PID PWM Valve

[deperkin]

Hello all!

I would like to see if anyone can help me with a project I am currently working on to control a proportional solenoid valve using a PID controller.
The valve is 0 to 30V.
I would like to understand how the PWM duty cycle works a little better.
from what I understand, the duty cycle is 0 to 100%, but do the values that I send for the duty cycle depend on whether I am using int's or longs, etc?

Or do I need to just convert the numbers to values between 0 and 100?

Basically I would like to adjust my PID gains appropriately in order to send the correct duty cycle for controlling pressure in a line. I have a reference pressure of 100psi and would like to maintain this pressure, limiting overshoot.

Right now I cannot post the whole code since it is almost 6000 lines long, but I would like to understand PWM a little better, and whether or not I need to scale my gains way down in order to get the correct duty cycles.

THank you as always and Best regards,

deperkin

[PCM programmer]

[FvM]

An important parameter, that has to choosen first, is the intended PWM frequency. This setting defines also the scaling of PWM duty cycle. Typically, the same range is given to the PID controller's manipulated value output. In case of an integer PID design, you should provide some additional (fractional) resolution in the integrator right of the LMN output's LSB.

[deperkin]

I am using a PIC18F2331.

Here is what I have setup so far for the PWM:

[deperkin]

I guess what I am really not understanding, to clarify, is that as of yet I am not sure what my maximum values will be in my error, particularly in my total error used for the integral control.
Without knowing these values how would I be able to scale for an appropriate duty cycle ??

Thanks again!

Deperkin

[deperkin]

what do you mean by LVM outputs least bit???

i know i will most likely need to scale the PID values, but is is tough at this point to see what error i will forsee, since i am actually not starting at the reference point, but well below the reference.

I am used to starting with zero error using PID, but in my case i have to start with an error of almost 100.

thanks again and you guys are great!

[FvM]

My remark regarding PID internal resolution is dedicated to fixed point PID designs. Cause you utilize float, there should be no problem of this kind.

In the discussion of maximum error signal, scaling of controller output (= PWM duty cycle in this case), I suggest a different approach. It's normal, that a controller output reaches limit values, cause most actors have electrical or mechanical limitations. But the integrator action should be stopped at this point. Otherwise you get large overshoots. With some actors, there may be a problem to set the controller numerical limit according to the actual actor end position.

A proportional valve actor must be clearly able to shut down the flow completely, the maximum output may be adjusted application dependant, but if the valve's orifice is choosen correctly, the controller output should go up to nominal solenoid current. This sets the duty cycle range.

The PWM frequency has a special meaning for proportional valve control. If choosen correctly, it can reduce mechanical hysteresis by forcing small vibrations of the valve. This results in PWM frequencies of a few 100 Hz, the optimal value is type dependant of course.

Another important parameter is the controller sampling frequency. The suitable range strongly depends on the control process (plant). For a pressure regulator, I would expect a few Hz up to several 100 Hz.

[deperkin]

Thank you again for your help, you are well knowledged here.

i am not too worried (at this point) with the frequency, but only the resultant RMS voltage coming from the PWM.

I was thinking for simplicity i could use:
setup_timer_2(T2_DIV_BY_16,100,1);

which would give me a freq of about 1238 Hz., but allowing me to look at the duty cycle as a percentage from 0 to 100... (correct)?

I could probably postscale this to lower the frequency if you think this would help me with hysteresis error... this will be worth trying if i have time (i only have two weeks to finish my project)...

i am really dealing with very low flowrates and i want to avoid any chattering of the valve.

we are boiling water to maintain a pressure...
my concern is that i will have initially large errors resulting in large duty cycles (thus opening my valve completely)... the way our system is set up is: the larger flowrate will lower the output temperature and keep the pressure lower, in effect keeping the error higher. What i would like to do is with a larger error make the duty cycle lower, thus trickling in the water and raising the temperature = lowering the error by raising the pressure.

am i making any sense??

Thank you all again!

[Ttelmah]

First, remember that the value you feed into the PWM, can be a 'short' (8bit) integer, or a 'long' (16bit) integer. The hardware PWM, allows values up to the timer count _plus one_ to be used. If the value used is a 16bit integer, two extra bits of scaling become available, using the individual cycles of the master clock, instead of the processor clock.
So:
If using 8bit integers to feed the PWM, setup_timer_2(T2_DIV_BY_16,99,1);
gives legitimate 8bit values from '0', to '100'.

However if using 16bit integers, setup_timer_2(T2_DIV_BY_16,24,1);
gives legitimate values from 0L to 100L, and gives you a possibly more friendly output rate.

Your frequency figure, suggests you are using 8MHz. For most PWM applications, unless there is significant output filtering present, you want to aim for a resulting output, that is above audio. Using a prescaler of /4, would take your frequency up out of audio. Might be worth considering....

Best Wishes

[FvM]

[deperkin]

Thank you again for all the good advice and help on this.

I will post again and let you know how it is working... time to do some testing!

regards,

deperkin