CMP Modeling and DFM AMC2008 Invited Talk September
CMP Modeling and DFM AMC-2008 Invited Talk September 23, 2008 Andrew B. Kahng, UCSD Kambiz Samadi, UCSD Rasit O. Topaloglu, AMD abk@ucsd. edu http: //vlsicad. ucsd. edu
CMP Process § Post-CMP wafer topography depends on metal density, individual feature widths and spacings § Long-range and short-range phenomena § Design manuals specify acceptable metal density ranges § “Dummy” fills inserted to make layout density more uniform § Else, CMP-related problems… slurry §Contains abrasives and chemicals Overremoval Puddling conditioner §A disk with diamond pyramids pad §Improves wafer removal rate Dishing Erosion Step height
BEOL Contribution to Variation (IBM) Parameter Delay Impact BEOL metal (Metal mistrack, thin/thick wires) -10% → +25% Environmental (Voltage islands, IR drop, temperature) 15 % Device fatigue (NBTI, hot electron effects) 10% Vt and Tox device family tracking (Can have multiple Vt and Tox device families) 5% Model/hardware uncertainty (Per cell type) 5% N/P mistrack (Fast rise/slow fall, fast fall/slow rise) 10% PLL (Jitter, duty cycle, phase error) 10%
Agenda § CMP fill, DFM, and design-awareness § Example questions § Opportunities for design-driven fill § What is still left on the table § Recap
CMP and Design for Manufacturability R, C Parasitics Topography CMP Depth of Focus Design Timing and Power Lithographic Manufacturability • CMP and Fill effects • Cu erosion and dishing change resistance • Fill helps planarity but changes capacitance • Topographic variation translates to focus variation for imaging of subsequent layers process window linewidth variation R, C variation • CMP impacts both IC parametrics and manufacturability
§ Given § A grid § A fill size § Number of fills to be inserted to meet target density § Output § Fill configuration that minimizes intra- and inter-layer coupling Interconnect Coupling (F) The CMP Fill Insertion Problem X% improvement
Current CMP Fill Insertion Approach § Layout density verified in fixed-size “windows” § Primitive fill insertion methods – e. g. : § Intersect array of potential fill shapes with empty space § Adjust sizes and spacings, or iteratively execute a ‘multi-pass’ heuristic, to improve density variation and reduce the number of fill shapes § Handled by either design house or foundry
Optimizers Have Improved (1998 -present) § Global optimization with millions of variables in large linear program – Kahng et al. 1998) § Optimization outcome very well-behaved § “Difficult” image sensor chip
Pre-/Post- Fill Densities Original Density Histogram (DD = 31%) min. Fill Density Histogram (DD = 15%) min. Var Density Histogram (DD = 13%)
Existing CMP Fill Insertion Approach § Layout density checked in fixed-size “windows” § Primitive fill insertion methods – e. g. : § Intersect array of potential fill shapes with empty space § Adjust sizes and spacings, or iteratively execute a ‘multi-pass’ heuristic, to improve density variation and reduce the number of fill shapes § Handled by either design house or foundry § Key issue: fill impact on timing, noise, power § Intralayer coupling: keep-off design rule defines minimum spacing between fill and interconnect § Larger keep-off less performance impact, but worse density control, more variation and performance impact… § Smaller keep-off better density control and less variation, but more capacitance, performance impact… § Conflicting goals !!! § Interlayer coupling: no design rules Key word: “design”
What Do We Want? § Objective for Manufacturability = Minimum Variation subject to upper bound on window density § Objective for Design = Minimum Fill subject to upper bound on window density variation For Manufacturability at 65 nm and below: § Multiple relevant planarization length scales: control density at § § § multiple window sizes N-layer BEOL stack: control density in a multi-layer sense Coupling, etch, OPC etc. : provide “staggered” fill patterns or wirelike (“track”) fill Mechanical stability in low-k: achieve (maximal) via fill Better CMP modeling: achieve smoothness of density Analog and mixed-signal variability: symmetric fill … § … all within a CMP model-driven framework
Example of Symmetric Fill (Analog Regions) Axis of Symmetry Analog Cell
Also Want Design-Driven Fill § Global optimization § CMP model-driven fill synthesis § Must tightly couple CMP model to parasitic extraction and timing analysis engines § Efficiency of design flow is an issue internal CMP model vs. signoff CMP model § Design-driven fill synthesis § Design concerns: timing, signal integrity, power § Concurrent analysis of fill impact on both topography and timing § New optimizations possible § Trade OPC cost for variability ? § Good design practices rewarded by reduced BEOL guardband in design ? § Fix hold time violations by inserting extra fill ? Layout, Design Data, Fill Constraints “Intelligent” Fill Uniform Effective Density +Step Height Objective Internal CMP Model Post-Fill Layout, Reports Signoff CMP Model
Example: Timing-Aware Fill § General guidelines § § Minimize total number of fill features Minimize fill feature size Maximize space between fill features Maximize buffer distance between original and fill features § Sample observations in literature § Motorola [Grobman et al. , 2001]: key parameters are fill feature size and keep-out distance § Samsung [Lee et al. , 2003]: floating fills must be included in chip-level RC extraction and timing analysis to avoid timing errors § MIT MTL [Stine et al. , 1998]: rule-based area fill methodology to minimize added interconnect coupling capacitance § Not a new concept, but only now reaching production design flows
M 2 Timing-Aware Keepout
Critical-Net Flow (Timing-, Power-Aware)
Example Questions (Design Flow) § Is CMP fill impact on dynamic power (CV 2 f) large enough to worry about? § Can CMP fill meaningfully improve timing robustness ? § Shortcut power/ground distribution networks with grounded fill less IR drop ? § Use fill to add extra capacitance to hold time critical paths more robust timing ? (And, additional decoupling cap? ) § What good layout design practices correspond to (can be incented by) reduced RC extraction guardband? § How tightly must CMP modeling be integrated into the design flow ? § Which tool (placer, router, physical verification, …) owns the CMP-related signoffs of performance and manufacturability ?
