Scalable Compact Models for Complex High-Speed Systems
Transcript of Scalable Compact Models for Complex High-Speed Systems
Scalable Compact Models for
Complex High-Speed SystemsComplex High-Speed Systems
Francesco Ferranti, Tom Dhaene
Department of Information Technology (INTEC),
Ghent University – IBBT
WMC: Advanced techniques for electromagnetic-based model generation
Introduction
Scalable Macromodels
OUTLINE
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 2
Numerical examples• Data-driven PMOR
• Model-driven PMOR
Conclusions
Introduction
Scalable Macromodels
OUTLINE
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 3
Numerical examples• Data-driven PMOR
• Model-driven PMOR
Conclusions
Introduction
input output
fluid dynamicselectronics
telecom
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 4
input outputout = f(in)
automotivechemistry biomodeling geology
Introduction
AerodynamicsAutomotive Electronics MetallurgyChemistry
Designvariables
width, temperature,
angle, frequency, ...
Responsevariables
lift, S-parameters,
pressure, stress, ...
Simulation Model
Fluent®, HSPICE®, CST®,
Comsol®, Abaqus®, ...
Costlyp2
Adaptive Modeling
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 5
OptimizationSensitivity
AnalysisPrototyping
CAD/CAM/CAE
Environment
Designvariables
Responsevariables
Cheap
Surrogate Model / Metamodel
Neural network, Kriging, SVM, rational function, spline,...
p3
p1
p3
p1
p2
Configurable infrastructure
Distributed Computing
Introduction
AerodynamicsAutomotive Electronics MetallurgyChemistry
Designvariables
width, temperature,
angle, frequency, ...
Responsevariables
lift, S-parameters,
pressure, stress, ...
Simulation Model
Fluent®, HSPICE®, CST®,
Comsol®, Abaqus®, ...
Costlyp2
Adaptive Modeling
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 6
OptimizationSensitivity
AnalysisPrototyping
CAD/CAM/CAE
Environment
Designvariables
Responsevariables
Cheap
Surrogate Model / Metamodel
Neural network, Kriging, SVM, rational function, spline,...
p3
p1
p3
p1
p2
Configurable infrastructure
Distributed Computing
Introduction
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 7
Introduction
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 8
Introduction
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 9
Design process
� several decisions
� materials
� geometrical dimensions
shape
Introduction
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 10
� shape
� constraints
� space
� cost
� performance
Simulators � implementation of models
� describe systems behavior
� help designers
Measurements
� post tuning
� verification
� help designers
Design process
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 11
Introduction
Scalable Macromodels
OUTLINE
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 12
Numerical examples• Data-driven PMOR
• Model-driven PMOR
Conclusions
A typical design process requires
� Multiple simulations (measurements)
� different design parameters values (e.g. layout features)
� design space optimization
� design space exploration
Scalable macromodels
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 13
� design space exploration
� sensitivity analysis
A typical design process requires
� Multiple simulations (measurements)
� computationally expensive (time and memory)
Scalable macromodels
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 14
� Can we do better?
