Interconnect Parasitic Extraction Speaker Wenjian Yu Tsinghua University

  • Slides: 78
Download presentation
Interconnect Parasitic Extraction Speaker: Wenjian Yu Tsinghua University, Beijing, China Thanks to J. White,

Interconnect Parasitic Extraction Speaker: Wenjian Yu Tsinghua University, Beijing, China Thanks to J. White, A. Nardi, W. Kao, L. T. Pileggi, Zhenhai Zhu

Outline Introduction to parasitic extraction n Resistance extraction n Capacitance extraction n n Inductance

Outline Introduction to parasitic extraction n Resistance extraction n Capacitance extraction n n Inductance and impedance (RLC) extraction 2

Introduction to Parasitic Extraction 3

Introduction to Parasitic Extraction 3

Introduction n Interconnect: conductive path n Ideally: wire only connects functional elements (devices, gates,

Introduction n Interconnect: conductive path n Ideally: wire only connects functional elements (devices, gates, blocks, …) and does not affect design performance n This assumption was approximately true for “large” design, it is unacceptable for DSM designs 4 Slides courtesy A. Nardi, UC Berkeley

Introduction n Real wire has: ¨ Resistance ¨ Capacitance ¨ Inductance n Therefore wiring

Introduction n Real wire has: ¨ Resistance ¨ Capacitance ¨ Inductance n Therefore wiring forms a complex geometry that introduces capacitive, resistive and inductive parasitics. Effects: ¨ Impact on delay, energy consumption, power distribution ¨ Introduction of noise sources, which affects reliability To evaluate the effect of interconnects on design performance we have to model them 5 Slides courtesy A. Nardi, UC Berkeley

Conventional Design Flow Funct. Spec RTL Behav. Simul. Logic Synth. Stat. Wire Model Front-end

Conventional Design Flow Funct. Spec RTL Behav. Simul. Logic Synth. Stat. Wire Model Front-end Gate-level Net. Gate-Lev. Sim. Back-end Floorplanning Parasitic Extrac. Place & Route Layout 6 Slides courtesy A. Nardi, UC Berkeley

From electro-magnetic analysis to circuit simulation Parasitic extraction / Electromagnetic analysis Panel with uniform

From electro-magnetic analysis to circuit simulation Parasitic extraction / Electromagnetic analysis Panel with uniform charge Filament with uniform current Thousands of R, L, C Model order reduction Reduced circuit 7

Challenges for parasitic extraction n Parasitic Extraction ¨ As design get larger, and process

Challenges for parasitic extraction n Parasitic Extraction ¨ As design get larger, and process geometries smaller than 0. 35 m, the impact of wire resistance, capacitance and inductance (aka parasitics) becomes significant ¨ Give rise to a whole set of signal integrity issues n Challenge ¨ Large run time involved (trade-off for different levels of accuracy) ¨ Fast computational methods with desirable accuracy 8

Resistance Extraction 9

Resistance Extraction 9

Outline Introduction to parasitic extraction n Resistance extraction n ¨ Problem formulation ¨ Extraction

Outline Introduction to parasitic extraction n Resistance extraction n ¨ Problem formulation ¨ Extraction techniques ¨ Numerical techniques ¨ Other issues Capacitance extraction n Inductance and RLC extraction n 10

Resistance extraction n Problem formulation ¨A simple structure L W H ¨ Two-terminal structure

Resistance extraction n Problem formulation ¨A simple structure L W H ¨ Two-terminal structure A + i V It’s a single R value ¨ Multi-terminal (port) structure Nx. N R matrix 11 - B i

Resistance extraction n Extraction techniques ¨ Square counting ¨ Analytical approximate formula n For

Resistance extraction n Extraction techniques ¨ Square counting ¨ Analytical approximate formula n For simple corner structure ¨ 2 -D n n n or 3 -D numerical methods For multi-terminal structure; current has irregular distribution Solve the steady current field for i under given bias voltages Set V 1 = 1, others all zero, flowing-out current Repeating it with different settings 12

Resistance extraction n Extraction techniques – numerical method to calculate the flowing-out current ?

