Résumé
Industrial testing of Analog-to-Digital Converters (ADCs) consists in evaluating the functional parameters of the component under test. By comparing the achieved performances to the tolerance limits given by the device specifications, the faulty instances can be separated from the fault-free ones. ADCs are characterized by two types of parameters: static and dynamic. Each set of parameters requires a dedicated test procedure (usually a statistical analysis and a spectral analysis, respectively). Consequently, the testing cost is becoming uppermost in the cost price of ADCs, and more generally of mixed-signal circuits. Therefore, reducing the ADC testing cost represents a critical issue for mixed-signal circuit testing. This thesis aims at studying whether a test procedure exclusively based on spectral analysis could lead to the evaluation of the whole set of ADC performances. We have hence investigated the correlations between static and dynamic parameters. The study is based on the simulation of an environment model for ADC testing. In a first approach, we have shown that each static error influence on the measured dynamic parameters is significant enough to allow redhibitory static errors detection through dynamic performance measurement. In a second step, we have evaluated the statistical efficiency to detect faulty instances for several alternative test flows using only spectral analysis. We have finally developed a software tool enabling one to adapt the statistical efficiency evaluation of each flow to a realistic test context.