Study of Floating Fill Impact on Interconnect Capacitance

1
Study of Floating Fill Impact on Interconnect
Capacitance

• Andrew B. Kahng Kambiz Samadi Puneet Sharma
• CSE and ECE Departments
• University of California, San Diego

2
Outline

• Introduction
• Foundations
• Study of Capacitance Impact of Fill
• Proposed Guidelines
• Validation of Guidelines
• Conclusions

3
Introduction

• Why fill is needed?
• Planarity after chemical-mechanical polishing
  (CMP) depends on pattern
• Metal fill reduces pattern density variation
• Stringent planarity requirements ? fill mandatory
  now
• Impact on capacitance
• Grounded fill
• Increases capacitance ? larger delay
• Shields neighboring interconnects ? reduced xtalk
• Floating fill
• Increases coupling capacitance ? significantly
  more xtalk ? signal integrity delay
• Increases total capacitance ? larger delay

4
Motivation

• Floating-fill extraction is complex
• Floating-fill capability recently added to
  full-chip extractors
• In past large buffer distance design-rule used
• Reduces coupling impact
• Density constraints cannot be met ? reduce buffer
  distance
• ? inaccuracy in capacitance estimation

• Grounded fill used despite disadvantages (e.g.,
  higher delay impact, routing needed)
• Designers use floating fill extremely
  conservatively
• ? Better understanding of capacitance impact
  needed

• We systematically analyze capacitance impact of
  fill config. parameters (e.g., fill size, fill
  location, interconnect width, etc.)
• Propose guidelines for floating fill insertion to
  reduce capacitance impact

5
Assumptions Terminology

• Same-layer analysis
• Fill affects coupling of all interconnects in
  proximity
• We study effect on coupling capacitance of
  same-layer interconnects
• ? simplifies analysis

• Usability not compromised because
• Coupling with same-layer neighbor large
• Validation multiple configs with different
  densities on different layers considered
• Fill insertion between two same-layer
  interconnects, increases coupling significantly
• Validation fill inserted everywhere
• ? Large fraction of coupling impact captured by
  same-layer analysis
• Synopsys Raphael, 3D field solver, used in all
  experiments

• Terminology
• Fill and coupling interconnects are on Layer M
  (layer of interest)
• ia and ib are interconnects of interest with
  coupling Cab
• We study increase in coupling ?Cab due to fill
  insertion
• Dimensions measured in tracks (0.3µ)

6
Outline

• Introduction
• Foundations
• Study of Capacitance Impact of Fill
• Proposed Guidelines
• Validation of Guidelines
• Conclusions

7
Foundation 1

• Experimental Setup
• Two interconnects on Layer M separated by three
  tracks
• Fill inserted on Layer M between two
  interconnects
• M1/M-1 density is set to 33
• 20 , 33 , 100 metal density for Layer M2/M-2
  tried

8
Foundation 2

• Experimental Setup
• Two interconnects on Layer M separated by three
  tracks
• M1 M-1 density is set to 33
• M2 M-2 assumed groundplanes
• Fill features inserted on Layer M at different
  locations

9
Outline

• Introduction
• Foundations
• Study of Capacitance Impact of Fill
• Proposed Guidelines
• Validation of Guidelines
• Conclusions

10
Fill Size

• Fill length (along the interconnects)
• Linear increase in ?Cab with Y-intercept

• Fill width
• Increases super-linearly
• Using parallel-plate capacitor analogy, 1/w
  relation expected

• Settings
• Interconnect separation 3 tracks
• Layers M-1/M1 have 33 density
• 2 track width, 1 track length

Guideline Increase fill length instead of width
11
Interconnect Spacing

• ?Cab decreases super-linearly with spacing
• For larger spacings (gt10 tracks), coupling with
  M-1 and M1 wires more significant
• Settings
• Fill size 2 tracks x 2 tracks
• Layers M-1/M1 have 33 density

Guideline Insert fill where wire spacing is large
12
Fill Location

• Y-axis translation
• Cab unaffected until fill close to an
  interconnect ending

• X-axis translation
• ?Cab increases linearly
• Capacitance between fill closer interconnect
  increases dramatically

• Settings
• Wire spacing 8 tracks
• Fill size 2 tracks wide, 4 long
• Layers M-1/M1 have 33 density

Guideline Center fill horizontally between
interconnects
13
Edge Effects

• Study two cases (1) two interconnects
  horizontally aligned, and (2) not horizontally
  aligned
• With Y-axis translation of fill, edge effects
  observed
• When fill no longer in Rab, ?Cab dramatically
  decreases
• Settings
• Layers M-1/M1 have 33 density
• Interconnect width 2 tracks
• Fill size 4 tracks long, 2 wide

Rab
Guideline Insert fill in low-impact region (
outside Rab)
14
Interconnect Width

• Change width of one interconnect
• Interconnect-fill spacing and interconnect
  spacing constant
• ?Cab increases rapidly, but saturates at 4
  tracks

Guideline Insert fill next to thinner
interconnects
15
Multiple Columns

• Vertically aligned fill geometries are said to be
  in a fill column
• Change number of fill columns in fill pattern
• Fill area is kept constant
• ?Cab reduces with number of fill columns
• Cf. Tran. Electron Devices 98 (MIT)
• Cf. VMIC-2004 invited paper (UCSD / UCLA)

Guideline Increase number of fill columns
16
Multiple Rows

• Horizontally aligned fill geometries are said to
  be in a fill row
• Change number of fill rows in fill pattern
• Fill area is kept constant
• ?Cab increases with number of fill rows
• As spacing between two fill rows decreases, the
  ?Cab decreases

Guideline Decrease number of fill rows and
inter-row spacing
17
Outline

• Introduction
• Background Terminology
• Study of Capacitance Impact of Fill
• Proposed Guidelines
• Validation of Guidelines
• Conclusions

18
Application of Guidelines

• Apply guidelines on 3 interconnect configurations
• Reasonable design rules assumed
• Configuration 1

• Guidelines applied
• Edge effects
• Maximize columns
• Minimize rows
• Centralize fill

19
Guidelines on Configuration 2

• Guidelines applied
• Wire width
• Minimize rows

20
Guidelines on Configuration 3

• Guidelines applied
• High-impact region
• Edge effects
• Wire spacing
• Minimize rows
• Centralize fill

21
Conclusions

• Coupling with same-layer neighboring wires
  significant and same-layer fill insertion
  increases it dramatically
• Systematically analyzed the impact of floating
  fill configurations on coupling of same-layer
  interconnects
• Propose guidelines for floating fill insertion to
  reduce coupling increase
• Ongoing work
• 3D extensions Impact on coupling of
  different-layer interconnects
• Timing- and SI-driven fill insertion methodology