Skip to Main Content
Negative bias temperature instability (NBTI) is limiting the lifetime of pMOSFETs, and it is often monitored by the shift of threshold voltage DeltaVt. Different techniques have been developed to extract DeltaVt, including the conventional extrapolation of the quasi-dc transfer characteristic I d ~V g and the more recent extrapolation of ultrafast pulse I d ~V g and the on-the-fly evaluation at stress bias. After the same stress, these techniques can produce a DeltaVt difference of up to one order of magnitude. The interpretation of this large difference is still controversial. The objective of this paper is to bridge the gap between the DeltaVt values extracted from these techniques. Degradation and recovery during measurement, measurement and truncation errors, and calculation of transconductance are all examined. After taking these factors into account, the gap in DeltaVt still cannot be filled, and hence, the effect of sensing Vg on DeltaVt must now be considered. It is found that | DeltaVt | increases with sensing | Vg |, and therefore, the popular assumption of DeltaVt being independent of sensing Vg is invalid. After taking both the effect of sensing Vg and recovery into account, the gap in |DeltaVt| is successfully bridged. The difference between the effect of sensing Vg and recovery is explored, and the results show that they are two different phenomena. This paper provides test engineers a method for determining the worst case DeltaVt under a given operation voltage.