Computer-implemented method for determining acceptability of a substrate

Abstract

This case discloses a dynamic modeling technique for detecting and classifying defects in assembled substrates, such as semiconductor substrates. One model includes a pixel-by-pixel allocation of pixel data based on images of the substrate to define a range of known acceptability for the substrate. This acceptability range can be defined between upper and lower thresholds. The model is dynamically updated upon acquisition of new image data of the substrate, and particularly new image data where an imaging factor has changed regardless of substrate acceptability. The model can be updated by shifting for one or more of the thresholds for one or more of the pixels.

Description

Technical Field

[0001] Invention Field

[0002] This disclosure pertains to the methods used to determine the acceptability and unacceptability of a matrix. Prior Technology

[0003] Background of the Invention

[0004] Semiconductor substrates are manufactured or assembled as part of the formation of semiconductor wafers or other types of integrated circuits (ICs). The components of the final IC can be incorporated into the substrate through a series of assembly steps. These assembly steps may include a deposition step of adding a thin film layer onto the substrate. The substrate may then be covered with a photoresist, and a circuit pattern of marking lines may be projected onto the substrate using photolithography. An etched process may then occur. During each assembly step, the tools used to perform that step may create defects or flaws on the substrate.

[0005] In some applications, a semiconductor substrate is composed of dies. A die is a block of semi-conductive material (e.g., silicon) on which a given functional circuit is assembled. For example, a functional circuit may take the form of a central processing unit. A functional circuit may be batch-manufactured on a single substrate. The substrate is then diced into dies, each die containing a copy of the circuit. Each die may be rectangular. A thin, non-functional spacing may be provided between the dies, which allows (e.g., with a saw) individual dies to be cut from a substrate without compromising the circuitry.

[0006] Grains can be inspected for defects such as scratches, voids, corrosion, and bridging. If a defect is detected, a decision must be made as to whether to discard or retain the substrate. This decision may be based on the type and severity of the defect detected, the type of grain, and other factors. Summary of the Invention

[0007] Invention Summary

[0008] Generally, this disclosure relates to improvements in modeling techniques used for defect inspection and classification of substrates. The ideas presented in this disclosure can be applied to any manufacturing process involving an assembled substrate where defects can be detected and classified based on images of the substrate. One example of such a substrate is a semiconductor substrate. One example of a semiconductor substrate is a semiconductor wafer or panel.

[0009] A semiconductor substrate can incorporate any number of different electronic devices, such as a light-emitting diode (LED). In some examples, the modeling perspective of this disclosure, as well as the defect inspection, detection, and classification perspective, can be applied to products containing semiconductor substrates, such as LEDs. That is, the methods of this disclosure can be used to determine which LEDs in a group are acceptable and which are unacceptable and should be discarded.

[0010] Some embodiments of this disclosure relate to computer-executable methods. Some embodiments of this disclosure relate to computer systems. Some embodiments of this disclosure relate to instructions executed by one or more processors and stored in a non-transitory computer-readable medium.

[0011] According to certain specific viewpoints of this disclosure, a computer-executable method for determining the acceptability of a substrate includes: generating a model representing image data obtained from reference images of a first substrate, each reference image including pixels relative to each of other reference images, the model defining an acceptable range of image values ​​for each pixel; receiving new image data obtained from a second substrate; and updating the model to generate an updated model, including modifying the acceptable range of image values ​​based on the new image data to generate a modified range.

[0012] According to other specific views of this disclosure, a computer-executable method for determining the acceptability of a substrate includes: generating a model representing image data acquired from a reference image of a complete die of a first semiconductor substrate, each reference image including pixels relative to each of other reference images, the model defining an acceptable range of image values ​​for each pixel; receiving new image data acquired from a complete die image of a second semiconductor substrate; determining, when the complete die image of the second semiconductor substrate is acquired and when the reference images are acquired, that an imaging factor unrelated to substrate acceptability is different; and based thereon: updating the model to generate an updated model, including modifying the acceptable range of image values ​​based on the new image data to generate a modified range.

