JP2008008713A - Comparison method of measuring data and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To compare accurately measured data of two measuring groups. <P>SOLUTION: A reliability test is performed relative to one-lot image sensors 11. Each electric characteristic of Ref lot wherein a test is not performed and tested Try lot is measured. A value showing high-order 5% of Ref measured data Xr of the Ref lot is defined as UpIR, and a value showing low-order 5% is defined as LoIR. A value showing high-order 5% of Try measured data Xt of the Try lot is defined as UpIT, and a value showing low-order 5% is defined as LoIT. The number (Zup) of Xt satisfying inequalities: UpIR<Xt≤UpIT and the number (Zlo) of Xt satisfying inequalities: LoIT≤Xt<LoIR are determined. Each ratio of Zup and Zlo to the number of all the Try measured data Xt is determined. After removing abnormal data deviated greatly from a mean value, the direction and the degree of deviation of an Xt distribution from an Xr distribution can be confirmed numerically. Hereby, comparison between both measured data can be performed accurately. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a measurement data comparison method and program for comparing measurement data of two measurement groups.

  Portable electronic devices, particularly digital cameras, are often used not only indoors but also outdoors. For this reason, semiconductor elements such as an image sensor built in a digital camera are susceptible to environmental changes such as changes in temperature and humidity. Therefore, the semiconductor element is required to have high reliability against such environmental changes, and a reliability test is performed on the semiconductor element.

  As the above-described reliability test, there are an environmental test in which a semiconductor element is placed in a specific environmental condition, an energization test in which an electrical load is applied under a specific environmental condition, and the like. And the measurement about various measurement items, such as a voltage measurement and a leakage current measurement, is performed with respect to the semiconductor element by which the reliability test was performed. The same measurement is performed for a semiconductor element that is not subjected to a reliability test. Next, the measurement data of the two measurement groups obtained for each measurement item are compared based on the difference between the average value and the variance, or whether or not they are within a predetermined standard value. As a result, it is determined whether or not the reliability characteristic value of the semiconductor element meets a prescribed requirement.

In addition to investigating the average value and the difference in variance, there is also a well-known method of making both measured data bitmaps and comparing both measured data based on whether the bitmap data match. (See Patent Document 1).
Japanese Patent Laid-Open No. 08-247782 (see FIG. 2)

  However, when comparing the measurement data of two measurement groups based on the average value or variance, if there are many abnormal data in the measurement data that deviate more than the average value, such abnormal data Will strongly affect the mean and variance values. As a result, it is not possible to accurately compare both measurement data. In addition, when the measurement data is converted into bitmaps and compared as described in Patent Document 1, if there are many abnormal data in the measurement data, the bitmap data is strongly influenced by the abnormal data. . As a result, it is not possible to accurately compare both measurement data.

  The present invention is intended to solve the above-described problems, and an object of the present invention is to provide a measurement data comparison method and program that can accurately compare measurement data of two measurement groups regardless of the presence or absence of abnormal data. And

  The present invention provides measurement data for comparing first measurement data included in one of two measurement groups measured in the same measurement environment for the same measurement item and second measurement data included in the other of the measurement groups. In the comparison method, the N (100> N> 50) percentile value of the first measurement data is obtained as the first upper limit reference value, and the (100-N) percentile value of the first measurement data is determined as the first lower limit reference value. And a second step of obtaining an Nth percentile value of the second measurement data as a second upper limit reference value and obtaining a (100-N) percentile value of the second measurement data as a second lower limit reference value. And counting the number of the second measurement data that is less than or equal to the second upper limit reference value and greater than the first upper limit reference value as the number out of the upper limit. A third step, a fourth step of counting the number of the second measurement data that is greater than or equal to the second lower limit reference value and smaller than the first lower limit reference value as a number out of the lower limit, and all of the second measurement data And a fifth step of displaying a ratio calculation result obtained by calculating a ratio of the number out of the upper limit with respect to the number of data and a ratio of the number out of the lower limit with respect to the total number of data.