Example Questions (CMP Modeling) Calibration data for each grid point: • X (um), Y (um) • Density • Cu thickness (A) • Dielectric thickness (A) • Optional: Pre-CMP Cu thickness, trench depth, barrier thickness, etc. Layout, Design Data, Fill Constraints What layout parameters must be comprehended by a CMP model? Test Layouts Signoff CMP Model (or silicon) Topography Predictions (or measurements) Intelligent Fill Uniform Effective Density +Step Height Objective Internal CMP Model Post-Fill Layout, Reports Signoff CMP Model Approximation of Signoff CMP Model How do we achieve a CMP model that is optimizable (fast, simple, accurate, …)? Are CMP processes and models stable enough to drive design flows?
Example Questions (Manufacturing Closure) § § Side view showing thickness variation over regions with dense and sparse layout. Top view showing CD variation when a line is patterned over a region with uneven wafer topography, i. e. , under conditions of varying defocus. How tightly do we need to connect OPC to post-CMP topography simulation ? What fill patterning strategies offer the best variability – mask cost tradeoff ?
Agenda § CMP fill, DFM, and design-awareness § Example questions § Opportunities for design-driven fill § What is still left on the table § Recap
Design- (Timing)-Aware Fill keep-off distance § Preserves performance while addressing density objectives § Shown: avoidance of fill on same/adjacent layers near a critical net § Timing-driven place & route creates natural “victims” for fill insertion when it leaves extra space around a critical net ! § Other issues: OPC, data volume, …
What We Leave on the Table: An Example § More sophisticated pattern synthesis guidelines exist but have not been automated §Place fills away from interconnects. §Example: Place fills to form a hour-glass shape §Minimize number of fills close to interconnects § § Want automation Want to account for circuit timing in fill insertion Want to account for interlayer coupling impact on timing Want to gain back the capacitance increase introduced by timingunaware (traditional) fills § Want power-aware fill for power-critical circuits § Next few slides: an ‘energy model’ heuristic for fill pattern synthesis
Adaptive Region Definition § Region-based instead of window-based fill insertion § Maximum-width empty regions identified between facing interconnects, using scanline algorithm Interconnect Region Grid rectangle Keep-off distance § After stripping out keep-off distances, a grid holding possible fill locations is formed § If orthogonal interconnect segments exist, disable overlapping grid rectangle locations
The Grid Model Utilizing Bonds § In this example, there are 36 rectangles with two fills in the grid shown below Region Grid rectangle Bonds incident to a location Interconnect Keep-off distance Fill Vertical bond Auxiliary frame Horizontal bond § An auxiliary frame is formed holding grid rectangles with bonds in between § Each bond has an adjustable energy § Originally considered physical analogy of electrons filling orbits… § When inserting a fill, bonds incident to a rectangle are summed up to find an energy; we find a minimum energy location to insert a fill
Energy Modeling in a Grid § Modeling of bonds indicates which location should be filled with higher priority § Model is flexible enough to satisfy target guidelines § Adjustable four-parameter model for vertical and horizontal bonds § Although we use linear models, second-order and more complex models can also be used Vertical model: Y Horizontal model: i : enumeration for a row of grid rectangle locations j : enumeration for a column of grid rectangle locations imid : middle row number Energies for vertical bonds jmid : middle column number , , , : fitting parameters X § Z axis gives the bond energy. Energies for horizontal bonds
Experimental Setup and Protocol § Cadence SOC Encounter v 5. 2 used for placement and clock tree synthesis and Nano. Route used for routing § Synopsys Star. RCXT 2006. 06 used for RC extraction § C++ code for proposed fill insertion methodology MFO (Metal Fill Optimizer). § Comparison against best available industry tools : Mentor Calibre, Blaze IF § TSMC 65 nm GPlus library § S 38417, AES, ALU and an industrial (microprocessor) testcase § Compare impact of fill algorithm on timing and power Fill Design Rules from Library Exchange File Sizes for Traditional Fill
Interlayer-Aware Fill Synthesis Flow 1. Place, synthesize clock network and route design 2. Extract SPEF parasitics from DEF 3. Run static timing analysis using SPEF file from Step 2 4. Use Perl scripts to obtain top critical net names 5. Check critical nets on neighboring layers for each net 6. Update energy values for bonds 7. Insert interlayer-aware fills §Add vertical bonds Slack Comparison