[0013] According to other specific views of this disclosure, a computer-executable method for determining the acceptability of a substrate includes: generating a model representing image data obtained from reference images of a first substrate, each reference image including pixels relative to each of other reference images, the model defining an acceptable range of image values ​​for each pixel; receiving new image data obtained from a second substrate; and based on the new image data: updating the model to generate an updated model, including modifying the acceptable range of image values; and determining whether the second substrate is acceptable or unacceptable by comparing the new image data with the model.

[0014] This invention provides a simplified approach to guiding the further description of the following embodiments. This invention is not intended to identify key or essential features of the claimed subject matter, nor is it intended to limit the scope of the claimed subject matter. Additional aspects, features, and / or advantages of these embodiments will be set forth in part in the following description, and will be more apparent in part from the description, or may be learned by practice of the disclosure. Simple Explanation of the Diagram

[0015] The following figures illustrate examples of non-restricted and non-exhaustive enumeration. In all the figures, common parts and features are indicated with the same numbers whenever possible.

[0016] Figure 1 schematically shows an example system for dynamically performing a modeled substrate acceptability determination method for a semiconductor substrate, according to the present disclosure.

[0017] Figure 2 schematically shows a reference image that can be used by the system of Figure 1 to construct a dynamic model for determining the matrix acceptability.

[0018] Figure 3 schematically shows pixel data of one of the pixels in the reference image of Figure 2.

[0019] Figure 4 schematically shows a complete and / or updated full grain model constructed using pixel data from Figure 3.

[0020] Figure 5 shows composite pixel data for one of these composite pixels in relation to the full grain model of Figure 4.

[0021] Figure 6 schematically illustrates an exemplary method for dynamically modeling the matrix acceptability determination method according to the present disclosure.

[0022] Figure 7 schematically illustrates an example method for updating model pixel data according to the method in Figure 6.

[0023] Figure 8 shows another exemplary method for dynamically modeling matrix acceptability determination in accordance with this disclosure. Implementation

[0024] Detailed Description of Preferred Embodiments

[0025] Inspection of a semiconductor substrate can be performed by a vision inspection device, such as a lens, for example, a high-resolution CCD lens. Image data captured by the lens on the substrate can be tested. Other types of imaging devices, including non-virtual imaging devices, can be used to acquire images that, according to this disclosure, are used to construct and update a dynamic substrate acceptability model. Image data of a candidate substrate can then be compared with image data of the initial or updated model. The model defines a substrate acceptability range as a function of image data. For example, the model can define a range of acceptable brightness values ​​for each pixel of an image of a die, or a portion of a die, a wafer, or a panel.

[0026] Figure 1 schematically shows an example system 100 for dynamically performing a modeled substrate acceptability determination method for a semiconductor substrate, according to the present disclosure.

[0027] The system 100 includes a computing device 202. The computing device 202 may be a server and / or other computing device that performs the operations discussed herein, such as modeling, model updating, defect detection, and the decision and classification operations described herein. The computing device 202 may include computing components 206. These computing components 206 include at least one processor 208 and memory 204. The memory 204 may include a non-transitory computer-readable medium. Depending on the specific configuration, the memory 204 (which, among other things, stores a modeling engine including instructions to perform the other operations disclosed herein) may be electrically dependent (such as RAM), non-electrically dependent (such as ROM, cache memory, etc.), or some combination of both. The computing device 202 may include one or more graphics processing units (GPUs) arranged together to accelerate model building and updating and / or applications for a model for defect detection and decision-making. Furthermore, the computing device 202 may also include storage devices (removable 210 and / or non-removable 212), including, but not limited to, solid-state devices, magnetic disks or optical disks, or magnetic tapes. Additionally, the computing device 202 may also have input devices 216, such as a touchscreen, keyboard, mouse, pen, voice input, etc., and / or output devices 214, such as a display, speaker, printer, etc. One or more communication connections 218, such as local area networks (LANs), wide area networks (WANs), peer-to-peer, Bluetooth, RF, etc., may also be incorporated into the computing device 202.

[0028] The system 100 may include an inspection device 102, which is operationally, for example, linked to and communicates with the computing device 202 via a network and the communication connection 218. The inspection device 102 may include an imaging device, such as a CCD lens that captures images of a semiconductor substrate. The computing device 202 receives such images and uses instructions from the modeling engine to perform the dynamic modeling, defect detection, acceptability determination, and classification operations described herein.

