Determining figures of merit from analog-to-digital converter output spectra

Abstract

Evaluation of the spectrum obtained from recorded output data after periodic, often sinusoidal excitation is a common method for gaining information on the dynamic performance of an analog-to-digital converter (ADC) under test. If all frequency components of the recorded data set have been sampled coherently, i.e., there is an integer number of periods in the sample, analysis of the recorded spectrum is comparatively straightforward. If coherent sampling is not possible, leakage in the spectrum occurs, the short-range part of which cannot be reduced by time-domain windowing. Period extraction and hybrid spectrum computation can be used to obtain low-leakage or even virtually leakage-free spectra from an incoherently recorded data set. Reliable information about the contents of both coherently and incoherently sampled data is provided by detailed line-oriented spectrum analysis, which is even possible if a certain amount of leakage is present in the spectrum. This analysis consists of line detection, line delimitation, splitting of spectral values and calculation of the individual power contents.