  There are a plurality of measurement items, and it is preferable that the method includes a sixth step of displaying the ratio calculation result of the measurement item whose ratio calculation result exceeds a predetermined value in a graph.

  A measurement data comparison program according to the present invention causes a computer to execute the measurement data comparison method according to claim 1 or 2.

  The measurement data comparison method of the present invention includes a first step of obtaining a first upper limit reference value and a first lower limit reference value of the first measurement data included in one of the two measurement groups, and included in the other of the measurement groups. A second step of obtaining a second upper limit reference value and a second lower limit reference value of the second measurement data, and a number of the second measurement data that is equal to or less than the second upper limit reference value and greater than the first upper limit reference value. A third step of counting as the number out of the upper limit, a fourth step counting the number of the second measurement data that is greater than or equal to the second lower limit reference value and smaller than the first lower limit reference value as the number out of the lower limit, And a fifth step of displaying a ratio calculation result obtained by calculating the ratio of the number out of the upper limit and the number out of the lower limit with respect to the total number of data of the second measurement data. On excluding the normal data, it is possible to indicate whether the distribution of the second measurement data to the distribution of the first measurement data is how much deviation in any direction numerically. Thereby, even if there are a lot of abnormal data in the measurement data, both measurement data can be compared more accurately than in the case where the average value or the variance is used for comparison as in the conventional case.

  In addition, since the measurement data comparison program of the present invention causes the computer to execute the measurement data comparison method according to claim 1 or 2, it is possible to accurately compare both measurement data in the same manner.

  Hereinafter, a reliability test is performed on the image sensors 11 arranged in a substantially grid pattern on a silicon wafer (hereinafter referred to as a wafer) 10 as shown in FIG. A method for comparing the electrical characteristics of the image sensor 11 and the electrical characteristics of the image sensor 11 not subjected to the reliability test will be described. Although not shown, each image sensor 11 provided on the wafer 10 has a light receiving surface that receives subject light, and this light receiving surface has a large number of photoelectric sensors that convert the subject light into electrical imaging signals. Conversion elements (not shown) are regularly arranged. In FIG. 1, ten image sensors 11 are provided on the wafer 10, but the number of image sensors 11 provided on one wafer 10 is not limited thereto.

  As described above, the reliability test performed on the image sensor 11 includes an environmental test in which the image sensor 11 (wafer 10) is placed in a specific environmental condition, and an electrical test performed on the image sensor 11 in a specific environmental condition. There are various tests such as an energization test that applies a dynamic load. In this embodiment, the electrical characteristics of the image sensor 11 for one lot for which the reliability test is not performed and the image sensor 11 for one lot for which the reliability test is performed are measured, and the obtained measurement is performed. Compare the data. In the present embodiment, for example, the image sensor 11 for one day production is defined as one lot.

  Specific measurement items of the electrical characteristics include the minimum value of the imaging signal output from the image sensor 11 (imaging signal minimum value), the maximum value of the imaging signal (imaging signal maximum value), the leakage current, the standby current, and the like. An example. These measurements are performed by the tester device 13 and the prober device 14 (see FIG. 2). The tester device 13 is a measuring device that measures the electrical characteristics of a semiconductor element such as the image sensor 11 via a prober device 14 described later.

  As shown in FIG. 2, the prober device 14 includes a probe card 20, a test head 21 to which the probe card 20 is mounted, a wafer chuck 24 that holds the wafer 10, and a moving mechanism 25 that moves the wafer chuck 24. I have. The prober device 14 is electrically connected to the tester device 13, and inputs an electric signal from the tester device 13 to the image sensor 11 via the probe card 20, and an input signal from the image sensor 11 to the probe card. 20 to return to the tester device 13.

