JPS63500976A - Automatic optical inspection method for objects such as integrated circuits - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] The present invention is an automated testing method for objects such as integrated circuits. Regarding equipment for testing small integrated electronic circuits, which is well known as do.

All integrated electronic circuits are made by depositing and diffusing thin layers on top and leading to connection pads. includes a substrate of semiconductor material on which is disposed a series of metal tracks . These integrated circuits are manufactured primarily by photographic techniques, which will not be discussed here, e.g. It has a diameter of approximately 100 mm and is supported by an orthogonal line network called a "cutting path". "Users" containing a large number of identical chips, e.g. up to 500, that are more isolated from each other. It is produced in the form of thin sheets or slices called waha.

After the wafers are finished, they are placed in a fixed frame suitable for stacking in special containers. It is fixed on a plastic film that is stretched over the frame. U for separating chips When cutting wafers, the wafer is usually kept on a plastic membrane. and the cutting is carried out along the cutting bath.

Although some tests are performed during manufacturing, other tests are performed after the chip is installed in the device. It should be done before the

Inspections performed before and after cutting are usually detrimental to the continued operation of the device. Electrical inspection using a probe is used to eliminate chips with defects that may cause examination and visual inspection under a microscope.

The most frequently observed defects are:

・Scratches on metal tracks; ・Pores in the passivation layer covering the metal; ・Ink stains on the metal connection pad ( ink from the area marked in front of the defective chip); - destroys the integrity of the protective ring. Cutting flaws that are destructive or dangerous; - Missing metal on the track; - Missing electrical inspection marks on the connecting butt; - Metallic bridge between two tracks; ・Oxide defects inside metal tracks; - Bridging between two tracks due to dust inside the passivation layer.

Optical inspection shows that the defect is not of the type that would result in a reject. Anomalies that are present from time to time are also detected. These include small (non-metallic) dust and - is associated with spots on the corn.

"Pellapercorn" often contains metal traps due to the polycrystalline nature of these layers. This is the name given to the irregularities on the surface of silicon or polysilicon.

Optical inspection performed by people using microscopes is a difficult and delicate task. It is. Moreover, it is not completely reliable, it depends on the operator, and even the same operator Depending on the degree of fatigue, the ratio of rejected products varies within a reasonable range of 1:10. current Today, the task tends to become more difficult due to increased circuit complexity.

Therefore, there is a need for improved and automated so-called "optical" inspection.

Automated optical inspection equipment designed to inspect masks used in chip manufacturing is already available. In this device, the image obtained by the microscope optical means is a video It is converted into an image, analyzed using appropriate software, and then inspected. Ma The problem with inspecting screens is that comparing a single black and white image with a standard image will detect all defects. It is relatively simple as it allows you to The problem is that the wafers are tested during fabrication. It becomes even more complex when it has to be When inspecting printed wafers, all listed defect items appear at the same time. reach the highest level. These defects consist, among other things, of several other defects. Almost invisible under lighting conditions that allow detection, and vice versa. be. This has the effect of complicating the equipment and increasing the time required for testing. Tarasu.

Currently used wafer inspection equipment generally does not carry the wafer under the microscope. With the help of automatic equipment handling the frame, which aligns it with the optical axis of the microscope, defects are detected. includes an optical system that includes an operator to detect and verify the Also, the width of the metal track Some machines use techniques that analyze video images to determine .

Inspection devices operated by direct comparison have also been proposed (SPIE Journal, George Hung, 1984, volume 449, pages 563-571. RGE Huang)'s paper "Robotic alignment inspection equipment for semiconductor processing (AR obotic Alignment and In5pection Syst em)　J)・In other words, this device has two separate optical systems that are aimed at each other. one is directed to the test object, the other to the standard object, and the two images are are combined and sent to the TV camera, and the signal is sent to the image processing circuit. It has become. Such a device has a large size of one wafer carrying the chips to be inspected. The two shall be placed apart from each other by a distance at least equal to the forced to use two different optical cents. Optical hand with the necessary accuracy in details Correct alignment must be maintained with sufficient precision to allow for combinations in stages. becomes a problem. Supports one or more integrated electronic circuits that avoid this difficulty A method for optically inspecting objects such as wafers is described in particular in U.S. Pat. No. 448 No. 1664. According to this method, a binary image, that is, a black-and-white image of the object In between, a comparison is made with a previously stored similar standard image and the recorded signal is To create a binary defect signal corresponding to the pixel when there is a discrepancy between the standard signal Thus, a specific binary image is created. European Patent Application No. 124113 is a class discloses a similar method.

