Issue link: https://iconnect007.uberflip.com/i/253291
February 2014 • SMT Magazine 45 Equation 1 Where n = the number of samples in the data set, xi is the value of a given data point within the data set, and x is the average value of the data set. In a normal population of data points (a symmetrically balanced bell-curve, or histo- gram, of test samples), 99.7% of all test data will fall within ±3σ about the mean value of the data population (see Figure 4). This indicates that for a sample size of 1000, for example, three sam- ples will then fall outside of this ±3σ range and could be considered an outlier or reject (yield loss). Process capability, known as Cp, indicates the spread of data relative to specification range. Cp is calculated based on the following formula: Equation 2 A process is said to be capable or in control when the Cp value is equal to or greater than 1.0. Using Cp, however, assumes that the cen- ter of the data set falls at the midpoint between the USL and LSL. Therefore, it is possible to have a highly repeatable process (low variation, tight distribution of data) but be completely off target (poor accuracy) as in the case of the left group of data points in Figure 2. Cpk, by comparison, considers both the spread of data within the specification range and location of the center of the data relative to a target that is midway between the USL and LSL. Cpk is defined with the following formula: Equation 3 Where x = mean of the sample data. Put another way, Cp is an indicator of how repeatably a system performs, and Cpk indi- cates how accurate and repeatable the system performs relative to USL and LSL. The process is said to meet specified capability when the Cpk value is equal or greater than 1.0. Conceptual differences between Cp and Cpk are shown in Figure 5. It is possible to have Cp > 1 (good re- peatability) but Cpk < 1 (poor accuracy) if the distribution of data is not centered at target. Figure 4: A normally distributed data set with percent of data at varying sigma levels. FEATUrE MaKING SeNSe OF aCCuraCy, rePeaTabILITy aND SPeCIFICaTION continues