign of a book-length subject in the offing.