Yield analysis method

By analyzing the physical layout of SRAM I/O and introducing parameters to classify and combine test results, the problem of high cost and long time in high-capacity SRAM yield analysis is solved. It achieves high accuracy and low cost in obtaining yield data and is applicable to SRAM yield analysis of various process nodes.

CN120951919APending Publication Date: 2025-11-14SHANGHAI HUALI INTEGRATED CIRCUIT CORP
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202511002341.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-07-18
Publication Date
2025-11-14

AI Technical Summary

Technical Problem

Existing SRAM yield analysis methods are costly, time-consuming, and inaccurate in high-capacity applications, making it difficult to represent the true yield level.

Method used

By analyzing the physical layout of SRAM I/O, a first parameter is introduced to classify the test results, and the test results of adjacent parameters are combined to obtain the required yield data, thus enabling yield analysis of multiple capacities in one test.

Benefits of technology

It reduces testing costs and time, improves the accuracy and representativeness of yield data, is applicable to SRAM yield analysis at different process nodes, and saves R&D costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120951919A_ABST
    Figure CN120951919A_ABST
Patent Text Reader

Abstract

The invention discloses a yield analysis method, which comprises the following steps of: 1, obtaining physical arrangement of each IO (Input / Output) of an SRAM (Static Random Access Memory); and 2, carrying out IO test on the SRAM and outputting an IO test result. A first parameter is introduced into the IO test, and the first parameter corresponds to the IO physical arrangement. And performing classified output of IO test results according to the first parameter. And step 3, according to IO physical distribution, IO test results belonging to different first parameters are combined to obtain required yield data. According to the invention, the required yield data can be obtained according to the process research and development requirements, the test duration is not increased, and the research and development cost can be reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to a semiconductor integrated circuit manufacturing method, and more particularly to a yield analysis method. Background Technology

[0002] In the semiconductor manufacturing industry, SRAM chips are one of the most commonly used tools for continuously monitoring the health of mature process platforms and developing new process platforms due to their compact and regular structure and manufacturing process that is closer to the design characteristics of logic products. SRAM yield analysis has always been a very critical part of the new process development and production process. Improving SRAM yield by improving process conditions is of great significance for speeding up the development process, completing the development of new process platforms, and ultimately achieving customer yield expectations.

[0003] However, with the continuous evolution of semiconductor manufacturing processes, the capacity of SRAM has also been constantly increasing, expanding from the early 2M capacity to the current mainstream 256M or even 512M capacity. Therefore, achieving breakthroughs and improvements in SRAM yield during early development has inevitably become a challenge. Figure 1 The image shows a table of data when the yield rate is 0% in all existing yield analysis methods. The 0% yield rate data cannot be used to intuitively analyze the differences.

[0004] There is an urgent need to find new analytical methods to address the above-mentioned new problems. The three existing improvement methods and their corresponding shortcomings are as follows:

[0005] The first approach involves designing SRAM chips of varying capacities, adopting different designs as the research and development phase progresses. However, the obvious drawback is the extremely high cost. Design costs multiply with the number of chips requiring design, and production requires replacing photomasks with each chip replacement, making the costs prohibitive.

[0006] The second method involves dividing the complete large-capacity SRAM into regions of different sizes for testing. The size of the test region can be adjusted as needed. However, due to the chip's operating mode and design scheme, the test time increases exponentially with the divided capacity, and the switching status of the divided region may affect the accuracy of the static leakage current test.

[0007] The third method involves splitting the complete large-capacity SRAM into the required capacity during testing and selecting one of them for routine testing. However, this method has a certain degree of randomness and is easily affected by design and leakage current in the early stages of R&D, making it very easy for the selected area to not represent the true yield level. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to provide a yield analysis method that can obtain the required yield data according to the needs of process development, without increasing the testing time, which is conducive to reducing the development cost.

[0009] To solve the above-mentioned technical problems, the yield analysis method provided by this invention includes:

[0010] Step 1: Obtain the physical layout of each I / O in the SRAM.

[0011] Step 2: Perform an I / O test on the SRAM and output the I / O test results.

[0012] A first parameter is introduced into the IO test, which corresponds to the IO physical arrangement.

[0013] The IO test results are categorized and output according to the first parameter.

[0014] Step 3: Based on the physical distribution of IO, combine the IO test results belonging to different first parameters to obtain the required yield data.

