SelfCompensating Design for Focus Variation - PowerPoint PPT Presentation

1 / 15
About This Presentation
Title:

SelfCompensating Design for Focus Variation

Description:

Across chip linewidth variation (ACLV) control is critical to ... Annular. Illumination type. 0.7. NA. 248nm. ?(wavelength) Parameters. 13% 11% 7. Edge Devices ... – PowerPoint PPT presentation

Number of Views:24
Avg rating:3.0/5.0
Slides: 16
Provided by: youngm2
Category:

less

Transcript and Presenter's Notes

Title: SelfCompensating Design for Focus Variation


1
Self-Compensating Design for Focus Variation
  • Puneet Gupta1, Andrew B. Kahng1,
  • Youngmin Kim2, Dennis Sylvester2
  • 1 Blaze DFM Inc., Sunnyvale CA
  • 2 EECS Department, University of Michigan, Ann
    Arbor

2
Outline
  • Introduction
  • Compensating Focus Variation
  • Linewidth Variation with Defocus
  • Experiments and Results
  • Conclusion and Future Works

3
Introduction
  • Within-die process variation has become one of
    the most important considerations in
    sub-wavelength regime.
  • Across chip linewidth variation (ACLV) control is
    critical to the timing and functionality of a
    design
  • Exploit of various RETs (Resolution Enhancement
    Techniques)
  • SRAF, OPC, PSM, and etc
  • Components of variation
  • Intrinsic variation (fabrication stage)
  • Random
  • Systematic
  • Dynamic variation (circuit operation)
  • Motivations
  • Focus is one of the major source of Leff
    variation
  • Traditional corner-case timing is very
    pessimistic in worst-case focus impact
  • Layout pitch and focus have very systematic
    interactions (e.g. Bossung plots)
  • Systematic through-pitch and through-focus
    variation can reduce timing uncertainty by up to
    40 (P.Gupta, H.Fook-Luen, Proc.DAC, 2004)

4
Compensating Focus Variation
  • Reducing systematic variation by OPC and AFs, but
    not completely.
  • Key Idea Isolated and dense lines retaining
    opposite behavior under varying defocus ?
    compensating for systematic variation in the
    design itself.
  • Compensation
  • Self-compensated cell layout compensation
    within-cell (e.g. one device iso and the other
    dense)
  • Self-compensated physical design compensation
    across cells (e.g. G1?G2, G1 iso gate, G2
    dense gate)

5
Self-Compensating Design Flow
  • Litho. Simulation and construction of CD look-up
    table (LUT) gives the Leff at different spacing
    (S) and focus (F) values
  • CD f ( Left Space, Right Space, Focus)
  • Library generation
  • 21 cells, 4 variants of each cell (original,
    iso, dense, and self-compensated cells)
  • Self-compensating design
  • Self-compensated cells
  • Optimization (self-compensated physical design)
  • Dense iso design
  • Original iso design
  • Optimization of delay versus area
    sensitivity-based approach to minimize area
    penalty while instantiating iso counterparts of
    dense cells in the circuit to meet timing
    constraints.

6
Linewidth Variation with Defocus
Parameters
13
11
  • 3 different range of space (criteria 4nm CD
    variation from nominal)
  • Dense 180nm 260nm
  • Self-compensated 280nm 360nm
  • Iso 360nm 400nm
  • (Note that 1st SB insertion point is at 420nm
    spacing)
  • In most-iso (i.e. 400nm), linewidth decreases
    11 from nominal at 0.4um defocus, while in
    most-dense (i.e. 180nm), linewidth increases
    13 from nominal

7
Edge Devices
Case1 no neighboring devices on either side
Case2 only one neighboring device
  • The distance from edge devices to the cell
    boundary for all cells is over 600nm ? the
    distance of two neighboring poly lines gt 1.2um
  • In case1, linewidth is insensitive to focus after
    two SBs are inserted (i.e. space gt 700nm) ?
    self-compensated
  • In case2, 180nm case behave like dense but,
    380nm case behave like self-compensated

8
Sample Cell Layout (NAND2x6)
9
Experiments and Results
  • Worst-case path delay of selected ISCAS85
    benchmarks with iso/dense/self-compensated
    libraries at 0.0 and 0.4um defocus level
  • In average, dense cells at 0.4um defocus give 15
    slower timing and iso cells at 0.4um defocus give
    9 faster timing than original library at 0.4um
    defocus
  • Average area overhead of the different cell
    versions (3.2 for dense, 17.4 for iso, and
    10.3 for self-compensated cells)

10
Optimization (Self-Compensated Physical Design)
  • Sensitivity-based optimization of delay versus
    area by instantiating iso counterparts of
    dense cells

While worst_slack is negative Calculate
sensitivities of all gates in the circuit Sort
sensitivities in non-increasing order Swap the
dense version with iso cell based on the
order of sensitivities Calculate new_delay of
circuit Update worst_slack
11
Optimization Results (c3540)
  • Black line below 0.0 slacks implicate timing
    failure over 0.23um defocus level in original
    library.
  • Self-compensated design, optimization (denseiso,
    originaliso) all satisfy the timing requirement
    throughout the defocus range

12
Area Penalty of Self-Compensating Design
  • 10 12 area penalty for self-compensated cell
    base design
  • 6 9 area penalty for the dense iso
    optimization
  • Less than 1 of area overhead in original iso
    optimization
  • 32 of dense gates need to replace to meet the
    timing in dense iso, but only 11 of original
    cells need to replace in original iso
    optimization

13
Distribution of Focus and Delay
  • Monte-Carlo simulation with 1000 trials
  • Normal distribution of focus with mean0.0um and
    3s0.4um
  • C3540 benchmark circuits with require time
    2.177ns is shown
  • 2 optimization strategies demonstrate appreciably
    tighter distribution than self-compensated cells
    option.

14
Conclusion and Future Works
  • A novel design technique to compensate for
    lithographic focus variation is proposed
  • More robust design to focus variation
  • Self-compensated cells
  • Optimization (i.e. Dense iso, Original iso)
  • Dense and iso optimization option shows 69 area
    overhead compare to 1012 in a self-compensated
    library base design. Only 1 area penalty by
    using original library and iso library
  • The impact of compensating design of more
    advanced technology will be even greater (e.g.
    65nm)
  • Layout generation for self-compensated, iso, and
    dense cells, and APR flow for more accurate
    timing and area analysis.
  • Joint optimization in the delay/area/leakage
    space ( iso fast but leaky, dense slow but less
    leaky)

15
Thank You!
Write a Comment
User Comments (0)
About PowerShow.com