Abstract
Integral non-linearity (INL) is the main static parameter of analog-to-digital converter. This paper presents a comparison between different INL test techniques based on INL estimation from the spectrum of the converted signal. The most common technique is based on polynomial fitting of the INL curve. This technique is well suited to the estimation of a smooth INL curve without sharp transitions. The new method described in the paper is based on a Fourier series expansion of the INL curve. We demonstrate that this new technique allows a more efficient INL estimation. The comparison between the two techniques has been realized thanks to a metrics that considers the uncertainty of production test measurements. Finally, we propose a first step of the study of implementation feasibility of the INL estimation technique. This study focus only on the optimization of required memory.