9D-6 Signal Analysis in Scanning Acoustic Microscopy for Non-Destructive Assessment of Connective Defects in Flip-Chip BGA Devices

Failure analysis in industrial applications often require methods working non-destructively for allowing a variety of tests at a single device. Scanning acoustic microscopy in the frequency range above 100 MHz provides high axial and lateral resolution, a moderate penetration depth and the required non-destructivity. The goal of this work was the development of a method for detecting and evaluating connective defects in densely integrated flip-chip ball grid array (BGA) devices. A major concern was the ability to automatically detect and differentiate the ball-connections from the surrounding underfill and the derivation of a binary classification between void and intact connection. Flip chip ball grid arrays with a 750 mum silicon layer on top of the BGA were investigated using time resolved scanning acoustic microscopy. The microscope used was an Evolution II (SAM TEC, Aalen, Germany) in combination with a 230 MHz transducer. Short acoustic pulses were emitted into the silicon through an 8 mm liquid layer. In receive mode reflected signals were recorded, digitized and stored at the SAM's internal hard drive. The off-line signal analysis was performed using custom-made MATLAB (The Mathworks, Natick, USA) software. The sequentially working analysis characterized echo signals by pulse separation to determine the positions of BGA connectors. Time signals originated at the connector interface were then investigated by wavelet- (WVA) and pulse separation analysis (PSA). Additionally the backscattered amplitude integral (BAI) was estimated. For verification purposes defects were evaluated by X-ray- and scanning electron microscopy (SEM). It was observed that ball connectors containing cracks seen in the SEM images show decreased values of wavelet coefficients (WVC). However, the relative distribution was broader compared to intact connectors. It was found that the separation of pulses originated at the entrance and exit of the ball array corresponded to the condition of the connector. The success rate of the acoustic method in detecting voids was 96.8%, as verified by SEM images. Defects revealed by the acoustic analysis and confirmed by SEM could be detected by X-ray microscopy only in 64% of the analysed cases. The combined analyses enabled a reliable and non destructive detection of defect ball-grid array connectors. The performance of the automatically working acoustical method seemed superior to X-ray microscopy in detecting defect ball connectors.