Such methods may be used to reduce irregularities such as irregularities from the edges of a metal strip deposited on an insulator. It is effective in detecting various types of defects. On the other hand, this method Sufficient reflected or diffused energy to change the value of the assigned binary signal - Other defects such as peeling of insulation or dirt on metal parts that do not affect the level It may be possible to avoid pitfalls.

It is therefore an object of the present invention to provide reflective or is a type of deficiency that does not necessarily cause the diffused energy level to fluctuate very much. Screening that clarifies defects and provides a large amount of data on all of the noted defects The objective is to provide an automatic inspection method and device for

To achieve this objective, the present invention includes one or more of the following steps. Provides a method for optically inspecting objects such as wafers carrying integrated electronic circuits It is. That is, a) Digital image of the object or part of the object to be examined, i.e. of the object limited by the coordinates; An image formed from one layer of digital signals corresponding to each pixel and each pixel of the object A record is made from the digital values corresponding to the reflected and/or diffused energy. Re, b) This digital image is created by combining the above objects to create a digitized differential image. By making a difference between the energy levels of the corresponding pixels of the body and the standard image, C) This image is compared with a pre-stored digitized standard image; Obtain a binary image containing defective signals representing pixels for which the detected signal and the standard signal do not match. is binarized in order to d) Parameters of each defect composed of several pixels showing a defect signal (Position, height 5 width 5 degree of elongation.

brevity), and e) If one of these parameters exceeds the predetermined limits, a failure decision will be made. determination is made and, if in doubt, the image is recorded for subsequent human decision-making. .

A basic feature of the invention is that in step b) the energy level between corresponding pixels is A comparison is made by making the difference between , which is converted into a binary differential image in step C) It lies in the fact that it becomes a digitized differential image that is converted.

This feature causes a significant number of defects or blemishes to remain outside the scope of the inspection. is eliminated, thus providing high reliability.

According to the advantageous method of operation, ・Defects constituting a “peraper cone”, i.e. defects on the metal or polycrystalline trunk Various defects that occur due to surface irregularities and affect a large number of pixels that are less than a predetermined value. is not dominated by parameter extraction or determination; - corresponds to different lighting conditions Several images are recorded and these images are compared not only to the standard image but also to each other. and especially in this case, the resulting image is recorded in visible light as well as a fluorescent image. The visible defects on the two images corresponding to the dust stains are recorded and the parameters extracted not be dominated by or decisions; ・Before recording the image, a standard image of the position and orientation occurring on the object is recorded and this The difference in position and orientation between the reference and the corresponding standard present on the standard image is measured. and the resulting image is moved mechanically or the two images are moved optically. or electronically moved to bring the recorded image to match the standard image. and preferably the difference in orientation between the two images is driven around an axis coaxial with the ray. This is corrected by rotating the Dove prism. Which is used Lines of interest of objects that intersect at almost right angles with high contrast in position and orientation even when Advantageously, the object is pre-positioned in the first position; The object was then transported to a second location for inspection, and the object was placed in the same manner at the relocation. Standard means are provided and means of transport are provided with associated means for reference. , and the standard means of transportation shall match the standard means of the above two positions one after another. The pre-positioning above the transport is maintained and advantageously frees any defects. at least some of the are detected at the first location; ・Use a probe to indicate on an object that it has a defect. If there is an ink mark, for example, made during a preliminary inspection such as an electrical inspection, the The detected defect corresponding to this mark is extracted and determined by its parameters. ・Recording should be done with an appropriate distribution of bright and dark backgrounds. made from at least one image obtained with a combined brightness; ・The pre-stored standard image can be used under the same conditions as the object image to be inspected. , an image of an object that was considered to be free of defects by the operator, or an object that was considered to be free of defects. To record images formed by juxtaposition of partial images corresponding to parts of the body. More can be obtained.

The invention is further illustrated in connection with a practical and non-limiting example illustrated by the following drawings. Detailed explanation.

FIG. 1 is a perspective view of an installation according to the method of the invention.

