Application of Signal Flow Network on Calibration Capacitive Rotary Encoder

The paper proposes an offline self-calibration scheme about establishing a signal flow network(SNF) to calibrate capacitive rotary encoder. This scheme proposes to simulate the flow of signals and store model parameter information in each node of the network. Unlike traditional optimization algorithms, the intermediate variables in the proposed solution are considered in the optimization pipeline, with the ability to converge fast and accurately. The proposed scheme no longer uses the traditional model linearization method. Instead, the method uses a nonlinear model to establish the network structure, ensures the independence of parameters, and uses an in-depth learning algorithm for improving the convergence speed as well as ability to a global optimal solution. According to the simulation results, the method proposed here is able to get good accuracy of identification, with a relative error of identification below 0.01‰. The validity of the method have also been verified in experiments and the error after the compensation is reduced to 13.02%. The reasons for the inconsistency between simulation and experiment were analyzed. Although the compensation effect is limited, it provides a new method to calibrate capacitive rotary encoder.