Probabilistic standard cell modeling considering non-gaussian parameters and correlations

Since semiconductor structure sizes reached the regime of 100~nm and below, global and local process variability have affected integrated circuit design. In the domain of digital circuits, enhancements to standard cell libraries were proposed to consider fluctuations in timing to ensure correct circuit functionality. However, further performance characteristics, such as dynamic energy and leakage power, are rarely taken into account, and there is nearly no method preserving the correlations between the performance characteristics. These drawbacks are addressed in this paper. We demonstrate a probabilistic standard cell modeling approach that treats the set of standard cell performance characteristics as a multi-dimensional random variable. In addition, the corresponding characterization setup is introduced. The application to a set of standard cells in a 28~nm technology shows that our modeling method supports probability distributions of multiple shapes and preserves correlations.