Browse Prior Art Database

Ovation Pattern Recognition System

IP.com Disclosure Number: IPCOM000049010D
Original Publication Date: 1982-Apr-01
Included in the Prior Art Database: 2005-Feb-09
Document File: 4 page(s) / 117K

Publishing Venue

IBM

Related People

Schmidt, S: AUTHOR

Abstract

The Ovation tool scans various MLC (multilayer ceramic) greensheet layers to guarantee accurate screening of conductor designs. Since there existed no capability to verify module post processor data prior to hardware testing, a set of algorithms and corresponding computer program (Ovation Pattern Recognition - OPAR) was developed as part of the OVATION checkmate system. OPAR was set up to recognize patterns formed in the individual layers so that they could be checked against the original input design.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 39% of the total text.

Page 1 of 4

Ovation Pattern Recognition System

The Ovation tool scans various MLC (multilayer ceramic) greensheet layers to guarantee accurate screening of conductor designs. Since there existed no capability to verify module post processor data prior to hardware testing, a set of algorithms and corresponding computer program (Ovation Pattern Recognition - OPAR) was developed as part of the OVATION checkmate system. OPAR was set up to recognize patterns formed in the individual layers so that they could be checked against the original input design.

The input is given as numerical control (NC) data- a flow of bits describing a 2-dimensional rectangular array of shapes. The output of the program is the shape type and its location. If a configuration is encountered which does not match predefined allowable shapes, the location of of the error is reported. The OPAR output is converted to GL1 for use in an independent verification of the originally post processed GL1. General Concepts

Figs. 1 through 4 give the types of shapes that must be recognized, validated, and tabulated. ON shapes correspond to the asterisks (ones), while OFF shapes are formed by blanks (zeros) of the NC data.

Some shapes may be comprised of a large number of bits. The first goal, therefore, is to identify a particular shape by looking at the smallest subset of bits needed to delineate that shape. Next, the shape is extended and verified to determine the location of its end points. Whenever possible, shape extension is accomplished by analyzing groups of bits rather than individual bits. For this reason patterns are tested horizontally in byte (8-bit) groups. If a pattern intersects any other shape, the common bits of the recognized pattern must be retained. Likewise, the non-intersecting bits are blanked out of the pattern matrix.

After a shape is identified, extended and the bits, which are not also parts of other shapes, are blanked, the scan resumes from the byte where the shape was originally identified. Shape Identification

Visualizing the bit stream as a 2-dimensional pattern matrix, the scan for a shape originates in the northwest corner and proceeds row wise (i.e., left to right, top to bottom). Upon encountering the first bit which is part of some shape, it is necessary to efficiently identify the shape type.

Referring to Fig. 5, it is assumed that a "hit" bit (*) has been located. This implies that the bits above and to the west are blank. The initial identification of a shape type is carried out by interrogating the immediate adjacent bits below and to the east.

Fig. 5 indicates the 16 possible ON shape configurations. It should be apparent that many of the identified shapes given in Fig. 5 are part of another shape type. It is, therefore, necessary to retain one or more of the 5 bits under consideration as pending for another pattern. The determination of which bits are pending is accomplished during the extension of the identified shape. The 4 bits checked are...