Resistance extraction n Extraction techniques – numerical method to calculate the flowing-out current ? Field solver ¨ Field equation and boundary conditions ¨ How Laplace equation inside conductor: divergence E Boundary conditions: : u is known port surface Normal component other surface: is zero; current can not flow out The BVP of Laplace equation becomes solvable 13

Resistance extraction n Numerical methods for resistance extraction ¨ Methods for the BVP of

Resistance extraction n Numerical methods for resistance extraction ¨ Methods for the BVP of elliptical PDE: ¨ Finite difference method n n Derivative -> finite difference: Generate sparse matrix; for ODE and PDE ¨ Finite n n n element method Express solution with local-support basis functions construct equation system with Collocation or Galerkin method Widely used for BVP of ODE and PDE ¨ Boundary n n element method Only discretize the boundary, calculate boundary value Generate dense matrix with fewer unknowns For elliptical PDE 14

Resistance extraction n Where are expensive numerical methods needed ? ¨ Complex n n

Resistance extraction n Where are expensive numerical methods needed ? ¨ Complex n n onchip interconnects: Wire resistivity is not constant Complex 3 D geometry around vias ¨ Substrate coupling resistance in mixed-signal IC 15

Resistance extraction n All these methods calculate DC resistance ¨ Suitable for analysis of

Resistance extraction n All these methods calculate DC resistance ¨ Suitable for analysis of local interconnects, or analysis under lower frequency ¨ High frequency: R of simple geometry estimated with skin depth; R of complex geometry extracted along with L n Reference W. Kao, C-Y. Lo, M. Basel and R. Singh, “Parasitic extraction: Current state of the art and future trends, ” Proceedings of IEEE, vol. 89, pp. 729 -739, 2001. ¨ Xiren Wang, Deyan Liu, Wenjian Yu and Zeyi Wang, "Improved boundary element method for fast 3 -D interconnect resistance extraction, " IEICE Trans. on Electronics, Vol. E 88 -C, No. 2, pp. 232240, Feb. 2005. ¨ 16

Capacitance Extraction 17

Capacitance Extraction 17

Outline Introduction to parasitic extraction n Resistance extraction n Capacitance extraction n ¨ Fundamentals

Outline Introduction to parasitic extraction n Resistance extraction n Capacitance extraction n ¨ Fundamentals and survey ¨ Volume discretization method ¨ Boundary element method ¨ Future issues n Inductance and RLC extraction 18

Capacitance extraction n Problem formulation ¨A parallel-plate capacitor n n Voltage: Q and –Q

Capacitance extraction n Problem formulation ¨A parallel-plate capacitor n n Voltage: Q and –Q are induced on both plates; Q is proportional to V The ratio is defined as C: C=Q/V If the dimension of the plate is large compared with spacing d, ¨ Other familiar capacitors interdigital capacitor 19 coaxial capacitor

Capacitance extraction n Problem formulation ¨ Capacitance exists anywhere ! ¨ Single conductor can

Capacitance extraction n Problem formulation ¨ Capacitance exists anywhere ! ¨ Single conductor can have capacitance n Conductor sphere ¨ N-conductor system, capacitance matrix is defined: Coupling capacitance Total capacitance Electric potential 20

Capacitance extraction n Interconnect capacitance extraction ¨ Only simple structure has analytical formula with

Capacitance extraction n Interconnect capacitance extraction ¨ Only simple structure has analytical formula with good accuracy ¨ Different from resistance, capacitance is a function of not only wire’s own geometry, but its environments ¨ All methods have error except for considering the Shield; whole chip; But electrostatic has locality character window Stable model ¨ Technique classification: n analytical and 2 -D methods C /unit length 2 -D method ignores 3 -D effect, using numerical technique to solve cross section geometry 21

Capacitance extraction n Interconnect capacitance extraction ¨ analytical and 2 -D methods ¨ 2.

Capacitance extraction n Interconnect capacitance extraction ¨ analytical and 2 -D methods ¨ 2. 5 -D methods Error > 10% ¨ Commercial n n tools lateral From “Digital Integrated Circuits”, 2 nd Edition, Copyright 2002 J. Rabaey et al. Task: full-chip, full-path extraction Goal: error 10%, runtime ~ overnight for given process 22

Capacitance extraction n Interconnect capacitance extraction ¨ Commercial tools(pattern-matching): ¨ Geometric parameter extraction n

