PI tunning with DSP

You can get to my article "PID Without a PhD" by going to

but I suspect that it's not designing or tuning a PI controller that's giving you fits, it's controlling an induction motor in general, am I correct?

Perhaps if you could give more detail on what you are trying to achieve it would help out. So here's some questions: How many phases? What's the power output rating of the motor you're controlling? Do you mean a real-true induction motor with no armature magnet or excitation? Are you trying to control the motor for a specific torque, a specific speed, to a position, what? What sort of feedback do you have?

I use 3 phase induction motor (370W,1.7A,230V,2 pole,1.46H)

yes, real-true induction motor with no armature magnet.

I would like to control for a torque, speed, and position. I use vector control method. For the inlet loop (the current control) I measure the 3 phase current and then control them. for the outlet loop i measure the speed of the motor then control them. but each loop has PI control, and my problem is in PI tunning in each loop. Because I use C programming in control action and have to be implemented in DSP, and the main problem is the right value (example : Kp value, and Ki value) to be implemented into DSP.

Assuming that you've got your vector control working correctly you should be able to use the procedures outlined in the article I mentioned ("PID Without a PhD",

That method will provide you with a good start, but you should use Bode plot analysis or some other more formal system identification and tuning procedure if you want to get things exactly right. Assuming that you either have a good MatLab/Simulink model of the motor or that you can run the motor by itself with an appropriate load (like a fan) I'd recommend tuning the inner loop first, then tuning the outer using the method outlined in "PID" for each one.

To model your software in Simulink just implement the equivalent discrete-time controller, keeping in mind that Simulink insists on correcting for the sample time in it's discrete integrators, which means that you either have to roll your own without the correction or keep in mind that your Ki value is scaled.