Re: Questions

Dear Ouisberg

I am glad you like our book. Sigurd has asked me to reply to some of
your questions so here goes.

> First question,
> I tried to change the controller scheme with 2 dof control for good time
> response but I don't know exactly how I should divide the controller
> with K1 and K2 shown in Figure 9.22 on your book. I need more
> explanation detaily.

I am not exactly sure what extra detail you require. The code for
calculating the controller is given on page 389.

You could also try looking at some of the original references eg
Hoyle et al (1991) and Limebeer et al (1993).

> Next one,
> I guess the low stiffness of the magnetic bearing using Hinf with ncf is
> due to integral windup.

This maybe so but it may be that your lop gain is rolling off too
quickly. You could try flattening out the loop gain in the region of
cross-over.

> Because the controller shows quite large gains to the stage but the
> stage would go mad from the beginning of the control action is turned
> on. The controlled output are saturated already as soon as the control
> starts.
> So I have no choice to use a little bit low gains. If the anti-windup
> can be implemented, the stage will get a high stiffness with high gains.
> I read a lots of literature and papers on the anti-windup and understand
> theorically(not perfectly).
> But,I can't find out how to implemet with matlab or on the actual
> system.
> I would like to get some examples.

A useful reference on anti-windup schemes is Edwards and
Postlethwaite, Automatica, 34(2), 199-210, Feb 1998.

> The last one,
> while I was designing a loop shaping with ncf, 'align' step was comming
> out.So I run the align command at that step. wow. It aligned whole
> singular value plots perfectly as the same one. I was suprised with the
> powerful algorithm and made a controller. The controller was converted
> at last with c2d command for implementation to the dsp control board. I
> was surprised once more because the output were mixed up. What happened?

Sounds like a sign error or if you are implementing this on a real
system perhaps your model is inaccurate.

Hope the above is useful.

Regards

Ian
-------------------------------------------
Professor Ian Postlethwaite
Head of Department
Department of Engineering
University of Leicester
Leicester LE1 7RH
UK

Tel: +44 (0)116 252 2547
Fax: +44 (0)116 223 1369
Email: ixp@le.ac.uk
Received on Thu Nov 04 17:02:58 1999