Browse Prior Art Database

Method to Distinguish TCM Chip Type

IP.com Disclosure Number: IPCOM000121530D
Original Publication Date: 1991-Sep-01
Included in the Prior Art Database: 2005-Apr-03
Document File: 2 page(s) / 54K

Publishing Venue

IBM

Related People

Nakano, H: AUTHOR

Abstract

This article describes an apparatus and method to distinguish chip type of TCM (Thermal Conduction Module) by employing neural networks as a pattern recognition method. The chip type can be distinguished by processing images of chip solder bumps with unique lighting which enhances the contrast between the substrate and solder bumps, because each type of TCM chip has a unique pattern of solder bumps.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 71% of the total text.

Method to Distinguish TCM Chip Type

      This article describes an apparatus and method to
distinguish chip type of TCM (Thermal Conduction Module) by employing
neural networks as a pattern recognition method. The chip type can be
distinguished by processing images of chip solder bumps with unique
lighting which enhances the contrast between the substrate and solder
bumps, because each type of TCM chip has a unique pattern of solder
bumps.

      As shown in the figure, the image of the chip solder bumps is
captured by a camera under lighting by a circular lamp which enhances
the contrast between the solder bumps and substrate.  After capturing
the image, the image is processed by an image processor including a
neural network emulator.  The network model used has three layers
(input, intermediate and output), and each layer has neurons.  Each
neuron has connections to all neurons which belong to the nearest
layer.  Each connection has its own strength.

      The following is the sequence of this method:
(A) Teaching Prepare all types of TCM chips and capture their images
as input patterns of the neural network.  Each image occupies 4096
neurons of the input layer.  The number of neurons of the output
layer must be equal to the number of chip types. The number of
neurons of the intermediate layer is twice that of the output layer.
The strength of each connection is set randomly (minimum -1 and
maximum 1) at first.  The teaching method is back propagation...