Study of bit error and thermal characterization of a successive approximation A/D converter

Abstract

The bit weight errors are derived in a real case by Walsh transforming the nonlinearity error measured for a Successive Approximation AID converter. The estimates of the bit errors obtained from the differential nonlinearity error and from the integral nonlinearity error are compared on the basis of the experimental conversion characteristic. To force the appearance of sizeable bit errors, a thermal characterization was performed in a wide temperature range. The operation of a Successive Approximation A/D converter (SAR-ADC) is based on the comparison of the analog input and a reference level which, during the conversion process, is changed following a binary search strategy in accordance with the comparison results. The reference level is provided by a built-in D/A converter with binary weighted voltages, currents or charges (l). The presence of D/A converter inaccuracies leads to errors in the whole analog-to-digital conversion process (2). It is therefore important to correlate the influence of the error in each binary weight with the input-output behaviour of the converter. It is known that the discrete Walsh transform (3) of the linearity error allows to estimate the binary weight errors 121. In 141 the theory is expanded and verified by simulations. In the present work this theory is applied to a practical case. Section I1 provides the basic background, while in the section I11 the experimental results are described. To facilitate the appearance of binary weight errors, the converter was stimulated in a wide temperature range. Finally, since bit weight errors may be estimated by Walsh transforming both differential and integral non linearity errors, a comparison between these methods wlis made, on the basis of experimental data and with the aid of computer simulations. 11. BACKGROUND A. The quantizer model