  The wafer 10 (image sensor 11) is set on the upper surface of the wafer chuck 24. The moving mechanism 25 moves the wafer chuck 24 in the X, Y, and Z directions. A large number of inspection probe needles 31 are attached to the probe card 20 on the lower surface side so as to be aligned with positions of input / output pads (not shown) of the image sensor 11 (see FIG. 1). The test head 21 is supported and fixed by a support member 33. Since the probe card 20 and the test head 21 are electrically connected, and the probe needle 31 of the probe card 20 contacts an input / output pad (not shown) of the image sensor 11 and is electrically connected. The tester device 13 and the image sensor 11 are electrically connected via the probe card 20.

  When the wafer 10 (image sensor 11) is held at a predetermined position of the wafer chuck 24, the moving mechanism 25 moves the wafer chuck 24 to a measurement standby position where the image sensor 11 to be inspected is located below the probe card 20. . Next, the moving mechanism 25 moves the wafer chuck 24 to an inspection position where an inspection probe needle 31 of the probe card 20 contacts an input / output pad (not shown) on the image sensor 11. Then, the tester device 13 transmits and receives an electrical signal from the probe needle 31 of the probe card 20 and performs measurement related to the measurement item selected from each measurement item of the electrical characteristics.

  Measurement data measured by the tester device 13 is sent to a personal computer (PC) 35. A measurement data comparison program 35a is installed in the PC 35. When the measurement data comparison program 35a is activated, the PC functions as a data analysis (comparison) device. As the measurement data comparison program 35a, for example, commercially available spreadsheet software can be used. When the measurement data is input from the tester device 13, the PC 35 creates a table in which the measurement data is collected by lot. Table 1 below shows an example of a table summarizing the measurement data of the imaging signal minimum value.

  As shown in Table 1, in this embodiment, a lot for which a reliability test is not performed is a reference lot (hereinafter referred to as a Ref lot), and a lot for which a reliability test is performed is a trilot (hereinafter referred to as a Try lot). It is said. These Ref lot and Try lot correspond to the two measurement groups of the present invention. And the measurement data of both lots are obtained in the same measurement environment. Here, the same measurement environment means that the same test apparatus (tester apparatus 13 and prober apparatus 14) is used, and there is no significant difference in temperature and humidity of the room where the test apparatus is installed. Thus, by performing measurement in the same measurement environment, it is not necessary to consider the influence of the measurement environment when comparing the measurement data of both lots.

  In the “Measurement number” column in Table 1, the order in which measurements were performed is entered. The Ref lot measurement data Xr is input to the “Ref measurement data” column, and the Try measurement data Xt of the Try lot is input to the “Try measurement data” column. Both of these measurement data Xr and Xt correspond to the first and second measurement data of the present invention.

  Table 2 below shows an example of a data analysis table obtained by analyzing the measurement data Xr and Xt of the Ref lot and the Try lot. First, basic statistics such as average value, maximum value, minimum value, N number, and variance are obtained for each lot.

As shown in Table 2, the average value (3.38 (V)) of all Ref measurement data is input in the “average value” column of the Ref lot, and the minimum of all Ref measurement data is input in the “minimum value” column. A value (0.30 (V)) is input, the maximum value (3.75 (V)) of all Ref measurement data Xr is input in the “maximum value” column, and all Refs are input in the “N _r number” column. The number of measurement data (53497) is entered, and the variance of all Ref measurement data (σ ² = 0.019) is entered in the “dispersion” column.

Similarly, the average value (3.47 (V)) of all Try measurement data is input to the “average value” column of the Try lot, and the minimum value (0. 31 (V)) is input, the maximum value (4.89 (V)) of all Try measurement data is input in the “maximum value” field, and the number of data of all Try measurement data is input in the “ _Nt number” field. (59859) is entered, and the variance (0.026) of all Try measurement data is entered in the “dispersion” column.

