Friday, December 16, 2011

On the Quality of Cleanroom Wipers

In assessing the quality of cleanroom wipers, it is vital to consider the consistency of their cleanliness as an integral property that allows the expected performance measures to be achieved.
Wipers are used in cleanroom environments to control contamination and minimize its effect on critical manufacturing processes. Whether one is manufacturing the next generation of semiconductor chips or the newest engineered vaccine, one must select a cleanroom wiper that is the best fit for the specific critical cleaning application. Various performance attributes are typically considered in selecting a cleanroom wiper. These may include sorption capacity/rate in addition to a series of measures of cleanliness of the wipers. Wiper cleanliness is a critical attribute because wipers carry the risk of leaving behind contamination through the burden that they might intrinsically carry into the cleanroom. A variety of testing protocols exist in order to assess the quality of cleanroom wipers. Specifically, one refers to the IEST Recommended Practices1 for evaluating the particulate, ionic, and non-volatile residue (NVR) burden carried by the wipers. However, these test results are typically reported as single data points, sometimes accompanied by an average value and/or a standard deviation. This is a statistically inadequate method to compare the cleanliness of wipers. We discuss here an alternative approach to assess the true cleanliness of cleanroom wipers.
Statistical Process Control (SPC) is typically used to control the physical, chemical, and contamination characteristics of wipers for each lot manufactured. Usually, SPC data are plotted by sample number as shown in Figure 1.
If one must compare the test results for several wipers, then Figure 2 is a typical result, which can be confusing and not very informative.
There arises a need to find a better way to represent all the available test results that are accumulated over a period of time on a given wiper.
Figure 1
Figure 2
Another method of comparing such large data sets is to convert them into a table of averages and standard deviations as shown in Table 1.
Summarizing the data in this manner assumes that the data exhibits a statistical normal distribution, which is often not the case in real world data sets. Since the data is summarized using just these two measures, much information is lost. Using average and standard deviation constitutes an incomplete representation of large data sets and an inadequate means to compare cleanroom wipers. How is the cleanliness of the wiper currently in hand represented by the average value? What if it were to be a lot dirtier than “average”? How would it compromise the integrity of the manufacturing environment? Once again, one recognizes the need to better represent large sets of wiper test data in a manner that is statistically relevant and unbiased.
Table 1
Singular “typical” values or average values and standard deviations inadequately represent the true quality of a cleanroom wiper. A more statistically unbiased method to evaluate many large sets of wiper testing data is through the use of Consistency Charts. These may also be referred to as “Box and Whisker Charts.”
The components that make up a Consistency Chart are:
  • Median – represents the median or middle value of a ranked data set. (Extreme values do not affect the median value as much as an average could be affected.)
  • Box – represents the range of values in which 50% of the data lie. If the median line is nearer to one end of the box, the data are skewed toward that end. A smaller box indicates that the data points are more similar to each other.
  • Whisker – the line at each end of the box, expresses a range of values in which 25% of the data set lies. A shorter whisker indicates that values within the whisker range are more similar to each other.
  • Outlier – indicates data points that are significantly different compared to the rest of the dataset. A five-point Box and Whisker chart shown in Figure 3 is a pictorial representation of large data sets that allows for straightforward evaluation and comparison of wiper test results with minimal loss of information. It is non parametric, in that no statistical distribution type is assumed.
Figure 3
Figure 4
A detailed statistical description of the construction of consistency charts from large data sets can be found elsewhere2,3 and is beyond the scope of this article.
A minimum of three data points are necessary to construct a Consistency Chart. However, when used to evaluate product quality such as test data on cleanroom wipers, it is advisable to have more than 15 data points to ensure statistically valid representation and comparisons.
Inspecting the chart below, the following can be observed:
  • Wiper 1 has the smallest box and the shortest whiskers.
  • Wiper 2 and Wiper 4 have similar medians.
  • Wiper 3 has the lowest median.
  • Wiper 3 has an outlier value shown by the asterisk above the whisker along with the longest whisker.
  • The median for Wiper 3 is nearer to the lower end of the box indicating many values that are similar and low, that is, 50% of the data values are between 2 and 12; however, the other 50% of the values range between 12 and 58.
  • Wiper 4 has the largest box and largest range in the data.
Summarizing these observations, Wiper 1 is most likely to be the cleanest among the four. In comparing Wipers 1 and 3, the following observation can be made:
  • The data set for Wiper 1 or the entire box and whisker chart lies within the box for Wiper 3.
  • Of the test results for Wiper 3, 25% are lower than those for Wiper 1. More than 25% of the rest of the results for Wiper 3 are higher than those for Wiper 1.
This presents us with an improved method to truly assess the cleanliness of cleanroom wipers than comparing “typical” values or the average values and/or the standard deviation which are by their very nature incomplete representations of the data set. Consistency Charts provide a true picture of the cleanliness of each wiper since they represent all of the data that is available. The true measure of the quality of a cleanroom wiper should be measured by the consistency of its cleanliness test over a period of time.
What truly matters in a critical cleaning operation is that each wiper in a bag, each bag in a lot, and each lot of a given wiper product is delivered to the cleanroom with the highest assurance of the expected cleanliness level. Consistency Charts offer a statistically unbiased approach to evaluate the variability of wipers from within a bag or lot, over time.
Selecting the best cleanroom wiper for a particular application requires an unbiased scientific assessment of the available data for any given wiper. Consistency Charts allow for such a statistically valid assessment of cleanroom wiper quality. The quality of a cleanroom wiper should be evaluated not through a typical or average value, but through the use of Consistency Charts that serve to indicate how frequently that typical value is achieved in practice over a period of time. The critical manufacturing processes in a cleanroom must not be susceptible to the variation in the cleanliness of the wiper that may be evidenced in the test data.
  1. “Evaluating Wiping Materials Used in Cleanroom and Other Controlled Environments,” IEST-RP-CC004.3, Institute for Environmental Sciences and Technology, Rolling Meadows, IL, 2004;
  2. John W. Tukey. “Exploratory Data Analysis.” Addison- Wesley, Reading, MA. 1977.
  3. Robert McGill, John W. Tukey, Wayne A. Larsen (February 1978). “Variations of Box Plots.” The American Statistician 32 (1): 12–16.

No comments: