Lecture 5-1: Introduction to Optimetrics and DSO license options ANSYS HFSS for Antenna Design
Introduction •
Optimization •
•
A process of finding a better or more suitable design instance among the possible design variations
Optimetrics is an add-on module module which provides numerous numerous analysis tools •
Parametric
•
Optimization
•
Sensitivity
•
Statistical
•
Tuning
•
Analytic Derivatives
•
Enables Enables ANSYS ANSYS DesignXplor DesignXplorer er Link
•
Allows centralized control of design iterations from one common interface
•
Optimetrics allows the user to: to: •
Automate parametric sweeps
•
Perform real time parameter tuning using Analytic Derivatives
•
Identify performance specifications to optimize
•
Perform sensitivity and statistical analysis on optimized model
•
Link to DesignXplore DesignXplorerr for – Optimization via a Surface Response using Design of Experiments (DOE) – Six Sigma Analysis
Robust Design using Optimetrics Tools Exploration of Design Space
Optimization
Parametric Analysis Manual overview of design space Find good nominal design Precursor to formal optimization Computations easy to parallelize with DSO •
• •
•
HFSS Optimization Minimize cost function to meet goal Several available algorithms •
•
“Optimal” Design (performance, manufacturability, etc.)
Analytic Derivatives Computes derivatives of SYZ parameters w.r.t. design variables Real-time tuning of SYZ parameters Explore the relative impacts of design variables on performance •
•
•
Sensitivity Analysis
DesignXplorer More robust than simple optimization Response Surface Fitting for Entire Design Space Optimization over Response Surface •
•
•
Six Sigma Analysis (DesignXplorer)
Robust Design i s MORE than simpl e optimization: i t is th e abili ty for a user to systematically explore an entire design space so as to arrive at an opti mal design
Using Optimetrics •
Process •
Create parameterized model
•
Define design parameters to vary – Model geometry, material properties, etc.
•
•
Perform analyses
Where can it be used? •
User may apply parameterization at all modeling stages – Geometry (size, shape, orientation, quantity, etc.) – Materials (lossless, complex, anisotropic, etc.) – Boundaries (impedance/conductance boundaries, linked boundary scan angles, symmetry or mode cases, etc.) – Solution setup – Post Processing Quantities (Port magnitude/phase, Deembedding, etc.)
•
Once model is parameterized, optimization can be performed toward an extensive array of cost functions – Circuit parameters (S, Z, or Y-parameters) – Antenna patterns (Directivity, gain, axial ratio, etc.) – Emissions – Derived field quantities (radiated power, etc.)
HFSS Optimetrics: parametric sweep example •
Opening the Project •
The project MicrostripBend.aedt was created and solved earlier in the training. •
If it is already open, Select the menu item File > Save As •
•
•
Click the Save button
Alternatively, browse to the folder containing the file MicrostripBend_optimetrics .aedt and select Open •
Select the menu item HFSS > Analyze All •
•
Filename: MicrostripBend_optimetrics
After analysis is complete save the project by selecting the menu item File > Save
Select the menu item HFSS > Design Properties to overview or edit variables in design •
Verify variables’ nominal values: •
Substrate Thickness: 1mm; TraceWidth: 2mm; TraceThick: 0.2mm
•
Click the OK button to close the window
Parametric Analysis Setup •
Parametric Sweep of Microstrip geometry •
•
Here we will change the value of the defined variables to see how the geometry of microstrip bend influences S-parameters
Create Parametric Sweep •
Select the menu item HFSS > Optimetrics Analysis > Add Parametric ... – Click the Add... button in the Setup Sweep Analysis window •
In the Add/Edit Sweep window: – Select the variable SubstrateThickness – Select Linear Step, Start: 0.5mm,
Stop: 1.5mm, Step: 0.25mm – Click the Add>> button – Select the variable TraceWidth – Select Linear Step, Start: 1mm,
Stop: 3mm, Step: 0.5mm – Click the Add>> button – Select the variable TraceThick – Select Linear Step, Start: 0.16mm,
Stop: 0.24mm, Step: 0.02mm – Click the Add>> button •
Click the OK button
Parametric Analysis Setup •
Number of variations •
We created parametric sweep that includes 5 values for each of 3 variables. The total number of variations is 5^3=125 – One can open Table tab of Setup Sweep Analysis window to look up t he requested variations
•
•
To reduce the total number of variati ons, the variables will be synchronized
Sync the variables •
Return to Sweep Definitions tab of Setup Sweep Analysis window – Using the Ctrl key, select all 3 lines – Click the Sync button •
Open Table tab to look up the requested variations, note that the total number of variations is 5
•
Click the OK button
– Click the OK button
Analysis Configuration •
High Performance Computing Configuration for DSO •
•
Parametric sweeps can be accelerated by solving multiple variat ions in parallel. DSO license option is required to run distribution for Optimetrics setup.