[0015] A further improvement is that, in step one, the physical distribution of the I / O is obtained by analyzing the layout of the SRAM.

[0016] A further improvement is that the electrical address count for each IO is 512k.

[0017] A further improvement is that each of the IOs is 8 bits.

[0018] A further improvement is that the number of I / Os corresponding to each of the first parameters is 8.

[0019] A further improvement is that the IO test results include the number of invalid bits.

[0020] A further improvement is that, in step three, the IO test results of adjacent first parameters are combined to obtain the required yield data.

[0021] A further improvement is that the required yield data includes multiple data points, obtained by combining the IO test results of adjacent first parameters in various ways.

[0022] A further improvement is that, in step three, all the required yield data are obtained at once by combining the IO test results of adjacent first parameters in multiple ways.

[0023] A further improvement is that the required yield data are yield data for SRAMs of various capacities.

[0024] A further improvement is that the capacity of the SRAM is 64k, 256k, 512k, 1M, 2M, 4M, 8M, 16M, 32M, 64M, 128M, 256M or 512M.

[0025] A further improvement is that when the yield data of the large-capacity SRAM is 0, the large-capacity SRAM is split into multiple small-capacity SRAMs, and the yield data of the small-capacity SRAMs is used to analyze the large-capacity SRAM.

[0026] A further improvement is that the yield data is used to monitor mature process platforms or for the development of new process platforms.

[0027] A further improvement is that the SRAM comprises one or more, formed on the same wafer or on multiple wafers.

[0028] A further improvement is that the SRAM has multiple cell modules, each cell module has multiple I / Os and has the same physical arrangement of cell module I / Os.

[0029] This invention obtains the physical layout of each I / O in the SRAM in advance, so that the I / O test results can be classified and output according to the I / O physical layout during I / O testing. Different types of I / O test results can be combined with each other. In this way, the corresponding combination of I / O test results can be formed according to the process development needs and the corresponding yield data can be obtained. Moreover, it can obtain all the required yield data at once, and the obtained yield data can represent the true yield level.

[0030] In addition, the IO test of the present invention can be completed in one test, so it will not increase the test time and helps to reduce the test cost.

[0031] This invention only requires a single capacity SRAM and does not require designing SRAMs of different capacities. Therefore, the design cost of the SRAM chip of this invention is low. At the same time, since no photomask is required during production, it also has the characteristic of low production cost.

[0032] Because the present invention has high yield data accuracy, short testing time, and low design and production costs of SRAM, it is beneficial to reduce R&D costs. Attached Figure Description

[0033] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0034] Figure 1 This is a tabular data chart showing the situation where the yield rate is 0% in all existing yield analysis methods;

[0035] Figure 2 This is a flowchart of the yield analysis method according to an embodiment of the present invention;

[0036] Figure 3This is a module partitioning diagram of an SRAM in the yield analysis method of this invention embodiment;

[0037] Figure 4 This is a layout analysis diagram of the IO physical arrangement in the yield analysis method of this invention embodiment;

[0038] Figure 5 This is a table showing the categorized output of the IO test results of the yield analysis method according to an embodiment of the present invention;

[0039] Figure 6 This is a table of yield data obtained by combining SRAMs of different capacities through the yield analysis method of this invention embodiment;

[0040] Figure 7 This is a table of yield data required to obtain 8M SRAM by combining the yield analysis method of this invention. Detailed Implementation

[0041] like Figure 2 The diagram shown is a flowchart of the yield analysis method according to an embodiment of the present invention; the yield analysis method according to an embodiment of the present invention includes:

[0042] Step 1: Obtain the physical layout of each I / O of SRAM101.

[0043] In this embodiment of the invention, the physical distribution of the I / O is obtained by analyzing the layout of the SRAM101. For example... Figure 3 The diagram shown illustrates the module partitioning of an SRAM 101 in the yield analysis method of this embodiment. The SRAM 101 has multiple unit modules 102, each unit module 102 having multiple I / Os with the same physical arrangement of the unit module I / Os. The physical arrangement of the I / Os of the entire SRAM 101 is composed of the physical arrangements of the unit module I / Os of all the unit modules 102. Therefore, by obtaining the arrangement of the unit modules 102 and the physical arrangement of the unit module I / Os of each unit module 102, the physical distribution of the I / Os can be obtained.

