Adaptive Synthesis Using Hybrid Genetic Algorithm And Particle Swarm Optimization For Reflectionless Filters

In recent years,reflectionless filters have gradually attracted the attention of researchers.As a new type of filter,this filter has the characteristic of maintaining a constant impedance over a frequency range wider than the passband,compared to conventional filters that usually exhibit reactance characteristics in the stopband.Thus,this filter dissipates the signal in the stopband inside the device.Therefore,this filter avoids distortion and intermodulation products caused by reflected signals when connected to a mixer.Meanwhile,the use of circulators is eliminated.In addition,when connected to an amplifier,the instability of the amplifier circuit is avoided by driving a constant load impedance.In addition,the reflectionless filter significantly reduces harmonic signals from nonlinearity.It increases the dynamic range of the system while improving linearity.However,compared to conventional filters,the design of reflectionless filters is more complex,the matching characteristics are affected by a variety of factors,and there is not yet a systematic design method that can be applied to a wide range of response characteristics.In this thesis,an adaptive synthesis methodology based on the hybrid of genetic algorithm and particle swarm optimization(HGAPSO)is proposed.This methodology is for designing reflectionless filters using lumped elements of capacitors,resistors,and inductors.The proposed synthesis starts from a pre-defined topology where each branch of the topology contains a small passive network composed of lumped elements.The proposed HGAPSO is used to trim the branches in the topology and to generate appropriate element values for the desired filter response.The proposed HGAPSO embeds a local search strategy based on random dimensional search and neighborhood search to improve its search capability.In addition,the optimization algorithm introduces classifier-based strategies and probabilistic methods to accelerate convergence and boost iteration.Based on HGAPSO,the synthesis algorithm is able to automatically determine the appropriate topology and component values so that the generated circuit satisfies the given constraints on the response.To accurately predict the actual filter response,this solution introduces a fine-tuning process based on the gradient descent method,in which the EM simulation results of the circuit layout and the parasitic parameter model of the aggregate components are considered.Several filter designs obtained by this synthesis method are presented in this thesis.The test results of the fabricated filters have response characteristics consistent with expectations and maintain good reflectionless performance over a wide range.