Why we use PID in digital controller design ?

[Sahil_khan]

Dear Guys

I have a very basic question that why we use PID algorithm in digital controller design ?

I mean cant we use some alternative approach e.g if i sense output voltage of a dc-dc converter and then accordingly i adjust my PWM.
Lets take an example :
Suppose Vout of a dc-dc converter is 22V and the desired Vout is 20V, when my controller senses then it decreases its PWM (lets say 10% decreases the ON time) then it senses the Vout and it finds the output is now 19.5V instead of 20V(20V is desired output volt) then it increases its PWM lets say 2% then it sense the output is almost 20V.

If this algo works then why do we always talk about PID implementations ?

Thank you all.

Regards
Sahil

 

[FvM]

It's not clear, if you are describing proportional behaviour (the duty cycle is set to a value according to voltage error) or integral behaviour (the duty cycle is increased or decreased each time step according to voltage error). But you should try the behaviour of both algorithms in a simulation.

You'll notice that neither a pure P (not reaching the set point) nor I implementation (most likely unstable) will do what you want. Combining both you have already a PI controller that can achieve moderate performance. Adding derivative action can further improve the dynamic behaviour.

In detail, the optimal controller depends on the control system properties. There are control systems, that work well with pure P or I controller.

P.S.: It should be also noted that digital control allows optimized controllers which exactly compensate the control system behaviour, e.g. dead beat controllers. But unfortunately, they are less robust than PID and need exact tuning. So PID is still a good first choice for basic digital controllers.

 

[Sahil_khan]

Thank you for your reply.

But I am sorry for me it is still not very clear. I do not understand what you mean

if you are describing proportional behavior (the duty cycle is set to a value according to voltage error) or integral behaviour (the duty cycle is increased or decreased each time step according to voltage error)

I mean if I increase and decrease my PWM pulse by pulse according to the voltage I sense from the feedback, in this case I can achieve steady state behavior ?

 

[FvM]

it is still not very clear

I'd say the same about your latest description. Do you mean

1) new PWM D.C. = previous PWM D.C. + k*voltage error

or

2) new PWM D.C. = constant + k*voltage error

1 is I controller, 2 P controller

 

[Sahil_khan]

Yes I am talking about the first one. i.e.

new PWM D.C. = previous PWM D.C. + k*voltage error

I mean if i use only this algo then the system should work and the error will be eliminated so then why we always talk about PID controllers in digital controllers.

 

[mtwieg]

Because there's more to control design than just reaching zero steady state error. You need to make sure the system is actually stable, and usually you care about transient response performance (rise time, overshoot, etc). Using full PID feedback allows you to have no steady state error, and also allows for adjusting other performance issues as well.

And PID is popular in particular because it's very easy to implement on a simple microcontroller.

 

[FvM]

O.K., you are talking about an I controller. Yes it can work for the application. But for most practical control systems the response to disturbances (e.g. load current steps) will be slow. With higher I gains, the step response becomes oscillating, due to PWM modulator deadtime and control system poles.