[0029] The inspection device 102 itself, and / or its external properties, may change over time. For example, a new inspection device 102 may be calibrated differently from a previous one. More generally, for a given die type, the die image data, and / or how that image data is acquired, may evolve over time or otherwise change due to external properties, such as factors unrelated to the actual acceptability or unacceptability of the die. For example, the imaging equipment used to acquire the image data may be repaired, replaced, or upgraded. In another example, environmental factors (e.g., background noise, lighting) present when the image data is acquired may change. In other examples, the calibration of the assembly equipment used to assemble the reference dies may change, the color or reflectivity of the material used to manufacture the die may change, and so on. Therefore, due to variables, such as those associated with one or more of the aforementioned exemplary factors, external attributes can also be regarded as a natural variation of pixel image data values. When examining defects that determine their acceptability, these factors do not affect the acceptability of a single chip, but they do affect the image data obtained from a single chip.

[0030] Using the modeling engine, the dynamic modeling performed by the computing device 202 can take into account the variability in the image data that does not represent the external properties of matrix defects.

[0031] Figure 2 schematically shows a reference image 12 that can be used by the system of Figure 1 to construct a dynamic model for determining substrate acceptability. This reference image 12 can be captured using the inspection device 102 (Figure 1). In some examples, the reference image 12 is a complete grain reference image, meaning it includes pixel data for a complete grain of a semiconductor wafer or panel. In other examples, the reference image 12 is a partial grain reference image, meaning it includes pixel data from a portion of a grain of a semiconductor wafer or panel. Multiple partial grain reference images of different portions of the same grain can be linked together to form a complete grain reference image.

[0032] The reference image 12 includes pixels 21, wherein each pixel represents a square in one of the square arrays on the reference image 12 in FIG2.

[0033] The reference image 12 is known to represent a non-defective grain or a portion of a grain. That is, the grain or grain portion from which the image was obtained to produce the reference image 12 is known to be acceptable, or to some extent it has any defects that are acceptable.

[0034] In some examples, certain reference images 12 used to generate a complete grain model are known to correspond to unacceptable grains or grain portions.

[0035] Figure 3 schematically shows pixel data 25 of one of the pixels 21 of the reference image 12 of Figure 2. The pixel data 25 can be any visual or other type of data generated on a pixel-by-pixel basis. One example of pixel data 25 is luminance. The pixel data 25 includes a single data point 27 corresponding to the pixel data obtained from a pixel 21. For example, the data point 27 can be a luminance value for a single pixel 21. The modeling engine of the computing device 202 of Figure 1 can be configured to derive the pixel data from each pixel 21 and receive the derived pixel data as input equivalent to an acceptable die or die portion.

[0036] Figure 4 schematically shows a complete full-grain model, which can be an initial full-grain model 10 or an updated full-grain model 14. With different external characteristics than the initial full-grain model used to construct it, an updated full-grain model 14 has been updated one or more times from an initial full-grain model based on additional image data acquired by the imaging device. Updating a full-grain model can be an iterative process, wherein the initial model is continuously updated to a previously updated model as relevant updated imaging data is generated by the inspection device 102 (Figure 1) during subsequent substrate manufacturing operations and feeds to the computing device 202.

[0037] The complete grain models 10 and 14 are composed of composite pixels 23, each of which is represented as a square in a square array of FIG4. Each composite pixel 23 is a composite pixel that includes relevant pixel data (e.g., brightness) obtained from the corresponding pixels of all the reference images 12 to generate the complete grain models 10 and 14.

[0038] Each composite pixel 23 represents a pixel data allocation from the reference image used to generate the composite pixel. In some examples, the pixel data allocation only represents pixel data indicating an acceptable grain or grain portion. In other examples, where some reference images used to generate the complete grain models 10, 14 include known unacceptable grains, the allocation represents some pixel data indicating an acceptable grain or grain portion, and other pixel data indicating an unacceptable grain or grain portion.

