JP2013068633A - Automated wafer defect inspection system and method for executing inspection - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an automated inspection system and a method for replacing a present manual inspection process.SOLUTION: An automated defect inspection system (10) is used for inspecting a patterned wafer, all wafers, a damaged wafer, a partial wafer, a waffle pack, MCM, and so on. The inspection system is intended for second optical wafer inspection and designed especially to inspect metallization defects such as scratch, void, corrosion and bridging, and defects such as a diffusion defect, a coat protection layer defect, a writing defect, a glass insulation defect, chip and crack from a notch, a solder projection defect, and a bond pad area defect.

Description

(Refer to related applications)
This application claims the benefit of US Provisional Patent Application Nos. 60/092923 and 60/092701, filed July 15, 1998.

(Background of the Invention)
(Technical field)
The present invention relates to a defect inspection system for the semiconductor industry. More particularly, the present invention relates to patterned wafers, all on a film frame, such as JEDEC trays, Auer boats, gel packs or waffle pack dies, multichip modules often referred to as MCMs, etc. For automated defect inspection systems for wafers, saw wafers, specifically metallization defects (scratches, voids, corrosion, bridging), diffusion defects, passivation layer defects, scribing Second optical wafers for defects such as defects, glass insulation defects, chip and cracks from notches, and bumps or bond pad area defects (eg, gold or solder bump defects or similar interconnect defects). Is intended for testing, it is designed. Specifically, the present invention is an automated defect inspection system, such as an integrated circuit, an LCD panel embedded with a photolithographic network, where the system is used as follows: Trained by observing multiple known good dies under the image head, resulting in a good die model, inspection recipes are entered into the system to define inspection parameters, dies are loaded, aligned and good Defect inspections can be performed where the defect is observed by the image head in comparison to the correct die model, and any assessment of the identified defects can be made, and the user accepts or outputs the report as necessary. obtain.

(Background information)
Over the past decades, semiconductors have grown exponentially in use and generality. Semiconductors, in effect, introduced the advancement of computers, electronics, and in general, by transforming many previously difficult, expensive and / or time consuming mechanical processes into simplified electrical processes that are simplified, Changed society. This boom in semiconductors is related to computers and electronics, more specifically whether it is in assembly lines, in laboratory test facilities, on desktop computers, or in home electronics and toys. Inspired by the insatiable desires of business and individuals about computers and electronics that are faster and more advanced.

  Semiconductor manufacturing has made significant improvements in the quality, speed, and performance of the final product and in the quality, speed, and performance of the manufacturing process. However, there is a continuing demand for faster, more reliable and higher performance semiconductors.

  One process that has evolved over the past decade is the semiconductor inspection process. The advantage of inspecting semiconductors throughout the manufacturing process is not only that defective wafers are processed to completion either by final inspection or due to failure in use, but there are various defects. This is obvious in that it can be removed in the process.

  A typical example of a semiconductor manufacturing process is summarized as follows. Bare whole wafers are manufactured. Thereafter, a network is fabricated on the bare whole wafer. The entire wafer with circuitry is then cut into smaller pieces known in the industry as dies. The dies are then processed into waffle packs and / or gel packs, well known in the art, or as substrate dies.

  It is well known today that various inspection processes are performed during this semiconductor process. Bare wafer inspection can be performed on bare whole wafers before the deposition of any layer that always forms a network, rather than after first being made from sand and / or after a long time after cleaning the wafer. . Defects that are inspected during bare wafer inspection include surface particles and surface imperfections or irregularities.

  During layer deposition (this is circuit construction), one or more first optical inspections can be performed on the entire wafer. The first optical inspection is an “in process” inspection of the wafer during network fabrication. This first inspection can be at less frequent intervals after each layer is deposited, or only once, during or after all depositions. This first optical inspection is usually at the submicron level in the range of 0.1 microns to <1 micron. This process is used to check for mask alignment or defects such as excess metal, missing metal, contaminants, etc. This first inspection is performed during the development of the wafer network.

  Once the entire wafer is at least completely deposited (all circuitry is fabricated thereon), the post first (or 1.5) inspection is performed on the fully processed wafer. In general, this is before the deposition of the passivation layer, but this is not necessary. Further, this post first inspection is generally prior to electrical inspection or probing of the entire wafer. This inspection is typically an optical inspection of 0.5 microns to 1 micron.

  One or more second optical inspections are performed after the entire wafer has been completely processed. The front end second optical inspection is performed immediately before or after this probing or electrical test to determine device quality, if probing is required after the entire wafer has been fully processed. Done. The back end second optical inspection picks up the die while applying ridges to the die or wafer, during or after cutting the wafer into the saw, and during or after dicing the wafer. , During placing in other packages such as trays or waffle packs or gel packs, or thereafter at various stages such as during or after placing the wafer on the substrate. This second optical inspection is typically at the 1+ micron level and is typically metal film defects (eg, scratches, voids, corrosion, and bridging), diffusion defects, passivation layer defects, scribing defects, glass Insulation defects, chips and cracks from notches, and probe or bond pad area defects.

  After the actual packaging, a third optical inspection is performed. This packaging includes at least one of the following: placing the die on the substrate, wire bonding the die, connecting the lead wires, connecting the ball to the flip chip, etc. At this point, inspection includes inspecting the ball grid array, lead linearity, wire bonding, ink marking, and defects for any package (eg, chips, cracks and voids). This third level of inspection is generally at the 5+ micron level.

  The focus of this semiconductor inspection industry is bare wafer and first optical inspection. Many industry leaders have developed, patented, manufactured and marketed first optical inspection systems to perform these inspections, including ADE, KLA, Tencor, Inspex, Applied, Orbit, etc.

  Often, this device is very expensive and large. In the first inspection stage, this cost and machine size issue is not as important as in the later inspection stage. Because only a relatively small number of organizations manufacture silicon wafers and therefore buy bare or cut wafers, and compare them to the vast majority of companies that process them into final chips, it is possible to inspect bare wafers. Because it is necessary. These are often expensive, large inspection devices that are not justified for smaller stores, etc., satisfy the needs in the second and third stages, and a very large number of smaller companies that finish process wafers. However, a more economical inspection device is required.

  To a lesser extent, some funds are spent on the third optical inspection, and several companies, including STI, View Engineering, RVSI, and ICOS, have developed systems for this purpose and are marketing these systems. .

  However, none of these systems address the specific and unique constraints of the second optics, and this area is largely ignored. In practical applications, the second optical inspection is performed slightly by manual inspection using a person and a microscope device. This manual process is inaccurate due to various factors, including operator perception, eye fatigue, fatigue, and different perceptions by different operators about the importance of discovery. In addition, smaller circuit shapes and higher throughput requirements are increasing the demands on semiconductor inspection at this second optical level, all of which further results in operator stress, eye fatigue, and sometimes lower quality. Become.

  In addition to the second optical inspection stage for the need for inspection for metal coating defects (scratches, voids, corrosion and bridging), diffusion defects, passivation layer defects, scribing defects, glass insulation defects, chips and cracks derived from cuts, etc. Uplift has become more important recently. This is due to the recent trend in the use of raised interface connections or flip-chips rather than leads, expanding the importance of second optical inspection, and thus there is a need for devices and systems that go beyond manual inspection.

(Object and summary of the present invention)
It is an object of the present invention to provide an automated inspection system that replaces current manual inspection processes.

  It is a further object of the present invention to provide a new state of the second optical inspection system technology.

