Dear Dr. Koteich,

I' d like to know:

1. The methods you suggest are better than Bode's method?

2. They really improve quickness of the system and overshot of transient curves?

Best wishes,

I. Dochviri

Hi,

As have been mentioned, the Bode plot is based on frequency domain analysis.

There are many approach based on time domain (if you are bounding yourself to linear systems) such as, root locus, pole placement, LQR, LQG.

Some of the approach will be more effective depends on the type of the system that you have and whether the system model is available of not.

Hi,

If we talk about LTI sytems, Bode maybe is the simplest and easiest in... simple systems, where the curves look close to stright lines and it is easy to see gaim margin, phase margn, bandwidth, etc.

As soon as the gain and phase plots start moving up-and-down, as it usually happens in realistic cases, Bode is just confusing. If the open-loop is unstable or just non-minimum phase, understanding the eventual behavior of the closed-loop becomes mission impossible.

Then you must go back to Nyquist theorem and Nyquist plots. However, as the gains may reach values as high as 10000 or even more, while you must see what occurs there yet also, mush more important, what occurs around the critical point -1, it is hard to work with Nyquist plots.

Here, Nichols plot, because of its logarithmic scale is of great help in practical applications. Also, it is not afraid of non-minimum-phase. Also, in real production, when you deal with more than one machine and want to test your controller design with all of them, it is the only toool that allows you to plot them all on one plot and see if the center region (around -1, which has become 0 dB, -180 degreees) remains clean and all lines go below it.

When I started, ploting a Nichols plot was just terrible. Now, however, with MATLAB tools, you just write nichols(sys), where sys is your system, and you get everything you want.