Capacitance extraction n Interconnect capacitance extraction ¨ Commercial tools(pattern-matching): ¨ Geometric parameter extraction n According to given process, generate geometry patterns and their parameters ¨ Build n n the pattern library Field solver to calculate capacitances of pattern This procedure may cost one week for a given process ¨ Calculation n of C for real case Chop the layout into pieces Pattern-matching Combine pattern capacitances ¨ Error: Cadence - Fire & Ice Synopsys - Star RCXT Mentor - Calibre x. RC pattern mismatch, layout decomposition 23

Capacitance extraction n 3 -D numerical methods ¨ Model actual geometry accurately; highest precision

Capacitance extraction n 3 -D numerical methods ¨ Model actual geometry accurately; highest precision ¨ Shortage: capacity, running time ¨ Current status: widely investigated as research topic; used as library-building tool in industry, or for some special structures deserving high accuracy n Motivation ¨ The only golden value ¨ Increasing important as technology becomes complicated ¨ Algorithms for C extraction can be directly applied to R extraction; even extended to handle L extraction 24

Capacitance extraction n Technology complexity ¨ Dielectric n n n Conformal dielectric Air void

Capacitance extraction n Technology complexity ¨ Dielectric n n n Conformal dielectric Air void Multi-plane dielectric ¨ Metal n n configuration shape and type Bevel line Trapezoid cross-section Floating dummy-fill They are the challenges, even for 3 -D field solver 25

Capacitance extraction n 3 -D numerical methods – general approach ¨ Set voltages on

Capacitance extraction n 3 -D numerical methods – general approach ¨ Set voltages on conductor; solve for Qi Solve the electrostatic field for u, then ¨ Global n method to get the whole matrix Classification Raphael’s RC 3 – Synopsys ¨ Volume discretization: FDM, FEM Q 3 D (Spice. Link) – Ansoft ¨ Boundary integral (element) method Fast. Cap, Hi. Cap, QBEM ¨ Stochastic method Quick. Cap - Magma ¨ Others – semi-analytical approaches 26

27 Slides courtesy J. White, MIT

27 Slides courtesy J. White, MIT

28 Slides courtesy J. White, MIT

28 Slides courtesy J. White, MIT

29 Slides courtesy J. White, MIT

29 Slides courtesy J. White, MIT

Capacitance extraction n Volume methods ¨ What’s the size of simulation domain ? ¨

Capacitance extraction n Volume methods ¨ What’s the size of simulation domain ? ¨ Two kinds of problem: finite domain and infinite domain ¨ Which one is correct ? both in most time 3 -D extraction is not performed directly on a “real” case n In the chopping & combination procedure, both models used n Because of attenuation of electric field, the results from two models can approach to each other Because of its nature, volume methods use finite-domain model n 30

inside alg. of Fast. Cap c Mo. M (method of moment) Method of virtual

inside alg. of Fast. Cap c Mo. M (method of moment) Method of virtual charge Indirect boundary element method 31 Slides courtesy J. White, MIT

Polarized charge 32 Slides courtesy J. White, MIT

Polarized charge 32 Slides courtesy J. White, MIT

33 Slides courtesy J. White, MIT

33 Slides courtesy J. White, MIT

34 Slides courtesy J. White, MIT

34 Slides courtesy J. White, MIT

35 Slides courtesy J. White, MIT

35 Slides courtesy J. White, MIT

Multipole expansion with order l l=0: 36 Slides courtesy J. White, MIT

Multipole expansion with order l l=0: 36 Slides courtesy J. White, MIT

37 Slides courtesy J. White, MIT

37 Slides courtesy J. White, MIT

38 Slides courtesy J. White, MIT

38 Slides courtesy J. White, MIT

Capacitance extraction n 3 -D numerical methods – direct BEM ¨ Field equation and

Capacitance extraction n 3 -D numerical methods – direct BEM ¨ Field equation and boundary conditions Laplace equation in dielectric region: divergence E Boundary conditions: : u is known conductor surface Neumann boundary: Finite-domain model, Nuemann boundary condition Inside alg. of QBEM 39 conductor

Capacitance extraction n 3 -D numerical methods – direct BEM ¨ Green’s Identity ¨

Capacitance extraction n 3 -D numerical methods – direct BEM ¨ Green’s Identity ¨ Free-space ¨ The Scalar field Green’s function as weighting function Laplace equation is transformed into the BIE: s is a collocation point is the fundamental solution of Laplace equation More details: C. A. Brebbia, The Boundary Element Method for Engineers, London: Pentech Press, 1978 40

