Scale factor in a PI controller

[seyyah]

How do you determine the scaling factor in a PI controller?
Lets talk over numbers. Say we have Kd and Ki, proportional and integral multipliers respectively. Say they can change between 0-100. What we control is the pwm duty cycle (shortly pwm) which can change between 0-500. And there is an error term which can be between 0-1000. Using these values i am trying to determine a scale factor. I think when the error is max the pwm must be max. If i can find the max error than i can find the scale factor. But error term always increases when an integral term exists because we always add the error to the previous error. This makes difficult to determine the max error. Can you please suggest me something. Is my idea true? Or is there a formal way of doing this?

 

[nicleo]

Do you have Matlab? If you do, you can refer to toolbox/powersys/powerdemo/psbdcdrive.mdl. Hopefully from there you can figure out how to adjust the Kp and Ki for your application.

 

[seyyah]

I doesn't tell me anything. i can do Ki 1 or 1 billion. How does matlab scales it?

 

[nicleo]

I think in your design, you use 'fixed' PWM frequency. The changing variable is the duty-cycle. In the model given in my previous post, the frequency of the switching signal (to driver or motor) is a variable as the model uses 'hyteresis controller'. Does your controller (PIC) monitor for overcurrent?

 

[seyyah]

yes it monitors.

 

[joanna_seczkowska]

I don't know what You control, but error is always as difference. However, You should calculate system gain. So You have to create open loop, and send range of input signals in whole range (PWM in your case). Note system output, and calcualate gain = So/Si. Then You should repeat experiment from Kp= 1/gain. In theory System gain multiplied by feedback gain (Kp) should = 1;

 

[usernam]

But error term always increases when an integral term exists because we always add the error to the previous error.

This is not necessarily true. If for example your output is greater than your reference value your error term will be negative. So you will deduct your current error from the previous error.

How do you determine the scaling factor in a PI controller?

To determine the sacle factors you first have to determine the transfer function of your system without the feedback loop. Then depending upon the required speed of response or steady state error you can determine the transfer function of the feedback loop by applying control theory.