How can I remove the few peaks before the overshoot, so I only have the overshoot peak and then it directly transition into steady state?

I built a buck converter ( circuit will be below):

The parameters are:

Vi = 12V, L = 33e-3, the R next to L =0.5, C = 680e-6 , R next to C= 5e-3, R = 10

This is the output voltage graph i got:

Currently I'm manually tuning the PI to achieve an overshoot of between 10% to 20% and a settling time of 0.01 to 0.1 these are no problem but I just can't remove the oscillations following the overshoot. Also when I change the value of D it mess up the whole graph. Is it possible to remove the oscillations so the overshoot follows directly into the steady state because no matter what I change p and I to it never removes the oscillation before the overshoot and changing d mess up everything

[Kevin Rogers]

The first commentator gave you the open loop transfer function. You can re-express the transfer function such that the denominator is (s^2 + 2*Delta*wn*s +wn^2). Adjust L&C such that:

Wn << Ws to give you the require attenuation, where Ws is the switching frequency. You have a 2nd order filter so the filter gain drops away at 40 db/decade.

Delta = 1 to give you critical damping.

This will remove the osciallations in the open loop response such that you are not depending on the PI controller to remove the oscillations but just to control the rise time and settling time.

[Kevin Rogers]

The PI filter will adjust the duty cycle to give you the required voltage output. D will be slightly higher than 2/3 to acount for source/series voltage drop.

[Samuel Pappalardo]

This method is very complex for me

[Samuel Pappalardo]

How will i know what to adjust L and c in order to get wn<<ws

[Kevin Rogers]

wn = 1/sqrt(LC)

2*delta * wn = 1/(RC)

For critical damping delta = 1

Therefor C = 1 / (2*wn * R).

Substitute C inot 1st equation to get L.

Answer 1

[Pat Gipper]

Hi Samuel. I found tuning the derivative to be difficult too. Eventually I settled in on P=3.5, I=850, D=0.01 and N=850. The default value of N was way too small. Here is the result.

[Samuel Pappalardo]

When i used your values i got this:

Answer 2

[Dr Narayanaswamy P R Iyer]

You have NOT mentioned anything about triangle carrier frequency. First obtain output voltage response without PI controller. Then apply Ziegler-Nicholas chart to tune PI controller. Refine this kp and ki value for desired response.

[Samuel Pappalardo]

Thanks but to apply the zieglar- Nicholas i need to know the TF. The carrier frequency would be 1Khz

would it be correct to say the Tf of my circuit would be:

If is correct i will just sub my values for R L C and Vi

Answer 3

[Dr Narayanaswamy P R Iyer]

TF is one method. You know the desired output voltage and duty-ratio D. Use this in the model to get open loop voltage response. Then apply Ziegler-Nicholas chart to obtain kp and ki. Then refine this PI controller parameters for desired output voltage closed loop response.

[Samuel Pappalardo]

How do i obtain open loop voltage reponse

My dersired output voltage is 8v and duty cycle should be 50%

Answer 4

[Dr Narayanaswamy P R Iyer]

For 8 V output, your duty-ratio should be 2/3 (0.67) for given input voltage.

[Samuel Pappalardo]

Yes you arre correct

how can i use these two values to obtain the open loop voltage

I'm unawre of the formula if there is one

Answer 5

[Zhao Wang]

Since you have built a Simulink model for the buck converter, you may want to consider tune the PID controller based on the actual model behavior. For a switching buck converter, you will need to conduct frequency response estimation to identify a linear system description at the desired operating point. Using the linear plant model, you can use the PID Tuner App (can be opened from the PID controller block) to tune PID controller gains to achieve the performance you want.

Here is an example about the workflow above: https://www.mathworks.com/help/slcontrol/ug/design-controller-for-power-electronics-model-using-frequency-response-data.html

[Samuel Pappalardo]

It never works for me everytime i use it saids plant cannot be linearsed, i even imported simulation data and still it doesnt work

[Zhao Wang]

I totally understand that the linearization error message will very likely appear in a switching circuit model. This linearization error is because of the discontinuities intrinsic in such power electronics models. There is an alternative way of tuning PID controllers automatically using the Closed-Loop PID Autotuner, as shown in the example: https://www.mathworks.com/help/slcontrol/ug/tune-pid-controller-in-real-time-using-closed-loop-pid-autotuner-block.html

Answer 6

[Pat Gipper]

My bad. Change I to 300.

[Pat Gipper]

But there are a few other things. I saturate the PID at +/-10Volts. And for the triangle wave I am using time values=[0 2e-4], output values=[0 10]. That will impact the overall gain.

[Samuel Pappalardo]

I made the chnages and got this

[Samuel Pappalardo]

Could you please email me your circuit

at [e-mail]

[Pat Gipper]

This is the model that was sent. It is using version R2021b Update 1. The gain constants are in the Model Workspace. Use Model Explorer to modify the constants.

Answer 7

[Antonino Riccobono]

Dear Samuel,

The difficulties you are encountering are understandable since when you tune a specific gain of your PID controller this will affect the other two. Therefore, manually tuning is not practical.

If you want to learn a systematic tuning workflow so that your feebback performance meets specific dynamic requirements, I invite you to consider the following training course:

Power Electronics Control Desing with Simulink and Simscape

Good luck,

Antonino