  It is a further object of the present invention to provide an automated defect inspection system such as pattern wafers, whole wafers, cut wafers, JEDEC trays, Auer boats, gel pack or waffle pack dies, MCMs and the like.

  It is a further object of the present invention to provide an automated defect inspection system specifically intended for and designed for the second optical wafer inspection.

  Metal film defects (scratches, voids, corrosion, bridging), diffusion defects, passivation layer defects, scribing defects, glass insulation defects, chips and cracks derived from notches, probe area defects, raised area defects and / or bond pad area defects It is a further object of the present invention to provide an automated defect inspection system for inspecting such defects.

  It is a further object of the present invention to provide an automated defect inspection system that eliminates or significantly reduces the need for manual microscopy of each die on each wafer.

  It is a further object of the present invention to provide an automated defect inspection system that observes even smaller circuit shapes in an accurate and quick manner.

  It is a further object of the present invention to provide an automated defect inspection system that provides higher throughput than manual inspection.

  It is a further object of the present invention to provide an automated defect inspection system that provides improved inspection quality and consistency.

  It is a further object of the present invention to provide an automated defect inspection system that provides improved process control.

  It is a further object of the present invention to provide an automated defect inspection system that has inspection recipes and that creates, copies, and edits such recipes to customize the system for user inspection requirements.

  It is a further object of the present invention to provide an automated defect inspection system that uses digital images to perform semiconductor wafer observations.

  It is a further object of the present invention to provide an automated defect inspection system that is trained by inspecting a good die so that once trained, the system detects changes from what it has learned.

  It is a further object of the present invention to provide a trainable automated defect inspection system.

  It is a further object of the present invention to provide an automated defect inspection system that develops a good die model and uses this model to inspect an unknown quality die.

  It is a further object of the present invention to use an excellent die model to provide an automated defect inspection system that includes a “excellent die” training process and a defect inspection process.

  It is a further object of the present invention to provide an automated defect inspection system that includes a “good die” training process, an inspection recipe creation process, and a defect inspection process.

  It is a further object of the present invention to provide an automated defect inspection system that includes a “good die” training process, an inspection recipe creation process, a defect inspection process, a defect evaluation process, and a report issuance or output process.

  It is a further object of the present invention to provide an automated defect inspection system that provides multi-dimensional adjustment of each wafer, substrate, or other device comprising a die to be inspected so that each die is uniformly positioned. is there.

  Providing an automated defect inspection system that provides x, y, theta (θ) adjustment of each wafer, substrate, or other device comprising a die to be inspected so that each die is uniformly arranged It is a further object of the present invention.

  It is a further object of the present invention to provide an automated defect inspection system that provides coarse adjustment, fine adjustment, and / or focusing of each wafer.

  It is a further object of the present invention to provide an automated defect inspection system that provides a “good die” modeling by observing a number of good dies and then developing the model.

  It is a further object of the present invention to provide an automated defect inspection system that provides defect inspection using an image head or camera for observing statically and properly aligned dies.

  It is a further object of the present invention to provide an automated defect inspection system that provides defect inspection using an image head or camera to observe dies that are still properly aligned while dynamically or moving.

  It is a further object of the present invention to provide an automated defect inspection system that provides defect inspection using an image head or camera to observe still properly aligned dies while dynamically or moving, where The strobe illumination is used to take a static picture of a die that moves dynamically.

  It is a further object of the present invention to provide an automated defect inspection system that provides an evaluation of a system in which defects have been detected, thereby eliminating the need for the user to view every die or every part of a die, instead Just observe the marked or noticed defects.

  It is a further object of the present invention to provide an automated defect inspection system that provides a method for accounting for die position or spacing drifting or irregularities.

  It is a further object of the present invention to provide an automated defect inspection system that provides a means for inspecting extended film frames in which dies are randomly spaced, rotated, drifted, etc. .

  It is a further object of the present invention to provide an automated defect inspection system that provides a method for measuring size, position, shape, shape and other features such as solder bumps, gold bumps, bond pads and the like.

  It is a further object of the present invention to provide an automated defect inspection system that provides a method for inspecting the quality of gold bumps, solder bumps, interconnects, etc., or probe marks on bond pads.

  It is a further object of the present invention to provide an automated defect inspection system that provides a method for detecting defects in bond pads, bumps or interconnects.

  Still other advantages and benefits of the present invention will become apparent to those of ordinary skill in the art upon reading and understanding the following summary and detailed description.

  The present invention thus satisfies these and other objectives, which relate to automated inspection equipment, systems and processes. Specifically, the present invention relates to full pattern wafers, cut wafers, broken wafers, partial wafers, and any type of wafer, die, gel pack die, waffle pack die on a film frame, multichip often referred to as MCM An automated method for inspecting any form of semiconductor wafer, including modules, JEDEC trays, Auer boats, and other wafer and die package configurations for detection, which includes (1) a plurality of known good wafers Training the model for good wafer parameters by optical observation, and (2) inspecting unknown quality wafers using this model.

  For example, the present invention provides the following items.

(Item 1) Automated method, all patterned wafers, notched wafers, damaged wafers, partial wafers, and any kind of wafer on film frame, die, gel pack die, waffle pack die, many called MCM An automated method for inspecting semiconductor wafers in any form for defects, including chip modules, JEDEC trays, Auer boats, and other wafer and die package shapes, which include the following steps:
Training the model for good wafer parameters via optical observation of a plurality of known good wafers; and inspecting an unknown quality wafer using the model;
Including the method.

  (Item 2) The automated inspection method according to item 1, wherein the step of training the model includes a step of aligning and observing a plurality of known good wafers via a camera.

  (Item 3) The automated inspection method according to item 2, wherein the step of training the model includes a step of performing gray scale measurement of each pixel in a lattice of pixels of each of the plurality of known good wafers.

  (Item 4) The automated inspection method according to item 3, wherein the step of training the model further includes the step of measuring the gray scale of each pixel from 0 to 255 scale.

  5. The method of claim 3, wherein training the model further comprises calculating an average and standard deviation of all the grayscale measurements for each pixel in the plurality of known good wafers. Automated inspection method.

  (Item 6) The step of calculating the average and standard deviation of all the grayscale measurements for each pixel of the plurality of known good wafers provides upper and lower limits, between which wafers of unknown quality Item 5 where the grayscale measurement at the equivalent pixel in is considered good, while the grayscale measurement at the equivalent pixel in the unknown quality wafer outside the upper and lower limits is at least of questionable integrity. The automated inspection method described in 1.

  (Item 7) The step of training the system as to whether test parameters, which are values indicating how much standard deviation is away from the average, are acceptable in defining the upper and lower limits. The automated inspection method according to item 6, further comprising:

  8. The automated inspection method of claim 1, further comprising the step of training the system with respect to wafer adjustment functions and parameters.

  (Item 9) Train the system on at least one of the following: wafer shape, wafer size, die shape on wafer, die size on wafer, number of die rows on wafer, number of die columns on wafer The automated inspection method according to item 1, further comprising a step of:

  (Item 10) The automated inspection method according to item 1, further comprising the step of training the system regarding inspection parameters.

  (Item 11) At least the following: defining how the substrate is selected from the containment, defining how the die on the wafer is selected for defect inspection, and if necessary For example, further comprising creating an inspection recipe to be used during inspection to perform one of the steps of defining how an inspection map is input or output. Item 2. The automated inspection method according to item 1.

