Integral non-linearity error measurement through statistical analysis of a multi-tone test signal

Conventionally, the integral non-linearity of an analogue to digital converter (ADC) is measured using a pure sine wave test signal [1]. The measured amplitude probability distribution function (APDF) of the ADC output is compared with the ideal distribution and the differences used to calculate integral non-linearity error (INL). Accuracy of this method is limited by the practical problem of generating a pure sine wave. Others have proposed random noise as an alternative to the sine wave test signal [2]. Unfortunately, a much larger number of samples is required for this method to achieve a given confidence level. This is due mainly to the need to be sure that they fit a Normal distribution with sufficient accuracy. The approach reported in the literature has been to fit a Tchebycshev polynomial to the INL result [3]. However, results measured using noise do not correlate well with those of sine wave tests. A new approach to testing will be described that makes use of a periodic signal derived from pseudo random binary sequences and offers the potential for increased accuracy in measuring INL.