This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Compensation Designer GUI

Other Parts Discussed in Thread: SFRA, POWERSUITE

Hi,

I would like to have some information about the way you choose a compensator in the Compensation Designer software from the SFRA Library. 

I have read the paper concerning this library and as it allows one to design a controller following a black box approach, without knowing the system, how should I pick the correct poles and zeros?

Thank you in advance,

Best regards,

Alexandre Jousse

  • Alex,

    Once you have measured the Plant, the compensation designer let;s you import that data as plant information.

    Now you can use any of the compensation style to tune your loop, typically users look for Bandwidth BW, Phase Margin PM and Gain Margin GM, of the open loop gain.

    You can proceed with a simple tuning process like PID, or many power supply engineers prefer using poles and zeroes. Typically the process is some what iterative in selecting the poles and zeroes. YOu will start with PID then if that does not meet performance criteria , try 2p2z, in some cases when there is resonance you may want a complex zero to damp it, if still does not meet requirements you may have to try 3p3z . The affect is pole and zeros affect the loop response.

    there are some great youtube tutorials on this: 

     

  • Hi Manish,

    I have a look at these tutorials that seem very well done.
    I was wondering if I had to add to the output of the controller the numerical gain of the PWM (which is 150MHz/100kHz+1 = 1500).

    Thanks

    Alex
  • Alex,

    1500 is the resolution not the gain, i.e. it has 1500 points,

    Gain will be Vdc

    regards
    manish
  • Hi,

    Thank you for your answers. 

    I'm a little bit confused though because when I calculate my controller the "old way" with my transfer functions and stuff for example in the current loop with a simple proportional integrator, I have a Kp and Ti that I have tested on my converter and that is stable. But when I do an SFRA on the inductor current of the converter and that I use my Kp and Ti on compensation deisgner, it says that the loop is unstable... And I am pretty sure that it is stable with the computations I have made.

    Thanks again for helping me,

    Best,

    Alex

  • Alex,

    Make sure you have the Frequency of the control loop set right,

    What does the SFRA GUI tell you? the SFRA measures the OL curve you can check to see if it is stable by seeing the SFRA GUI ??

    If you can provide some graphs that will help in advising you, 

    Regards

    Manish

  • Hi Manish,

    Here are the graphs I have in open loop by inserting the sinus injection in the alpha of the PWM:

    NB: Here, I got all the data by looking directly into the memory of the SFRA registers. My converter is a Boost with a duty cycle of 0.85. The input voltage is 10V.

    Moreover, on the plant gain response, the resonant frequency is almost identical to the theoretical bode diagram. Thus, I believe that the frequency of the control loop is set right.

    Thank you for your time,

    Regards,

    Alex

  • Alex,

    I think the compensation designer may be getting confused because of the a-typical OL curve. Here are my observations:

    You should have some Gain (positive gain) in a typical power supplies, if it does not see that it will mark it unstable.

    Also i am not sure if your OL curve crosses twice in the Gain which a typical power supply will not do.

    Also for a positive gain you have 180 degree phase which means it is unstable.

    You can import this data in MATLAB, i am quite sure it will classify it as unstable as well..

    Following videos may help you understand some more,

    www.youtube.com/watch

    www.youtube.com/watch

    Also if you are into TI literature

    www.youtube.com/watch;list=PLAgYbpzfyR1vgNOEuWN35Y8TSWO3al9Tv
  • Hi,

    So I am not sure that I understand everything. I am running the SFRA library in an open loop so basically why are there two different bode diagrams? The plant response has no controller at all and it is just an open loop..

    Moreover, when I try compensator coefficients that the GUI find stable, they are not on my converter.

    Thanks again

    Alex

  • When operating in open loop, the open loop gain measurement will be incorrect as the loop is not closed.

    We capture this in

    www.ti.com/.../spruhz5a.pdf

    under Case Study, Page 42, bullet 35
  • I can't find why I have the transfer function of the form:

    1.539e04 s - 1.472e08
    -----------------------
    s^2 + 4707 s + 3.839e07

    with the experimental SFRA data where I should have this type of TF with a Boost converter:

    a - b * s
    -----------------------
    s^2 + c * s + d

    Because this is the reason why I have a 180° phase at a positive Gain..
  • Alex,

    I am probably not able to comprehend your question, 

    The compensation designer GUI does not extract the compensator from the SFRA data. It is purely to design compensator based on the PLANT data that SFRA collects. It ignores the OL data. 

    On the Compensation designer GUI, you can enter the poles and zeroes and these will then be multiplied with the Plant SFRA data to give you the open loop gain plot. 

    If the question is related to that you curve fitted the SFRA data to get the TF, and it does not match the modeled form. Then the only confirmation we can give is that SFRA data will be correct to what you will measure even with an external frequency response analyzer. Please note when comparing with the external analyzer you should compare OL gain curves. 

    Also just looking at your TF you do not have zoh accounted for in the model, this can change the number of pole and zero you need to curve fit to get an accurate estimate. Also the mapping from bode-> p&z is not one to one and hence a visual check is much better than checking the pole zero location itself. 

  • Hi Manish,

    Does the Compensation Designer GUI's calculation of Phase Margin account for phase margin erosion due to the digitalization process, related to sample frequency and crossover frequency, or should we budget that in ourselves as extra beyond the displayed Phase Margin number? The expression I've seen for estimating this effect is "phase_margin_erosion = 360degrees * F_co / Fs", where F_co is the crossover frequency and Fs is the sample frequency.

    Thank you,

    Dave
  • David,

    Depends on how you use the compensation designer, if you use the plant in the compensation designer as measured by SFRA it will already have the phase margin erosion accounted for as the plant is measures in circuit, this is one of the advantages of using SFRA for plant measurement.

    Another option supported in powerSUITE is to use the model of a particular solution like Buck, PFC etc, this option is only available from the solution adapater and also accounts for the phase margin erosion. In that it is modeled as a zoh delay.
  • I see, that helps. Thank you, Manish!