� Yes
� By scalable macromodels
discretisation by
simulators
modeling
model
PDE
Scalable macromodels
real world
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 15
Model-driven PMOR
sX= AX+BUY=CX+DU
U(s,g) Y(s,g)
H(s,g) U(s,g) Y(s,g)
scalable macromodel
PDE
Scalable macromodels
discretisation by
simulators
modeling
model
PDE
real world
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 16
Data-driven PMOR
data
simulations
Model-driven PMOR
sX= AX+BUY=CX+DU
U(s,g) Y(s,g)
PDE
H(s,g) U(s,g) Y(s,g)
scalable macromodel
Scalable macromodels
measurements
modeling
model
PDE
real world
discretisation by
simulators
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 17
Data-driven PMOR
data
Model-driven PMOR
sX= AX+BUY=CX+DU
U(s,g) Y(s,g)
H(s,g) U(s,g) Y(s,g)
scalable macromodel
PDE
PMOR concepts
Two design space grids are used in the modeling process
� estimation grid
� validation grid Design space
Scalable macromodels
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 18
N
n
ngg 1
)()(
==
discretisation by
simulators
modeling
model
PDE
Scalable macromodels
real world
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 19
Model-driven PMOR
sX= AX+BUY=CX+DU
U(s,g) Y(s,g)
H(s,g) U(s,g) Y(s,g)
scalable macromodel
PDE
PRIMA
LAGUERRE SVD
FAST TBR
HYBRID TBRModel Order Reduction
Interpolation
Least Squares
SYSTEM MATRICES
TRANSFER FUNCTIONS
Model-driven PMOR
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 20
A B
C D
HAr Br
Cr Dr
Hr
EIGENSPACE PROJECTION
MOR
PROJECTION MATRICES
PRIMA
LAGUERRE SVD
FAST TBR
HYBRID TBRModel Order Reduction
Interpolation
Least Squares
SYSTEM MATRICES
TRANSFER FUNCTIONS
Model-driven PMOR
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 21
EIGENSPACE PROJECTION
PROJECTION MATRICES
A B
C D
HAr Br
Cr Dr
Hr
A B
C D
HAr Br
Cr Dr
Hr
A B
C D
HAr Br
Cr Dr
Hr
A B
C D
HAr Br
Cr Dr
Hr
PRIMA
LAGUERRE SVD
FAST TBR
HYBRID TBRModel Order Reduction
SYSTEM MATRICES
TRANSFER FUNCTIONS
Model-driven PMOR
Interpolation
Least Squares
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 22
EIGENSPACE PROJECTION
PROJECTION MATRICES
model-driven PMOR
accuracy
efficiencyHr(s,g) stability guaranteed
passivity guaranteed
A B
C D
HA B
C D
HA B
C D
HA B
C D
H
Data-driven PMOR
discretisation by
simulators
modeling
model
PDE
real world
measurements
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 23
Data-driven PMOR
data
simulations
Model-driven PMOR
sX= AX+BUY=CX+DU
U(s,g) Y(s,g)
PDE
H(s,g) U(s,g) Y(s,g)
scalable macromodel
VECTFIT
MULTIPORT
PASSIVITY
STABILITY
Rational Fitting
Methods
SYSTEM MATRICES
POLES/RESIDUES
TRANSFER FUNCTIONS
Data-driven PMOR
Interpolation
Least Squares
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 24
data-driven PMOR
Hr(s,g)
NOISY DATA
accuracy
efficiency
stability guaranteed
passivity guaranteed
scattered data
POLES/RESIDUES
Data-driven PMOR
VECTFIT
MULTIPORT
PASSIVITY
STABILITY
Rational Fitting
Methods
SYSTEM MATRICES
POLES/RESIDUES
TRANSFER FUNCTIONS
Interpolation
Least Squares
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 25
Hr(s,g))()())()((),(
1
r gDgBgAIgCgH +−=−
ss
∑=
−
=
P
p pass
1
r)(
)(),(
g
gQgH
scattered data
data-driven PMOR
NOISY DATAPOLES/RESIDUES
Introduction
Scalable Macromodels
OUTLINE
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 26
Numerical examples• Data-driven PMOR
• Model-driven PMOR
Conclusions
Introduction
Scalable Macromodels
OUTLINE
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 27
Numerical examples• Data-driven PMOR
• Model-driven PMOR
Conclusions
Data-driven PMOR
VECTFIT
MULTIPORT
PASSIVITY
STABILITY
Rational Fitting
Methods
SYSTEM MATRICES
POLES/RESIDUES
TRANSFER FUNCTIONS
Interpolation
Least Squares
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 28
Hr(s,g))()())()((),(
1
r gDgBgAIgCgH +−=−
ss
∑=
−
=
P
p pass
1
r)(
)(),(
g
gQgH
scattered data
data-driven PMOR
NOISY DATAPOLES/RESIDUES
3D example: Interconnection structure
� Mobile phone application
� 10 lines (20 ports)
� etched ground plane
Data-driven PMOR example
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 29
� CPU time model estimation
� CPU time model validation Design space
Data-driven PMOR example
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 30
Poles
Data-driven PMOR example