  Conventionally, both measurement data were compared based on whether there was a difference in the average value or variance of both lots (both measurement data), or how much difference there was. There are cases where there are a large number of pieces of measurement data (hereinafter referred to as abnormal data) that deviate greatly from the average value. For example, the minimum value of the Ref measurement data is 0.30 (V), which is larger than the average value (3.38 (V)). Further, the minimum value of the Try measurement data is 0.31 (V), the maximum value is 4.89 (V), which is also largely deviated from the average value (3.47 (V)). Therefore, if there are a large number of such abnormal data in the measurement data, the abnormal data strongly affects the average value and the variance value. As a result, there is a possibility that the measurement data cannot be accurately compared. Therefore, in the present invention, when comparing both measurement data, the comparison is performed in consideration of abnormal data. Hereinafter, the measurement data comparison method of the present invention will be specifically described.

In the present embodiment, in addition to the basic statistics described above, for example, a value indicating the upper 0.5% (99.5 percentile value) of the Ref measurement data is obtained as the Ref upper limit reference value (UpIR), and for example, the lower order of the Ref measurement data. A value indicating 0.5% (0.5 percentile value) is obtained as the Ref lower limit reference value (LoIR). Here, the percentile value, in which the number of the values following Ref measurement data Xr is represented how occupies percentage of the total Ref measurement data number (N _r number). For example, UpIR of the present embodiment (99.5 percentile) is 3.64 (V), the number of Ref measurement data Xr which becomes 3.64 (V) below, 99 _{N r} number (53947). Occupies 5%. Further, a Loir (0.5 percentile value) 3.07 (V), the number of Ref measurement data Xr which becomes 3.07 (V) or less accounts for 0.5% of the _{N r} number (53947) .

  Similarly, a value indicating the upper 0.5% of the Try measurement data (99.5 percentile value) is obtained as the Try upper limit reference value (UpIT), and a value indicating the lower 0.5% of the Try measurement data (0 .5 percentile value) is determined as the Try lower limit reference value (LoIT). In this embodiment, UpIT is 3.77 (V), and LoIT is 3.09 (V).

  As shown in FIG. 3, the distribution diagram of the Ref measurement data Xr and the Try measurement data Xt has a shape close to a normal distribution diagram. The horizontal axis of both distribution diagrams is measurement data (imaging signal minimum value), and the vertical axis is the number of data. In the measurement data comparison method of the present invention, a numerical value indicates in which direction and how much the distribution of the Try measurement data Xt excluding the abnormal data is deviated from the distribution of the Ref measurement data Xr excluding the abnormal data. . That is, the numerical value indicates how much the distribution of the Try measurement data Xt satisfying LoIT ≦ Xt ≦ UpIT deviates in which direction with respect to the distribution of the Ref measurement data Xr satisfying LoIR ≦ Xr ≦ UpIR.

  Specifically, the number of Try measurement data Xt satisfying UpIR <Xt ≦ UpIT is counted as a number outside the upper limit (Zup). In addition, the number of Try measurement data Xt satisfying LoIT ≦ Xt <LoIR is counted as the number out of the lower limit (Zlo).

  As shown in FIG. 4, in this embodiment, UpIR <UpIT. For this reason, Zup is obtained by counting the number of Try measurement data Xt in the area A1 indicated by hatching in the figure, and Zup = 6498 (see FIG. 2). Further, as shown in FIG. 5, in this embodiment, since LoIR> LoIT, Zlo = 0 (see Table 2).

  In this embodiment, UpIR <UpIT, but for example, when UpIR> UpIT as shown in FIG. 6, Zup = 0. In the present embodiment, LoIR <LoIT. For example, as shown in FIG. 7, when LoIR> LoIT, Zlo counts the number of Try measurement data Xt in the area A2 indicated by the oblique lines in the figure. Is required. Although not shown, if UpIR <UpIT and LoIR> LoIT, the number of Try measurement data Xt in area A1 is counted to obtain Zup, and the number of Try measurement data Xt in area A2 is obtained. Zlo may be obtained by counting.

