UCD3138: I have some questions about the application document "UCD3138 control theory" from TI

Figure 10 and 11 are used to show the effect of KP, KI on frequency response, the value in the figure is for illustration purpose, may not be accurate.

I understand that you mean that the examples and diagrams are only schematic diagrams and are not completely accurate values.
I still have a problem. In the actual debugging, taking ki=0.0216 as an example, should I not expand ki by 10 times, because when ki=0.216, the two zeros of fz1 and fz2 are imaginary, in the bode diagram Can't know where their frequency is?
Or, when designing, I should ensure that the zeros are all calculated, right?

Expand ki 10 times to 0.216, theoretically calculate that the fz1 and fz2 zero frequencies are imaginary,
then in fact, will they appear in the bode diagram? What is the impact on the frequency response?

Hello, when debugging, should we guarantee that the formula {[Kp(1-a)+Ki(1+a)^2]^2-8(1-a)KiKd} is greater than 0?That is, no imaginary solution.

Is the system unstable when “delta= {[Kp(1-a)+Ki(1+a)^2]^2-8(1-a)KiKd}” there is an imaginary solution?

No matter the pole/zero are real values or complex with imaginary, as far as you have enough phase margin and gain margin, the system will stable. Our GUI can give you choice to select KP, KI, Kd, or real pole/zero, or complex pole/zero, and it will show the bode plot for you with your choice.

Regards,

Bosheng

  

Thank you, I understand it. The solution to the stability of the system is not the real or imaginary solution of the transfer function, but the margin of the actual frequency response.

I still have a question.

I calculated the calculations on pages 31 and 32 and got the following results:

The result of my calculation is different from the result in the document. Is it a calculation error or is there a clerical error in the data in the document?

I still have a question. The figure below is Figure 30 on page 30 of the document:

I think this should be the main power bode diagram without compensation.

I used matchcad to draw the amplitude-frequency curve according to the parameters given in the document, as follows:

The blue line in the figure is similar to the document. The transfer function is shown in the figure G(f).

The black line in the figure is different from the document. The transfer function is shown in the figure G1(f).

The difference between the two is whether the function is multiplied by the input voltage Vin.

I guess your document is not multiplied by Vin, and I see that some reference books are multiplied by Vin.

Does the main power bode map multiply by Vi, which curve is correct for blue and black?

I also have a problem.

I checked the 4.2.3 example on page 32 of the document.

In the example, Ko=10000，Ki=0.025, Kp=1.59, Kd=101.4,

But I calculate it in matchcad as follows:

The kd value is 4 times worse. I don't know why?

Hello

I still have a question I would like to ask.

When I calculate Kp, Ki, Kd, a, these values are mostly decimal or single digits, and the values in the actual registers are large integers of hundreds or thousands.

I think there should be a conversion factor inside.

Is this replacement factor a table 19 in the document? As shown below:

I checked the case of Figure 43 as follows:

The result of my calculation is that when Ko=1000*2*pi, the kp, ki, and kd values in the PID are consistent.
When Ko = 1000, the values of R1, R2, C1, C2, and C3 are in agreement.

Why is this? Why is Ko different, and it happens to be 2pi？

Is the three values PRD, KCOMP, and SC in the above formula set in the firmware?

All three are UCD3138 registers:

PRD is PWM period

KCOMP is used to calculate the PWM pulse width

SC is used to scale up or down of PID result.

Please refer UCD3138 technical reference manual for details

Thank you! Let me see the manual.
There are also two questions about the study and the main power modeling problem.