Efficient Evaluation of the Time-Dependent Threshold Voltage Distribution Due to NBTI Stress Using Transistor Arrays

For the design of reliable ICs the application of accurate transistor aging models is indispensable. However, especially for modern deeply-scaled technologies modeling NBTI induced aging poses significant challenges as transistors can not be characterized individually but need to be described as a statistical ensemble. We make use of a dedicated array structure enabling the collection of a large data-set at reasonable experimental time. Note that by using an extensive data-set the statistical confidence can be maximized. We present a modeling approach based on physical considerations to describe the aging of SRAM-sized transistors under consideration of statistical effects. Our approach applies a modified defect-centric model under consideration of RTN to describe the time-dependent V
th
distribution. Furthermore, we introduce a simple method to extract parameters describing the distribution. Additionally, we show that our approach provides great flexibility regarding stress conditions as well as scalability to different transistor dimensions, which makes it ideal for applications such as circuit simulation.