Automatic On-Wafer Probe Positioning System Based on Convolutional Neural Network Model

In this work we present an automatic probe positioning system based on a commercial station environment and employing open loop piezo actuators for probe control. The proposed work utilizes open source convolutional neural networks (CNNs) to identify and track the probe and the landing pad. Benchmarking of the various software modules composing the automatic station is presented in terms of their speed, and absolute position tracking control. Moreover, a procedure to define <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$t$</tex> he “zero-landing” <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$h$</tex> eight isp roposed a nd u sed to evaluate the overall system landing accuracy. The proposed system demonstrates the ability to position the probe on IC aluminum pads with a standard deviation of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$1.31 \mu \mathrm{m}$</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$0.96 \mu \mathrm{m}$</tex> on the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$X$</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$Y$</tex> axes, respectively.