pid control with pwm output

[ferkar]

hi everyone,

i would like make a pid controller. i simulated the code that i wrote in a simulation program,İSİS. on the screen i saw everything seemed well, i mean when look at oscillscope the pwm duty ratio varies between 0% to %100 in 10bit configuration. However when i test the code on the real system i realized that there was big dead band, about %50.

now i need some help to understand where the real problem is software or harware.

could anyone tell me where ı should look for digital pid controller flow chart?
thanks indeed...

[Ttelmah]

Obvious thing to look at, would be something silly with the variables used to control the PWM. Remember that CCS, 'switches mode', and is an 8bit variable is used, feeds this to just the top eight bits of the PWM control, while if a 'long' is used, give the 10bit output.
There are a couple of very good PID examples, in the Microchip site, under 'servo' control. AN523, gives in particular the process logic diagram for a digital PID loop. I posted sections of a much modified PID program here some time ago. The big problem is keeping the loop processing _fast_. Otherwise 'windup' behaviour in the control is liable to happen, with several loops occuring, before a response is seen, resulting in overcorrection. A lot also depends on the resolution of your encoder setup, and how often changes are likely to be seen.
If you can't find something to help, I'll try to trim the parser, and other stuff out of my code, and post it.

Best Wishes

[ferkar]

thank you interested in,

i check an523 on microchip site but i disappared. in other sources i read so much about pid. i need a programming support. if you want to see my code i could send it to check if any missing or failure lines..

thanks

[mmprestine]

here is a good reference

http://www.embedded.com/design/embedded/4211211/PID-without-a-PhD

http://www.embedded.com/ContentEETimes/Documents/Embedded.com/2000/f-wescot.pdf


here is code based on it that works well....

[s_mack]

fascinating. I'm having a difficult time understanding some of your code though. In particular:

[FvM]

Obviously, the code is rather a principle example than a functional application. As you already found out, part of the code is missing, e.g. initialization of parameters.

There are however more basic issues. As you are looking for fixed point PID code, you may want to check how scaling of product terms is achieved in the code. Unfortunately, it isn't at all. I can't see, how you get a functional PID code this way.

When you are using e.g. int16 variables and parameters, you have to define a scaling, too. You can have e.g. 4 bits left of the implicite decimal point, this is usually called a fractional number format. The multiply int32 result has to be scaled accordingly. To scale back to your your fractional format, you have to drop 4 MS and 12 LS bits. But saturation rather than overflow should be performed on the MSB side.

As another disadvantage of the PID example, it uses an explicite integrator limit, that typically doesn't correspond to the actual manipulated value's limits. But a poor windup suppression is still better than none.

Regarding your pedal and speed setpoint discussion, it isn't specific to the present PID code in any respect. Basically a car's pedal control doesn't involve a feedback controller, it directly sets the drive power.

[mmprestine]

Did you read the article "PID Without a PhD" that was posted with the sample code?

[s_mack]

Yes, of course...

my problem isn't so much the PID theory as it is getting it to work without floats.