JP2937541B2 - Quality level verification method for electronic components - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[Object of the Invention]

[0002]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to the inspection of electronic components such as semiconductor products, and more particularly to a means for verifying the quality level of products.

[0003]

2. Description of the Related Art Semiconductor device manufacturing is broadly divided into a pre-processing step of introducing and diffusing impurities into a wafer to produce circuit components such as devices and resistors, and a post-processing step of assembling a single device. As an inspection step, a die sorter test after the pre-processing step, a characteristic inspection step after the final step, and a storage inspection are performed.

However, at the manufacturing site, various inspections are performed at specific steps in order to improve the yield of products. However, the equipment cost required for the inspection steps cannot be greatly increased, and a reasonable inspection system or inspection is required. At present, there is a need for a method.

Inspection contents of semiconductor products can be roughly classified into two types. The first is that the output characteristic value with respect to the input of the product to be inspected represented by the AC characteristic / DC characteristic inspection shown in FIG. This is a method of discriminating good / defective products depending on whether they are within the standard values between the standard values.

As shown in FIG. 1B, the second method is a method of discriminating a non-defective product from a defective product based on whether an output pattern of a product to be inspected represented by a function inspection matches a target pattern.

[0007]

As described above, a semiconductor product is usually inspected by a combination of two types of measurement contents (measurement items). For example, when one measurement item is determined to be defective. The process of selecting the products to be inspected as defectives is performed on a lot basis, and all the products are inspected for each product type.
On the other hand, a standard yield is determined for each product type depending on a process capability of manufacturing a semiconductor product, and is referred to as a standard yield.

[0008] As a method of operating quality assurance in a factory,
In addition to inspecting each product and sorting out good and defective products,
The quality assurance inspection will be strengthened if the yield of each lot is less than a certain level compared to the standard yield.

In addition, the inadequate point in the manufacturing process which influences the characteristics is repaired by feedback and the management state is maintained, so that a stable good product can be obtained.

However, such a yield comparison method has a disadvantage that it is not possible to directly detect which test item is insufficient.

According to the present invention, a determination algorithm is provided for each of two types of inspection contents of a semiconductor product, and the quality level of a lot to be inspected is determined for each inspection content of a measurement item. The purpose is to judge.

[Structure of the Invention]

[0013]

A first characteristic evaluation step for obtaining a characteristic of a plurality of electronic components of the same kind from a distribution capability of each measurement item and a process capability index obtained from a standard value; Based on the measurement results, a second characteristic evaluation step obtained from the count values of non-defective / defective products for each measurement item of the electronic component, and the results of the first characteristic evaluation step and the second characteristic evaluation step, in advance, for each product type Statistical standard evaluation data indicating the characteristics of the product lot obtained from the manufacturing process, and the inspection lot characteristic evaluation data obtained by the same calculation method as the statistical standard evaluation data for the inspection lot and the relative comparison of the characteristics of the inspection lot. This control step is characterized by the quality level verification method for electronic components according to the present invention.

[0014]

In the method of the present invention, standard data indicating the manufacturing process capability for each product type is compiled and registered for each measurement item for each lot, and the data is compared with the total measurement data of the inspected lot to determine each measurement item. Was provided. Here, the determination algorithms for the two types of inspection contents that can be roughly classified are: By comparing the margin ratio of the distribution with the standard based on the process capability index as a distribution management evaluation that can be evaluated as a characteristic value, 2. A method is adopted in which inspections whose distribution cannot be managed as characteristic values by pattern comparison are determined by comparison based on the frequency of defective products.

[0015]

An embodiment according to the present invention will be described with reference to FIG.

Basically, standard characteristic data of a lot manufactured in a normal control state is registered in a computer for each measurement item, and this is compared with the measurement result of the measurement item of the lot to be measured. Is based on the idea of giving a judgment result.