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 31
Output
Data-driven PMOR example
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 32
Output
Data-driven PMOR example
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 33
� Design optimization
� minimum S, minimum α
� |S13(s,S,α)| <= -22 dB over [0 - 20] GHz
Data-driven PMOR example
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 34
� Design optimization
� minimum S, minimum α
� |S13(s,S,α)| <= -22 dB over [0 - 20] GHz
� initial values
Data-driven PMOR example
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 35
� Design optimization
� minimum S, minimum α
� |S13(s,S,α)| <= -22 dB over [0 - 20] GHz
� initial values
� optimal values
Data-driven PMOR example
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 36
� optimal values
Data-driven PMOR example
� Design optimization
� minimum S, minimum α
� |S13(s,S,α)| <= -22 dB over [0 - 20] GHz
� initial values
� optimal values
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 37
� optimal values
Introduction
Scalable Macromodels
OUTLINE
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 38
Numerical examples• Data-driven PMOR
• Model-driven PMOR
Conclusions
PRIMA
LAGUERRE SVD
FAST TBR
HYBRID TBRModel Order Reduction
SYSTEM MATRICES
TRANSFER FUNCTIONS
Model-driven PMOR
Interpolation
Least Squares
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 39
EIGENSPACE PROJECTION
PROJECTION MATRICES
model-driven PMOR
accuracy
efficiencyHr(s,g) stability guaranteed
passivity guaranteed
A B
C D
HA B
C D
HA B
C D
HA B
C D
H
3D example: Spiral inductor
Model-driven PMOR example
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 40
Original order = 801
Reduced order = 91
3D example: Spiral inductor
Model-driven PMOR example
Ly= 0.46 mm Ly= 0.93 mm
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 41
3D example: Spiral inductor
Model-driven PMOR example
Lx= 0.63 mm
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 42
Introduction
Scalable Macromodels
OUTLINE
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 43
Numerical examples• Data-driven PMOR
• Model-driven PMOR
Conclusions
Conclusions
input output
fluid dynamicselectronics
telecom
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 44
input outputout = f(in)
automotivechemistry biomodeling geology
Conclusions
AerodynamicsAutomotive Electronics MetallurgyChemistry
Designvariables
width, temperature,
angle, frequency, ...
Responsevariables
lift, S-parameters,
pressure, stress, ...
Simulation Model
Fluent®, HSPICE®, CST®,
Comsol®, Abaqus®, ...
Costlyp2
Adaptive Modeling
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 45
OptimizationSensitivity
AnalysisPrototyping
CAD/CAM/CAE
Environment
Designvariables
Responsevariables
Cheap
Surrogate Model / Metamodel
Neural network, Kriging, SVM, rational function, spline,...
p3
p1
p3
p1
p2
Configurable infrastructure
Distributed Computing
Conclusions
AerodynamicsAutomotive Electronics MetallurgyChemistry
Designvariables
width, temperature,
angle, frequency, ...
Responsevariables
lift, S-parameters,
pressure, stress, ...
Simulation Model
Fluent®, HSPICE®, CST®,
Comsol®, Abaqus®, ...
Costlyp2
Adaptive Modeling
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 46
OptimizationSensitivity
AnalysisPrototyping
CAD/CAM/CAE
Environment
Designvariables
Responsevariables
Cheap
Surrogate Model / Metamodel
Neural network, Kriging, SVM, rational function, spline,...
p3
p1
p3
p1
p2
Configurable infrastructure
Distributed Computing
Scalable macromodels
Multiple design variables
Compact models
Conclusions
H(s,g) U(s,g) Y(s,g)
scalable macromodel
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 47
Compact models
Efficient design activities (excellent speed-ups)
� Multiple simulations (measurements)
� Design space optimization, exploration, sensitivity analysis
Scalable macromodels
Time-domain simulations
• Non-linear drivers and receivers
Stochastic modeling
Conclusions
H(s,g) U(s,g) Y(s,g)
scalable macromodel
WMC: Advanced techniques for electromagnetic-based model generation IMS2012, Montreal, June 17-22, 2012 48
Stochastic modeling
• impact of manufacturing tolerances
Models from measurements
• noise to handle
Applications in different domains