APPLYING SYSTEM-IDENTIFICATION USING COMMERCIALLY AVAILABLE SOFTWARE AND HARDWARE

System identification methods are now incorporated into several commer cially available products, including two MATLAB Toolboxes and a number of frequency response analyzers. This paper reports on the accuracy o btained with the two MATLAB Toolboxes, when used with idealized noisy data, and also on their performance when used with real measurement da ta The performance of the curve-fitting algorithms on two commercial F FT analyzers is also compared using two applications for illustration. It is found that while the MATLAB Toolboxes perform almost equally we ll when used in the applications examined, the hardware analyzers are less comparable. One of the analyzers gives reliable results, but resu lts from the other can prove quite difficult to interpret in the noise -free case, and can be erroneous when a comparatively small amount of noise is present on the measured output signal.