Configuring HPC Settings •
From the Analysis Options Toolbar, select Local configuration
•
Click the Edit Active Analysis Configuration button Analysis Options Toolbar
•
Selected Analysis Configuration Edit Active Analysis Configuration In the Analysis Configuation window, change the following: – Uncheck Use Automatic Settings – Tasks: 2 – Cores: 4 – RAM Limit(%): 80
•
Open Job Distribution tab of Analysis Configuration window
Note: Additional machines can be added to the configuration to further accelerate solutions. Each machine can be used to solve 1 or more task.
Analysis Configuration •
Job Distribution Tab •
•
Make sure that “Optimetrics Variations” line is checked in Enable Distribution Types box
Preview Distribution Setup box •
•
•
Note that for nominal design, HFSSdesign1:Setup1, frequency sweep will be distributed if Frequency option is checked in Distribution Types box From the drop down menu, select HFSSdesign1:ParametricSetup1 Note that Optimetrics variations will be distributed for parametric sweep
Note: Parametric sweep contains 5 variations. One of it is the same as nominal design and has already been solved. The parametric sweep will run 4 variations. To speed up parametric sweep, one can solve 2 or 4 variations in parallel. While considering the number of tasks defined in the Analysis Configuration, consider the number of requested variations and license options available.
Note: If an analysis does not contain a parametric sweep, the solution will distribute frequency points if a frequency sweep has been specified. One can preview the distribution by changing Setup in the Preview box
Distributed Solve Option (DSO) •
Distributed Solve •
Distributed Solve is a productivity enhancement option that accelerates solution times for model variations by leveraging modern computer resources – This option offers a near-linear speed up over conventional simulation sweeps by distributing and simultaneously solving across a network of computers
•
Requires a Distributed Solve license – Common license for Ansys Electronics Desktop designs and Maxwell
DSO distributes and simultaneously solves design variations across networked processors
HFSS Optimetrics: parametric sweep example
•
Analyze Parametric Sweep •
•
In the Project Manager window, select Optimetrics > ParametricSetup1, right click and select Analyze Note that 2 variations run in parallel
Note: One can choose “Save Field And Mesh” option from Options tab of Setup Sweep Analysis. If fields are saved, one can inspect it by selecting HFSS > Results > Apply Solved Variations
Viewing Results •
Create Reports for all solved variations •
Select the menu item HFSS > Results > Create Terminal Solution Data Report> Rectangular Plot – Solution: Setup1: Sweep – Domain: Sweep – Trace Tab: •
Category: Terminal S Parameter
•
Quantity: St(Trace_T1, Trace_T2),
•
Function: dB
– Families Tab •
Make sure all values are included
– Click New Report button – Click Close button •
Save and close the project
Analytic Derivatives in Optimetrics •
Capabilities •
Compute the derivatives of SYZ parameters with respect to project and design variables for HFSS designs
•
Eliminates need to solve multiple variations with small differences and numerical noise – More efficient and more accurate
•
•
Provides real-time tuning of reports to explore effects of small design changes
•
Improves derivative-based optimization methods (SNLP)
Example: MicrostripBend •
3 design variables: thickness of substrate, thickness and width of trace
•
Solve for the derivatives of many variables at once
HFSS Optimetrics: analytic derivatives example •
Opening the Project •
The project MicrostripBend.aedt was created and solved earlier in the training. •
If it is already open, Select the menu item File > Save As •
•
•
Click the Save button
Alternatively, browse to the folder containing the file MicrostripBend_derivatives .aedt and select Open •
Select the menu item HFSS > Analyze All •