[0039] Figure 5 shows composite pixel data 20 and 24 for one of the composite pixels 23 in the full grain model of Figure 4. Composite pixel data 20 is based on composite pixel 23 in an initial full grain model 10 (Figure 4). Composite pixel data 24 is based on composite pixel 23 in an updated full grain model 14 (Figure 4).

[0040] Each set of composite pixel data 20, 24 is a pixel data allocation from the corresponding pixel of the lower reference image. The composite pixel data 20 is a pixel data allocation from the initial reference image set. The composite pixel data 24 is a pixel data allocation from the initial reference image set plus one or more updated reference images generated by external attributes.

[0041] The modeling engine of the computing device 202 in Figure 1 can be configured to combine pixel data from each pixel 21 of each reference image to generate the pixel data allocation of the corresponding composite pixel.

[0042] Figure 6 schematically illustrates an exemplary method for dynamically modeling the matrix acceptability determination method according to the present disclosure.

[0043] Referring to Figure 6, the initial complete grain model 10 can be constructed from multiple substrate reference images 12, wherein pixel-by-pixel image data from these multiple images are combined to generate the model 10. It can be understood that the construction, modification, and application of a pixel-by-pixel-based model is only one exemplary implementation of the system 100 (Figure 1). In other examples, other aspects of this art can be implemented based on different image units, such as a group of pixels, a portion of a pixel, a spatially defined region of an image without pixel identification, etc.

[0044] In some examples, the substrate reference image 12 is a complete grain reference image, and the model 10 is a model of a complete grain. A complete grain reference image is an image that includes image data of grain integrity for a given semiconductor component, such as a specific type of functional circuit.

[0045] In some examples, each complete grain reference image 12 is a composite image of one of several other images linked together. For example, each other image may be an image of a different portion of the grain. These other images are linked together to form a complete grain reference image.

[0046] For all grains in the complete grain reference images 12, the acceptability or unacceptability of the corresponding grains is known. At least some grains in the complete grain reference images 12 are known to be acceptable. That is, the grains are defect-free or have one or more defects that are not significant enough to make the grain unacceptable. In some examples, at least some grains in the complete grain reference images 12 are known to be unacceptable.

[0047] Because multiple complete grain reference images 12 are used to generate the initial model 10, and because these complete grain reference images 12 contain different image data between different images, the initial model 10 includes an image data allocation for each pixel of the combined image corresponding to one combination of the reference images 12. The number of complete grain reference images 12 can be 10, 100, 1,000, 10,000, or higher. Therefore, for each pixel, the initial model 10 can represent a histogram 20 of image data values ​​(Figures 5, 7). The histogram 20 for each pixel defines a total number N (N is a positive integer) of reference images used to generate the histogram 20, an average value µ of the image data values ​​for the total number N, and a standard deviation σ of the image data values ​​for the total number N.

[0048] The reliability of the initial model 10 can be defined as either its under-enclosure or over-enclosure capabilities. The initial model 10 is over-enclosed to the point that it cannot identify an unacceptable (or fatal) defect in one of the candidate grains examined and compared with the initial model 10. The initial model 10 is under-enclosed to the point that it identifies an acceptable grain examined and compared with the initial model 10 as unacceptable due to a fatal defect that does not actually exist.

[0049] The viewpoints disclosed herein can be used to improve the reliability of the initial model 10 by dynamically updating the model 10 to generate the (etc.) updated model 14.

[0050] To reduce the impact of variability in external properties of the model, the initial model 10 is updated with new data. This new data may be image data with known acceptability (e.g., the corresponding grains are known to be acceptable or unacceptable), and possess one or more external properties that distinguish the new data from the initial model 10. As a result, by updating the initial model 10 with the new data, the updated model 14 becomes more reliable than the initial model.

[0051] In particular, this new data can be used to dynamically modify the total, mean, and / or standard deviation and critical values ​​of the initial model, thereby modifying the acceptable range to a more reliable range for detecting defects based on the updated inspection tool or procedure. For example, for each pixel, the acceptable range can be defined between upper and lower acceptable thresholds, and it can itself be correlated with the standard deviation or some other statistical measure.

