This paper presents optical design, proof-of-principle experiments, and post-signal processing for a wafer microcrack detection system. Cylindrical lens and Solid Immersion Lens (SIL) are used for the optical module. Near-field probe array is also investigated for future implementation. For post-signal processing, Probabilistic Neural Network (PNN) is used in order to identify vibration-induced deviation. Derived Short-Time Discrete Wavelet Transform (STDWT) is developed in order to identify microcrack.