Graphics Script Provides Quick Classification Continued...
Commercial data plotting packages do not use this scheme of a color histogram with the colors keyed to locations of measurements on a semiconductor wafer. A decade ago, the author began using a custom program developed by D. Elsaesser, S. Dudley and J. Sewell who implemented the scheme in a FORTRAN wafer mapping program on a Digital Equipment Corp. VAX/VMS supermini computer. Unfortunately, their program used color and plot routines from a proprietary Software Terminal Interface library that has never been ported from the VAX/VMS format. The author began looking for a personal computer based data analysis and technical graphical program with powerful graphics capabilities and a scripting language that would make it possible to automate the development of these scripts.
The author discovered Origin v 5.0, which is a Windows-based data analysis and technical graphics software package from Microcal Software, Inc., located in Northampton, Massachusetts. This software package provides several crucial advantages that make it ideally suited for this application. Origin utilizes an extremely powerful programming language, called LabTalk that provides access to virtually every function in the program. It provides a wide range of graphing capabilities, allowing the user to adjust virtually any parameter of the graph simply by clicking on it with a mouse. Finally, Origin provides a wide range of statistical capabilities.
The author wrote a LabTalk script to perform the necessary analyses and create color plots. The data is typically stored in an ASCII file, each datum consisting of the X- and Y-coordinates of the measurement and a value for the measurement. Running the script leads first to a choice box for selecting the data file to be plotted. The drive, subdirectory, and file name are selected and double clicked. Very shortly, the script asks for minimum value, maximum value, and granularity. The user enters a general range of values to be plotted and the granularity, typically about a hundredth of the difference between the highest and the lowest values. The script then uses Origin's built-in black and white histogram routine to manipulate and plot the data. This choice process can be refined for two minutes to define the best values. The size of the resulting histogram plot and worksheets is then reduced to make space for display of the color histogram and plot.
Next, the script asks for the number of color bins which must be a multiple of fourteen (the number of colors used). If a non-multiple of fourteen is entered, the script corrects this to the next larger number that is a multiple of fourteen. After asking if a standard (blue-to-red) rainbow or a reverse rainbow (red-to-blue) plot is wanted, the color version of the histogram is calculated and plotted. The script then asks for the sample identification and type of measurement which are printed as a label on the histogram along with the mean, standard deviation, number of values within the chosen range, total number of values input, and the date the plot is made. The script next asks for the size of the area to be plotted. For a plot of 7,000 measurements on a two-inch wafer or 17,000 measurements on a three-inch diameter wafer, a value of "3" works well. For 20,000 measurements on a 100 mm wafer and 16,000 measurements on a 150 mm wafer, a value of "4" is better. The last entry is the wafer diameter in millimeters.