FIG. 2 is a diagram of the optical column of the first station.

FIG. 3 is a diagram of the optical column of the second station.

FIG. 4 is a diagram of the main light source of the second station.

FIG. 5 is a diagram of a fluorescent light source.

FIG. 6 is a diagram of the control and inspection station and image processing electronics.

The equipment shown automatically attaches a wafer onto a stretched plastic membrane within a frame. This is for dynamic inspection.

After bonding, the wafer may be cut by an automatic machine or may be left as a single piece of wafer.

Avoid cutting the plastic membrane.

The frame is of a type that can be cut automatically and is made of metal or plastic material. It consists of a ring.

Its outer periphery has four faces and a V-shaped notch that defines the frame reference on either side of one of them. It has a notch and a rectangular notch.

The frames are transported in containers containing, for example, 10 or 25 frames one on top of the other.

The equipment shown is designed to be installed at the beginning of an assembly line. this is exploration Accepts a container of electrically inspected wafers between the needles and inspected for cutting It delivers a container of wafers, which would otherwise remove defective chips. Unable to assemble good chips to obtain various devices.

The illustrated installation preferably includes the following equipment on a common chassis 1:

・A container to be inspected, a container for inspected wafers, and a container for wafers to be discharged. Receiving container and receiving station - delivery station 2; ・Detects pre-marked chips and certain large defects, as well as center and It is called "Macrovision" station 3, which pre-positions the wafer in the azimuth. station; ・Referred to as "Microinspection" Station 4, which detects and identifies defects. station to do. Considering the speed of each station, the equipment is divided into four to eight stations. Conveniently includes an ``Icroinspection'' station; ・Station 5 for marking defective chips; ・Frame processing assembly 6; - Control and inspection stations; ・In most cases, they are grouped and related to memory 'A'Jl or program media. connected electronic circuits.

In the above embodiment, the receiving station 2 is permanently fixed. In other words, the container or frame will not be moved.

This may be due to improper positioning of the container or the presence of a container that does not conform to the intended configuration. contain devices to prevent or signal. This also means that the container is in the proper position. Contains a device that holds the

Furthermore, if the container is observed during wafer inspection and the container is not filled accordingly; It is also equipped with a device that emits a signal.

All these devices are well known to experts in this field, so we will not discuss them in detail. There is no need to describe it in detail.

The macrovision station accepts a frame that carries a wafer, and the frame contains a flat plate. The clamp, which is connected to the pressure source, is positioned by the device and the support 8 Fixed. This pressure holds the film on the plate, so the wafer 9 is fixed. Ru. The flat plate itself has two directions X.

It is carried by a rotary plate which is carried by a table that moves in the Y direction. The flat plate is on the back perpendicular to the optical axis of the observation device described. The positioning of the frame is determined by the This is done by way of example using the two notches mentioned above. The wafer is placed against the observation device. Errors in the positioning of the wafer after the wafer is fixed to the flat plate, or chips that the wafer contains The error for more accurate positioning of the tip is ±1 mm in the direction of translational movement and rotational The direction is expressed as ±6°.

The purpose of pre-positioning is to use tens of micrometers in the direction of translational movement, and The aim is to reduce this error to the order of a few minutes in the direction of rotation.

For this purpose, the cutting path is used as a reference in the case of uncut wafers; For already cut wafers, cutting marks are used. This macro vision station Other tasks include detecting ink marks made during preliminary electrical testing. It is. This detection does not require as great a precision as the marking, and the There is no need to check for already recognized typhoons, and there is no need to check marks. Sequences can also be avoided, making it easier to use the most demanding macrovision stations. Transit time can be shortened.

Similarly, at a Macrovision station, the land of the cut and the chip garter are Detection of large defects such as flakes that damage the ring (or even the absence of it) ・During electrical inspection, it is also possible to detect marks on the connection band caused by the probe. Wear.

The optical device includes an optical column associated with a video camera type sensor.

This optical column (see Figure 2) has magnifications of ×2 and ×0.3, and Two switchable lenses for determining and detecting defects or marks 10.11 , a removable filter 12 for detecting ink stains, a light source 14, and a lens 1. of the first group to produce bright field illumination by sending a beam of light to the mirror through 6 Illumination comprising an optical fiber 15 and a second group of optical fibers 18 for producing dark field illumination device 13. A moving screen 19 provides a combination of brightfield and darkfield illumination. Either combined or bright field illumination only can be obtained.

