TESTING INL OR DNL―IS THERE A TRADE-OFF?

摘要

Traditionally, production testing of data converters is based on measuring either the Integral or the Differential Nonlinearity (INL or DNL), respectively and computing the other. Doubts about this practice are raised as examples show that some error mechanisms in a given circuit have a characteristic INL-signature while others have a distinctive DNL-signature. In this work we address the question "should one measure INL or DNL" We extend the Linear Error Mechanism Modeling Algorithm to use both INL and DNL measurements simultaneously (if so required) to measure a device under test (DUT) with maximum accuracy given a fixed number of measurement points.