•
Filename: MicrostripBend_derivatives
After analysis is complete save the project by selecting the menu item File > Save
Select the menu item HFSS > Design Properties to overview or edit variables in design •
Verify variables’ nominal values: •
Substrate Thickness: 1mm; TraceWidth: 2mm; TraceThick: 0.2mm
•
Click the OK button to close the window
HFSS Optimetrics: analytic derivatives example •
Add Mesh Linked solution setup •
• •
In the Project Manager window, right click on Setup1 and select Add Meshed Linked Solution Setup Note that Setup2 appeared in Project Manager In the Project Manager window, right click on Setup2 and select Properties.. – Popped up window, General Tab: note that Maximum Number of Passes is set to 1 – Advanced Tab: note that Import Mesh option is checked to use the last adaptive mesh from Setup1 – Derivatives Tab: check boxes to use all variables
•
Click the OK button to close the window
HFSS Optimetrics: analytic derivatives example •
Add sweep for Mesh Linked solution setup •
In the Project Manager window, expand Setup1, right click on Sweep and select Copy
•
right click on Setup2 and select Paste
•
Setup2 appeared in Project Manager
•
In the Project Manager window, right click on Setup2 and select Properties.. – Popped up window, General Tab: note that Maximum Number of Passes is set to 1 – Advanced Tab: note that Import Mesh option is checked to use the last adaptive mesh from Setup1 – Derivatives Tab: check boxes to use all variables
•
Click the OK button to close the window
HFSS Optimetrics: analytic derivatives example •
Analyze Setup2 •
•
In the Project Manager window, right click on Setup2 and select Analyze
Create Reports for tuning S-parameters •
Select the menu item HFSS > Results > Create Terminal Solution Data Report> Rectangular Plot – Solution: Setup2: Sweep – Domain: Sweep – Derivative: – Trace Tab: •
Category: Terminal S Parameter
•
Quantity: St(Trace_T1, Trace_T2),
•
Function: dB
– Click New Report button – Continue on the next page
HFSS Optimetrics: analytic derivatives example •
Create Reports for tuning S-parameters •
Change the option for Derivative line – Solution: Setup2: Sweep – Domain: Sweep – Derivative: All – Trace Tab: •
Category: Tune Terminal S Parameter
•
Quantity: TuneSt(Trace_T1, Trace_T2),
•
Function: dB
– Click Add Trace button – Click Close button •
Note that there are 2 traces in the plot that coincide – Edit curves color and style by double clicking on traces in Curve Info box in the plot
HFSS Optimetrics: analytic derivatives example •
Tuning Plot • •
•
Select the menu item HFSS > Results > Tune Reports … Move the scroll bars in the Report Tuning window to predict the performance for various values of defined variables Click the Close button
Note: The predicted response is based off the nominal solution and partial derivative that was computed during the solution process. Analytic Derivatives can be used before any optimization to more quickly narrow the solution space by testing how individual parameters will affect the antenna performance.
Robust Design using Optimetrics Tools Exploration of Design Space
Optimization
Parametric Analysis Manual overview of design space Find good nominal design Precursor to formal optimization Computations easy to parallelize with DSO •
• •
•
HFSS Optimization Minimize cost function to meet goal Several available algorithms •
•
“Optimal” Design (performance, manufacturability, etc.)
Analytic Derivatives Computes derivatives of SYZ parameters w.r.t. design variables Real-time tuning of SYZ parameters Explore the relative impacts of design variables on performance •
•
•
Covered by examples above
Sensitivity Analysis
DesignXplorer More robust than simple optimization Response Surface Fitting for Entire Design Space Optimization over Response Surface •
•
•
Six Sigma Analysis (DesignXplorer)
Refer to HFSS Online Help and Customer Portal to explore other analysis types enabled by Optimetrics
