An IIR filter approach to fast convergence spectral analyzers

The use of the cascading form realization of an infinite-impulse response (IIR) linear time-invariant (LTI) filter in the estimation of the autocorrelation function from a stream of data samples provides a direct access to a variety of fast and slow convergence recursive power-spectrum estimators. These estimators are generated by first transforming the data into the spectrum domain and then propagating the result through the filter structure. Each of the filter poles and zeros alters the statistical properties of the spectrum estimate in a different manner and, therefore, provides a wide class of estimators which vary in performance and characteristics. A large number of classes can be generated as a result of using different order of cascade of the filter poles and zeros. A discussion is presented of the cascading structure of the IIR LTI filter from the power-spectrum-estimation perspective. The effect of the cascading order on the differences among the estimators associated with the filter is examined.<>