As shown in Table 2, when Zup and Zlo is sought, it obtains all Try measurement data number ratio of Zup and Zlo for _{(N t} number) (ratio calculation results), respectively. The ratio calculation result (Pup) of the number out of the upper limit and the ratio calculation result (Plo) of the number out of the lower limit are obtained by the following formulas (1) and (2), respectively. In this embodiment, Pup = 10.9% and Plo = 0.0%.
・ Pup = Zup ÷ N _t number × 100
・ Plo = Zlo ÷ N _t number × 100

  By obtaining Pup and Plo as described above, the distribution of Try measurement data Xt excluding abnormal data is shifted in the direction in which the minimum value of the imaging signal is increased, compared to the distribution of Ref measurement data excluding abnormal data. I know that. Furthermore, it can be seen that the measurement data Xt shifted in this direction is 10.9% of the total Try measurement data Xt. Thus, by performing a reliability test, it is possible to confirm numerically how much the distribution of the measurement data of the imaging signal minimum value is shifted in which direction.

  When the imaging signal minimum values Pup and Plo are obtained, other measurement items (imaging signal maximum value, leakage current, standby current, etc.) are also measured using the above-described tester device 13 and prober device 14, and similarly. What is necessary is just to obtain Pup and Plo. When a PC is used as the PC 35, the above-mentioned UpIR, LoIR, UpIT, LoIT, Zup, Zlo, Pup, Plo can be easily calculated by using commercially available spreadsheet software as the measurement data comparison program 35a. Can be done.

  When Pup and Plo are obtained for all the measurement items, as shown in FIG. 7, the value of Pup is displayed as a bar graph for the measurement item in which Pup exceeds 5%, for example. Further, as shown in FIG. 8, the value of Plo is similarly displayed as a bar graph for measurement items where Plo exceeds 5%, for example. By referring to the bar graphs of FIGS. 7 and 8, it is possible to easily confirm the measurement items in which the deviation of the distribution of the measurement data becomes large. Such a graph can also be easily created by using commercially available spreadsheet software as the measurement data comparison program 35a.

  Next, a flow when comparing the measurement data of the Ref lot and the measurement data of the Try lot will be described using the flowchart of FIG. When the measurement item to be compared is determined to be, for example, “minimum imaging signal value”, the imaging signal minimum value of the Ref lot and the Try lot is measured using the tester device 13 and the prober device 14 (see FIG. 2). The Ref measurement data Xr and Try measurement data Xt obtained by this measurement are input to the PC 35 that has been started up the measurement data comparison program 35a. Then, both measurement data Xr and Xt input to the PC 35 are collected in a table as shown in Table 1 above.

  Next, after obtaining the above basic statistics (see Table 2), the Ref upper limit reference value (UpIR), the Ref lower limit reference value (LoIR), the Try upper limit reference value (UpIT), and the Try lower limit reference value (LoIT) ) And ask. As described above, UpIR and LoIR are the 99.5th percentile value and the 0.5th percentile value of the Ref measurement data, respectively. UpIT and LoIT are the 99.5th percentile value and the 0.5th percentile value of the Try measurement data, respectively.

When UpIR, LoIR, UpIT and LoIT are obtained, the number of Try measurement data Xt satisfying UpIR <Xt ≦ UpIT is counted to obtain Zup. Further, Zlo is obtained by counting the number of Try measurement data Xt satisfying LoIT ≦ Xt <LoIR. Then, determine the ratio of Zup Pup to all Try measurement data number _{(N t} number), and a ratio Plo of Zlo for _{N t} number. Similarly, Pup and Plo are obtained for other measurement items.

  When Pup and Plo are obtained for all the measurement items, the Pup value is displayed as a bar graph for the measurement item for which Pup is 5% or more, and the Plo value is similarly used for the measurement item for which Plo is 5% or more. Is displayed as a bar graph. In addition, when performing a reliability test other than the environmental test, for example, an energization test in which an electrical load is applied under a specific environmental condition, the above-described operation may be repeated.

