Information source model of A/D converters to characterize digital spectral analyzers based on averaged smoothed periodograms

Highly accurate spectral estimates on noisy multitone signals can be performed by digital spectral analyzers, realized, today, through a dedicated hardware architecture or, otherwise, by means of object-oriented software-modules. The accuracy characterization of the PSD estimation algorithm has been made possible deriving the expression of the rms error which occurs after averaging estimates of overlapped periodograms, smoothed with flat-top windows. The global noise of the A/D conversion has been evaluated and accounted for through effective bits, which are suitable to give the accuracy of estimate. A proper modelling of the A/D conversion block, on the counterpart, leads to define the figure of merit or effective information, more suitable to characterize the A/D block in terms of information rate