A THz Waveguide Bandpass Filter Design Using an Artificial Neural Network

Area of Science:

• Electrical Engineering
• Electromagnetics
• Terahertz Technology

Background:

• Waveguide filters are essential components in high-frequency communication systems.
• Designing filters for the terahertz (THz) frequency range presents unique challenges due to component size and precision requirements.

Purpose of the Study:

• To design and fabricate a 300 GHz waveguide bandpass filter with specific performance characteristics.
• To explore the application of artificial neural networks (ANNs) in optimizing filter geometry for desired frequency responses.

Main Methods:

• Utilized coupling matrix synthesis to design a 6-pole Chebyshev filter.
• Employed an artificial neural network (ANN) to determine optimal filter geometries based on frequency response targets.
• Fabricated the filter using computer numeric controlled (CNC) milling for high precision.

Main Results:

• The fabricated filter operates in the 276-310 GHz range.
• Achieved an insertion loss of less than 3 dB.
• Demonstrated a return loss better than 17 dB, indicating efficient signal transmission and minimal reflection.

Conclusions:

• The study successfully demonstrates a high-performance 300 GHz waveguide bandpass filter.
• The integration of ANNs offers an effective approach for optimizing complex microwave and THz component designs.
• The achieved performance metrics are suitable for advanced THz system applications.