How to Run SynMatrix-HFSS AI Optimization

Before Running Optimization
  1. Download “SynMatrix APP” from our website under “e-Library”
  2. Synthesize the coupling matrix


Intelligent Tuning

Go to Intelligent Tuning and select the optimization method.


Intelligent Tuning: AI Optimization (Steps)

Step 1: Load the HFSS file and setup the dimension mapping rules.



Step 2: Setup the HFSS simulation parameters and obtain the initial response


Step 3: Select the optimization method and click “Start”


Setting Up The Optimization Parameters

drawing drawing

  • Trial Modification:
    • Physical dimension change values used in initial training to build the preliminary slope data
    • Inappropriate values will result in inaccurate non-linear over-fitting effects and decrease optimization efficiency
  • Max Changes:
    • The maximum physical dimension changes allowed during the auto-run process
  • Convergent Criterion:
    • The threshold to stop the run process.
  • Iteration Pass (AI method) or Simulation Times (space mapping):
    • Iteration Pass: The maximum simulation group for automatic AI optimization runs; each group run includes the frequency, coupling or transmission zeros optimization
    • Simulation Times: The maximum simulation times for space mapping. The initial simulation times is defaulted as training times plus the extra 10 times simulation

For space mapping only

  • Fully Train:
    • SynMatrix will train each parameter in the coupling matrix and build up the deviation matrix
    • Full deviation matrix build up with co-coupling effects consideration;
    • Suitable for complicated case with quasi-linear coupling performance
  • Smart Train:
    • SynMatrix will analyze the coupling matrix and only train the key parameters by considering the coupling interaction effects
    • Intelligently detects the key training elements from user defined topology
    • Suitable for traditional CT & CQ topology type;

AI vs. Space Mapping – Which one to use?

Key to success

  • Accurate matrix extraction
    • Dispersive effect
    • Wideband application
    • Capture range
    • Simulation points
  • Accurate EM simulation
    • Iteration pass Vs. Mesh quantities
    • Narrow BW Vs. Wide BW
    • Sweep type;
  • Appropriate mapping rules
    • Independent variables


Application scenarios

  • Space Mapping:
    • Non-sensitivity or moderate sensitivity zone with quasi-linear response
    • Minimum EM fields interaction
      • Ex: Coaxial cavity with quasi-TEM;
      • Rectangular WG with TE01
      • Narrow BW application for ML
  • AI:
    • More generic: used for initial target searching
    • Human experience incorporated: handle the non-linear response
  • AI + Space Mapping:
    • Apply the AI + Space mapping hybrid method to handle complex cases
  • Limitations:
    • Inaccurate simulation results
    • Inappropriate trial steps
    • Non-independent variables
By accepting our cookies, you consent to the use of cookies in accordance with our cookies policy.