DPT Aware PEX Flow Evaluation – Preliminary Report

Download Report

Transcript DPT Aware PEX Flow Evaluation – Preliminary Report

Double Patterning-Aware Extraction and Static Timing
Analysis Flows For Digital Design Sign-Off in 20/14nm
Tamer Ragheb, Steven Chan,
Adrian Au Yeung, and Richard Trihy
Design Methodology CAD Team
June 2-6, 2013
Outline
 Why Double Patterning (DPT)?
 DPT Mask Misalignment Modeling in Parasitic Extraction
 Different DPT Extraction Flows
 Parasitic Extraction/STA Analysis on P&R Design Blocks
 Recommended 20/14nm Extraction/STA Block-Level Sign-Off Flow
July 17, 2015
2
Why Double Patterning (DPT)?
 20nm needs 64nm BEOL min. Pitch for scaling
 Delay in readiness of next generation lithography (NGL)
 BEOL min. pitch = 2.λsource.k1 / NA
 λsource =193nm / NA=1.35 / k1 is the process coefficient =0.25 (difficult printability)
 Min. Pitch ~ 72nm with one mask  Solution: Double Patterning Technology
 DPT utilizes the existing lithography intelligently
 Most commonly adopted approach is a litho-etch, litho-etch (LELE)
 Pros: Achieves 64nm pitch needed / Relaxes k1 coeff (better and more reliable litho)
 Cons: Can’t guarantee 100% accuracy in overlaying the two masks
Decomposition
One Mask
July 17, 2015
Two Masks
3
DPT Misalignment Modeling Techniques:
Mask Shift
 Two common approaches for DPT misalignment modeling:
1. Mask shift flow
 Actually shifts one mask with respect to the other mask in XY directions
 Pros: Accurate if the mask misalignment on silicon is known (value & direction)
 Cons:
 Requires coloring or decomposing the design
 Requires designer to specify exact mask shift amount and direction for each DPT layer
 Determining worst case impact on timing requires 2N different shifts extractions per RC
corner, where N is the number of DPT layers
July 17, 2015
4
DPT Misalignment Modeling Techniques:
DPT Corner
 Two common approaches for DPT misalignment modeling:
2. DPT corner flow
 Models mask misalignment as change in dielectric constant (ER_VS_SI_SPACING)
 Pros:
 Bounds mask misalignment effect
 Supports both colorless (non-decomposed) and colored (decomposed) flows
 Requires almost no changes to existing sign-off flows
 Cons: Can be pessimistic with respect to a real mask shift (assume change in space on both
sides)
July 17, 2015
5
Different DPT-aware Extraction Flows
Full or
selective
coloring
July 17, 2015
6
Outline
Why Double Patterning (DPT)?
DPT Mask Misalignment Modeling in Parasitic Extraction
Different DPT Extraction Flows
 Parasitic Extraction/STA Analysis on P&R Design Blocks
 Recommended 20/14nm Extraction/STA Block-Level Sign-Off Flow
July 17, 2015
7
DPT Extraction/STA of P&R Design Blocks
 SIMD multi-media engine from a high-frequency CPU core
 Std cell count ~150K / Freq > 1.25GHz / with > 85% std cell utilization
 Routing on M2-M7 (M1 is reserved for standard cells only)
 Three different BEOL stack options used
 3Mx: 3 DPT levels (baseline for comparison)
 3Mx_dense: 3 DPT levels but denser than 3Mx
 6Mx: 6 DPT levels
 For each of the 3 BEOL variants, we ran 13 different Extraction/STA flows
 Analyzed change in capacitance distribution / overall impact on timing & Frequency
Extraction
Flow
Corner/Shift
Decomposition
Colored_DPmax
DPT corner
DPmax
Fully-decomposed
Colorless_DPmax
DPT corner
DPmax
Non-decomposed
Shifts 1-4
Mask shift
Fully-decomposed
Shifts 5-8
Mask shift
Colored_DPmin
DPT corner
4 Typical shifts (±x, ±y)
4 Maximum shifts (±x,
±y)
DPmin
Colorless_DPmin
DPT corner
DPmin
Non-decomposed
No DPT modeling
Traditional
Traditional
Non-decomposed
July 17, 2015
Fully-decomposed
Fully-decomposed
8
Findings from Capacitance distributions
1) Direction of mask shift is NOT important