Power-Aware Fill § Alter flow to handle interconnect switching power criticality 1. 2. 3. 4. 5. 6. 7. Place, synthesize clock network and route design Extract SPEF parasitics from DEF Compute interconnect switching power using SPEF file from Step 2 Use Perl scripts to obtain top power-critical net names Check critical nets on neighboring layers for each net Update energy values for bonds Insert power-aware fills
Timing Slack Results §Timing slacks shown §Less negative (towards the right) is better §Proposed Metal Fill Optimizer (MFO) outperforms intelligent fill (IF) variations
Post-Fill Topographies and Histograms §Core 1 of industrial testcase Traditional fill §We obtain a histogram with a single peak MFO fill
Agenda § CMP fill, DFM, and design-awareness § Example questions § Opportunities for design-driven fill § What is still left on the table § Recap
Recap: What’s on the Table § Example of a physically-motivated, simple heuristic § Testbed with 65 GP process and fill design rules, leading-edge commercial tools § Automation of fill insertion guidelines and intuitions § Large testcases including an industrial (u. P) testcase § Interlayer layout awareness utilized for first time § Timing-aware and power-aware fill options § Can reduce fill impact on timing § by up to 85% for 30% pattern density § by up to 65% for 60% pattern density § Significant value is left on the table by today’s CMP fill methodologies
Recap: Example Open Questions § Design Flow § Is CMP fill impact on dynamic power (CV 2 f) large enough to worry about? § Can CMP fill meaningfully improve timing robustness ? § Can good (layout) design practices correspond to (can be incented by) reduced RC extraction guardband ? § How tightly must CMP modeling be integrated into the design flow ? § CMP Modeling § Which layout parameters are necessary to feed a CMP model? § How do we achieve a CMP model that is optimizable (fast, simple, accurate, …)? § Are CMP processes and models stable enough to drive design flows? § Manufacturing Handoff § How tightly do we need to connect OPC to post-CMP topography simulation ? ? § What fill patterning strategies offer the best variability – mask cost tradeoff ?
Thank you!
Religious Questions in BEOL DFM § Should CMP fill be owned by the routing / timing closure tool or by the DRC § § § / PG tool? § Answer: proper fill is best achieved today post-layout by a tool that maintains the signoff Must fill be “timing-driven”, or is “timing-aware” sufficient? § Answer: “Timing-aware” is likely sufficient through the 45 nm node Are CMP and litho simulations for “more accurate parasitics and signoff” really necessary? § Answer: Probably not. CDs and thickness variations are “selfcompensating” w. r. t. timing. Guardbands are reasonable. There is a big mess with existing calibrations of the RC extraction tool to silicon. If two solutions both meet the spec, are they of equal value? How elaborate must cost functions and layout knobs be for EDA tools to understand via yield / reliability, EM, etc. ? . . .
“Intelligent” Fill Goals for 65 nm and Beyond § True timing- and SI-awareness § § § § Driven by internal engines for incremental extraction, delay calculation, static timing/noise analysis § Open Question: is this done by the router? Or post-layout processing? True multi-layer, multi-window global optimization of effective density smoothness and uniformity § Recall: millions of “tiles” – can we optimize all fill on all layers simultaneously? Analog fill, capacitor fill, via fill Floating, grounded and track fill Standalone, ECO, and ripup-refill use models Supports thickness bias models (CMP predictors) Key technology for managing BEOL variability and enhancing parametric yield
Conclusions: Futures for CMP/Fill in DFM § Goal: Design convergence § Integrate design intent and physical models § CMP simulation + fill pattern synthesis + RCX + timing/SI driven § Performance awareness § Maintain timing and SI closure § “Multi-use” fill: IR drop management, decap creation § Device layer: STI CMP modeling / fill synthesis, etch dummy § Topography awareness § Close the loop back to RCX, fill pattern synthesis, OPC guidance § Intelligent fill pattern synthesis § Minimum variation and smoothness in addition to density bounds § Handle MANY constraints at once: multi-window, multi-layer, etc. § Optional mixing of grounded and floating fill § Mask data volume control (e. g. , shot-size aware, compressible fill)
- Slides: 37