Automated Synthesis of Non-Foster Impedance Matching Circuits

Active non-Foster circuits can cancel out the reactance of a high-Q load and are a potential solution for wideband impedance matching of electrically-small antennas. Due to the inherent instability and non-linearity of non-Foster circuits, the design of such circuits is time consuming and relies heavily on the designers' experience. To make non-Foster circuits more accessible to the society, we propose an automated tool for synthesizing stable non-Foster circuits in real time. The proposed tool employs three machine learning models for non-Foster circuit synthesis: artificial neural networks, k-nearest neighbors, and Gaussian process. The proposed tool can be used to synthesize (i.e., reverse design) non-Foster circuits with a desired negative capacitance within the −385 pF to −52 pF range. To verify the accuracy of the proposed tool, a non-Foster circuit was synthesized (input: target capacitance = −196 pF) and fabricated using commercial-over-the-shelf components with values closest to the synthesized parameters (output). The simulated capacitance of the synthesized circuit is −195.78 pF (error = 0.11%). The measured capacitance of the prototype is −201.41 pF (error = 2.76%).