Browse Prior Art Database

Explicit test case group rule naming for Design Rule Checking program verification tool Disclosure Number: IPCOM000132331D
Original Publication Date: 2005-Dec-07
Included in the Prior Art Database: 2005-Dec-07

Publishing Venue



As semiconductor processes have become more complex and the Design Rule Checking (DRC) programs that verify the geometries on the masks used to manufacture have become commensurately complex, it has become necessary to rigorously test DRC programs against large numbers of standard test cases to verify correct behavior. To this end a shape based testing methodology was invented within IBM that is described in patents US06063132 and US06732338 and which is called generically The Shapediff Methodology. Previous implementations of IBM's shapediff methodology as described in patents US06063132 and US06732338 were limited in several respects. Firstly, the implementation required an adherence to a strict column and row based format for all test layouts used to verify DRC programs. It was difficult to "simply" add real layout data which exposed DRC program issues. It was not possible to mark exact error locations, but rather only indicate groups of shape expecting a fail. There was no mechanism to handle inevitable differences in the DRC tools used. Finally, no external mechanism was available to handle rule naming differences between DRC programmers. Herein we describe methods to handle each of these shortcomings in turn. A method which permits the definition of an explicitly named rule region which does not adhere to the normal shapediff row and column methodology. A method to exactly and implicitly indicate the regions where fails are expected for a particular error. A method to easily handle DRC tool differences while maintaining a single test case library, and finally a method to indicate rule name synonyms external to the program by means of a simple file.

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 33% of the total text.

Page 1 of 13

Explicit test case group rule naming for Design Rule Checking program verification tool

Within the existing shapediff methodology the test case data is set up in rows where each row corresponds to a particular rule name in the DRC program. Nothing crosses the Y axis and groups of failing shapes are on the left side of the y axis at x=0 and sets of shapes you want to pass the rule are on the right side of this axis. The image below is a simplified set of test case data for green shape width, GR500, and green shape space, GR502. Failing cases are on the left and passing cases are on the right.

<Figure 1>

Shapediff groups the test cases by looking for gaps first vertically then horizontally. In the figure


[This page contains 1 picture or other non-text object]

Page 2 of 13

below two of the gaps used to determine rule rows and test case groups are shown. The gap at left is sufficient to break all green shapes into two rule rows. The gap at right is sufficient to break the green shapes in the rule row GR502, and to the right of the y axis at X=0, into two passing test case groups.

<Figure 2>

In this next image the test case groups are circled in yellow.


[This page contains 1 picture or other non-text object]

Page 3 of 13

<Figure 3>

The row of test cases is associated with a particular rule name through the use of a text label appearing somewhere in the row. This text is usually on a level called ALPHA though any name can be used so long as you are consistent.

<Figure 4>

Working through a real example we can look at the creation of rule rows and rule shape groups for a small set of regression layout data. We start with the following regression data.


[This page contains 2 pictures or other non-text objects]

Page 4 of 13

<Figure 5>

The shapediff program looks for vertical breaks or separations in active or design levels. These breaks must be greater than or equal to a predetermined value. Shapediff finds the following rule rows in our data. Shapediff looks for the rule row label at this point.

<Figure 6>


[This page contains 2 pictures or other non-text objects]

Page 5 of 13

With the rows in hand, the shapediff program now looks for horizontal breaks again using the active levels within each row. Shapediff finds the following rule shape groups in the data. The rule row label is propagated to each of these rule shape groups.

<Figure 7>

Shapediff reads in the error shapes from the DRC tool and checks that errors were generated over shape groups to the left of the y axis at x=0. Further that no errors were generated to the right of the y axis at x=0. Basically that the DRC tool is doing what it is supposed to do.

Often it will be the case that some particular problem layout will appear on a customers part. We will work to capture these structures as part of the common regression library. Let us start with the test case below - rule rows already shown.


[This page contains 1 picture or other non-text object]

Page 6 of 13

<Figure 8>