| Summary: | 碩士 === 國立中興大學 === 機械工程學系所 === 101 === The surface topography is important technology in the semiconductor and optical component manufacturing. Now, the Phase Shifting Interferometry is application of reconstructed image and display the true height. The effectively display the true height for Phase Shifting Iterferometry must be take the exact phase. There are two important points, 1. The accuracy of the phase shift 2. The interference wave front is harmonice or not.If we want to use the low cost of piezoelectric materials in place of high precision and cost of PI controller, we have to handle the hysteresis of piezoelectric materials and the environmental vibration or air turbulence of phase-shift mechanism take images so long time. That is why PZT can’t exactly movement for take the exact phase, and affect the accuracy of the phase shift.
We can use the software-based algorithms to estimate the phase shifting approximately, and display the high accuracy of images and the real height efficiency.
Here we use PCA (principal component analysis) and LSM (least squares method), and it is using the concept of principal component analysis of statistical analysis.
First we take off the high correlation to reduce the dimension, and find out the eigenvectors and characteristics values for the optimization of phase distribution. It can be estimated the whole phase in this way, and we can use the least-squares iteration for residual errors.This algorithm doesn’t have any backgrounds, the amplitude and phase shift modulation limitations, so that calculation volume is very low and can be used in a very large image, and can be random phase shift, at least three steps.
We use matlab to write the PCA +LSM algorithm and by simulation and experimental verification. In simulation, we have established the use of oblique plane algorithm to display the true height. In this simulation results, the value of phase shift is around 0.02 in this algorithm and be effectively display the true height. The experiments, we use the USAF 1951 target validated by experiments found only once in accordance with iteration algorithm. That is ineffective, noise interference effect even more, and about 30 nm wave pattern. Then we added to convergence conditions iterative to improvement algorithms to effectively reduce the noise interference and reduce estimation time constraints will reduce the wave pattern around 10 nm, estimated time reduced by a 1.29 seconds to 1.14 seconds. By the white light interferometer test, this specimen height of about 50 nm, after this algorithm is built from the rear, can get about 50 nm height.
This article uses temporal phase estimation, which easily affected by external factors, such as environmental vibration disturbance, PZT hysteresis makes the phase shift position can not be accurately reach, though with the software method is a correction, but there is still noise interference, the future can go spatial phase estimation algorithm research, and time to compare the phase estimation, will be affected by external factors, dry interference noise reduction without compromising image accuracy. In the synthetic wavelength reconstructed image areas, there is still residual noise, algorithms, or can be expected to improve software-based filtering method, the residual noise reduction.
|
|---|