  (Item 12) The method according to item 1, further including a step of defect evaluation following completion of the inspection, whereby only dies that do not fit a good wafer model are evaluated.

(Item 13) An automated system for all pattern wafers, cut wafers, damaged wafers, partial wafers, and any type of wafer, die, gel pack die, waffle pack die on the film frame, often referred to as MCM An automated system for inspecting a substrate, such as a chip module, JEDEC tray, Auer boat, and any form of wafer, including other wafer and die package shapes, for defects, including:
Wafer inspection plate;
Means for providing a wafer to the inspection plate;
A wafer conditioning device for aligning each and every wafer provided on the test plate at the same x, y, z and θ positions;
Visual inspection device for visual input of multiple known good quality wafers during training and for visual inspection of other unknown quality wafers during inspection, and
A computer system that provides processing and memory capabilities to develop a good quality wafer model by taking a grayscale measurement for each pixel on a plurality of known good quality wafers. A computer system that calculates and averages and standard deviations for each pixel location therefrom, and compares other unknown quality wafers relative to the model;
A system comprising:

  (Item 14) The automation system according to item 13, further comprising a focusing mechanism for focusing an image of at least a part of the wafer so as to be seen by a camera.

  (Item 15) The automation system according to item 13, wherein the visual inspection device is a camera.

  (Item 16) The automation system according to item 13, wherein the visual inspection device is a CCD camera.

  (Item 17) An item further comprising a device for entering parameters for entering parameters and other constraints or information, including sensitivity parameters, shape, die size, die shape, die pitch, number of rows, and number of columns 13. The automation system according to 13.

  (Item 18) The automation system of item 13, further comprising a display for displaying the scene viewed by the visual inspection device at the present time or at a previously stored time.

  (Item 19) The automation system according to item 13, further comprising a marking head for placing a mark on or adjacent to a pixel of the wafer outside the model.

  (Item 20) Further comprising a frame and a hood, on which the wafer inspection plate, the wafer adjustment device, the visual inspection device, and the computer system are mounted, and the hood defines an inspection area on the inspection plate. 14. The automation system of item 13, which is for enveloping.

FIG. 1 is a perspective view of one embodiment of the system. FIG. 2 is a flowchart of the entire process. FIG. 3 is a more detailed flowchart of one step of the process shown in FIG. FIG. 4 is a more detailed flowchart of one step of the process shown in FIG. FIG. 5 is a more detailed flowchart of one step of the process shown in FIG. FIG. 6 is a more detailed flowchart of one step of the process shown in FIG. FIG. 7 is a more detailed flowchart of one step of the process shown in FIG. FIG. 8 is an overall perspective view of a system similar to that shown in FIG. 1 at different angles. FIG. 9 is a front view of the wafer top plate and optics. FIG. 10 is a left front perspective view of a portion of an inspection station including a wafer top plate and optics. FIG. 11 is a right front perspective view of the top of the inspection station. 12 is a right side perspective view of the top of the inspection station shown in FIG. FIG. 13 is an enlarged view of the optics and wafer top plate. 14 is a side view of the wafer top plate and optics of FIG. FIG. 15 is a left side perspective view of the inspection station shown in FIGS. FIG. 16 is an enlarged view of the wafer top plate and the x, y, and θ adjusters. FIG. 17 is a partial perspective view of the top portion of the inspection and wafer handling station. FIG. 18 is an enlarged view of the wafer handling and wafer top plate portion of the present invention. FIG. 19 is a side view of the wafer handling station. FIG. 20 is a partial view of the dark field option of the present invention. FIG. 21 is an enlarged view of a dark field laser as a dark field option.

(Description of Preferred Embodiment)
The automated defect inspection system of the present invention is generally indicated at 10, which is best shown generally in FIGS. 1 and 8 (detailed in FIGS. 2-7 and 9-21) and in one embodiment. And used to find die defects on the pattern wafer W, as well as all wafers, notched wafers, broken wafers, film frames, gel pack dies, waffle pack dies, MCMs, JEDEC trays, Auer boats, and Other uses are contemplated, including inspecting any type of wafer as well as die package configuration of other wafers (however, all of these uses will be generally referred to hereinafter as inspection of wafer W). The system inspects many types of defects including (but not limited to): metal film defects (scratches, voids, corrosion, bridging, etc.), diffusion defects, passivation layer defects, scribing defects, glass insulation. Defects, chips and cracks derived from notches, probe or bond region defects (eg, loss of probe marks, discoloration, metal loss and probe bridging), poor diffusion, vapox, etc. This system is also used in addition or alternatively to inspect interconnects or bumps (gold or solder bumps) for defects or other features such as size and shape.

  The system and process generally includes a system training step (step A), creating an inspection recipe (step B), inspecting a die or wafer based on this training and recipe, as shown in FIG. (Step C), if desired, includes a multi-step process of defect evaluation (Step D) and, if desired, a defect reporting step (Step E). A system 10 for performing this purpose generally includes a wafer test plate 12, means shown as 14 for providing a wafer to the test plate, the same x, y, and θ positions or x, y, z and Wafer adjustment device 16 (xy-theta or xyz-theta aligner), focus mechanism 18, camera 20 or a good die during training to align each wafer at theta position. Other visual inspection devices for visual input and for visual inspection of other unknown quality dies during inspection, parameters and sensitivity parameters, shape, die size, die shape, die pitch, Parameter input device 22 for entering other constraints or information such as number of rows, number of columns, current or any previously saved period, depending on the camera Processing and memory to store display 24, computer system 26 or entered good die from which to see the field of view seen, develop models from them, and compare and analyze other dies in comparison to the model Other capable computer-like devices, marking head 28, frame 30, hood 32, control panel 34, and system parameter display 36 are provided.

  More specifically, the system 10 and associated processes are as follows. The training process (step A) initially shown in FIG. 2 and shown in more detail in FIG. 3 includes (1) defining and / or training the adjustment functions and parameters in the computer system 26 for use during training. Process (and process of storing) (where all this is shown in process A1), (2) wafer and / or die shape, wafer and / or die size, die pitch, number of rows, number of columns Storing all such information in the computer system 26 (where all this is a process) for use during the process of defining (and entering into the computer system) and training and / or testing. (3) device 1 to define how the ideal die will look based on the normal features observed. Train the system about what the “good die” contains by observing a plurality of known good dies with the camera 20 and forming a model in the computer system 26 (where this (All shown in step A3), (4) how close an unknown quality die must match a good die model that is considered good (ie what difference from the exact model is still good) Setting inspection parameters that are values that indicate to the computer system 26 what is acceptable to be considered a die (where all this is shown in step A4), (5) this training model and its function , Storing parameters and the like in computer system 26 (shown in step A5).

  The step of creating an inspection recipe (step B) includes the step of creating a new recipe (the step of creating B is skipped if a previously defined recipe is to be used). The process of creating a new recipe includes (1) defining how the wafer W is selected from a cassette or other containment (herein all shown as process B1), (2) A process that defines how the dies on each wafer W should be selected for defect inspection (herein this is shown in its entirety as process B2) (often the dies are in sequential or similar order) (3) A mandate that specifies how the defect inspection map file is input and output (where this is step B3). And (4) storing this recipe (where this is step B4).

