| Computer-Aided Design tools play an important role in the efficient design of high-performance digital systems. In the past, time, area and testability were the main concerns of the design community. With the growing need for low-power electronics, power analysis and low-power synthesis have also become major design concerns.; The objective of this thesis is to provide the theoretical foundations of an integrated framework for power analysis of digital CMOS circuits. To this end, the problem of power estimation is addressed from a probabilistic viewpoint and the following contributions are made: (1) For the class of static power estimation techniques, the previous work done on switching activity estimation is extended to explicitly account for complex spatiotemporal correlations which occur at the primary inputs when the target circuit receives data from real applications. More precisely, using lag-one Markov Chains, two new concepts—conditional independence and signal isotropy—are brought into attention and based on them, sufficient conditions for exact analysis of complex dependencies are given. It is shown that the relative error in calculating the switching activity of a logic gate using only pairwise probabilities can be upper-bounded. Relying on the concept of signal isotropy, approximate techniques with bounded error are proposed for estimating the switching activity. (2) For the class of dynamic power estimation techniques, an approach which reduces the simulation time by orders of magnitude is proposed using the paradigm of sequence compaction. Given an initial input sequence, we compact this sequence into a much shorter one such that the new data represents a good approximation as far as total power consumption is concerned. To this end, a hierarchical Markov model is introduced as a flexible framework for capturing not only complex spatiotemporal correlations, but also the dynamic changes in the sequence characteristics. Furthermore, a family of variable-order dynamic Markov models is presented to handle the case of sequential circuits. These Markov models provide an effective way for accurate modeling of external input sequences that affect the behavior of finite state machines. |