[0052] New data used to update model 10 to an updated model 14 may be supplied from any of a number of sources. One exemplary source of such new data is a new full-grain reference image 16 known to represent an acceptable grain but having one or more external properties not yet considered in the initial model 10.

[0053] Another exemplary source of this new data is a candidate defect image 18. The candidate defect image 18 may be an image of a portion of a die for acceptability checking using an existing model (i.e., using the initial model 10 or the updated model 14). Pixels of the image 18 are compared with corresponding pixels of the full die model, such that the portion of the die captured in the image 18 is aligned and mapped to the corresponding portion of the full die model. Furthermore, the candidate defect image 18 itself can be used to update the model to an updated model 14. In some examples, the candidate defect image 18 is used to update the model only if it determines (e.g., using models 10, 14) that the portion of the die represented in the image 18 is acceptable. In other examples, the candidate defect image 18 is used to update the model regardless of whether the portion of the die represented in the image is determined to be acceptable.

[0054] The model can therefore be updated each time it receives a new complete grain reference image 16 or a new candidate defect image 18. This can occur periodically, for example, at least once per hour, at least once per minute, or at least once per second, or more frequently or less frequently, providing the high dynamic thresholds of the models 10, 14. The latest version of the model can then be used to perform defect detection and defect evaluation to provide an updated model output 19 for the next candidate grain. That is, the image data obtained from a candidate grain is compared with the corresponding allocation of the latest iteration of the image data of the complete grain models 10, 14, and based on the layout of the candidate grain data relative to the allocation, the computing device 202 (FIG. 1) determines whether it is an acceptable grain (if the candidate grain image falls within the acceptable range of the data allocation) or an unacceptable grain (if the candidate grain image falls outside the acceptable range of the data allocation).

[0055] Figure 7 schematically illustrates an example method for updating model pixel data according to the method in Figure 6.

[0056] The model update pertains to the display of image data for a single pixel of the model, corresponding to a specific location within a grain. New data 22 is used to modify the composite pixel data allocation 20 (or the initial specific pixel image histogram) corresponding to one of those pixels in the initial model, to generate an updated composite pixel image data histogram 24 for that data. Due to this added data, such as a different total, a different mean, and / or a different standard deviation, the updated pixel image data 24 may have different statistical properties.

[0057] The shape of the allocation changes (e.g., skews) from the composite pixel data 20 to the composite pixel data 24. This shape change is caused by importing the pixel data from the updated images. Some or all of the updated reference images correspond to acceptable grains or grain portions. In some examples, some of the updated reference images may correspond to unacceptable grains or grain portions.

[0058] These composite pixel data assignments 20 and 24 can be any visual or other type of data assignment. One example of the assigned pixel data 20 and 24 can be luminance. In the case of assignment 20, the pixel data is derived from an initial reference image, and in the case of assignment 24, it is an assignment of individual data points derived from the initial plus updated reference image.

[0059] In some examples, each allocation corresponds only to acceptable grains. In other examples, each allocation includes a subset of allocations corresponding to one set of acceptable grains and a subset of allocations corresponding to one or more unacceptable grains. For example, in the initial allocation 20 falling between the upper and lower thresholds 31, or in the updated allocation 24 falling between the upper and lower thresholds 33, acceptable grains or grain portions may correspond to grain data.

[0060] As described above, the new data 22 can cause a skew in the data allocation and / or a shift of one or more of the critical values, such that a matrix, with respect to the initial allocation, falls outside the range of the critical values ​​31, and is therefore deemed unacceptable according to the initial model; with respect to the updated allocation, it falls within the range of the critical values ​​33, and is therefore deemed acceptable according to the updated model. Conversely, the new data 22 can cause a shift of one or more of the critical values, such that a matrix, with respect to the initial allocation, falls within the range of the critical values ​​31, and is therefore deemed acceptable according to the initial model; with respect to the updated allocation, it falls outside the range of the critical values ​​33, and is therefore deemed unacceptable according to the updated model.

[0061] The data values ​​corresponding to these threshold values ​​31 and 33 can be selected to optimize the overall system (e.g., maximizing the reliability of the matrix acceptability determination method while minimizing processing time and resources). For example, the range between the threshold values ​​can be selected to correspond to one (or two or more) standard deviations of the peak values ​​of the allocation, wherein unacceptable grains or grain portions correspond to pixels falling outside the acceptable range of the allocation.

