Filter coefficient (N) in PID block?

Good evening folks!

While tuning my quadrotor model in Matlab I found that the Matlab PID function had a Filter coefficient (N) term in the PID equation. However I have not encountered such a term in my typical PID programs. How would one implement this term into the PID equation and what would be some simple C code to add this term into the derivative? I tried generating C code from my model but got a lot of header and C program files with nothing close to simple.

My current PID code line in C is:

U2 = KP * phi_error + KI * phi_errorsum + KD * (PHI_RATE_desired-PHI_RATE);

Many Thanks, Robert

Answer 1

[Arkadiy Turevskiy]

The code would look like that.

    FilterCoefficient = (Kd * u - Filter_DSTATE) * N;
    y = (Kp * u + Integrator_DSTATE) + FilterCoefficient;
    Integrator_DSTATE += Ki * u * 0.01;
    Filter_DSTATE += 0.01 * FilterCoefficient;

HTH.

Arkadiy

[Robert]

Awesome!

Why are Filter_DSTATE and Integrator_DSTATE scaled by 0.01?

Thank you!! Robert

[Arkadiy Turevskiy]

0.01 is the sampling time I specified in the block before generating code. You can set the sampling time to whatever you want and the code will of course be updated accordingly.

[Marzieh Hosseini]

Hi,

Is it the same code if I want to implement PID in Matlab code, while I am controlling a device directly with Matlab and I got real time information in my Matlab?

[Andre Elliot]

Hi what are u, N, y and the meaning of DSTATE? nice day.

Answer 2

[Shashank Prasanna]

Hi Robert, this link should give you a better idea of what N means:

http://www.mathworks.com/help/simulink/slref/pidcontroller.html#br9ejg0-18

If you have Simulink Control Design toolbox, then you can use the built in pidtune functionality to tune the PID.

[Brian Candiloro]

Can this filter term be realized in a physical opamp differentiator circuit or only in code?

Answer 3

[Robert]

Thank you for the response, but I saw that link and do not quite understand it entirely, especially the filter block.

So in essence should the code look something like this:

Derivative PID:

Derivative_gain*(set_rate - N*acutal_rate)?

I am not quite sure where the filter (1/s) comes into play.

Thank you

Answer 4

[Robert]

Any ideas?

[Seng Tak Goh]

Hi Robert,

In essence, the filtered derivative is doing a first order filtering on the change of the inputs, scaled by D gain.

The 1/s integrates the previous output scaled by N term, then adding the input change scaled by D * N to produce the output.

If you multiply the D term of the continuous form by s, you will see it better how the filtering comes into play.

In discrete form:

y[n] = D*N*(u[n]-u[n-1]) + (1-N*Ts)*y[n-1];