  The inspection process (process C) shown as defect inspection includes (1) a process of inputting a wafer identification code (if desired), and a process specified as process C1, and (2) a process of selecting a recipe defined in process B (Here, this selection step is step C2), (3) Product setting selection step and input step (this is step C3), (4) Wafer test using wafer providing means 14 Loading wafers onto plate 12 (where the loading step is step C4), (5) adjusting each wafer to the same x, y, and θ positions or x, y, z, and θ positions. Using the wafer conditioning device 16 to align the wafers of the wafer test plate 12 using the defined and / or trained process A1 adjustment functions and parameters (all of which are processes). (Shown as C5), (6) if desired, focus the camera 20 on the wafer W (this is shown as step C6), (7) to align the camera with the first die or other part Collecting the image of the wafer W using the camera 20 by moving the plate 12, opening the shutter and observing and recording the die or part thereof by allowing the camera to observe and record the image, plate Moving the plate 12 to align 12 with the camera and another die or part thereof, observing and recording another die or part thereof, all of the wafer dies or parts desired to be observed Repeat these steps until all are observed and recorded (all of which are shown as step C7), (8) step C At the same time, determining which defects of a given die are located based on the “good die” model of step A3 and the acceptability of step A4 (all of which are step C8) And (9) creating a defect map of wafer W that is all collection of all images of the die including all defects found (all of which are step C9).

  Alternatively, step C7 can be replaced with the step of collecting an image of the wafer W using the camera 20 by continuously moving the plate 12 to scan across the die on the wafer. By this step, the wafer is irradiated with strobe light in the order corresponding to the moving speed of the plate, so that each die is strobed at the exact time under the camera 20. This allows a stop and advance (go) procedure for aligning the camera with the first die during continuous acquisition of the image, the process of observing and recording the die, moving the plate 12 to another There is no need for aligning the die, observing and recording this other die, and repeating these steps until all dies on the wafer are observed and recorded.

  When the defect rating D is desired and desired (this is generally at the end of defect inspection on a given wafer W. At this point, the classification of the defect is often desired. This includes the following steps: (1) a step of loading the defect map created in step C9 (this reloading step is referred to as step D1), and (2) defects to be evaluated. Step of selecting (or step of evaluating all defects on the wafer in order) (referred to as step D2), (3) The plate 12 is moved, whereby the wafer W is moved to a specific die having defects on the surface. Positioning to a proper position under the camera 26 (these are all steps D3), (4) the user observes and classifies the defect, and the user of the system 10 observes it. A process of observing and classifying defects (these are all processes D4), (5) a process of repeating processes D2 to D4 (process D5) until all defects desired by the user are evaluated and classified And (6) storing the classified defect map (referred to as step D6), and alternatively or additionally, storing defect information in any of a number of other database formats. Process or other management and evaluation.

  The defect reporting step E includes the step of outputting or printing out the data stored in the database format in step D6 if desired and desired. This data can then be analyzed or otherwise used to perform statistical or other analysis on defect type, defect frequency, defect location, etc. This analysis is useful for manufacturers of wafers W to allow them to focus on areas with defects.

  The above steps and sub-steps are a basic description of the system and process of the present invention. The following description is a more detailed description of various components and systems and details of the processes they perform.

  The wafer test plate 12 is a rotating stage that includes a universal interface platform that uses a vacuum, all of which provide a flexible interface for wafer and die package fixation. Wafer test plates include all wafers, notched wafers on film frames, dies in gel packs, dies in waffle packs, MCMs, JEDEC trays, Auer boats, and wafer and die packages. Other arrangements and configurations are defined for quick installation and inspection.

  The means (referred to as 14) for providing the wafer to the test plate is manual (in which the user moves the wafer from the cassette or magazine to the test plate 12) or in the embodiment of the drawing. As shown, it can be either automatic. In an automatic embodiment, the wafer providing means 14 has a robotic arm that moves from a first position (where the wafer W is first grasped from a magazine or cassette) to a second position (where: Wafer W is positioned on wafer test plate 12 for inspection). After inspection, the robot arm pivots to return the wafer W from the second position on the test plate 12 to the first position, where the wafer W is placed back in or on the magazine or cassette. .

  The robot arm, in the illustrated embodiment, is a two part arm having two sections, the first section pivots around a central support point, and the second section is the end of the first section. Swirl around. In one embodiment, the robot arm is surrounded by at least one cassette receiver (two in FIG. 1), optionally a wafer pre-conditioner, and an inspection station, which have a plurality of receivers therein. A standard wafer transfer cassette in which wafers are stacked, the wafer preconditioner provides wafer preconditioning or coarse adjustment, and the inspection station includes wafer top plate 12, xy-theta adjuster 16, optical A part 18 and a camera 20 are provided.

  The wafer adjustment device 16 for adjusting all the respective wafers at the same x, y, z, and θ positions includes a rotary motor, a ball screw, a motor driven directly or by a belt, a worm gear or other gear, an actuator. A precision system of hydraulic equipment, push rods, vacuum or other mechanical or electrical equipment to move this rotary stage straight or angled to the exact desired position .

  The same adjustment mechanisms and processes used during testing are used during training. In the illustrated embodiment, in particular, the wafer adjustment device is a 2-D x, y and θ adjustment process, which is optionally combined with z-height control. In particular, in one embodiment, it is a process of coarse adjustment of 2-D x and y (followed by fine theta (θ) adjustment), all of which are focus mapping processes (previously generated height or (Use a map of focus) (this process is done before and to determine the height or z, and thus to make sure that the wafer is in focus). Basically, coarse adjustment uses a pattern that is located approximately in the center of the wafer. This pattern is trained to know and predict the position of x and y on it, so that the orientation of this pattern and the x and y (2-D line) of the wafer is at least that It can be seen that this is a rough adjustment. This orientation is performed using stage 12. Thereafter, a fine adjustment is performed using a pattern near the periphery of the wafer. This pattern is trained to know that a correction for the θ (rotation) adjustment has been obtained. This adjustment is also performed using stage 12. In both cases, the camera finds the pattern and the adjustment mechanism moves the wafer until the wafer is adjusted.

  Focus mapping or z-orientation adjusts the camera and / or camera arm distance and then focuses as described below and / or by changing the objective and / or the camera focus. Implemented by combining. The adjustments made are based on a wafer height map. From this height map, the focus is defined using pre-programmed points on the wafer.

  The focus mechanism 18 is an optical imaging mechanism, and includes a large number of optical components for different inspection analyses. A microscope turret with a motor allows the imaging optics to be selected from a number of options. For example, multiple optical components (eg, 3 or 5 optical components) can be supplied, and typical options include 1.25x, 2.5x, 5x, 10x, 20x, 50x Double and 100x objective lenses are mentioned, but any other objective lens is contemplated. A motorized microscope turret and a separate objective lens provide a means of optical magnification selection.

  The camera system 20 or other visual inspection device is for visual input of good dies during training and for visual inspection of other unknown quality dies during inspection. The camera system can be any type of camera capable of high resolution inspection. An example of one part of such a camera system is a 3-CCD inspection camera, which is used to capture dies or other images during defect analysis.

  One example of a camera system 20 contemplated by the present invention is a two-camera system, where one camera is an inspection camera and the other camera is an observation camera. The inspection camera is a high resolution CCD camera, which provides a high resolution grayscale image for inspection. The viewing camera is a high fidelity color image camera for visual evaluation in finding defects, for example, at 758 × 582 pixel resolution, or 1008 × 1018 pixel resolution, or other known pixel sizes. Furthermore, by observing the camera, a high quality color image is provided for the operator to evaluate the defect.

  Computer controlled optical components are subject to the use of a long working distance microscope objective to provide the low distortion images required for accurate defect detection. A number of magnifications can be automatically selected based on the inspection recipe defined by the user as described below.