[0062] The computing device 202 (Figure 1) detects and classifies defects in a candidate substrate or determines whether the candidate substrate is acceptable by comparing image data of the candidate substrate with a threshold value of the image data allocation of corresponding pixels in the current iteration of the model. In some examples, the defect determination and classification (or non-determination) can be performed based on a pixel-by-pixel comparison between the candidate defect image and the current iteration of the full grain model.

[0063] The initial modeling and updates of this disclosure can be computationally efficient due to dynamic model updates, such as a continuous or near-continuous model update operation. Therefore, for example, the model does not need to be recalibrated every time a new inspection device is used or some other external property changes. For example, the model does not need to be recalibrated when the light intensity at a grain inspection site changes due to an upgrade or recalibration of an inspection tool.

[0064] Furthermore, in some paradigms, limited data (or data types) are tracked to perform dynamic modifications to the model as described herein, as well as defect detection and non-detection, which can further increase computational efficiency. This type of data can be limited to, for example, the mean, the standard deviation, and the total, without requiring or needing to process calculations of other data from one or more images of the candidate matrix.

[0065] Furthermore, because the model is dynamically updated to take into account the variability in external attributes, the testing of such external attributes (e.g., background lighting test, noise test, high contrast test, low contrast test, etc.) can be reduced or eliminated, simplifying the overall procedure for detecting and classifying defects and determining the acceptability of the substrate.

[0066] Additional advantages of the embodiments disclosed herein include shorter substrate inspection times. Shorter inspection times can be due to, for example, simplified inspection procedures with fewer steps and tests.

[0067] Because improvements do not actually affect the variability of substrate acceptability inspection, an additional advantage of the embodiments of this disclosure includes greater reliability in the determination of fatal defects in the substrate. This can reduce unnecessary repairs and / or replacements of substrate assembly and inspection tools, reduce unnecessary discarding of acceptable substrates, and increase the output of acceptable substrates.

[0068] Figure 8 shows another exemplary method 300 for dynamically modeling matrix acceptability determination in accordance with the present disclosure.

[0069] In some embodiments, not all steps of method 300 are required. In some embodiments, one or more steps of method 300 may be performed in a different order than those shown. The steps of method 300 may be performed by the computing device 202 (FIG. 1) of system 100.

[0070] In step 302 of method 300, image data is received by the computing device 202. This image data may include a full-die reference image and / or partial-die reference images. In some examples, these partial-die reference images may be concatenated to generate a full-die reference image. Each reference image may be marked to indicate whether the corresponding die or die portion is acceptable or unacceptable. These reference images may be obtained on a bare substrate or in a device in which they are incorporated, such as a substrate in an LED.

[0071] In step 304 of method 300, the initial defect detection model is constructed by incorporating pixel-by-pixel image data from the complete grain reference image. The generated model is a complete grain model consisting of composite pixel data allocated to each pixel of the grain, which has known acceptability or unacceptability.

[0072] In step 306 of method 300, the new image data is received by the computing device 202. The new image data may be derived from a candidate defect image or from a new complete grain reference image. The new image data may be marked to indicate whether the corresponding grain or grain portion is acceptable or unacceptable. The new image data has one or more external properties that differ from the corresponding external properties of the image data received in step 302. The new image data may be obtained from a bare substrate or a device in which the substrate is incorporated, such as a substrate in an LED.

[0073] In step 308 of method 300, the model is updated using the received new image data. The updated model may have an adjusted image data allocation for one or more pixels of the die, indicating an adjustment range or threshold for the acceptability or unacceptability of the substrate (or a device incorporated into the substrate, such as an LED) due to the data allocation skew caused by the new image data.

[0074] These steps 306 and 308 can be repeated one or more times in an iterative process that updates the model and keeps it up-to-date by changing the external properties of the substrate inspection device and procedure.

[0075] In step 310 of method 300, a candidate substrate is examined and imaged using an imaging device. The candidate substrate is one from which a determination is made as to whether the substrate is acceptable or unacceptable and should be discarded. The candidate substrate may be a bare substrate or a substrate already incorporated into a device, such as a substrate in an LED.

