Graphically Analyzing Yield Loss

Laura Peters, Senior Editor -- Semiconductor International, 10/1/1999

In our last two columns on Integrated Yield Management ( August SI, p. 46. and September SI, p.46), we covered the use of wafer zone-based yield analysis to analyze equipment yield loss. In the third stage of Y_E analysis, the user performs graphical analysis on each unit of the worst-performing process equipment. In this column, we illustrate how the various graphical tools work.

The box plot (Fig. 1) is the most basic of the four graphical analyses shown. It shows the relative yields and spread of data for each equipment unit as compared to the other equipment units at an operation.

The confidence interval graph (Fig. 2) shows the 95% confidence level of the medians shown in the box plot. If the tails of the confidence interval plots overlap each other, the means are not significantly different at the 95% confidence level.

The probability plot (Fig. 3) is the most complex of the four reports in this analysis. It shows the relative yield trend for all the wafers in a lot and comparatively describes differences in the yield distribution from one equipment unit to another.

The trend plot (Fig. 4) shows lot yield over time. When a sudden change in equipment performance occurs, a significant shift change is shown on the trend plot, allowing prompt investigation of the cause.

Fig 1 A box plot compares the yield performance of different places of the same type of process equipment.	Fig 2  Illustration of the 95% confidence level of the medians shown in the box plot (Fig 1). Error bars overlap if the difference between two means is not significant at the 5% level.
Fig 3  This plot comparatively depicts the differences in the distribution of yield for different equipment units.	Fig 4  The time line history of means for each lot. A jump in data indicates a sudden change in equipment performance.