  Computer controlled illumination is integrated with and within the inspection camera and optics, thereby completing the wafer imaging process. Alternatively, the illumination system can be attached to the camera and optics as long as the illumination system is operating with the camera. In a strobe illumination environment, as described herein, illumination occurs simultaneously or substantially simultaneously with the camera shutter (which in one embodiment is a high speed electronic shutter mechanism). Or, in an environment where there is no stroboscopic irradiation, irradiation is typically continuous or as required.

  Irradiation is obtained by any known irradiation means such as high-intensity light, laser, fluorescent lamp, arc discharge lamp, incandescent lamp and the like. The illumination angle can be bright field only, dark field only, or both bright and dark field variations.

  Bright field illumination involves illuminating the substrate from above, where the illumination system is typically adjacent to or part of the camera mounted directly above the substrate. . As shown in FIG. 1, this irradiation has an orientation of about 90 ° with respect to the substrate. In the illustrated embodiment, the bright field illuminator is adjacent to the camera and functions in harmony with the camera. This bright field illumination is very effective in illuminating a flat or relatively flat object on the substrate. This is because this light is generally reflected back to the camera. In contrast, a 3-d object on the substrate reflects light at an angle, causing the light to be at an angle away from the camera. As a result, flat objects appear bright in the camera, while 3-d objects appear dark.

  Dark field illumination is often used in conjunction with bright field to "brighten" a 3-d object, or just to view a 3-d object simply. Dark field light is provided at a low angle relative to the wafer top plate 12. Dark field illumination reflects light to the camera at an angle of approximately 90 ° to the substrate when dark field light is introduced at a low angle into a 3-d object on the substrate, but when the object is flat. Works opposite to bright field in that it reflects light at a low angle along the perimeter facing the introduction of light. Therefore, dark field light illuminates a 3-d object brightly, but does not irradiate a flat object so much.

  The system 10 with a dark field option includes a plurality of illuminators spaced around the periphery, as best shown in FIGS. In one embodiment, four illuminators are used and each is equally spaced from each other with a 90 ° separation. In another embodiment (a pair is shown, FIGS. 20-21), eight illuminators are used with a 45 ° separation, and in the alternative shown in the drawings for these, the illuminator is 90 They are paired together at a separation of ° and have twice the irradiation power at a given angle.

  If the system comprises both bright and dark field illumination, the user has the option to use one or the other or both. This provides a significant option. For example, if inspections are performed on dies that tend to have only flat objects on them, the bright field illuminates these objects well and is very sufficient for this type of inspection. Alternatively, dark field may be sufficient if inspection is performed on dies that tend to have 3-d objects. However, in many cases (for example, those with gold ridges that are generally very flat but very rough and tend to contain 3-d masses protruding therefrom), the combination of the two is often advantageous. It is. In this example, bright field illumination indicates the presence of any defects such as scratches and the presence of ridges, while dark field illumination indicates lumps and rough surfaces on the ridges. Without a dark field, this ridge appears as a dark image. Once the dark field is introduced, the mass is identified as a white spot on the ridge.

  Dark field also assists in defect classification. This is because bright-field light does not distinguish between particles or defects protruding from the surface and those buried or scratched on the surface. Dark field illumination distinguishes between these protruding defects and buried defects.

  In one embodiment, the system 10 includes a bright field illumination system that is physically located adjacent to or physically incorporated into the camera and for objects illuminated thereby. Provides bright field illumination from above. In another embodiment, the system 10 includes a dark field illumination system that is around the periphery of the wafer top plate 12 at a different low angle (eg, an angle between 1 ° and 10 °) from the top plate. Be placed. In yet a further embodiment, both bright field illumination from above the object and dark field illumination from around the periphery of the object are provided. As indicated above, illumination such as that provided by brightfield and darkfield illumination systems can be achieved with a white light source (eg, incandescent, fluorescent, or other similar gas encapsulated or similar electrical light). May be provided by any known radiation source, such as, or by a laser or similar device.

  The parameter input device 22 is for inputting parameters and other constraints or information. These parameters, constraints and information include sensitivity parameters, shape, die size, die shape, die pitch, number of rows, number of columns, and the like. Any form of input device (including keyboards, mice, scanners, infrared or radio wave transmitters and receivers, etc.) is considered sufficient.

  The display 24 is for displaying the field of view seen by the camera at that time or any previous saved period. This display is preferably a color monitor or other device for displaying the color display format of the image viewed by the camera 20 to view the user's field of view or the image stored in memory. is there. This monitor, or another adjacent or other monitor, can be used to view a grayscale inspection image of the camera 20 being used by the system 10. This display 24 is used to show the images observed by the camera 20 during the examination. In addition, a system parameter display 36 can also be used to display other information, such as system parameters, if the user desires.

  The computer system 26, or other computer with processing and memory capacity, has a defect to save the input good die, generate a model from that die, and based on defect filtering and sensitivity parameters. It is for comparing or analyzing other dies against that model to determine whether or not. The computer system 26 is also used to perform all other mathematical and statistical functions and all operations. In one embodiment, computer system 26 is in a parallel processing DSP environment.

  A marking head 28 is provided to mark certain dies, such as those that are defective. In one embodiment, the marking head is a die inking mechanism. This can be used so that each die can be inked after inspection, or all defective dies can be inked, or all defective dies can be evaluated and / or classified Can it be inked after? Inking can also be used in a “forced inking” mode, so that pre-specified dies (such as all dies at the edge of the wafer) can be used for electrical or visual inspection. Regardless of being inked.

  An air knife is provided to clean the wafer prior to inspection, if necessary. The air knife is basically a conduit of some design through which air can be injected, where the conduit comprises one or more orifices or outlets. This air is expelled from orifices selectively and relative to the wafer on the conduit to blow away dust and other particles from the wafer prior to evaluation. This eliminates erroneous defect determination.

  These systems and components are part of the system 10 and are used for performing defect inspections. This defect inspection is briefly described above and will now be described in detail below.

  The entire training process A is described in further detail below.

  Step A1 defines the step of determining and / or training (and storing in the computer system) the adjustment characteristics and parameters in the computer system 26 for use during training. This conditioning technique is a two functional process when performed in steps A3 and C5 as described below to determine and inspect a good die. That is, physical adjustment and image adjustment. At this point, it defines which parameters should be used during these physical and image adjustments. These parameters include defined markers (eg, those required during physical adjustment), as well as separate elements and buffers (eg, those required during image adjustment). The actual physical and image adjustment steps occur during steps A3 and C5, as described below.

  Step A2 defines (and inputs to the computer system) the wafer and / or die shape, wafer and / or die size, die pitch, number of rows, number of columns, etc., and all such information Is stored in the computer system 26 for use during training and / or examination.

  Step A3 is the step of training the system as to what the “good dies” contain, adjusting a plurality of known good dies via the device 16 and those via the camera 20 By observing the die, as well as forming a model in the computer system 26 to determine how the ideal die should look based on the observed common properties, elements, ranges, etc. A good die is very likely to have process variations and is defined as a die that has no defects except what may actually be; however, all these process variations are not defects, Rather it has been regarded as an acceptable variation. Preferably, the entire full spectrum of acceptable random deviations is provided by this training of a good die set, typically up to 20, 30, or 100, shown in the system during training. However, neither minimum nor maximum is required. However, the results become more accurate as the pool grows. This is because better and more diverse models are created. Thus, color drift and contrast shifts, as well as numerous other deviations, are part of this training set. Basically, the system 10 performs die inspection by studying a known good set of dies provided by the user.

