Key-pin-position and Gaussian-L2-distance based rigid registration for accurate and fast location of ball grid array chip

Abstract

The large number and various arrangements of solder balls in a Ball Grid Array (BGA) of an electric chip poses significant challenges on its accurate and quick location to the Printed Circuit Board (PCB). To solve these problems, a key-pin-position and Gaussian L2 distance based rigid registration method was proposed. The key idea was to use the pins only at the corners (key positions) to build up data point which was used to obtain the offset and rotation angle of the BGA relative to the specified place by the Gaussian-L2-distance-based rigid registration method. Gaussian function was used to model the actual solder balls’ positions and the specified positions, the Expectation and Maximization algorithm was used to solve the iteration process of the rigid registration. It was experimentally found that the maximum positioning error in three directions (X, Y and rotation angle) is only 1/12 pixels, which significantly outperforms Samsung’s method with a positioning error of 0.68 pixels and the state-of-art method with 0.7 pixels. The location time for a BGA with more than 200 balls is no more than 200 ms which is less than half of that of a commercial device (Samsung SM482). Experimental results demonstrate the effectiveness of this method in handling BGA chips with varying numbers and arrangements of solder balls.