| Summary: | Exhaustive testing of high-resolution analogue-to-digital converters (ADCs) is confounded by the great number of parameters which need to be determined to evaluate performance in both the analogue and digital signal domains. Such complex procedures are time-consuming and, in a commercial context, expensive. As a result tests have been evolved which, although speedier and more commercially viable, cannot adequately exercise all the parameters of interest, commonly employing only a single sinusoid (single tone), or occasionally two tones as the test signal. This work reviews many of these test methods and summarises their respective advantages and deficiencies. Discussion follows upon what might be required from an 'ideal' test method within the commercial constraints and the need to test each ADC under less atypical signal conditions. Parallels are drawn with established analogue signal testing techniques and the slew-rate distribution and amplitude probability density function of the test signal are identified as being important issues. It is argued that a three tone test signal can satisfy many of these requirements. The processing of the codes obtained at the output of the converter under test, the application of the Fast Fourier Transform and digital signal processing technology are discussed in order to derive a potential new test method. The experimental test signal comprised 3 high-purity synthesised sinusoids carefully arranged to permit the direct application of the Fast Fourier Transform without the deleterious effects resulting from the need to introduce an intermediate windowing function. Experimental results are presented for a range of 12 and 14 bit ADCs, together with results obtained for the same devices from the established 'servo-loop' and 'histogram' methods. Finally the three sets of results are discussed and the basis of a standard method of three tone testing is proposed.
|
|---|
