sisotool graphical tuning problem with frd discrete model

Hello! I have the frd model which created by

   G=frd(A,[1e1:1:1e5],Ts);

and compensator

   C=(5.724*z-5.615)/(z-1)
  Ts=1.428e-005;

System: http://i.piccy.info/i5/94/51/1035194/system.gif

In sisotool graphical tuning this system appears incorrect http://i.piccy.info/i5/48/51/1035148/sisotool.gif but in ltiviewer it is correct http://i.piccy.info/i5/53/51/1035153/ltiviewer.gif

Thanks

Answer 1

[brag]

With another FRD model sisotool works even worse http://i.piccy.info/i5/00/29/1092900/Untitled-1.gif right window(LTI viewer is true)

we can see a big gain offset and incorrect phase response in Graphical Tuning Utility(left window)

What is wrong?

Answer 2

[Craig]

Hi Brag,

This is a bug in the SISO Design Tool that has been fixed for the upcoming release R2011b. One way to work around this issue in the short term is that you could loop-shape the frequncy response with a continous time controller by setting the Sample Time of the FRD model to be 0. Once you are finished loop shaping use the c2d command (note: matched or tustin method will problably give best frequency domain match) to convert your controller to to the appropriate sample time.

[Adeyemi Adepegba]

It doesnt allow you set a sample time of zero.

Answer 3

[Adeyemi Adepegba]

Hi Craig,

When is R2011b going to be released ?

Mathworks doesn't mention anywhere about this bug, how did you find out about this. I have done a substantial number of designs in discrete time using SISO tool and the simulations results seemed to be correct, are you saying I have to discard this. My system is such that I have to do all my designs in the discrete domain as the system has a peculiarity such that it produces an unstable zero which is not present in the continuos time. If I use the sampling idea you suggested, the practical implementation would still fail.

[Walter Roberson]

Mathworks does not publicly give out concrete release dates before the actual release, but _traditionally_ their "b" release is early September. For example R2010b was released September 3, 2010.

[Craig]

Hi Adeyemi,

This issue only affects the bode/nichols graphical editors plots when the plant model is defined as a discrete time FRD model. If the plant model is defined as either a discrete or continuous zpk, ss or tf model, you will not run into this issue.