PWM and dither

A 'dither' technique typically increases the PWM reolution by one or two bits depending upon your tolerance for dither ripple. What is your dither spec and is it the same?

Cheers

My specs about the dither are the frequency (lower than the Pulse Width Modulation), the duty cycle fixed on 50% and the spec on the amplitude current (which still I need to understand). I'm sorry but I could not catch what you meant.

Thanks, Pasquale.

IE

Yes, that's correct. I need a PWM to control the valve current plus anothes signal called "dither" to overcome the stiction. I should implement it but actually I don't know how to consider the dither.

Thanks, Pasquale.

You have some error signal controlling the PWM output. You can add a dither signal of the appropriate amplitude and frequency to that error signal, such that the valve is kept in constant motion.

Do you means directly by programming the PWM output of the microcontroller ?

Ok I will have an average signal because the load will damp it out. For this reason I'll have some error in my feedback control that i cannot correct but it's not what I want. Or better I should have a lower frequency PWM and use your way to do in order to make the valve moves around its setpoint value.

depending upon your tolerance for dither

I think he's suggesting an actual dithering of the valve position. IE the valve actually vibrates around its setpoint. Possibly to overcome stiction?

Robert

Ciao Pasquale, a word of caution about using dithering, this is based on my experiences of eons ago. Dithering is used to overcome the effect of "stick-slip", a highly non-linear behavior that can cause instability in mechanical closed loop position systems. To be effective dithering has to actually vibrate the mechanical elements that are part of the control loop. The question is: how much has to be the mechanical displacement and at what frequency? Both these questions have a great relevance on the life (wear characteristic) of the mechanical elements and their performance as far as the long-term stability of the loop. For the injection of the dithering parameters, I leave it to the others that have replied to your posting. But said that, if your PWM is based on a fixed frequency, why not "kill two birds with one stone"? In other word why not use the PWM frequency as the dithering frequency? Can be done? Buona fortuna con il tuo progetto. Gene

Lets assume you had an interrupt handler for each starting ON duty. and the timer value of lets say 500 is to much and 499 is not enough to get the width you need for the precise value. What you would do in this case is to have a variable in the background that would give you a counter or a bit image to use to oscillate the value between

You'll want to add in a dither signal, probably square. I wouldn't call it "PWM", because a simple 50% duty cycle square wave at the appropriate frequency will do fine.

Any application of this principle gets pretty specific pretty quick, but here's a paper on implementing dither to overcome friction in motor/gear circuits:

I would make sure that I understood the minimum amount of dither with which I could get by, to reduce wear and tear and power consumption. If wear and tear and power consumption is a consideration I would also consider implementing a deadband in my controller, and turning the dither off (or letting it trail down) whenever my error signal was within the deadband.

Don't dither the output directly then, dither the setpoint. Remember you want the output to change. The other posters suggestion is for when you want a fixed output but eith higher resolution than the PWM provides. I've used that technique to provide higher resolution control when the timer facilities only allowed a relatively coarse PWM. There was also no closed loop feedback on the PWM in that case, I simply set an output.

Robert

By just adding in the appropriate sample of the digitized dither waveform to each update of the PWM value.

Let me try if I understood well. You mean that I can use an high frequency PWM (in order to control the current) varying its duty cycle and sometimes (depending on the dither frequency and dither the duty cycle that are both fixed) also add a dither signal to my high frequency PWM ? Is it possible to do that in a micro ?

Grazie mille, Pasquale.

Just use a low (100Hz or so) PWM frequency to begin with. Sure, the valve will vibrate a tiny bit, but that will stop it sticking. This works for hydraulic valves at least, I'm sure it will work for your application as well.

Cheers,

Al

Some valves recommend 100% dither by driving with PWM at the dither frequency, if the OPs does not then you can only presume it would be too much.

I have seen valve specifications recommending dither frequency adjustable between 70 and 250Hz although they didn't bother to mention how or why to adjust the frequency. I suspect they want the dither adjusted to stimulate a mechanical resonance in the valve sp just saying about 100Hz doesn't really cut it.

Dithering is recommended for some proportional valves to avoid the hysteresis due to the valve getting stuck.

Just set the PWM frequency low enough (~100Hz) so the mechanical vibration of the valve will prevent it from getting stuck. They specify the operating frequency range in the datasheets.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

Let me try if I understood well. You mean that I can use an high frequency PWM (in order to control the current) varying its duty cycle and sometimes (depending on the dither frequency and dither the duty cycle that are both fixed) also add a dither signal to my high frequency PWM ? Is it possible to do that in a micro ?

Non esattamente. Taking into account that I don't know anything about your application, my suggestion is the following: If you choose a low enough PWM frequency, you will automatically inject a dithering effect, that is the pole created by the mechanical inertia is not enough to suppress the PWM base frequency. Non so se ti e' chiaro. Gene

Epsilon Rho ha scritto:

Mi è chiaro. Sure I understood what you mean but the control problem I'm trying to focus on is different. On the point of view of the controller design it's better to consider the PWM and the dither as different signal. This will produce a good tracking performance using an high frequency PWM and to make the non linearities in the system lower by adding a dither. It's like to superimpose the dither and the PWM. That is what I understood from therical applications but I still have a big gap to get in the pratics.

Grazie mille, Pasquale.

Pasquale, stai operando su un circuito di posizione in retroazione negativa? Gene