The images received by the video camera 20 are processed by real-time coefficientization of the images and rapid transfer of the images to storage. to a processing system that performs storage and normal arithmetic and logical operations between the two images. It will be done.

This system is based on Imaging Technology J. rAP512J, rFB512J, rALU51 marketed by The electronic circuit board assembly designated 2J and rHF512J is currently advantageously made. ing.

The analog processing processor 1AP512 receives analog video signals from standard cameras. This signal is then counted into 512 points per line and 256 gray levels. These data are transferred to the image storage circuit board F through a special video bus. Sent to B512. This corresponds to basic image storage, i.e. 256 gray levels. represents 512 x 512 pixels coded.

The computer circuit board ALU512 has two "operating" ports for each of the 512x512 points. The standard histogram circuit board HF512 provides real-time (TV A histogram of gray levels in a scan) and an image with a given gray level. Calculate the coordinates of a point in the image.

For this device, the following operations indicating the positioning scan sequence can be continued.

- Mechanical positioning of the center of rotation of the support frame according to the optical axis of the camera; ・Two vertical references 1viI (cutting path or cutting marks) in the field of view Iterative exploration of this surrounding area to find out; ・Bringing the two reference lines to the axis of vision Mechanical centering control X, Y, O; maximum displacement performed is ±1 mm/±6° It is a stand. This centering ensures that the clamping mechanism is aligned with the frame of the wafer on the support. Initial uncertainty in positioning chip lithography on wafers Loss of sexuality; Machines along the X or Y axis of the table towards the edge of the wafer (in the 50mm range) target movement; ・Calculation of the exact position of the intersection of two reference lines displayed in the field of view. At this stage the machine - towards the opposite edge of the wafer (on the order of 100 mm) mechanical movement of the table along the same axis; two cutting marks displayed in the field of view; Calculating the exact location of the intersection of the arcs; - Mechanical angular centering (on the order of 15") and centering in x, y translations.

In the case of uncut wafers, searching for the cutting path is simply done in the bright field.

For cut wafers, cut marks can be cut in a “mixed” field of view (i.e. a mixture of bright field and dark field). It is preferable to determine the

The 2 magnification lens used in this step of the operation has a 10 micrometer pixel size. It corresponds to the size.

For output, a minimum magnification of 0.3 is used to allow for larger pixels and a more extended field of view. I am responding. In the 50 mm range, there are only a few 100 mm (4 inch) wafers. It is possible to be covered by four pulses.

Using the data processing equipment described above, the map of the wafer will contain marked chips and missing chips. It is established along with the location of the depression, eliminating the need for re-examination. micro inspection The tion station is generally similar to the macrovision station, but , only the main differences, which are superior in terms of fineness of optical analysis, are discussed below. Ru.

The first difference is due to positioning. Macrovision station is the center First, determine the orientation.

At this time, be careful to maintain the required positioning of the frame while However, even the slightest displacement cannot be avoided. obtained Accuracy is determined by positioning in two stages: at the center chip of the wafer; Also cut the wafer using the average optical magnification to determine the center position of all of each chip in the case of wafers, or of several chips in the case of uncut wafers. Determining the center position of the group now has to be done at high magnification. ing. Especially in the positioning mentioned last, the scope of inspection is mostly related to the dimensions of the chip. positioning with the same precision as with a Macrovision station. Therefore, a cut that is outside the area over its entire length cannot be It is not possible to use paths or cut marks. Therefore, regarding the chip itself, It is necessary to use a predetermined prominent line.

In the current state of the example, it is preferable to perform this predetermining manually; This must be well defined in the chip layout, including two orthogonal lines. , and select isolated patterns. Workers use image processing methods The work is subsidized by using it. For example, a worker may On the image of the chip, you can clearly see that it is isolated and has sufficient length. Only this line can be displayed. such that it cannot be used as a standard Suppose you have a storage device that provides an image that contains many short lines. The positioning has little involvement with the edge of the chip, and for better positioning against this It should be noted that it is better for the reference line to be near the edge of the chip.