[0076] In step 312 of method 300, the candidate substrate image data is compared with a complete grain model of composite pixel data, for example, based on a pixel-by-pixel correspondence, to determine the position of the candidate substrate image data in the allocation of image data for each composite pixel of the complete grain model. It determines whether the candidate substrate image data for each pixel falls within the acceptable range (and is therefore acceptable) defined by the allocation of the corresponding composite pixel of the complete grain model, or falls outside the range (and is therefore unacceptable).

[0077] In step 314 of method 300, the candidate substrate (and in some cases, the device into which the candidate substrate is incorporated (e.g., an LED)) is classified based on the pixel-by-pixel results of the comparison in step 312. For example, based on the number and placement of pixels determined to be unacceptable in step 312, the overall candidate substrate or device, or at least the grains of the candidate substrate, are classified as acceptable or unacceptable. In some examples, this additional classification may be generated for a given detected defect based on the comparison in step 312. For example, certain defect types may be known to have certain signatures or fingerprints in their image data. The computing device 202 (FIG. 1) can access a library of such signatures and fingerprints, and by comparing the pixel data of a candidate substrate that detects a defect with the library, the defect type can be determined by the computing device 202, the tool that generates the defect, the manufacturing process in which the defect occurs, etc. All information regarding the detected defect, including whether it makes the die acceptable or unacceptable, can be provided to one of the users of the computing device 202 via a user interface.

[0078] If the candidate substrate (or device incorporated into the candidate substrate) is determined to be unacceptable, an alert may be issued regarding this impact (e.g., via a user interface), and the candidate substrate or device may be discarded, and any indicated repairs or tool replacements may be performed.

[0079] If the candidate substrate (or the device incorporated into the candidate substrate) is determined to be acceptable, the substrate may proceed to the next production or distribution step.

[0080] In some embodiments, image data from the candidate substrate image is used to further update the model, particularly if the inspection equipment or procedure for the candidate substrate includes one or more different external attributes.

[0081] While specific embodiments have been described herein, the scope of this technology is not limited to these specific embodiments. Furthermore, when different examples and embodiments can be described separately, such examples and embodiments can be combined with each other to implement the technology described herein. Those skilled in the art will recognize that other embodiments or improvements are possible within the scope and spirit of this technology. Therefore, the specific structure, action, or medium described is disclosed as an example only. The scope of this technology is defined by the following claims and any of their equivalents.

[0082] 10: Initial Fully Grained Model 12: Reference Image 14: Updated Full Grain Model 16: Full-grain reference image 18: Candidate Defect Images 19: Updated model output 20: Composite pixel data (allocation), histogram 21: pixels 22: New Information 23: Composite Pixel 24: Updated composite pixel image data (histogram) 25: pixel data 27: Single data point 31, 33: Critical values 100: System 102: Inspection device 202: Computing device 204: Memory 206: Computing Components 208: Processor 210: Portable storage device 212: Non-removable storage 214: Output device 216: Input device 218: Communication Connection 300: Demonstration Method 302, 304, 306, 308, 310, 312, 314: Steps N: Total µ: average value σ: Standard deviation

Claims

1. A computer-executable method for determining the acceptability of a substrate, comprising: generating a model representing image data obtained from a reference image of a first substrate, each reference image including a first pixel relative to each of other reference images, the model defining an acceptable range of image values ​​for each first pixel; receiving new image data obtained from a second pixel of an image of a second substrate; (i) identifying an imaging factor that does not affect the acceptability of the second substrate; (ii) determining that a second image value defined by the second pixel is affected by the imaging factor; and updating the model based on (i) and (ii) to generate an updated model, including modifying the acceptable range of image values ​​based on the second image value to generate a modified range.

2. The computer-executable method of claim 1, wherein each reference image includes image data representing a complete grain of one of the first substrates, the first substrates being a semiconductor substrate.

3. The computer-executable method of claim 1 or 2, wherein the new image data represents a complete grain of one of the second substrates, the second substrate being a semiconductor substrate.