  Adjustments can include either physical adjustments or image adjustments, or both. Physical adjustment basically involves a specific position marker, each wafer, die, or sub-section of the die (which is used as a starting point from which the wafer and die are positioned and adjusted). Of inputting on or around. In step A1, these markers were defined.

  Physical adjustment adjusts each and every wafer, die, etc. via wafer adjustment device 16 by looking for and adjusting these positioning markers to the same x, y, and θ positions. A test plate 12 is included. In use, the system takes an entire photo or image of these wafers, dies, or subsections thereof, and looks for specific positioning markers. This system uses a hunting method to find markers. Once one or more specific positioning markers have been identified and found to be in any other position or orientation than expected, the wafer test plate 12 can be translated or rotated in a manner that The wafer, die or subsection is spun, turned, adjusted or otherwise moved in the x, y, and θ directions.

  The system can also perform image adjustment. During step A1, separate elements and buffers were defined as being necessary for image adjustment.

  This image adjustment may also be referred to as software adjustment. This is because the software actually performs this adjustment by adjusting the captured image rather than physically moving the wafer or die. This image adjustment is performed on each section of the wafer, such as each die, during one or both of good die modeling and unknown quality die inspection. This is often necessary. This is because each captured image can be slightly displaced as compared to an adjacent image or a common position on another wafer. The actual process of image placement basically ensures that all images at a particular location imaged are adjusted. That is, when superimposed, these image features are aligned rather than having an offset or twist, so that only the defects are noticeable.

  If image adjustment is performed as required in steps A3 and / or C5, a separate element on the die from which the image is turned or moved to “square up” can be It includes the process of searching for a camera. This separate element is generally an element that is large enough so that its defects are not a problem. This element must also have a distinct shape. If this separate element is in the position where it was predicted to be there, the image is lined up and no image adjustment is necessary; but otherwise, a separate element is found. , And the image must be adjusted.

  Hunting of discrete elements in image adjustment can be performed on the entire die. However, this is expensive and time consuming. As a result, sophisticated adjustments can alternatively be implemented.

  For smart adjustments, a buffer is defined in the image. This buffer can be “wiggled”, ie moving or twisting the image. This buffer is typically the x and y amount of movement that is expected. This buffer is then used to define the area around the predicted position of the distinct element to be searched for the distinct element. Once a distinct element is found, the x and y distances away from the position of the distinct element at which the distinct element was predicted is the x and y the entire image has been moved to adjust the image. The distance in the direction.

  The step of observing is to use an image of the wafer W (a known good wafer) to move the plate 12 using the camera 20 and move the camera to the first image (this is the whole wafer, part of the wafer, Collecting, and then viewing and recording the image by aligning with a die, which may be part of a die). Thereafter, moving the plate 12 to align the camera with another image, observing and recording this other image, and repeating these steps until all images on the wafer are observed and recorded. Is included. An alternative step C8 involves continuous movement and strobe illumination, as described below. In each case, this is then repeated for a plurality of known good dies or wafers. This is because observing the pool of wafers is necessary to create a good die model.

  The actual defect inspection algorithm is calculated from a collection of “good die” sets of images. An image or images are taken for each good die of the good die set. Each image is composed of pixels (eg, approximately 1000 × 1000 array or grid of pixels, but any number can be used). For each identical pixel in all of the good die images, ie for each common x, y coordinate (which is a pixel), the mean and standard deviation are calculated in pixel values, which The grayscale value for a given pixel. Thus, in a group of 30 good dies, each die is an array of 1000 × 1000 spots (each called a pixel) (a total of 1 million spots), as used in the above example. The average and standard deviation of the number of gray scales for each pixel in the x, y coordinates of all 1 × 1, 1 × 2, 1 × 3, etc. up to 1000 × 1000 is calculated. That is, the mean and standard deviation are calculated for each of the million dies, such as using gray scale measurements for pixel 1 × 1 in all 30 dies for pixel 1 × 1.

  In one embodiment, the number of gray scales for each pixel is used to calculate the mean and standard deviation, and these are within a certain range. One example is a 256 scale scheme, where one end, such as 0, in this 256 scale scheme represents an image that is dark, black, or shaded, and 255 in the same 256 scale scheme. The other end represents a white or shaded image.

  For a type of die, the collection of all means (for all pixels) is effectively a complete die of that type and essentially defines a good die model. The collection of all standard deviations for a type of die is practically acceptable as adjusted for sensitivity and filtering as described below, inside which the die is good And on the outside, the die is suspected of being defective.

  Step A4 is the step of setting inspection parameters, and how these parameters must closely match a good die model in order for an unknown quality die to be considered a good die (ie, This is a value that indicates to the computer system 26 how much the difference from the exact model is acceptable to still be considered a good die. Some of such inspection parameters are defect sensitivity, minimum defect contrast and defect filtering.

  In the illustrated embodiment, defect resolution depends on optical magnification. Selecting a higher magnification results in a smaller image field of view. Depending on the magnification selected, multiple images may be required for inspection of a single die, or multiple images may result in a single image being suitable. The die size and optical magnification are entered in step A2. However, it is noted that smaller defect resolution results in more imaging per die, and therefore more time for defect inspection of the same quality die. Alternatively, a camera with adjustable resolution can be provided, whereby this adjustment feature controls sensitivity rather than image size.

  Defect sensitivities involve the user-defined magnification of the mean and standard deviation calculated to define a known good die model as described above. The defect sensitivity is described in more detail below in step C7.

  Minimum defect contrast involves an absolute limit defined by the user at the upper and lower limits defined from the mean and standard deviation. Minimum defect sensitivity is also described in more detail below in step C7.

  Defect filtering involves statistical or data filtering and includes area, size, problem area and / or clustering filtering, and connection and / or reduction factor filtering. This filtering allows the user to have defects of size other than those in critical areas where it is desirable to be of sufficient size or shape or otherwise acceptable and therefore not be labeled as defective. Items found to be found can be filtered out. In the illustrated embodiment, defect filtering is provided for each inspection recipe or round. This allows the performance of the system to be optimized for the user application. This defect filtering feature uses defect position and shape information such as shape, size, xy coordinates, etc. to automatically determine if the defect requires further evaluation and classification by the operator To do. One example is as follows: Any defect larger than a certain size can be positively determined to be a defect that is not subject to further evaluation. Additionally or alternatively, any defect smaller than a certain size is filtered out as not being a defect, even if it is outside the scope of the “good die” model. Further, in the evaluation process of the process D4, there may be a region where the operator's evaluation is necessary during that time. Similarly, unusual shapes, positions, configurations, placements, etc. can be filtered from a “good die” model. Defect filtering is further defined below.

  Step A5 is a step of storing the training model and its features, parameters, and the like in the computer system 26.

  Step B of creating an overall inspection recipe includes the steps of creating and storing an inspection recipe for each type of item to be inspected (ie, wafer, die, etc.). A vast number of inspection recipes can be created, copied, and edited, allowing the user to customize the inspection process.

  Step B1 defines how the wafer W is selected from a cassette or other containment vessel. Step B2 defines how the dies on each wafer are selected for defect detection. Step B3 defines how the defect detection map file is input and output. Step B4 stores this recipe.

