CN107340487A - A kind of method checked test system and be in actual processing ability under stable state - Google Patents

Abstract

The invention discloses a kind of method checked test system and be in actual processing ability under stable state.Method includes：4site is carried out by J750 test machines and surveyed, obtains initial data；Judge whether initial data is continuity data；If initial data is continuity data, judge whether initial data meets normal distribution；If initial data meets normal distribution, determine whether the testing process of test system is stably and controllable based on control figure；If testing process is stably and controllable, process capability index is calculated based on initial data；The actual processing ability of test system is determined according to process capability index.The present invention breaks through conventional analysis method, by calculating its C_PKValue, check the actual processing ability that measuring system is under stable state.First by C_PKApply in integrated circuit testing field, analyze wafer-level test data, check the capability of process, will combine theory and practice together, to ensure product quality, the stability of test provides a kind of effective means.

Description

A kind of method checked test system and be in actual processing ability under stable state

Technical field

The present invention relates to a kind of integrated circuit testing field, more particularly to one kind to check that test system is under stable state The method of actual processing ability.

Background technology

One ATE is (referred to as：ATE several devices) can be measured simultaneously, after either disk either encapsulates Chip will often use multi-point sampler；When shipment amount is huge simultaneously, it is also necessary to which more test machines are tested.This is just needed To confirm whether each test result between test machine and test station be consistent, to avoid various before production test Abnormal even accident.

Traditionally whether exact p-value result unanimously can be first in the case where ensuring that ATE calibrations pass through, repeated multiple times measurement Simultaneously statistical measurements respectively test whether position stablizes to same wafer to check same board.Checking two has same resource distribution Board when, tested often through by same wafer on this two test machines, compare two wafer figures of generation, count yield Difference, whether the tube core and defective tube core that statistical test passes through have saltus step etc..

Process capability index is (referred to as：C_PK) it is the index that modern enterprise is used to represent process capability.Process capability index is A kind of method for representing processing procedure level height, Essential Action are the height for reflecting process yield.Process capability can just be stablized by force Production mass it is high, the high product of reliability.The research of process capability is the degree for confirming that these characteristics are up to specification, with Ensure that the yield of processing procedure finished product, can be as the foundation that processing procedure persistently improves on desired level.

The content of the invention

Check that test system is in actual processing energy under stable state the technical problem to be solved in the present invention is to provide one kind The method of power.

In order to solve this technical problem, check that test system is under stable state the invention provides one kind and actually add The method of work ability.In the case where ensuring that measuring system is stable, C is calculated_PKValue, stable state is in so as to investigate measuring system Under actual processing ability.

Break through in the past by analyses such as the average to test result or yields come identification chip measuring stability, using C_PKCome Reflect the capability of process.First Application analyzes wafer-level test data in integrated circuit testing field.Use MINITAB 16 is analyzed, and Capability Sixpacks instruments in quality tool calculate C_PKObtain visual knot Fruit.

Brief description of the drawings

Fig. 1 is 100 specific test datas；

Fig. 2 is data analysis scheme；

Fig. 3 is test of normality result；

Fig. 4 is control figure analysis result；

Fig. 5 is C_PKResult of calculation；

Fig. 6 is C_PKFive classification standards；

C after the improvement of Fig. 7 process capabilities_PKResult of calculation.

Explanation of nouns

site：Usual professional refers to a chip position on wafer a site；

Multi-site concurrent testings：Multi-site parallel test techniques refer to can be simultaneously to more on a test machine Individual IC chip is automatically detected, and multiple chips can be connected to simultaneously by the probe card specially designed and produced On pin so that test machine can carry out the test of multiple chips simultaneously, and record the test result of multiple chips, pass through unit The quantity of increase chip under test improves the throughput of test machine in time, reduces test machine slack resources, improves test and set Standby utilization rate, greatly reduces energy consumption, reduces the floor space of test factory building.

Embodiment

From a Teradyne J750 ATE, ensure that calibration passes through.It is chosen on Teradyne J750 The magnitude of voltage of the output pin for a product that success is developed is data source, and the magnitude of voltage test specification upper limit is 3.8V, under It is limited to 3.0V.This product takes 4site and surveys pattern.ATE tests 5 pieces of wafers altogether, randomly selects the 5 of 4 site Secondary test amounts to 100 data and carries out subsequent analysis.100 data are shown in Fig. 1.Data analysis scheme is shown in Fig. 2.

