Description
This study details the design of a filter and an impedance-matching network for a long-range wireless System on Chip (SoC) by following a methodology based on load-pull analysis results from a well-known supplier to reduce losses and obtain maximum RF performance. The resulting design parameters are used to get optimizable responses to reduce the transceiver RF network's reflection coefficient and loss imbalance. These responses were compared with classical optimization methods applying general gradient-based algorithms and the available downhill simplex Nelder-Mead. The structure of this essay begins by presenting the basic concepts applied at the foundation of this case study such as impedance matching, gradient-based optimization procedures, and the minimax formulation for circuit optimization. Following the definition of the project, specifications to work on the 868MHz Industrial, Scientific, and Medical radio band (ISM band) and the applied guidelines of the long-range SoC for impedance matching of the Power Amplifier (PA transmitter) and Low Noise Amplifier (LNA receiver) paths based on the optimal impedance data. This section presents a series of simulation results based on the reflection coefficient to demonstrate the effects on the impedance-matching network due to the addition of filters and transmission lines. These results include further specifications in the design to mitigate the effects of unwanted frequencies based on reflection coefficient responses. To continue with the formulation of an objective function that will serve to apply the classical optimization followed by the presentation of optimization results. It is intended to implement classical optimization to overcome the disturbances of the impedance matching caused by the addition of mandatory filters to the PA and the balun circuit of LNA paths, nonetheless, it is unavoidable to tune the lumped component values in the actual PCB, by employing laboratory equipment such as a spectrum analyzer to confirm output power. This timeconsuming task could be eased by creating fine models using advanced circuit simulators that typically include a good enough set of algorithms for optimization. However, it was decided for this project to implement basic algorithms of classical optimization methods since the stand-alone optimization algorithms make it possible to customize cost functions from different simulators’ analysis responses.