  The entire inspection process C is called a defect inspection, and is an advanced unique digital image analysis technique for semiconductor wafer inspection. In step A3, the system first performs a wafer inspection after first examining a known good die of a user-supplied set as described above. This method of learning and checking is more powerful than traditional template or model conformance checking. It should be noted that even random changes in a known good die can be determined to be acceptable, and this is not the case with traditional template or model fitting. In short, this robust approach to wafer inspection functions similarly to a human operator without fatigue and other problems.

  In step C1, a wafer identification code is input if desired. This provides a way to identify each wafer for later defect evaluation and so on, where the wafer mapping occurs is required. The wafer identification code can be any known identification system, such as alphanumeric characters, bar codes, 2-D matrix codes.

  In step C2, the recipe defined in step B is selected. Step C3 selects product settings if desired.

  Step C4 loads the wafer onto the wafer test plate 12 using the wafer providing means 14. This loading onto the wafer test plate can be manual loading or the use of an automated system in which the wafer is automatically transferred from the cassette or magazine along with the die thereon to the inspection area. This automated system allows the removal of all manual operations.

  Step C5 adjusts the wafer on the wafer test plate 12 using the wafer adjustment device 16 to adjust each and all wafers to the same x, y, z and θ positions and is defined in step A1. Adjust using adjusted and / or trained adjustment functions and parameters. This has been described in detail above. This is because similar physical and image adjustment processes are used here, as used to adjust a known good wafer to form a good die model.

  It is often necessary to focus the camera 20 on the wafer W if it is not already in focus. This occurs if necessary during or after step C5 and is the z-direction of the wafer as defined in the height map.

  Step C6 uses the camera 20 to move the plate 12 to align the image of the wafer W with the first image, which can be the entire wafer, a portion of the wafer, a die, or a portion of a die. Collect and observe and record the image, and then move the plate 12 to align the camera with the other image, observe and record the other image, and perform these steps all on the wafer. Repeat until the image is observed and recorded. Alternative step C6 includes continuous motion and strobe irradiation as described below.

  Step C7 is concurrent with Step C6 and determines where the defects on the given die to be observed are located based on the “good die” model of Step A3 and the tolerances and parameters of Step A4. The process of carrying out is included. Basically, each pixel on an unknown quality wafer is observed, whereby defect sensitivity and filtering are used with a “good die” model to make the pixel and / or any group of pixels “good” or suspicious To decide.

  Differences between the initially abnormal or “good die” model and the image are discovered and then sensed and filtered. To simplify this determination, upper and lower level values are determined based on average and standard deviation calculations and user defined sensitivity and absolute limits for each pixel on each die. The observed image is then one that includes connection factoring, reduction or noise reduction factoring, and statistical or data filtering for area, size, blob identification such as regions of interest, and / or interactive filtering. Filtered using the above types of filter techniques. After filtering, suspicious defect areas are identified. Basically, defect sensitivity and minimum defect contrast are used and once the sensitivity is taken into account, the upper and lower level values, which are the actual adjusted standard deviations on either side of the average, are calculated. Stipulate. Thereafter, filtering is often used to better identify true defects.

  In one embodiment, the defect sensitivity is basically a plurality of user-defined standard deviations. Through actual analysis of good and bad dies, the user defines multiple standard deviations that most accurately define all defects but do not falsely define good dies as defects. An example is as follows. Assume three known good dies with grayscale values of 98, 100 and 102. This average is 100 and the standard deviation is +/- 2. The user defines the defect sensitivity as 5 through the inspection information. The upper and lower limits can then be 110 and 90, respectively.

  In one embodiment, this minimum defect contrast is similar to a user-defined absolute limit. In the above example, the user via information notices that a grayscale measurement with a minimum contrast of 15 is not a defect. This minimum defect contrast is therefore set to 15 and as a result, the upper and lower limits must instead be 115 and 85.

  In a preferred embodiment, after each pixel of unknown quality wafer or die is observed, an inspection image is created using simple image subtraction. The inspection image essentially subtracts the grayscale measurement of the inspection pixel (eg 98 for that pixel) from the good die upper limit, eg 110, or the good die lower limit eg 90 is again 98. Is obtained by subtracting from the gray scale measurement value of the inspection pixel to obtain a binary good or bad display. These upper and lower limits are preferably sensed. If this number is positive, it is colored black (or white) as being within this range, and if this number is negative, it is colored white (or black) as being outside this range . A binary black and white image is obtained. This image, because of its simplicity, allows it to be filtered at a much faster rate compared to storing the actual 256 color image. Alternatively, full color images such as 256 colors can be used if sufficient memory is available and optimal speed is not important.

  In one embodiment, one or more subsequent filters are used on the binary black and white image. Image processing functions such as connection factoring and reduction factoring can then be used individually or at once. Statistical filtering or data filtering for spot identification may also be performed individually or at a time.

  Connection factoring includes a “close” operation. This identified pixel, white in the above example, is expanded and then faded, doubled and then faded, or any other known combination. This connects or fills the defects in order to filter out small defects or acceptable irregularities.

  Decreasing factoring includes an “open” operation. This identified pixel is faded and then expanded or double faded and then double expanded or any other known combination. This reduces noise.

  Blot analysis includes identifying a blot in a binary black and white image. Once recognized, various parameters are identified, including, for example, size, position, region, etc., such as x size and y size. Statistical or data filtering is then performed on the parameters of this spot.

  Such statistical or data filtering includes region filtering, size filtering, region of interest filtering, and interactive defect classification filtering. Region filtering discards spots that are smaller than a given region. Size filtering discards spots that are of a predetermined x or y size or smaller. Target area filtering allows the user to define locations on the die that are less important or less important, and by doing so, make any defects above it irrelevant. Finally, interactive defect classification includes clustering of identified pixels that are in close proximity (where the distance defining proximity is user defined) but not touching.

  Basically, an unknown quality die is examined by observing the image and comparing each pixel with its mean and standard deviation via an upper and lower limit. Sensitivity and filtering also allows compensation for factors that are more or less important to the user. That is, if any one of a given observed pixel in an unknown quality die is outside the upper and lower limits when sensed and filtered, the die is defective and The defect spot is inked or otherwise written down as described.

  This step C8 creates a defect map of the wafer W, which is a collection of all defect data for all dies, and is stored in a data file. In the preferred embodiment, this is a binary black and white image.

  As an alternative to the above inspection process, an alternative process C6 collects an image of the wafer W using the camera 20 by moving the plate 12 continuously to scan all dies on the wafer. A process whereby the wafer is illuminated by a strobe light in a sequence correlated to the moving plate and each die is strobed under the camera 20 at the correct time. Basically, a short illumination pulse of light on the moving plate effectively “freezes” the image. This allows continuous collection of images, viewing and recording of the die without the need for stopping and operating procedures to align the camera with the first die, and moving the plate 12 to move the camera to the other Align with the die, observe and record the other dies, and repeat these steps until all dies on the wafer are observed and recorded, etc.

  All defect evaluation steps D are generally at the end of defect inspection for a given wafer W. This is because defect classification is often desirable in this respect. The defect inspection or detection process of this Step C is all automated and rapid, so once completed, the user can manually set parameters, filters, sensitivity, etc., not all dies or wafers for defects. Only defects found on the basis can be inspected. Considerable time is saved.