Direct BEM for Cap. Extraction v Discretize domain boundary • Partition quadrilateral elements with

Direct BEM for Cap. Extraction v Discretize domain boundary • Partition quadrilateral elements with constant interpolation • Non-uniform element partition • Integrals (of kernel 1/r and 1/r 3) in discretized BIE: s • Singular integration • Non-singular integration j t • Dynamic Gauss point selection • Semi-analytical approach improves computational speed and accuracy for near singular integration Wenjian Yu 41

Direct BEM for Cap. Extraction n Write the discretized BIEs as: , (i=1, …,

Direct BEM for Cap. Extraction n Write the discretized BIEs as: , (i=1, …, M) Compatibility equations along the interface • Non-symmetric large-scale matrix A • Use GMRES to solve the equation • Charge on conductor is the sum of q For problem involving multiple regions, matrix A exhibits sparsity! Wenjian Yu 42

Fast algorithms - QMM n Quasi-multiple medium method n In each BIE, all variables

Fast algorithms - QMM n Quasi-multiple medium method n In each BIE, all variables are within same dielectric region; this leads to sparsity when combining equations for multiple regions v 11 u 12 q 21 v 22 u 23 q 32 v 33 3 -dielectric structure s 11 s 12 s 21 s 22 s 23 Population of matrix A s 32 s 33 n n Make fictitious cutting on the normal structure, to enlarge the matrix sparsity in the direct BEM simulation. With iterative equation solver, sparsity brings actual benefit. Wenjian Yu QMM ! 43

Fast algorithms - QMM n QMM-based capacitance extraction n Make QMM cutting Then, the

Fast algorithms - QMM n QMM-based capacitance extraction n Make QMM cutting Then, the new structure with many subregions is solved with the BEM Environment Conductors z Time analysis n while the iteration number dose not change a lot x y Master Conductor n Z: number of non-zeros in the final coefficient matrix A A 3 -D multi-dielectric case within finite domain, applied 3 2 QMM cutting Guaranteed by efficient matrix organization and preconditioned GMRES solver Wenjian Yu 44

Fast. Cap vs. QBEM Contrast Fast. Cap QBEM Formulation Single-layer potential formula Direct boundary

Fast. Cap vs. QBEM Contrast Fast. Cap QBEM Formulation Single-layer potential formula Direct boundary integral equation System matrix Dense for single-region, otherwise sparse Matrix degree N, the number of panels A little larger than N Acceleration Multipole method: less than N 2 QMM method -- maximize the matrix operations in each matrixsparsity: much less than N 2 operations in Other cost n vector product each matrix-vector product Cube partition and multipole expansion are expensive Efficient organizing and storing of sparse matrix make matrix-vector product easy Resemblance: n boundary discretization n stop criterion of 10 -2 in GMRES solution n similar preconditioning n almost the same iteration number Wenjian Yu 45

Sparse blocked matrix each block is dense 46 Slides courtesy J. White, MIT

Sparse blocked matrix each block is dense 46 Slides courtesy J. White, MIT

Quick. Cap Fast. Cap FFTCap QBEM Raphael 47 Slides courtesy J. White, MIT

Quick. Cap Fast. Cap FFTCap QBEM Raphael 47 Slides courtesy J. White, MIT

Capacitance extraction v Future issues o Improve speed and accuracy for complex process o

Capacitance extraction v Future issues o Improve speed and accuracy for complex process o Make field solver suitable for full-chip or full-path extraction task o Parallelizability o Rough surface effect – stochastic integral equation solver o Process variation (multi-corner) # pattern becomes larger o Consider DFM issues (dummy-fill, OPC, etc) 48

Capacitance extraction Reference [1] W. Kao, C-Y. Lo, M. Basel and R. Singh, “Parasitic

