Method and device for detecting non-unimodal distribution of measurement data

Abstract

The invention discloses a method for detecting the non-unimodal distribution of measurement data, comprising the following steps: collecting measurement data capable of being detected; selecting a minimum data value, a second minimum data value, a median minimum data value, a median maximum data value, a second maximum data value and a maximum data value in the measurement data; making a first regression line for data between the minimum data value and the second minimum data value; making a second regression line for data between the median minimum data value and the median maximum data value; making a third regression line for data between the second maximum data value and the maximum data value; calculating the ratio Delta of a difference value between a maximum value and a median value to a difference value between the maximum value and a minimum value among an intersecting point x1 of the first regression line and an X-axis, an intersecting point x2 of the second regression line and the X-axis and an intersecting point x3 of the third regression line and the X-axis; and comparing the Delta with a detection critical value C alpha so as to confirm whether the measurement data is in non-unimodal distribution or not. The invention also provides a device for detecting the non-unimodal distribution of measurement data. The invention improves the accuracy of control in a counting process.

Description

technical field [0001] The invention relates to a method and a device for detecting non-unimodal distribution of measurement data. Background technique [0002] In the industrial production process, in order to find out the problems in the process in time, eliminate the problems or reduce the losses caused by the problems, it is necessary to carry out quality management control on each process, which includes testing each process. For example, the manufacturing process of semiconductor integrated circuits is a planar manufacturing process, which combines photolithography, etching, deposition, ion implantation and other processes to form a large number of various types of complex devices on the same substrate, and integrate them interconnected for full electronic functionality. Among them, a problem in any step of the process may lead to the failure of the production of the circuit. Therefore, in the prior art, the production results of each step of the process are often det...

AI Technical Summary

Problems solved by technology

However, when actually collecting the measurement data in the production process, there will be a phenomenon that the measurement data is a non-unimodal distribution (non-normal distribution), such as figure 2 The bimodal distribution shown

Measurement data with non-unimodal distribution will produce large errors when applied to analyze defects in the manufacturing process, which will affect the accuracy of statistical process control

[0007] Due to the lack of ready-made or similar statistical methods or means for the measurement data of non-unimodal distribution in the existing technology, it cannot be detected in time, and then large errors will be generated when analyzing defects in the manufacturing process, which will affect the accuracy of statistical process control sex

Images