[0044] like Figure 4 The diagram shown is a layout analysis diagram of the IO physical arrangement in the yield analysis method of this invention. The region of one IO is shown in large brackets 103. In some embodiments, the electrical address quantity of each IO is 512k, and the bit length of each IO is 8 bits. Each bit of each IO is determined by adding a bit number to DQ, such as DQ0 to DQ7 for the first IO.

[0045] Figure 4 As shown, a physical layout of one of the unit modules includes 8 IOs, with each bit corresponding to DQ0 to DQ63, totaling 64 bits. Figure 3The image shows 32 of the aforementioned unit modules 102. In other embodiments, the number of unit modules 102 can be changed according to actual needs.

[0046] In this embodiment of the invention, the SRAM101 includes one or more, which are formed on the same wafer or on multiple wafers.

[0047] Step 2: Perform IO testing on the SRAM101 and output the IO test results.

[0048] A first parameter is introduced into the IO test, which corresponds to the IO physical arrangement.

[0049] The IO test results are categorized and output according to the first parameter.

[0050] In this embodiment of the invention, the number of I / Os corresponding to each of the first parameters is 8. Since each I / O has 8 bits, the first parameter has 64 bits of data.

[0051] like Figure 5 The table shown is a classification output table of the IO test results of the yield analysis method according to an embodiment of the present invention. Figure 5 In this context, the first parameter is represented by DQ00, DQ01 up to DQ63, respectively, corresponding to... Figure 4 The IO bits included in the physical arrangement of the unit module IO in the above are DQ0 to DQ63.

[0052] Figure 5 In this context, "wafer" refers to a wafer, and the corresponding data represents the wafer number. Figure 5 In this context, all wafers are numbered 1, indicating that they are identical. Diex and Diey represent the x and y coordinates of the SRAM chip on the wafer, respectively. Typically, a combination of Diex and Diey corresponds to... Figure 3 One of the unit modules 102.

[0053] The IO test results include the number of failed bits. Figure 5 In the table, the data represents the number of invalid bits.

[0054] Step 3: Based on the physical distribution of IO, combine the IO test results belonging to different first parameters to obtain the required yield data.

[0055] In this embodiment of the invention, the required yield data is obtained by combining the IO test results of adjacent first parameters.

[0056] The required yield data includes multiple parameters, which are obtained by combining the IO test results of adjacent first parameters in various ways.

[0057] All the required yield data can be obtained at once by combining the IO test results of adjacent first parameters in multiple ways.

[0058] The required yield data refers to the yield data of the SRAM 101 with various capacities. In some embodiments, the capacity of the SRAM 101 is 64k, 256k, 512k, 1M, 2M, 4M, 8M, 16M, 32M, 64M, 128M, 256M, or 512M. Figure 6 The figure shown is a table of yield data of SRAM101 with different capacities obtained by the yield analysis method of this embodiment of the invention through combination. Figure 6 The document shows that the SRAM101 has capacities of 4M, 8M, 16M, 32M, 64M, 128M, 256M, and 512M. The yield is not 0%, which allows for analysis during process development.

[0059] In some embodiments, when the yield data of the large-capacity SRAM 101 is 0, the large-capacity SRAM 101 is split into multiple small-capacity SRAMs 101, and the yield data of the small-capacity SRAMs 101 is used to analyze the large-capacity SRAM 101.

[0060] like Figure 7 The figure shown is a table of yield data required to obtain 8M SRAM101 by combining the yield analysis method of the present invention. The yield of 8M SRAM101 is 0%, and yield analysis cannot be performed. In this case, the 8M SRAM101 can be split, for example, the yield of 8M SRAM101 can be analyzed by using the yield of 64K to 1M.

[0061] In this embodiment of the invention, the yield data is used to monitor mature process platforms or to develop new process platforms.

[0062] This invention provides an embodiment of the invention that pre-determines the physical layout of each IO of SRAM101. This allows for the classification and output of IO test results based on the IO physical layout during IO testing. Different types of IO test results can be combined with each other, thus forming a combination of IO test results according to process development needs and obtaining corresponding yield data. Moreover, it enables the acquisition of all required yield data at once, and the obtained yield data can represent the true yield level.

[0063] In addition, the IO test in this embodiment of the invention can be completed in one test, so it will not increase the test time and helps to reduce the test cost.

[0064] The embodiments of the present invention only require the use of a single SRAM101, without the need to design SRAM101s of different capacities. Therefore, the SRAM101 chip design cost of the embodiments of the present invention is low. At the same time, since no photomask needs to be added during production, it also has the characteristic of low production cost.