Capacitance extraction Reference [1] W. Kao, C-Y. Lo, M. Basel and R. Singh, “Parasitic extraction: Current state of the art and future trends, ” Proceedings of IEEE, vol. 89, pp. 729 -739, 2001. [2] Wenjian Yu and Zeyi Wang, “Capacitance extraction”, in Encyclopedia of RF and Microwave Engineering , K. Chang [Eds. ], John Wiley & Sons Inc. , 2005, pp. 565576. [3] K. Nabors and J. White, Fast. Cap: A multipole accelerated 3 -D capacitance extraction program, IEEE Trans. Computer-Aided Design, 10(11): 1447 -1459, 1991. [4] Y. L. Le Coz and R. B. Iverson, “A stochastic algorithm for high speed capacitance extraction in integrated circuits, ” Solid State Electronics, 35(7): 1005 -1012, 1992. [5] J. R. Phillips and J. White, “A precorrected-FFT method for electrostatic analysis of complicated 3 -D structures, ” IEEE Trans. Computer-Aided Design, 16(10): 10591072, 1997 [6] W. Shi, J. Liu, N. Kakani and T. Yu, A fast hierarchical algorithm for threedimensional capacitance extraction, IEEE Trans. Computer-Aided Design, 21(3): 330 -336, 2002. [7] W. Yu, Z. Wang and J. Gu, Fast capacitance extraction of actual 3 -D VLSI interconnects using quasi-multiple medium accelerated BEM, IEEE Trans. Microwave Theory Tech. , 51(1): 109 -120, 2003. [8] W. Shi and F. Yu, A divide-and-conquer algorithm for 3 -D capacitance extraction, IEEE Trans. Computer-Aided Design, 23(8): 1157 -1163, 2004. 49

Inductance Extraction 50

Inductance Extraction 50

Outline n Basic ¨ Two laws about inductive interaction ¨ Loop inductance n Onchip

Outline n Basic ¨ Two laws about inductive interaction ¨ Loop inductance n Onchip inductance extraction ¨ Partial inductance & PEEC model ¨ Frequency-dependent LR extraction Fast. Henry n Inductance or full-wave extraction with BEM ¨ Maxwell equations & assumptions ¨ Boundary element method 51

Two inductive laws n Ampere’s Law Magnetic field created by: currents in conductor loop,

Two inductive laws n Ampere’s Law Magnetic field created by: currents in conductor loop, time-varying electric fields Curl operator Integral form (derived via Stokes’ Law): For 1 D wire, field direction predicted with right-hand rule Displacement current density AC current flowing through capacitor 52

Two inductive laws n Ampere’s Law (cont’d) Current density Displacement current density For IC,

Two inductive laws n Ampere’s Law (cont’d) Current density Displacement current density For IC, the second term is usually neglected 0. 13 m technology: Transistor switching current: 0. 3 m. A minimal spacing of conductor: 0. 13 m maximal voltage difference: 2 V minimal signal ramp time: 20 ps 0. 13 0. 26 Decouples inductive and capacitive effects in circuit Quasi-static assumption: 53

Two inductive laws n Faraday’s Law Time-varying magnetic field creates induced electric field This

Two inductive laws n Faraday’s Law Time-varying magnetic field creates induced electric field This induced electric field exerts force on charges in b Magnetic flux Eind is a different field than the capacitive electric field Ecap: How about curl? Both electric fields have force on charge ! 54

Two inductive laws n Faraday’s Law (cont’d) Induced voltage along the victim loop: Orientation

Two inductive laws n Faraday’s Law (cont’d) Induced voltage along the victim loop: Orientation of the loop with respect to the Eind determines the amount of induced voltage. Magnetic field effect on the orthogonal loop can be zero ! That’s why the partial inductive couplings between orthogonal wires becomes zero 55

Loop inductance n Three equations All linear relationships Relationship between time-derivative of current and

Loop inductance n Three equations All linear relationships Relationship between time-derivative of current and the induced voltage is linear as well: Mutual inductance; self inductance if a=b There are inductors in IC as components of filter or oscillator circuits; There also inductors not deliberately designed into IC, i. e. parasitic inductance 56

Onchip interconnect inductance n Parasitic inductive effect ¨ An example ¨ Ringing behavior ¨

Onchip interconnect inductance n Parasitic inductive effect ¨ An example ¨ Ringing behavior ¨ 50% delay difference is 17% n 0. 18 width, 1 length Model magnetic interaction ¨ “chicken-and-egg” problem Far end of active line # Possible current loop: O(N 2) Generate L coefficients for all loop pairs is impractical ! O(N 4) “Many of these loop couplings is negligible due to little current; but in general we need to solve for them to make an accurate determination” 57

Onchip interconnect inductance n Partial inductance model ¨ Invented in 1908; introduced to IC

