A Power Modeling and Characterization Method S Standard Cell Library
Wen-Zen Shen, Jiag-Yang Jou
Abstract
Wen-Zen Shen, Jiag-Yang Jou
Abstract
In this paper, we propose power consumption models for complex gates and transmission gates, which are extended from the model of basic gates proposed in [I]. We also describe an accurate power characterization method for CMOS standard cell libraries which accounts for the effects of input slew rate, output loading, and logic stare dependencies. The characterization methodology separates the power consumption of a cell into three components, e.g., capacitive feedthrough power, shortcircuit power, and dynamic power. For each component, power equatioh is derived from SPICE simulation results where the netlist is extracted from cell's layout. Experimental results on a set of ISCAS'85 benchmark circuits show that the power estimation based on our power modeling and characterization provides within 7% error of SPICE simulation on average while the CPU time consumed is more than two orders of magnitude less.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
In this paper, we propose power consumption models for complex gates and transmission gates, which are extended from the model of basic gates proposed in [I]. We also describe an accurate power characterization method for CMOS standard cell libraries which accounts for the effects of input slew rate, output loading, and logic stare dependencies. The characterization methodology separates the power consumption of a cell into three components, e.g., capacitive feedthrough power, shortcircuit power, and dynamic power. For each component, power equatioh is derived from SPICE simulation results where the netlist is extracted from cell's layout. Experimental results on a set of ISCAS'85 benchmark circuits show that the power estimation based on our power modeling and characterization provides within 7% error of SPICE simulation on average while the CPU time consumed is more than two orders of magnitude less.
Key concepts: Netlist, Spice, Standard cell, Dynamic demand, Benchmark (surveying), Power (physics), Electronic engineering, Computer science