[0065] Because the yield data of the embodiments of the present invention is highly accurate, the testing time is short, and the design and production cost of SRAM101 is low, it is beneficial to reduce R&D costs.

[0066] The method of this invention utilizes existing testing equipment to measure electrical properties such as voltage and current. During testing, new IO test parameters are introduced by setting the chip pins, and different IO data are output separately. Then, combined with chip layout analysis, the number of failed bits of the required IO output is grouped and combined to finally realize the yield calculation of different SRAM capacities.

[0067] The method of this invention is applicable to all SRAM yield testing and analysis. It can obtain SRAM yields of different capacities for analysis by splitting and combining them when the yield is generally low in the early stage of R&D, without increasing the testing time.

[0068] Unlike previous SRAM chip testing at 40nm and above process nodes, advanced process nodes, due to the doubling of SRAM capacity and higher process complexity, face the first hurdle in project development: achieving zero yield. In the initial stages before reaching zero yield, existing methods struggle to analyze the impact of various complex process conditions on yield. However, the method of this invention, employing different module combinations in stages to create a suitable yield analysis scheme, not only saves testing time but also significantly accelerates project development, yielding substantial benefits in the early and mid-stages of the project. This invention's method can also be extended to testing other process nodes.

[0069] The method of this invention can be used in SRAM yield analysis at any process development stage. It can combine various suitable yield data schemes according to the needs of the development stage, effectively improving the readability of data in the early stage of development and reducing the overall development cost.

[0070] Compared to existing traditional SRAM testing and analysis, the method of this invention offers more flexible testing conditions and saves more time, effectively improving the benefits of yield analysis.

[0071] The present invention has been described in detail above through specific embodiments, but these are not intended to limit the invention. Many modifications and improvements can be made by those skilled in the art without departing from the principles of the invention, and these should also be considered within the scope of protection of the present invention.

Claims

1. A yield analysis method, characterized in that, include: Step 1: Obtain the physical layout of each I / O in the SRAM; Step 2: Perform I / O testing on the SRAM and output the I / O test results; A first parameter is introduced into the IO test, and the first parameter corresponds to the IO physical arrangement; The IO test results are categorized and output according to the first parameter; Step 3: Based on the physical distribution of IO, combine the IO test results belonging to different first parameters to obtain the required yield data.

2. The yield analysis method as described in claim 1, characterized in that: In step one, the physical distribution of IO is obtained by analyzing the layout of the SRAM.

3. The yield analysis method as described in claim 1, characterized in that: Each I / O has 512k electrical addresses.

4. The yield analysis method as described in claim 3, characterized in that: Each of the aforementioned IOs is 8 bits.

5. The yield analysis method as described in claim 4, characterized in that: The number of I / Os corresponding to each of the first parameters is 8.

6. The yield analysis method as described in claim 1, characterized in that: The IO test results include the number of invalid bits.

7. The yield analysis method as described in claim 1, characterized in that: In step three, the IO test results of adjacent first parameters are combined to obtain the required yield data.

8. The yield analysis method as described in claim 1, characterized in that: The required yield data includes multiple parameters, which are obtained by combining the IO test results of adjacent first parameters in various ways.

9. The yield analysis method as described in claim 8, characterized in that: In step three, all the required yield data are obtained at once by combining the IO test results of adjacent first parameters in multiple ways.

10. The yield analysis method as described in claim 9, characterized in that: The required yield data are yield data for SRAMs of various capacities.

11. The yield analysis method as described in claim 10, characterized in that: The SRAM capacity is 64k, 256k, 512k, 1M, 2M, 4M, 8M, 16M, 32M, 64M, 128M, 256M or 512M.

12. The yield analysis method as described in claim 10, characterized in that: When the yield data of the large-capacity SRAM is 0, the large-capacity SRAM is split into multiple small-capacity SRAMs, and the yield data of the small-capacity SRAMs is used to analyze the large-capacity SRAM.

13. The yield analysis method as described in claim 1, characterized in that: The yield data is used to monitor mature process platforms or for the development of new process platforms.

14. The yield analysis method as described in claim 1, characterized in that: The SRAM may include one or more, formed on the same wafer or on multiple wafers.

15. The yield analysis method as described in claim 2, characterized in that: The SRAM has multiple unit modules, and each unit module has multiple I / Os with the same physical arrangement of unit module I / Os.