Research On The Molten Silicon Liquid Level Detection System Based On FPGA

Silicon single crystal is a key raw material in the field of integrated circuit manufacturing and a pillar of the modem information industry.With the rapid development of large-scale integrated circuit technology,the demand for single crystal silicon is increasing while the quality requirements are higher.The Czochralski single crystal furnace is the key technical equipment for growing single silicon crystal.During the growth of single crystal silicon,the change of molten silicon liquid level will cause defects inside the crystal.In order to ensure the quality of single crystal silicon,the molten silicon liquid level must be kept at the set,so real-time high-precision detection of the molten silicon level is an important prerequisite for liquid level control system.Due to the high temperature and sealed environment and special mechanical structure of the Czochralski furnace,a set of molten silicon liquid level detection system based on FPGA is designed according to the laser triangulation method.This system uses Cyclone IV series FPGA as the core controller,which consists of clock control module,laser module,video acquisition module,data storage module,data processing module and display module.The clock control module provides clock for the whole system to meet different requirements.First,a laser mounted on one side of the single crystal furnace projects a laser line to the molten silicon surface of the crucible,and the CMOS camera on the other side captures the liquid surface image containing the laser light;Then,the image data output by the image acquisition module is written into the DDR2 memory to complete storage of one frame of image;In order to improve the processing precision of the algorithm,it is necessary to preform grayscale and binarization on the image to obtain a binary image containing only the laser line,and calculate the laser center by the centroid method.The center of the laser line is subj ected to Kalman filter,and the filtered laser position is converted into the liquid level of the single crystal furnace according to the linear relationship between the liquid level and the displacement of the laser line;Finally,the estimated liquid level is displayed on the digital tube,and the liquid surface video is displayed on the LCD screen in real time.In addition to the centroid method,this paper proposes two laser center localization algorithms.One is a laser center localization algorithm based on Bayesian estimation.Bayesian estimation is used to fuse all laser spot coordinates to obtain the optimal value as the laser center.This algorithm solves the problem that laser spot loss causes positioning difficult.The other is a laser center localization algorithm based on the MUSIC algorithm,which converts the laser line centering problem into a spectral estimation problem,and establishes the far-field and near-field models to estimate the tilt angle and offset of the laser line,respectively,and then achieve the purpose of laser strip center positioning.Finally,the hardware system is tested.The experimental results show that the system can detect the molten silicon liquid level of the single crystal furnace in real time,and can display the video on the LCD screen and display the liquid level result on the digital tube to meet the liquid level measurement requirements of the single crystal furnace.The simulation of the proposed laser center localization algorithm based on Bayesian estimation method and MUSIC algorithm is carried out to verify the feasibility of the algorithm.