Aggressive scaling of CMOS circuits in recent times has lead to dramatic increase in leakage currents. Previously, sub-threshold leakage current was the only leakage current taken into account in power estimation. But now gate leakage and reverse biased junction band-to-band-tunneling leakage currents have also become significant. Together all the three types of leakages namely sub-threshold leakage, gate leakage and reverse bias junction band-to-band tunneling leakage currents contribute to more than 25% of power consumption in the current generation of leading edge designs. Different sources of leakage can affect each other by interacting through resultant intermediate node voltages. This is called loading effect and it leads to further increase in leakage current. On the other hand, sub-threshold leakage current decreases as more number of transistors is stacked in series. This is called stack effect. Previous works have been done that analyze each type of leakage current and its effect in detail but independent of each other. In this work, a pattern dependent steady state leakage estimation technique was developed that incorporates loading effect and accounts for all three major leakage components, namely the gate leakage, band to band tunneling leakage and sub-threshold leakage. It also considers transistor stack effect when estimating sub-threshold leakage. As a result, a coherent leakage current estimator tool was developed. The estimation technique was implemented on 65nm and 45nm CMOS circuits and was shown to attain a speed up of more than 10,000X compared to HSPICE. This work also extends the leakage current estimation technique in Field Programmable Gate Arrays (FPGAs). A different version of the leakage estimator tool was developed and incorporated into the Versatile Place & Route CAD tool to enable leakage estimation of design after placement and routing.
Leakage current is highly dependent on the steady state terminal voltage of the transistor, which depends on the logic state of the CMOS circuit as determined by the input pattern. Consequently, there exists a pattern that will produce the highest leakage current. This work considers all leakage sources together and tries to find an input pattern(s) that will maximize the composite leakage current made up of all three components.
This work also analyzes leakage power in presence of dynamic power in a unique way. Current method of estimating total power is to sum dynamic power which is ½&#;CLVDD2f and sub-threshold leakage power. The dynamic power in this case is probabilistic and pattern independent. On the other hand sub-threshold leakage is pattern dependent. This makes the current method very inaccurate for calculating total power. In this work, it is shown that leakage current can vary by more than 8% in time in presence of switching current.
Identifer | oai:union.ndltd.org:UMASS/oai:scholarworks.umass.edu:theses-1079 |
Date | 01 January 2007 |
Creators | Rastogi, Ashesh |
Publisher | ScholarWorks@UMass Amherst |
Source Sets | University of Massachusetts, Amherst |
Detected Language | English |
Type | text |
Format | application/pdf |
Source | Masters Theses 1911 - February 2014 |
Page generated in 0.014 seconds