Phase Unwrapping Algorithm Of High-resolution Wafer Visual Measuring System

High-resolution wafer visual measuring system plays an important role insophisticated wafer process design. In this field, common-path phase-shiftinginterference microscopy system is widely used because it has many outstandingadvantages like simple construction and strong resistance against environmental disrupt.However, phase-shifting may introduce a phase truncation when using arc tangent tocalculate. It makes the figure discontinuous to reconstruct its topography. Phaseunwrapping algorithm is used to revise this error and correctly develop its phase value.There are two kinds of phase unwrapping algorithm now. They are path-followingmethod and minimum-norm method. But these traditional methods can only be usefulwhen sample surface is continuous or relatively flat. They will introduce big error ifthere is a lot of noise or the sample surface is discontinuous. So, it’s necessary todevelop a new method for the common-path phase-shifting interference microscopysystem which can overcome those two shortcomings.First of all in this thesis, the differential phase-shifting interference microscopysystem will be set up based on illuminating by partially coherent light source (LED). Bycollimating and filtering, LED’s coherence is improved. Not only can the improvedbeam satisfy the condition of coherent light for measurement, but also the stray lightcoming from scattering can’t interfere. Coherent noise and speckle noise get to be wellrestrained. And the measuring precision of the system is improved.Next, a new phase unwrapping algorithm is developed. The sine and cosine filterwill be used to improve noise restrain and keep the phase truncation. Then the methodof image segmentation and differential integral is used to divide invalid pixels andresidues respectively. Then the phase derivative variance quality map is born. Flood fillmethod is used whose turn is decided by the quality of every point. And the new methodwill combine Goldstein method and quality-guide method which can restrain noise. Itcan also be used in those samples whose surface is discontinuous. Then a user operationinterface will be designed to fulfill visual and instant requirement.Lastly, both simulated image and tested image are calculated by this new method.It will prove directly how LED restrains noise compared with laser. And then the resultwill show the priority and applicability of the new method. Also, system’s parameterwill be given according to the measured data.