Evaluation of the Iddq Signature in devices with Gauss-distributed background current

Abstract

In the last decade, the single-threshold IDDQ approach made way for more elaborated techniques like Delta-IDDQ and adaptive IDDQ. Due to increasing background currents, however, also these methods are beginning to have problems to distinguish between good and bad devices. A good evaluation algorithm for IDDQ takes all known information about 'good' and 'bad' parts into account, i.e. it 'knows' how the IDDQ signatures of good and bad parts look like. Unfortunately, not only do the signatures of good parts differ significantly, but the signatures of bad parts differ even more. Moreover, since IDDQ faults are more often than not non-fatal (not impairing the functionality), it is frequently hard to say if a device is really 'good' or bad'. There are two kinds of information, however, which are known without referring to a certain process or IC type: one is the model of the IDDQ fault, and the other is the statistical distribution of the background-IDDQ. Using this information, an estimator with higher discrimination capability than the traditional Delta-IDDQ-approach is created. Measurement results form several lots of a 180 nm chip are presented..