Base 11! The line determination is done on a reference chip and is generally applied to all chips of the same type. It is effective against

The procedure involves using a macrovision station to determine the center and orientation differences from the desired location. The same is true for the version.

Another difference with Macrovision is that the angular deviation is small, so the wafer can be rotated mechanically. Instead of moving the optical device, for example, the mass of the mechanical assembly supporting the wafer is It is clearly possible to use a low doped prism to rotate the image on the sensor. This is due to the fact that it is possible.

On the other hand, chips by chip inspection are inspected every minute positioning sequence. As with many chips, it undergoes many displacements in the X and Y directions. optical analysis Another consequence of the fine detail is that the depth of field is very shallow, which is This is smaller than the wafer flatness error. This is called "autofocus" is removed by an automatic focusing device, which is well known under the name ” and does not need to be described again. It will be destroyed. However, in order to improve performance (classification speed), this autofocus This is done by moving the optical element, which moves the inspected wafer. Please note that this is not due to the movement of Ha.

Figures 3 and 4 show details of the optical column.

It has the same structure as Macrovision as a whole, but for the sake of simplicity, for example, the magnification Class fQ1 parts are given the same reference numerals even if they are completely different.

Differences can also be seen in terms of illumination, with interferometric imaging modes providing better performance in non-interferential imaging modes. A much larger defect contrast and a larger transfer function value can be obtained than when using a conventional method. the other smell According to analytical theory, typical edge differential effects (interference rings, interference stripes) occur; This complicates the image and makes it difficult to judge. In addition, silicon particle size control It tends to make the last part stand out, which can cause problems.

After many tests, we found that the illumination had a coherent factor close to 0.5. Selected. It is also possible to provide means for varying this factor.

Another important difference is in terms of lighting. This is because fluorescent lighting requires another high power source. This is due to this. This power source is designated by the reference numeral 21, and the optical fiber is designated by the reference numeral 22. Ivar and the “Gukuro Ink Cube” numbered 23 are related to this (No. 5 figure).

The autofocus device includes a laser diode 24, a diverging lens 25, and a two-image limit diode. 26. The signal emitted from this diode changes the optical focus device 27. control position. Switchable lens 28 allows for various adjustments without changing focus. We are trying to get a good magnification.

The prism 29 is switchable at start-up and can be changed to a visual inspection. child This is only installed in one of the microinspection stations, but It is also suitable for both.

Reference numerals 30 and 31 indicate color and neutral switching with various densities. Each possible filter is shown, tested using various colors, and This allows the signal level to roughly match the characteristics of the sensor.

Reference numeral 32 indicates a doped prism whose rotation axis is aligned with the optical axis of the column.

The optical columns just mentioned are bright-field WF illumination, dark-field illumination, and bright-field-dark field mixed illumination. Inspect chips using different colors with bright and fluorescent lighting. I am trying to make it possible to do it without any problems.

In more complex and expensive deformations, lenses are used in the optical path of the radiation emitted by the object. Later, it is compatible with the filter that selects the luminous flux of various colors, and this luminous flux is transferred to another record. By installing a beam divider to send to the seaber (video camera), simultaneous It is also possible to carry out inspections. Bright field illumination and mixed field illumination are different colors. - can be produced from a source. Image processing is performed on the same principle as above.

Marked stations are under electrical control on chips that are certified as defective. It is provided for depositing an ink stain having a diameter of about 0.5 mm on the surface. Te The cable is the same type used in the Macrovision station. , the problems encountered when transferring from the microinvection station are very small. Since it is small, there is no need to perform positioning again through image processing. The table is a macrobiotic defective chips whose location on the wafer was recorded during the vision and microvision stages. according to a control signal to place and place under the fixed ink deposition device. Can move to X and Y. Marked sections indicate the structure of the station. means and methods known or understood by experts in the field are required. It is. Therefore, there is no need to go into further detail. Set the wafer to IF and apply ink It is also possible to move the device.

In Figure 1, the marked units connected to the Macrovision station are tion is shown. In fact, the table is common to both stations. be.

The time of the frame passing through the macro vision station is the time of the micro inspection shorter than the time required to pass the station, a single macrovision Compared to the four microvision stations, Macrovision stations are not constantly crowded.