  In step D1, the defect map created in step C9 is loaded. Step D2 selects a defect to be evaluated (or sequentially evaluates all defects on the wafer). Step D3 moves the plate 12 to position the wafer W so that the specific defect is properly positioned under the camera 20. Process D4 is a defect user observation and classification process that allows the user of system 10 to observe and classify the observed defects. Any number of classifications are possible, and the classification is user defined. Step D5 is a repeat of steps D2-D4 until all defects have been evaluated and classified. Step D6 is a step of saving the classified defect map and, alternatively or additionally, saving the defect information in a database or other number of other formats for management and evaluation. .

  The entire defect reporting step E is a step of outputting or printing out data stored in the database format. This data can then be analyzed or used in other ways to perform statistical or other analyzes such as defect type, defect frequency, defect location, etc., which is useful for wafer W manufacturers And allows manufacturers to concentrate on the defect mounting area. This step E provides complete and effective data analysis because it reports data in multiple formats including graphic, tabular, and actual image displays. Data arranged in a tabular format allows numeric values to be easily correlated with other values such as electrical formats. Graphical data displays quickly display trends that would otherwise be difficult to see.

  System 10 is based on standard computer technology, such as Pentium Pro or similar computer platform, and allows many different communication options, for example both in serial and network form. For example, the system includes a TCP / IP configuration and may alternatively include SEC-II / GEM or other computer industry standard protocols.

  System 10 can also be used to perform inspections using drift maps. This is useful when individual dies of the wafer W are cut on the film and then extended as necessary to pick them up and remove, which is a well known technique. The problem here is that since the film material stretches unevenly, orthogonality can be lost during stretching and the die moves in different directions. The approximately square or rectangular cutting dies are now oriented in all different directions, and the rows of such dies are no longer linear, but rather wavy or otherwise directional. When this large stretch and loss of orthogonality occur, a drift map and drift process are added to compensate for this. This step is typically inserted before scanning.

  In one embodiment, a frame grid is made for the purpose of defining the possible placement of each die. It is known to stretch film-cut wafers that have been transported to make it easier to pick up each die without damaging adjacent dies. However, this stretching process is typically not uniform and results in a non-directional die. This drift map predicts the stretched position of each die using the stiffness before cutting and the starting point of the die that is known for pitch.

  To create a drift map, marks or dummy dies are placed on the wafer every nth (eg every 10th). Using the machine's field of view, the system 10 looks for the mark at its possible location and then looks around if it is not found. Once the actual position is found, the machine field of view advances to the next possible position of the mark and the process is repeated. Once all the marks are found, the pitch is calculated assuming consistent properties between the marks. Using this pitch and knowing the original position of each die prior to making a cut, a drift map is created that accurately predicts the position of this die.

  System 10 may also incorporate the use of autofocus functions. Such a function is based on a sharpness calculation where the image is scanned at each predetermined image point. The sharpness calculation is then used to find the exact focus. In order to save time, this can be done only in every nth image.

  In short, the basic procedure of operation is as follows, except for automated wafer transfer and wafer mapping options. The operator or user must first train the system as to what is a “good die”, ie create a good die model or select an existing good die model. As noted above, this inputs position markers and aligns each die so that multiple known good dies are properly imaged from the exact same x, y, z and θ positions and The process to be used is included. In addition, wafer and / or die geometry, size, pitch, number of rows, number of columns, etc. must be entered prior to good die imaging. Multiple good dies are then each aligned and viewed by a CCD camera so that the computer system looks at ranges such as pitch, color, angle, position, etc. by grouping common properties. A “good die” model is then formed. Basically, the system 10 performs wafer inspection by looking at the user-provided good die known settings. In general, it is preferred that at least 20 or 30 dies are provided, but a minimum or maximum is not required. Inspection parameters are also set to show how close and certain characteristics of a “good die” model must be matched for an unknown quality die to be a good die Is done. These include sensitivity parameters and defect filters.

  The user must also create or select a pre-stored inspection recipe. This includes how wafers W are selected from a cassette or other containment, how they are selected for defect inspection on each die of each wafer W, how defect inspection map files Information about whether or not is input and output.

  The system 10 is now ready to inspect an unknown quality die. If the required identification code is used when wafer mapping is activated, it must be entered at this point. Thereafter, a wafer W (or a cut wafer, or a gel pack die, or a waffle pack die, etc.) is placed in the inspection area, particularly the wafer inspection plate 12 (which is under the inspection camera). This is accomplished using the wafer providing means 14. Since the “good die” has been loaded using the adjustment functions and parameters defined and / or trained in step A1, the wafer adjustment device 16 then puts the wafer in the same x, y, z and Align at the θ position. The desired magnification is then selected and then the camera 20 is focused.

This system in turn moves the camera 12 by moving the plate 12 to align the image of the selected area (first die position) of the wafer W with the selected area, such as the first die position. In use, it is ready to be collected, so that the entire image, part of the wafer, die, or part of the die is taken and its first image is taken and then the image is observed and recorded. Automated defect inspection and bond pad analysis are performed on the digital image of the die. If this die is inked, it is automatically identified (mapped) as an “inked die”. If the die is not inked and no defects are found, then the system collects and stores detailed information about each defect such as the defect location, size, and shape on the die.

  The plate 12 is then moved to align the camera with another selected area (which can be the next adjacent or non-adjacent area) and the image (second die) adjacent to the first image. Position) on the wafer. Basically, this plate is indexed to the next die position under the inspection camera. This second die position is then observed and recorded. These steps are repeated until all images on the wafer are observed and recorded. Simultaneously with these image viewing steps, defect sensitivity and filtering are used in conjunction with “good die” model observations to determine whether the difference between the original anomaly or “good die” model and this image is the actual defect. Determine whether or if they should be filtered out. A defect map for wafer W is then created in the computer system from a collection of all defects on all dies, including all defects found thereon.

  In other embodiments, rather than moving the plate in a gradual process, the plate is moved continuously during the strobe exposure. The sections of the wafer are then scanned by synchronizing the camera with the stroboscopic illumination as the camera is properly positioned over each section of the moving substrate, with the stroboscopic illumination occurring simultaneously with image acquisition through the camera.

  At the end of defect inspection on a given wafer W, defect classification is often desired. Each recorded defect is manually evaluated by the user where the plate 12 has been moved to the location on the wafer W where the particular defect is located so that the user can observe and classify this defect. . This is then repeated for all defects. This classified defect is then saved as a classified defect map.

  The wafer is then removed and another wafer is loaded for inspection. This removal and new loading can either be done manually or can be performed automatically.

  Accordingly, the present invention provides simplified, efficient, safe, inexpensive and effective devices, systems, and methods, as will be appreciated by those skilled in the art and above, which achieve all of the stated objectives. It provides the elimination of difficulties encountered with prior art devices, systems and methods, and solves problems and obtains new results in the field.

  In the foregoing description, certain terms have been used for simplicity, clarity and understanding, but such terms are used for descriptive purposes and are not intended to be understood as broadly intended. The necessary limitations do not imply beyond that of the prior art.

  Moreover, the description and illustration of the invention is by way of example, and the scope of the invention is not limited to the exact details shown or described.

  Having described the functions, discoveries and principles of the invention, the manner in which the invention is understood and used, the features of this construction, and the advantageous new and useful results obtained, the novel and useful structures The devices, elements, arrangements, parts, and combinations are set forth in the appended claims.

  The preferred embodiment of the present invention, the best mode figure to which the applicant intends to apply the principles, is described in the foregoing description, shown in the figure, particularly pointed out, and in the appended claims. be written.

  Like numbers refer to like parts throughout the drawings.

Claims (1)

Invention described in this specification.

Images