Onchip interconnect inductance n Partial inductance model ¨ Invented in 1908; introduced to IC modeling in 1972 ¨ Definition: magnetic flux created by the current through the virtual loop which victim segment forms with infinity ¨ Loop L is sums of partial L’s of segments forming loop Sij are -1 if exactly one of the currents in segments i and j is flowing opposite to the direction assumed when computing Lij, partial 58

Onchip interconnect inductance n Partial inductance model ¨ Partial inductance is used to represent

Onchip interconnect inductance n Partial inductance model ¨ Partial inductance is used to represent the loop interactions without prior knowledge of actual loops ¨ Contains all information about magnetic coupling n PEEC model ¨ Include partial inductance, capacitance, resistance ¨ Model IC interconnect for circuit simulation ¨ Has sufficient accuracy up to now A two-parallel-line example 59

Onchip inductance extraction n To calculate partial inductance ¨ Formula for two straight segments:

Onchip inductance extraction n To calculate partial inductance ¨ Formula for two straight segments: Assumptions: current evenly distributed ¨ Analytical solution is quite involved even for simple geometry ¨ Numerical solution, such as Gaussian quadrature can be used, but much more time-consuming ¨ How about high-frequency effects ? n n Skin effect; proximity effect Path of least impedance -> least loop L 60 Signal line & its return

Inductance extraction n Related research directions ¨ Design n Limited current loop; inductive effect

Inductance extraction n Related research directions ¨ Design n Limited current loop; inductive effect is reduced, or easy to be analyzed (calculating partial L is costly) Simplify the problem ¨ Use n n solution to cope with inductive effects partial inductance (PEEC model) Consider issues of circuit simulation Inductance brings dense matrix to circuit simulation; both extraction and simulation is expensive, if possible No good locality as C Approaches of matrix sparsification ¨ Inductance n n extraction considering high-frequency Beyond the onchip application MQS, EMQS, full-wave simulation 61 No L explicitly; just Z

Frequency-dependent LR extraction n High frequency consideration ¨ nonuniform current distribution affects R ¨

Frequency-dependent LR extraction n High frequency consideration ¨ nonuniform current distribution affects R ¨ Extract R and partial L together ¨ Capacitive effects analyzed separately (MQS) ¨ Due to the interaction of magnetic field, values of L and R both rely on environments, like capacitance ¨ Problem formulation: Terminal pairs: Impedance extraction: Mutual resistance becomes visible at ultra-high frequency 62

Frequency-dependent LR extraction n Fast. Henry of MIT ¨ Two assumptions: MQS; terminal pairs

Frequency-dependent LR extraction n Fast. Henry of MIT ¨ Two assumptions: MQS; terminal pairs with known current direction ¨ Partitioned into filaments, current distributed evenly Simplified PEEC Solve circuit equation ! 63

Frequency-dependent LR extraction n Fast. Henry of MIT ¨ Nodal analysis: ¨ Avoid forming

Frequency-dependent LR extraction n Fast. Henry of MIT ¨ Nodal analysis: ¨ Avoid forming Z-1: ¨ Mesh-based approach A: incidence matrix Inverse of a dense matrix ! Much larger system M: mesh matrix GMRES can be used to solve this system, with given Vs 64

Frequency-dependent LR extraction n Fast. Henry of MIT Multiple right-hand sides solve: ¨ Multipole

Frequency-dependent LR extraction n Fast. Henry of MIT Multiple right-hand sides solve: ¨ Multipole acceleration; preconditioning techniques ¨ To 30 pins 35 pins ¨ Application: package, wide onchip wires (global P/G, clock) Field solver ! ¨ Shortage: computational speed model inaccuracy; substrate ground plane 65

Problems of Fast. Henry n Segments are 1/3 actual width Lossy substrate discretization ¨

Problems of Fast. Henry n Segments are 1/3 actual width Lossy substrate discretization ¨ Current direction is not clear ¨ Ground plane Huge # of unknowns ! ¨ Multilayer substrate n With frequency increase ¨ Filament n n # increases to capture skin, proximity effects Used only under MQS assumption How to improve ? ¨ Surface integral formulation (BEM) 66 Computational expensive !