The movement required by the inked station table is is no larger than the movement of the station table, so both stations It is more convenient to use the same table. The actual ink deposition device is Macrovision. It is placed slightly toward the optical axis side of the optical column. Common table water at X or Y The horizontal movement width is sufficient so that the frame can be processed at any station. I can say that there is.

Of course, if the number of microinspection stations increases, two stations will be added. It is more convenient to completely separate the

The full processing functions are as follows.

- Removal of the frame from the container located at the receiving-feeding station; ・Transfer this frame to the macrovision station; ・Transfer the frame to the macrovision station Transfer the frame to the selected micro-inspection station; Transfer the key to the station; ・Transfer the frame to the receiving and sending station; Furthermore, insert this frame into a suitable container.

In the embodiments described above, the receiving and unloading stations are fixed, i.e. and does not move due to container movement, macro vision, micro inspection With printing and inking, the station moves with very little radius.

All of this processing is performed by a Hayabusa device called the ``Robo 7 To Server''. example For example, the receiver can be used for loading and unloading and inserting frames placed in the unloading station. Obviously, different constructions in other devices are also possible.

The robot-server is attached to the wrist, which can also be secured by electromagnetic locking. It has an arm that carries a gripping device. The first of these grasping devices is , placed in a basket, and in the model currently in use, they are spaced 6.35 mm apart from each other. Grasp the desired frame without touching the container itself, which is holding the overlapping frames. It consists of a grip that is provided for holding.

The robot-server is installed from a support within the container or from an inspection station. To lift the frame from the table, the grip is given vertical movement and the equipment is Translation horizontal movement in two orthogonal directions to use various stations and rotational motion about the vertical axis. These stations are lined up If the stations are placed in a circle, no rotational movement is necessary and the stations are arranged in a circular If so, rotational movement and radial translational movement are sufficient.

Due to the inaccuracies caused by the movement of the robosotober, the Macrovision Before tightening the frame with the device, it is necessary to roughly position the frame. child For tightening, the device holds two vertical cylindrical positioning pins and the robot -The server places the frame on these two pins, and one pin enters the V-shaped notch in the frame. , so that the other end enters the rectangular notch. Robot-server equivalent joints is bendable in advance. The frame is held by pins and the tightening on the table is done by the mounting. Due to the positioning, everything is not moving.

After pre-positioning at the macrovision station, the microinjection During transfer to the subtraction station 1. It is appropriate to retain the frame. child In order to do this, the robot-server lifts the frame on the device and tightens it while it is being transported. It is equipped with a glyph par with a suction cup that can hold the frame. x, y, z The reference plane is integrated with the suction cup grip, and the cooperative x, y, and z reference planes are macroscopic. installed at the vision station and all microinspection stations It is being In fact, the x, y, and z reference planes integrated with the suction cup grip are constructed using glyph par. and the cooperative x, y, z reference planes are configured with corresponding shapes.

The sequence of operations during transfer is as follows.

- Positioning the server on the opposite side of the Macrovision station; - Orientation of the unloaded gripper towards the macrovision station; - Lower glyph par to load level; ・Forward movement of the gripper; ・Allows the wrist to bend; ・Take local X, Y, Z reference planes with the gripper and apply the suction cup to the frame; ・Grasp the frame with a suction cup; ・Untighten the frame; ・Retract the gripper and lift the frame to eliminate the local reference surface; ・Fix the wrist; ・Retract the tip completely; ・Raise the gripper; ° Move robot-server to micro-inspection; ・Fix the frame in the same order as above.

Real-time image processing systems are as just described from a hardware perspective. The system from a software perspective is described below.

The device includes two sections (see Figure 6):

・General microcomputer; ・Specific image section.

This microcomputer was designed around the following IN locations 40 multipaths We provide standard circuit boards.

・IN location r802 related to the r80287J type mathematical collaborative processor It is controlled by the 86J type microprosensor 42 and is manufactured by IN Okisha. circuit board 41; ・0.5 MB expansion RAM storage circuit board 44゜This mass storage device has the following It consists of two readable and writable disk devices.

・Compatible floppy diskette 46; ・Fixed 200 MB hard disk 47゜Specific image section uses standard imaging technology (1?IAGINGT) ECHNOLOGY) Circuit board AP512 and FB512 and ALU512 and HF 512, consisting of circuit board 49.50.50 (second) and 50 (third) Ru.

