Abstract:
The power estimation in electronic system designs (ESD) has become a significant problem. To keep a balanced approach between the estimation accuracy and speed is the key challenge in the system design. In this paper we present a power analysis method at high abstraction level. We focus on the statistically power dissipation sources of the individual ESD blocks and interconnects/buses of the system. Our hybrid technique combines the regression analysis and look-up-table (LUT) approaches with the non-linear behavioral function. Our algorithm generates input binary signals with five statistical characteristics that help to estimate the average power dissipation. For the power function, we performed Monte-Carlo zero-delay simulation and observed the exact steady-state power behavior of the system. Our power model estimates an accurate power with 10.16% average error. The results are compared with a commercial power estimation tool for the validity of our proposed approach.