Fundamentals of BEM n Inside alg. of Fast. Imp of MIT Maxwell’s equations are

Fundamentals of BEM n Inside alg. of Fast. Imp of MIT Maxwell’s equations are not in dispute ¨ Governing equations Two other equ’s usually known: Faraday’s law Ampere’s law ¨ Constitutive equation for conductor Inside each conductor: Vector identity: 67

Fundamentals of BEM n Equation in each conductor Vector Holmholtz equ. Classification of PDE

Fundamentals of BEM n Equation in each conductor Vector Holmholtz equ. Classification of PDE ? General solution: With 68

Fundamentals of BEM n Equation in the homogeneous medium Hold anywhere A: Magnetic potential

Fundamentals of BEM n Equation in the homogeneous medium Hold anywhere A: Magnetic potential Sum for all conductors 69

Fundamentals of BEM n Boundary conditions Contact is artificially exposed surface NC, C C

Fundamentals of BEM n Boundary conditions Contact is artificially exposed surface NC, C C Hold due to assumption of no charge accumulation No transversal component of current into contact C Here is surface charge density NC Totally 8 state variables: C 70 Full wave simulation

Fundamentals of BEM Fullwave analysis Discretization & unknown setting Equation formulation 71

Fundamentals of BEM Fullwave analysis Discretization & unknown setting Equation formulation 71

Fundamentals of BEM n Full wave ¨ Complete n Maxwell’s equations (no assumption) Electro-Magneto-Quasistatics

Fundamentals of BEM n Full wave ¨ Complete n Maxwell’s equations (no assumption) Electro-Magneto-Quasistatics (EMQS) ¨ Consider RLC ¨ Ignore the displacement current n in medium equ. in conductor equ. Magneto-Quasistatics (MQS) no RL ¨ Ignore the displacement current ¨ Consider Three modes all are widebanded; they behave differently at high frequencies 72

Fast. Imp n Frequency-dependent multiple kernels Algorithms in Fast. Imp ¨ Integral n n

Fast. Imp n Frequency-dependent multiple kernels Algorithms in Fast. Imp ¨ Integral n n calculation Singular, near-singular integral p. FFT algorithm Direct, interpolation, convolution, projection ¨ Scaling n n size u is small Improve condition number ¨ Preconditioning n Preconditioned GMRES 73

Fast. Imp n Experiment results ¨A ring MQS analysis 74

Fast. Imp n Experiment results ¨A ring MQS analysis 74

Fast. Imp Length: 2 cm; separation: 50 um n Experiment results ¨ Shorted Cross-section:

Fast. Imp Length: 2 cm; separation: 50 um n Experiment results ¨ Shorted Cross-section: 50 x 50 um 2 transmission line ¨ EMQS, fullwave, 2 D balanced T-Line ¨ MQS, Fast. Henry 75

Computational Results: Various Practical Examples Slides courtesy Zhenhai Zhu, MIT

Computational Results: Various Practical Examples Slides courtesy Zhenhai Zhu, MIT

Computational Results: Various Practical Examples 10 x 3 Buses Stacked 9 -turn Stacked 8

Computational Results: Various Practical Examples 10 x 3 Buses Stacked 9 -turn Stacked 8 -turn Circular spirals Rect. spirals 9. 5 min Fast. Imp 340 Mb 68 min 642 Mb 54 min 749 Mb 160 min Iterative 19 Gb 750 min 19 Gb 590 min 22 Gb 136 days 19 Gb 100 days 19 Gb 168 days 22 Gb LU Slides courtesy Zhenhai Zhu, MIT

Inductance extraction Reference [1] M. W. Beattie and L. T. Pileggi, “Inductance 101: modeling

Inductance extraction Reference [1] M. W. Beattie and L. T. Pileggi, “Inductance 101: modeling and extraction, ” in Proc. Design Automation Conference, pp. 323 -328, June 2001. [2] M. Kamon, M. J. Tsuk, and J. K. White, “Fasthenry: a multipoleaccelerated 3 -D inductance extraction program, ” IEEE Trans. Microwave Theory Tech. , pp. 1750 - 1758, Sep 1994. [3] Z. Zhu, B. Song, and J. White. Algorithms in Fastimp: a fast and wideband impedance extraction program for complicated 3 -D geometries. IEEE Trans. Computer-Aided Design, 24(7): 981 -998, July 2005. [4] W. Kao, C-Y. Lo, M. Basel and R. Singh, “Parasitic extraction: Current state of the art and future trends, ” Proceedings of IEEE, vol. 89, pp. 729 -739, 2001. [5] http: //www. rle. mit. edu/cpg/research_codes. htm (Fast. Cap, Fast. Henry, Fast. Imp) 78