Statistically, shift value is more important

Shift direction is important for specific paths (2N different combination/shift value)
2) Both colored and colorless DPT corners/flows have similar results
3) DPT corners bound ALL shifts
4) DPT effect increases by increasing routing density & #of DPT layers
Max. Shifts
Typical Shifts
July 17, 2015
9
Findings from STA Results: WNS
 Using the freq change in the slowest path “IP Freq”
1) Direction of mask shift is important when we study just one path in STA
2) Both colored and colorless DPT corners have similar results (timing difference
within noise)
3) DPT corners bound most mask shifts (all Typical shifts & most Max. shifts)
Worst Negative Slack Path
July 17, 2015
10
SI-aware Routing & Hold Time analysis
1) DPT effect decreases by applying
SI-aware routing

~50% reduction in DPT effects

All our results with SI-aware routing
Stack
DPT effect w/o SIaware routing
DPT effect w/ SIaware routing
3Mx
~X%
<0.5X%
3Mx_dense
~Y%
<0.7Y%
6Mx
~Z%
~0.5Z%
2) DPT effect on Hold time is minimal due to:

Short data paths – not much DPT effect

Clock skew induced race conditions possible

But DPT effect on clock skew is very small

Most clock routes are not on DPT layers

Clock is routed with 2w/2s NDRs
– less impact
July 17, 2015
11
Outline
Why Double Patterning (DPT)?
DPT Mask Misalignment Modeling in Parasitic Extraction
Different DPT Extraction Flows
Parasitic Extraction/STA Analysis on P&R Design Blocks
 Recommended 20/14nm Extraction/STA Block-Level Sign-Off Flow
July 17, 2015
12
Extraction Corners Recommendations
 5 traditional corners with mask_shift enabled
 Cmax / Cmin / Nominal / RCmax / RCmin
 Enable mask shift to analyze DPT effect on specific paths if needed
 Recommended 4 DPT corners for most P&R designs and flows
 4 DPT corners expand the BEOL space to account for mask misalignment
 CmaxDPmax / RCmaxDPmax / CminDPmin / RCminDPmin
CmaxDPmax
Cmax
C
RCmaxDPmax
RCmin
Nominal
RCmax
RCminDPmin
Cmin
CminDPmin
R
July 17, 2015
13
Extraction and STA Flow Recommendations
 Extraction Flow: Use colorless flow with DPT corners – no change
needed to existing CAD flow
 STA Flow: No change needed to existing CAD flow – small impact on
setup time due to DPT PEX corners, almost no impact on hold time
Recommend Option 1a
for most designs
July 17, 2015
14
Q&A
July 17, 2015
15
DPT Extraction of Simple Structures
 Using an interdigitized simple dense interconnect structure:
 Varying #of lines @min pitch from 2-20 lines
………
 Mask shift flow is our golden reference of accuracy
 Gives exactly same results as Manual layout shift
 Shifts “E2” mask in the GDS layout to the right/left relative to “E1” mask
 DPT corner flow: uses DPmax and DPmin extraction corners
 Compare DPT corner flow to the mask shift flow to analyze extraction accuracy
July 17, 2015
16
DPT Extraction of Simple Structures: Results
 Observations:
 For #of interconnects >4, DPmax and DPmin provide good bounds
 DPmin corner bound not as tight on symmetric dense structures
due to corner cases
 Case of 2 lines: Usually does not exist in real designs due to metal fill
 Case of 4 lines: Still one edge is direction sensitive but well bounded by DPmax & DPmin
Y-axis is
NOT %cap
July 17, 2015
DPmax and DPmin
are good bounds
17