Researches On The Single-Pole Magnetic Encoder Based On HALL Sensor

With the rapid development of economy and technology,it usually requires a feedback system to realize intelligent control in a modernized control system.As the most common angle position sensor in a feedback system,the angle and position information of the rotor can be output according to the rotation of the rotor,so as to achieve the purpose of controlling the rotor.The commonly used encoder mainly includes the rotating transformer,the photoelectric encoder,magnetic encoder.As more and more areas of the requirements of the diagonal position detection sensor has been improved,the rotating transformer is difficult to meet the requirements of high accuracy and gradually eliminated by the market,though photoelectric encoder has the advantages of high accuracy,but it is difficult to working under hostile environment.However the magnetic encoder as an emerging encoder industry,is valued by more and more researchers.Therefore,how to improve the performance of magnetic encoder will become a crucial problem on the process of encoder technology research in future.In view of this,this paper use HALL sensor and single pair of magnetic poles as sensor components of magnetic encoder and drum element.At the same time it has carried on theoretical analysis and in-depth research on generating process and processing procedure of magnetic encoder signal so that make the magnetic encoder reaches a certain accuracy and resolution,and achieves stable output.On the basis of calculation and simulation of magnetic field signal,the design of the signal source structure model of the magnetic encoder is completed,and the signal source data acquisition of the magnetic encoder is realized by the corresponding hardware processing circuit.For improving the magnetic encoder resolution and output precision,it has analyzed the traditional signal processing algorithm,introduced a kind of magnetic encoder segmentation algorithm based on least square law linear fitting combine with error compensation by neural network.The specific contents as follow:In the research of magnetic encoder signal source model,how to obtain high quality of original voltage signal is one of the researching key.At first,this paper calculated the theoretic magnetic field signal of radial magnetization of the single pole pair cylindrical permanent magnet,and carried on the simulation analysis of magnetic field signal by Ansoft Maxwell software.According to the simulation of the magnetic gap characteristic and magnetic induction distribution results,not only to develop the structure of the signal source and the air gap distance of HALL sensor,and the kind of signal acquisition program compared with the same magnetic encoder,there are advantages like smaller,cost lower,structure and manufacturing process more simple,error by air gap and phase deviation generated less and so on.Then,completing the peripheral hardware circuit design of voltage zero setting,filtering,A/D conversion,serial communications,and other functions based on the processor chip of STM32.Finally,the experimental platform is built up and made acquisition of the magnetic encoder signal by the upper computer software.Linear HALL sensor converted the changing magnetic field signal to changing voltage signal so that achieve the purpose of output,then needed to convert voltage analog signals to digital signals and transmitted digital signals to the processor in order to obtain the required angle output through the corresponding shaft Angle calculation.Based on the analysis of signal processing algorithms such as the arc tangent algorithm,table look-up algorithm,Kalman filtering based algorithm,CORDIC algorithm,etc.,and combined with the regular feature of cosine analog signal generated from sensor in signal source,this paper designed a kind of method that complete the correction of non-linear signal with margin calculation through cosine signal with phase difference of 90° first,then made linear fitting to the corrected function based on least square method to realize the initial resolving treatment of magnetic encoder signal.The experimental result show that the signal subdivision algorithm designed not only to avoid complex calculations effectively reduce the operation load of the processor,it is also capable of more efficient,stable and complete axis angle solver,and ultimately effective in achieving the purpose of the software segment,finally the resolution of the hall magnetic encoder is 16384p/r.Finally,in view of the overlarge output Angle error of linear fitting,this paper compensated angle error with RBF neural network and BP neural network algorithm according to the feature of artificial neural network,respectively.With angle values as input layer data,error value as output layer data,completed the network building,two types of neural network in the same training samples and testing samples to complete the training and prediction of network.The experimental results show that both RBF neural network and BP neural network can effectively applied in angle error compensation system of the magnetic encoder.Finally to a certain extent,the error of the subversion algorithm based on the least square method is improved,which makes the accuracy of the output of the hall magnetic encoder reached 0.09°.