Optimizing a Compact Ring Coupler with Neural Network Modeling for Enhanced Performance in Radio Frequency Applications

This paper presents the design and optimization of a compact 900 MHz hybrid ring coupler using lumped reactive components, aimed at achieving harmonic suppression and size reduction for Radio Frequency (RF) applications. Traditional hybrid ring couplers rely on quarter-wavelength transmission lines, resulting in large size device and limited harmonic rejection. To address these challenges, a novel coupler structure was developed that replaces long transmission lines with composite branches, significantly reducing device dimensions while enhancing performance. In the proposed coupler, instead of the six conventional 90-degree lines, six compact networks composed of microstrip lines, three inductors, and one capacitor are used. The inductors have values of L1, L1, L2, and the capacitor has a value of C. These four parameters significantly influence the coupler’s performance; thus, they were selected as inputs for the applied neural network, with the scattering parameters S11, S12, S13, S14, and frequency considered as the five output parameters. The dielectric constant (Ɛᵣ) of the substrate is 2.2, and the substrate material is RT/duroid 5880 with a thickness of 20 mils. By feeding the neural network model with these parameters as inputs, the coupler’s output response was predicted and analyzed, enabling the selection of optimal component values. Optimal responses were obtained with L1 = 10.1nH, L2 = 2.3nH and C = 2.1pF, which allows the coupler to operate effectively at 900 MHz. At this operating frequency, the values are S11 = −32.6dB, S12 = −3.05dB, S13 = −3.03dB, and S14 = −45.9dB, indicating excellent coupler performance.