4. The computer-executable method of claim 1 or 2, wherein the new image data represents a complete grain smaller than one of the second substrates, the second substrate being a semiconductor substrate.

5. The computer-executable method of claim 1 or 2, wherein the acceptable image value range is defined between a first threshold and a second threshold, and wherein modifying the acceptable image value range includes moving at least one of the first threshold and the second threshold.

6. The computer-executable method of claim 1 or 2 further comprises: determining whether a defect exists in the third substrate by comparing image data representing at least a portion of a third substrate with the updated model.

7. The computer-executable method of claim 6, wherein one of the defects in the third matrix is ​​detected, the method further comprising: classifying the defect as acceptable or unacceptable.

8. A computer-executable method as requested in claim 1 or 2, wherein the model defines a plurality of images, an average image data value of one of the plurality of images, and a standard deviation of the image data values ​​of the plurality of images for each of the updated models.

9. The computer-executable method of request item 8, wherein the standard deviation defined by the model defines the acceptable range of image values, and the standard deviation defined by the updated model defines the modification range.

10. The computer-executable method of claim 1 or 2, wherein different images of the reference images are acquired using different imaging devices.

11. The computer-executable method of claim 10, wherein the imaging device used to acquire the new image data is different from or calibrated differently from all imaging devices used to acquire the reference images.

12. The computer-executable method of claim 1 or 2, wherein when the new image data is acquired, a state of environmental conditions presented at the second substrate is different from a state of environmental conditions presented at the first substrate when each of the reference images is acquired.

13. The computer-executable method of claim 1 further comprises: receiving another new image data obtained from a third substrate; and updating the model again to generate another updated model, including modifying the acceptable range of image values ​​again based on the other new image data to generate another modified range.

14. A computer-executable method as claimed in claim 1 or 2, wherein the model includes a plurality of data assignments, each data assignment being relative to composite pixel data from a corresponding pixel of one of the reference images.

15. The computer-executable method of request 14, wherein each data allocation includes one allocation of pixel brightness.

16. A computer-executable method as in request 1 or 2, wherein at least some of such first bases are known to be acceptable.

17. A computer-executable method such as request 1 or 2, wherein the second base is known to be acceptable.

18. The computer-executable method of request item 1 further includes: using the modified scope to determine whether a defect exists in a third matrix.

19. The computer-executable method of claim 18, wherein the third base system is incorporated into a light-emitting diode (LED) device.

20. The computer-executable method of claim 19, wherein the first base system is incorporated into a light-emitting diode (LED) device.

21. A computer-executable method as claimed in claim 1, wherein (i) includes determining that the imaging factor is external to the second substrate.

22. A computer-executable method for determining the acceptability of a substrate, comprising: generating a model representing image data acquired from a reference image of a complete die of a first semiconductor substrate, each reference image including pixels relative to each of other reference images, the model defining an acceptable range of image values ​​for each pixel; receiving new image data acquired from a second pixel of a complete die image of a second semiconductor substrate; (i) identifying an imaging factor that does not affect the acceptability of the second semiconductor substrate; (ii) determining that a second image value defined by the second pixel is affected by the imaging factor, the imaging factor not affecting the acceptability of the second semiconductor substrate; (iii) determining that the imaging factor is different when the complete die image of the second semiconductor substrate is acquired and when the reference images are acquired; and updating the model based on (i), (ii) and (iii) to generate an updated model, including modifying the acceptable range of image values ​​based on the new image data to generate a modified range.

23. The computer-executable method of claim 22 further comprises: determining that one of the grains of the second semiconductor substrate relative to the complete grain image of the second semiconductor substrate is acceptable.

24. The computer-executable method of claim 22 or 23, wherein at least one of the reference images of the complete grain of one of the first semiconductor substrates includes a plurality of interconnected images of a portion of the complete grain.

25. The computer-executable method of claim 22 or 23 further comprises: determining whether the third substrate is acceptable or unacceptable by comparing image data representing at least a portion of a third substrate with the updated model.

26. The computer-executable method of claim 18, wherein (i) includes determining that the imaging factor is external to the second semiconductor substrate.

27. The computer-executable method of request item 18 further includes: using the modified scope to determine whether a defect exists in a third matrix.

Images