For two kinds of data types of continuous type and discrete type, different analysis methods can be selected.To continuous data Speech, first has to check whether data meet normal distribution, control map analysis is used to the data for meeting normal distribution, if obtaining Journey is stablized controlled, then can calculate C_PKValue.The Capability Sixpacks works in MINITAB16 softwares can be used Tool obtains visual C_PKValue.If data do not meet normal distribution, to check whether that other distributions can be met by conversion, The data for not meeting any distribution are then calculated using formula.For discrete data, for based on bi-distribution or being based on Poisson distribution has different computational methods.

This example data are continuous data, are specifically analyzed using following methods：First, check whether data meet normal state point Cloth, examined here from Anderson-Darling, be a kind of inspection based on empirical cumulative distribution function.Weighed by P values Amount, general P values then think approximate normal distribution when being more than 0.05, P values give tacit consent to normal distribution when being more than or equal to 0.1.By 100 numbers According to importeding into MINITAB16 softwares, software can export feedback result, see Fig. 3.P values are 0.639, it is seen that this group of data fit Normal distribution.

Secondly, to meeting the data of normal distribution using control map analysis, check flow whether stablize it is controlled, from monodrome Control chart for variables (I-MR control figures) instrument, equally 100 data are imported into MINITAB16 softwares, software can export Feedback result, see Fig. 4.All test values are in upper and lower two control lines (UCL/LCL) as can see from Figure 4, therefore, stream Journey does not have abnormity point, stable, controlled.

3rd, if obtain process stablize it is controlled, can start calculate C_PKValue.Specific calculation procedure is seen below, wherein, USL represents upper specification limit, and LSL represents lower specification limit：

1) average value of 100 data is calculated：

2) standard deviation of 100 data is calculated

3) specification tolerance：T=USL-LSL

4) specification central value：

5) process accuracy：

6) processing procedure precision：

7) process capability index：

Cpk=Cp (1- | Ca |)

It can be obtained using the Capability Sixpacks instruments in MINITAB16 softwares simultaneously more intuitively visual Change result.Fig. 5 is result figure.As a result 6 parts are divided, wherein, it is to be in slave mode that Xbar, R and tendency chart, which are used for verification process,. Bar chart and normal probability plot are used to verify data Normal Distribution.C can be directly read from can try hard to_PKValue.

The C that will be obtained_PKValue refers to C_PKFive classification standards judged, C_PKFive classification standards see Fig. 6.Here The actual processing ability of test machine can be made a decision according to five classification standards, check the height of process capability.For processing procedure Ability it is relatively low need propose Improving Measurements, improve processing procedure.Process capability it is high continue to keep.This example C_pkIt is worth for 1.24, reference C_pkFive classification standards, it is known that this test machine is in two level, ability reduces still, it is necessary to be operated and quality improving Variation, improve quality.

Look back whole testing process, find in whole test process can with improved part, consider personnel, equipment, This five part of material, method, environment, by FMEA, key factor being screened, the exploration card for finding to use during test is oxidized seriously, Cause loose contact or test result bad.After being operated by card grinding, 100 data are extracted again and are divided with face method Analysis, can obtain new C_pkIt is worth for 1.45, with reference to C_pkFive classification standards, belong to one-level, and situation is preferable, it is seen that effective to improve Processing procedure.Visual result can be obtained using MINITAB16, sees Fig. 7, the yield of product can be protected.

Claims (3)

1. A kind of 1. method checked test system and be in actual processing ability under stable state, it is characterised in that methods described bag Include：

   4site is carried out by J750 test machines and surveyed, obtains initial data；

   Judge whether the initial data is continuity data；

   If the initial data is continuity data, judge whether the initial data meets normal distribution；

   If the initial data meets normal distribution, determining whether the testing process of the test system is stable based on control figure can Control；

   If the testing process is stably and controllable, process capability index is calculated based on the initial data；According to the process capability Index determines the actual processing ability of the test system.
2. 2. according to the method for claim 1, it is characterised in that described to judge whether the initial data meets normal state point Cloth, including：

   Determine the p value of the initial data；

   When the p value within a preset range when, determine that the initial data meets normal distribution.
3. 3. according to the method for claim 1, it is characterised in that the test that the test system is determined based on control figure Whether flow is stably and controllable, including：

   According to the Raw Data Generation control figure；

   Determine the monodrome and moving range value whether in two control lines of UCL and LCL of the control figure.