1. First, the measurement items (inspection contents) whose output corresponding to the input, such as DC inspection and AC inspection, are represented as characteristic values and whose distribution can be managed, are registered as registered standard data with the process capability index (margin ratio of characteristic distribution with respect to the standard). , The lot to be measured, the inconvenient lot, or the rank of the lot from the viewpoint of the quality level. 2. Similarly, if the result of a function inspection or the like is only two values, that is, good or bad, the characteristic distribution cannot be evaluated. Therefore, judgment results are totaled in lot units or sampling based on frequency values (count values), and both (standard The relative verification between the lot data and the data of the lot to be measured is performed, and the same verification is performed based on a certain standard.

From these results, a relative comparison with the standard lot is performed for each measurement item (inspection content) of the lot to be measured, and the process capability index is expressed by Cp = (upper limit−lower limit) / 6σ.

As shown in FIG. 2A, the distribution of the measurement items of the standard lot whose distribution can be controlled is a normal distribution,
FIG. 2B shows a calculation example when the distribution of the measurement results of the lot to be measured deteriorates as clarified in FIG. 2B.
In an example where k = 1.5 but deteriorated, Cpk = 0.8. Therefore, a relative comparison yields 0.8 / 1.5 = 0.51.

On the other hand, in the example using the frequency value shown in 2, the standard defective product frequency value / the total number of standard lots = 10/500 =
Although 0.02 is obtained, the number of defective products to be inspected / the total number of inspected lots = 20/500 = 0.04, and the relative comparison is 0.02 / 0.04 = 0.5.

Here, the quality rank is A = 0.8 or more, and B =
If 0.8 to 0.6 and C = 0.5 or less, both ranks 1 and 2 are ranked C.

As described above, the relative comparison with the standard lot is performed for each measurement item of the inspection lot.

[0023]

At the end of the lot inspection, the following results can be calculated by the present invention.

[0024]

[Table 1]

In the table, A: 0.8 or more, B: 0.8-0.
6, C: 0.6 or less.

Such a verification method is merely an example, and, for example, a method of determining a rank and a ranking may be performed according to circumstances on the operation side. That is, the user (U
Ser), product characteristic standards, and the like, and cannot be determined in any way.

Therefore, as one operation method, the overall ranking of the lot is made to be the lowest rank of rank C. When setting the rank, the method of the present invention is used. However, since the reference value is different depending on, for example, the person in charge of inspection, the person in charge of manufacturing, or the user side, absolute values are not given.

According to such a method, the present invention is incorporated as a supplementary function of the inspection apparatus by verifying the quality and the degree of fabrication at the time of manufacturing a product such as a semiconductor element (when the inspection of the inspection lot is completed). In addition to discriminating non-defective / defective products from each other), a verification result is obtained for each measurement item. As a result, quality assurance of mass-produced products can be obtained, and inspection data in lot units can be obtained.

In the recent production of semiconductor devices, there is no one who takes care of everything from design to inspection, and because the division of labor system is in widespread use, information that leads to each technical department is needed. From this point of view, the method of the present invention can be judged to be valuable, and it is also a great advantage that the inspection result can be obtained not in a single product but in lot units.

[Brief description of the drawings]

FIG. 1 shows each measurement item of a semiconductor product, and FIG. 1a shows an inspection method for expressing an output special value with respect to an input in an analog manner. FIG. 1B shows an inspection method for selecting non-defective / defective products by evaluating an input pattern.

FIG. 2 is a calculation method for verification showing an algorithm in a basic mechanism of the present invention, and the calculation method is shown in FIGS. 2a and 2b corresponding to FIGS. 1a and 1b.

Claims (1)

(57) [Claims] 1. A first characteristic evaluation step which is obtained from a measurement result of a plurality of electronic component characteristics of the same kind from a process capability index obtained from a distribution characteristic and a standard value for each measurement item, and an electronic characteristic is obtained from the measurement result. A second characteristic evaluation step obtained from the count values of good and defective products for each measurement item of the component, and a result of the first characteristic evaluation step and the second characteristic evaluation step are obtained in advance by a manufacturing process for each product type from the results of the first characteristic evaluation step and the second characteristic evaluation step. It is characterized by comprising a control step consisting of statistical standard evaluation data indicating the characteristics of the product lot, and a relative comparison between the inspected lot characteristic evaluation data and the inspected lot characteristics obtained by the same calculation method as the statistical standard evaluation data. Quality level verification method for electronic components