TMS320F28377D: Notch filter problem when the input data is not stable

and I use SFRA to make sure the notch response is correct. I have check the signal frequency.

Xiao,

I am unable to distiguish between the use case

1. When you say it worked

2. When you say it does not work

Can you elaborate on that ?

If you are trying to say that in an open loop situation when the notch filter is not part of the control loop it is working that kind of verifies the notch filter implementation.

However if you use the notch filter in control loop it is causing issues? that could be due to interactions from the closed loop??

can you clarify?

Manish:

1. both situations are in closed loop condition

2. the notch is on the voltage feed back path

3. when it worked, the DC voltage is stable, that means the DC vol ref and DC vol feed back is stable, around 600V

4. when it doesn't work , DC vol ref and DC  vol feed back is not stable , they are both increasing, cause at this time

the voltage ramp function is working, DC voltage will increase 540V to 600V.

you can see in the beginning the notch output is not correct compare to the real feed back

I don't know whether I make myself clear......

At last, you remind me something important: I don't update the PI controller coefficients of the voltage loop...

The coefficients I use is what I use SFRA to design without notch before...

Do I need to redesign the vol loop PI coefficients with notch in the feedback path?

This is data flow.

I also checked Grid Connected Inverter project again. 

1. I find that it seems the notch filter is afterwards the PI controller (I'm not sure about this) in the project, and I think the notch filter should be

in front of the vol PI controller.

2. Does the "cntl2p2z_LeadLag_coeff" this part use for the compensation of the notch? should I add it？

I think you are having a settling time issue with the notch filter.

You might have to check step response of the notch filter in MATLAB and then see what coefficient you need to tune for this.

I myself have not done this exercise.

you mean the coefficient of the notch or the coefficient of vol loop?
I think due to your function the notch out frequent point and the c1 c2 is confirmed the coefficients is fixed...

Xiao,

What i mean is that you may have to simulate this in matlab or other cad tools to check the step response. 

-manish 

step response for the notch only or for the control loop with the notch?

steps response of the notch only, because what you are showing is the step response of the notch

Hi Manish:

I have done some matlab work about the notch filter which I used as you said:

this is the step response of the filter used in the project:

which is the same as I test before.

And I also use the filter designer tool in matlab. I try to design a commom notch filter to see if the step response is the same.

and this is the step response of the design notch filter:

It is like the notch filter I used in the project.

I don't understand why you want to see the step response of the notch filter.

Because if the type of the filter is determined, Butterworth, Bilinear transformation, and the notch frequency, the filter is determined.

which you can only change the c1 and c2(in your function). Under these circumstances, I don't think the notch filter will cause some 

problems.

So basicly I don't know why you said it need to see the step response of the notch filter, or I should tune the c1 c2 para to make the notch

filter not aggressive?

Thanks for you reply, Manish.

Yes one approach can be to have c1 and c2 not very aggressive, that will reduce the oscillation during a step response.

very thanks for your help:
first I answer your questions:
1. X-axis is the data sequence, like I record 1024 or 512 data. It's not the time. Y-axis is scale, IQ(24), 16777216 is IQ(1.0).
2. Yes
3. Yes. 1-33 represents this ramp
4. I think this is not important, because at that time, I don't know what the step response of notch filter is like. And after that I do some
work in this thread, I don't know why there is another thread. You can check this:

e2e.ti.com/.../778416
I use matlab do some work, and I know the step response of notch filter is just like this .

5. sorry I don't know what you mean.

Yes, I think so. This is the IIR notch filter, it just like this.
But I have another problem about the notch filter. Maybe you can give me some advice or should I ask a new question:
The voltage loop open loop shear frequency or cut frequency is 275Hz. And there are several interference signal is 296.9Hz, 156.3Hz and 46.88Hz. I find that 296.9Hz is disappear, but the 156.3Hz and the 46.88Hz still have a half scale of without the notch. Besides, the scale is twice or more bigger than there is no notch filter.

The notch filter is after the voltage PI controller.
simply say when use the notch, the scale 156.3Hz and 46.88Hz increase first and go through the notch the scale get a little smaller.
This make the notch filter effects influenced. I don't know why the scale will increase, this is notch filter right?
 Is this ralated with the open loop gain? Because these two signal is lower than 275Hz. Maybe you can help me about that.

And I want to change the position of the notch filter. I want to put it on the feedback path, not after the PI controller, so what do you suggest?

Anyway, you could check the link above to see if all these are right(about the step response of the notch filter), than I will close these two questions.
At last, your support is greatly appreciated. Thank you very much, you and Manish. This community gives me a lot of help. Thaks for all your guys.

Xiao,

Thanks for the answers.

1. I am not sure why the other in-band interference frequencies increase in amplitude with the notch filter in-place than without the notch filter. It doesn't sound like anything the 296Hz notch filter is itself doing. It must be something in the overall system.
2. I thought you already had the notch filter in the feedback path, based on the block diagram. So where do you want to move it to? After the Voltage loop block?

Thanks,
Sira

Hi Sira:
1. 296Hz notch filter will only increase the amplitude of 296Hz interference. As you said 296 is out-band, so there is nothing wrong. I make the loop frequency 100Hz, and the 156Hz notch filter works right too. I am not sure where is the problem yet.
2. I have moved the notch filter after the Voltage loop block very early, cause I think this is easy to control the limit of the voltage order. But now I'm not sure where the problem is, so I will put it back on the feedback path.
I will do two things next :
1. in-band SFRA test of the loop with notch filter, I just test the 296Hz before(and that's OK), not involve the two in-band notch filters.
2.put the filter on the feedback path to see the difference.
Thanks for you reply Sira.

Xiao,

I will move this thread to closed, if you have follow up questions please start a new thread.

-Manish