Dynamic Testing of A/D Converters Using the Coherence Function

The problem of characterizing an A/D converter is important to manufacturers and end users. The measured characteristics provide the means to compare devices and make design choices. This paper presents a procedure for testing a pair of A/D converters that sample the same analog signal. Since both A/D converters are measuring the same signal, the portion of the measured signals that is incoherent provides an estimate of the total noise of the two devices, from which the A/D converter effective number of bits and signal-to-noise ratio can be calculated. The removal of the coherent signal requires the calculation of the coherence function, which is computed from auto- and cross-spectral densities estimated by the Welch periodogram method. Examples of the procedure are provided for various input signals, input levels, sampling rates, and (simulated) interchannel delays. The measurement examples show that for white-noise inputs, the technique is independent of input-signal level, spectral data window, and sampling rate, provided the inputs are simultaneously sampled. With sine-wave inputs, the technique can be compared to the signal-to-noise and distortion ratio test, and it is shown that coherent removal eliminates the need for an ultrapure sine-wave source. The use of data windows is shown to be detrimental, since sidelobe energy is suppressed causing loss of coherence. The technique simplifies the setup and programming requirements for A/D converter testing because it can be applied using any input signal and is based on widely accepted processing techniques. End users and manufacturers alike should have little difficulty obtaining the programming tools needed to perform the analysis