Application Of Phase Measurement Profilometry In 3D Solder Paste Inspection

With the further development and application of Surface Mount Technology(SMT),precision electronics manufacturing is developing towards high density and high performance.But the on-line quality inspection of the product has become the weak link of the SMT production line,especially the first process of the production line: the solder paste printing process,and more than 80% of the defects of the patch installed electronic products are caused by the quality defects of the solder paste printing.Based on the research of the latest solder paste inspection(SPI)at home and abroad,the key technologies of SPI based on phase measuring profilometry(PMP)are studied in this paper.(1)establish a three-dimensional measurement system based on PMP to complete the three-dimensional detection of solder paste.First,the sinusoidal grating projection is performed,and then the phase is obtained by the four-step phase shift algorithm,then the phase difference is obtained by the appropriate phase unwrapping algorithm.Finally,the polynomial fitting algorithm is used to complete the phase-height system calibration.The system can quickly and accurately complete the 3D reconstruction of the solder paste with high detection precision.The average detection error can reach 1 ?m through the verification of three party authentication calibration block.(2)An improved phase difference algorithm is proposed,which determines the detection range through the width of the grating stripe,on the basis of the original phase unwrapping algorithm,the effect of the "drawing line" is removed so that the phase difference can be quickly and accurately obtained.The experiment shows that the algorithm can not only meet the uniqueness and accuracy of the phase unwrapping in the PMP algorithm,but also meet the requirements of the real-time performance of the solder paste detection in the SMT line.(3)The detection speed of the system is improved.The main work has two aspects: first,the system algorithm uses C++ and CUDA hybrid programming,which can greatly improve the speed of the algorithm,espically the calibration algorithm accelerates obviously and accelerates more than 1000 times;second,we study the algorithm of FOV lensallocation,which can detect all the regions to be tested with the least number of shots,and improve the detection speed of the system.