  As described above, in the present invention, a numerical value representing how much the distribution of the Try measurement data Xt excluding the abnormal data is deviated from the distribution of the Ref measurement data Xr excluding the abnormal data is used. Thus, the measurement data Xr and Xt are compared. Thereby, when there are many abnormal data in the measurement data Xr, Xt, the measurement data Xr, Xt can be compared more accurately than in the case where the average value or variance is used for comparison as in the prior art. .

  In the above embodiment, UpIR and UpIT are values indicating the upper 5% of the measurement data Xr and Xt (95th percentile value), and LoIR and LoIT are values indicating the lower 5% of the measurement data Xr and Xt (5 However, the present invention is not limited to this. The N (100> N> 50) percentile values of the measurement data Xr, Xt may be UpIR and UpIT, and the (100-N) percentile values of the measurement data Xr, Xt may be LoIR and LoIT.

  In the above embodiment, values are displayed as bar graphs for measurement items with Pup and Plo exceeding 5%. However, the present invention is not limited to this, and Pup, The value of Plo may be displayed as a bar graph. Moreover, the values of Pup and Plo may be displayed using a graph other than the bar graph.

  In the above embodiment, the case where the measurement data of the image sensor 11 that is not subjected to the reliability test and the measurement data of the image sensor 11 that is subjected to the reliability test is described as an example. The present invention is not limited to this, and the present invention can be applied when comparing measurement data of two measurement groups measured in the same measurement environment for the same measurement item.

It is a perspective view of a silicon wafer. It is a side view of a probe apparatus. It is the distribution map of the measurement data which measured the imaging signal minimum value of Ref lot and Try lot. It is the enlarged view which expanded and displayed the skirt by the side of UpIR and UpIT of both distribution in FIG. It is the enlarged view which expanded and displayed the skirt by the side of LoIR and LoIT of both distribution in FIG. It is the enlarged view which expanded and displayed the UpIR of the both distributions when UpIR> UpIT and the skirt on the UpIT side. It is the enlarged view which expanded and displayed the LoIR of both distribution when LoIR> LoIT, and the skirt on the LoIT side. It is the bar graph which displayed the value of Pup for every measurement item. It is the bar graph which displayed the value of Plo for every measurement item. It is the flowchart figure which showed the flow at the time of comparing the measurement data of a Ref lot, and the measurement data of a Try lot.

Explanation of symbols

10 silicon wafer 11 image sensor 13 tester device 14 prober device 35 PC
35a Measurement data comparison program

Claims (3)

In a measurement data comparison method for comparing first measurement data included in one of two measurement groups measured in the same measurement environment with respect to the same measurement item, and second measurement data included in the other of the measurement groups,
A first (N) (100>N> 50) percentile value of the first measurement data is obtained as a first upper limit reference value, and a (100-N) percentile value of the first measurement data is obtained as a first lower limit reference value. Steps,
A second step of determining the Nth percentile value of the second measurement data as a second upper limit reference value and determining the (100-N) percentile value of the second measurement data as a second lower limit reference value;
A third step of counting the number of the second measurement data that is less than or equal to the second upper limit reference value and greater than the first upper limit reference value as an off-limit number;
A fourth step of counting the number of the second measurement data that is greater than or equal to the second lower limit reference value and smaller than the first lower limit reference value as an off-lower limit number;
And a fifth step of displaying a ratio calculation result obtained by determining a ratio of the number out of the upper limit with respect to the total number of data of the second measurement data and a ratio of the number out of the lower limit with respect to the total number of data. How to compare measured data. There are a plurality of the measurement items,
The measurement data comparison method according to claim 1, further comprising: a sixth step of displaying the ratio calculation result of the measurement item whose ratio calculation result exceeds a predetermined value in a graph.   A measurement data comparison program that causes a computer to execute the measurement data comparison method according to claim 1.

Images