Testing high-resolution digitizers using conventional signal sources

This paper presents a method to evaluate the performance of high-resolution digitizers using signal sources that may be of lower specification than the device under test. The method is based on the best fitting of a straight line to a set of measures of the signal to noise ratio, excluding the harmonic distortion, obtained for several amplitudes of the test signal. Theoretical derivations and both computer simulated and experimental data are included. It is shown that the effective number of bits estimated using this method is not affected by the leakage effects that occur when other frequency domain tests are used and the input signal is sampled along a non-integer number of cycles.