The image circuit board is connected to the microcontroller by MULT I BUS40 at the bottom of the container. connected to the computer.

The ProSensor controls the image circuit board through its control and status registers.

The video analog signal emitted by the TV camera 20 has a 256 gray level. It is converted into digital data that represents the data. In this way, the matrix of 512 x 512 points The digitized image in the form of a box is loaded into the random access memory of the circuit board 512. is coded. This stored image can be permanently displayed on the TV monitor by circuit board AP512. displayed on the screen. In reality, this circuit board receives the digital signals emitted by FB512. Convert the number. The flow of digital data exchanged between two circuit boards is dedicated and rapid Pass the bus.

Therefore, the software uses 5 12X512 codes coded at 256 gray levels. Process the image of points. All calculations between any two image storage devices It is possible to perform mathematical and logical operations.

Software can be divided into three groups:

・“Data processing” utility; ・Basic image program; ・Specific inspection program.

Data processing utilities enable operators to: Ru.

・Create an operator program; ・Storing data on a hard disk or diskette; ・Print the results of alphanumeric characters or black and white images.

The basic image program is concerned with the following matters.

- acquisition of images with the possibility of horizontal and/or vertical zooming by a factor of 172 or 2; -Transferring the image to the image storage surface; - Calculation of image gray level: cross section along video line, global histogram totogram, spatial signature and various display routines; - Write protection for Graphics Proft equipment and pixels.

The specific inspection program enables:

・Image binarization; ・Extraction of cutting paths or cutting marks from images; ・Implementation by electrical tests Isolation of defects due to ink stains and cut size; ・Generation of a specific line on the chip; ・Automatic centering of the image with respect to the reference system; ・Bright field and/or dark field and Detection and extraction of defects present in images using colorimetric and/or colorimetric analysis; metal areas; Sorting out “peso bar cones” from;・Acquisition of direct drawing method for metal trunks; - Determination of degree of defect.

In addition to each of the electronic image processing devices mentioned above, control and inspection devices are also available for operator intervention. stages, ie essentially a control panel 48, a TV camera 51 and a microinstrument as described above. and an eyepiece in at least one optical column of vection.

On the other hand, the intervention by the operator is a routine operation, and involves a group of frames, one frame, or Identification of a single wafer (this operation can be done by marking the frame or by coating the wafer) intervention in case of operational defects or breakdowns (can also be carried out in data processing equipment) , commands to start and stop equipment operation, element when the characteristics of the wafer or frame change modifications to the mechanical and optical structures that must be carried out with great care to achieve the required accuracy. Adjustment of component parts, etc.

Two types of interventions have been further specified, the first of which is reference chip selection. It's about. In theory, the chip pattern as established by the designer can be used as a reference, but this operating method is suitable for chip circuit regeneration and Deformations due to deformation and formation processes must be taken into account. Therefore, Based on chips selected for inspected wafers that are considered defect-free It is preferable to use it as This is because changes in manufacturing conditions are compatible with Chi Nobu. stored in storage unless the change requires the generation of a new standard. This chip is used to check all chips of the same type. This is data regarding. If necessary, add two or three chips and the top of each chip. A standard is constructed using good parts.

The second specific intervention of the operator is to This concerns the selection of reference lines.

These reference lines are also stored in memory so that changes in manufacturing conditions may require other reference lines. It will be used until it is recognized as such. As mentioned above, the selected reference line originates at the same location. It can be used for a variety of chips as long as it remains raw.

As mentioned above, among the parameters of each defect considered for the removal decision are: The location of the defect is also included.

A defect that has no effect and does not affect the metal area does not constitute a defect. removed as . On the other hand, defects in metal parts or in the immediate vicinity are The defect appears to correspond to a defect that would result in removal according to the criteria of the "List". .

To determine whether the defect affects or is affecting the metal area The binary image of the metal area is taken as a starting point, i.e. each pixel is either "white" or "black". It is used as an image with the following state. This image is an image of the chip manufacturing mask. It can be equivalent. However, in the inspection equipment itself, the “dark field” illumination of the reference chip It is preferable to remove defects from the image. Thus, all inherent deformation defects This is not only due to the manufacturing process, but also due to the characteristics of the optical components of the device. This is caused by diffraction phenomena caused by lighting methods.

