Defocus-Aware Leakage Estimation and Control ( http://vlsicad.ucsd.edu )

1
Defocus-Aware Leakage Estimation and Control(
http//vlsicad.ucsd.edu )
Puneet Sharma (sharma_at_ucsd.edu) Advisor Prof.
Andrew B. Kahng Jointly with Mr. Swamy
Muddu Electrical Computer Engineering
Introduction
Defocus-Aware Leakage Estimation
Defocus-Aware Gate-Length Biasing

• Previously proposed gate-length biasing (Gupta et
  al. 04) selectively increases the gate length of
  devices to reduce leakage at the cost of device
  delay increase.
• Only the devices on non-critical paths are biased
  for minimum circuit delay penalty.
• Biasing a device can prevent biasing other
  devices so order in which devices are biased is
  important. Bias in order of decreasing
  sensitivity ( leakage improvement/delay
  penalty).
• ? Accurate estimation of leakage is important.
• We add defocus awareness by substituting
  leakage estimation in gate-length biasing with
  our defocus-aware estimation.
• Results

Leakage power is one of the most critical issues
for ultra-deep submicron technology. Subthreshold
leakage is dependent exponentially on linewidth
and therefore variation in linewidth translates
to a large leakage variation. A significant
fraction of variation in linewidth occurs due to
systematic variations involving focus and pitch.
In this work, we propose a new leakage estimation
methodology that accounts for focus-dependent
variation in linewidth. The ideas presented in
this work significantly improve leakage
estimation and can be used in existing leakage
reduction techniques to improve their efficacy.
We modify the previously proposed gate-length
biasing technique to consider systematic
variations in linewidth and further reduce
leakage power. Our method reduces the spread in
estimated leakage between worst-case and
best-case process corners by up to 56.
Defocus-awareness improves leakage reductions
from gate-length biasing by up to 7.

• Methodology
• Create Bossung LUT using several parallel line
  patterns of varying pitch and for defocus values
  in the range -20020200nm.

Compute Sensitivity
Select Bias Cell
Unbias if slack not met
Timing Analysis
Defocus-Aware Leakage Estimation
Background

• Leakage decreases exponentially with linewidth
  (gate length).
• Better linewidth estimation ? better leakage
  estimation.
• Linewidth variation is partly systematic. Using
  defocus and line pitch, 30 of gate-length
  variations can be modeled.
• Defocus and pitch determine systematic linewidth
  variations.
• Recent advances in simulation of
  chemical-mechanical planarization (CMP) of
  shallow trench isolation (STI) layer allow
  accurate topography prediction. Topography
  variation contributes significantly to defocus.
• Defocus can be predicted from layout analysis.
• Can calculate line pitches from layout analysis
  (with LVS).
• ? Core idea Layout Analysis ? Linewidth
  Prediction ? Accurate Leakage Estimation

Bossung Plot Linewidth is a function of pitch and
defocus.
Conclusions and Ongoing Work

• Exponential dependence of leakage on linewidth.
• Pessimism in linewidth variation ? large leakage
  pessimism and over-design.
• Need to model systematic linewidth-variation
  components for improved leakage estimation.
• Our leakage estimation methodology reduces
  leakage spread by half and can estimate leakage
  of all devices with better accuracy.
• Leakage optimization benefits from improved
  leakage estimation e.g., up to 7 more leakage
  reduction from gate-length biasing.
• Future work
• Model other systematic variations such as lens
  aberrations.
• Consider delay impact of systematic variations
  within leakage optimization.