The binary image of the metal area is converted into binary by the ``pickup area'' due to the ``expansion'' of the metal area. used to generate images. At this time, the logical comparison is a binary image of each chip defect. and a binary image in the "pink-up area".

Defects within metal areas in “pink-up areas” are brought to the interpretation process. This is a potential defect.

bswll, @. −＾−−Regret. Gang PCT/FR86100307

Claims (15)

[Claims] 1. a) a digital image of the object or part of the object to be examined, defined by coordinates; An image formed from a set of digital signals corresponding to each pixel of the body and each pixel of the object A record is made from a digital value corresponding to the energy reflected and/or diffused by the b) comparing the digital image with a pre-stored digital standard image; the stage in which c) Determine the binary defective signal corresponding to the pixel where the recorded signal and the standard signal do not match. a step in which a binary differential image is created by creating; d) Parameters of each defect made up of several pixels giving a defect signal (position, height, width, elongation, conciseness) is determined; e) A rejection decision is made when one of the above parameters exceeds a certain predetermined limit. Steps where images are recorded in order to make or, in case of doubt, the next decision. and step b) extracts the difference in energy level between corresponding pixels. A comparison is made by a wafer carrying one or more integrated electronic circuits on which differential images are generated; A method of optical inspection of objects such as. 2. Defects constituting "petspercorns", i.e. on metallic or polycrystalline tracks occurs due to the unevenness of its surface and affects a large number of pixels that are less than a predetermined value. The defect does not depend on the extraction or determination of parameters, as claimed in claim 1. Product circuit inspection method. 3. Several images corresponding to different lighting conditions are recorded and the images are used as a standard image. The methods as claimed in claim 1 are compared not only with each other but also with each other. 4. An image is recorded in visible light and a fluorescent image is recorded to correspond to dust stains. The defects visible on both images are not due to parameter extraction or determination. The method as requested in scope 1 of the request. 5. Before recording the image, a standard image of the position and orientation present on the object is recorded, Differences and position and orientation between these standards and the corresponding standards present on the standard image is measured so that the object is mechanically moved or one of the two images is optically or electronically moved to match the recorded image with a standard image The method as claimed in claim 1. 6. The difference in orientation between the two images above rotates the doped prism around an axis coaxial with the ray. The method as claimed in claim 5, modified by. 7. Position and orientation standards intersect at virtually right angles with high contrast to the line of interest of the object. A method as claimed in claim 5, comprising: 8. A pre-positioning of the object is carried out in a first position, after which the object is moved to a first position for inspection. are sent to two positions, both positions are provided with standard means placed similarly, and There are standard means associated with the means of transport, and the standard means of the above means of transport are: Standard means of the above two positions so that the above pre-positioning is maintained even after transportation. As claimed in one of claims 1 to 7, which are brought one after another to coincide. How to do it. 9. 8. At least some of said defects are detected at said first location. in a manner similar to that requested by. 10. A probe is used to indicate on an object that it has a defect. For example, if there are ink marks made during a preliminary inspection such as an electrical inspection, If the detected defect corresponds to the mark, its parameters are not extracted or determined. Claims 1 to 9 are designed to result in a judgment of failure. It's like charging one person. 11. The recording is performed using illumination with a selected proportion of light and dark backgrounds. one of claims 1 to 10, which is made from at least one image obtained by in a manner similar to that requested by. 12. The recording was obtained with illumination with partially coherent light. A method as claimed in one of claims 1 to 11, produced from one image. 13. The pre-stored standard image is subject to the same conditions as the image of the object to be inspected. an image of an object that is considered to be defect-free by the operator or that is defect-free under the An image formed by juxtaposing partial images that each correspond to a part of an object thought to be obtained by recording an image, as claimed in one of claims 1 to 12. Method. 14. A binary standard of the metallic or polycrystalline zone of the object to account for the location of each defect. Starting with an image, we create a “big up zone” by expanding the metal or polycrystalline zone. A binary image is created that has a big up zone, and the image that has a big up zone is in the genus or polycrystalline zone or enlarged by the combination of big-up zones Various types of defects depending on whether they are located inside or outside the big-up zone. In order to investigate and analyze the A method as claimed in one of claims 1 to 13. 15. How a binary standard image of a metal or polycrystalline zone is described in claim 13 A method as claimed in claim 14, obtained in