JPS62245388A - Apparatus and method for identifying or recognizing known/unknown part including defect by automatic inspection of object - 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] [Industrial Application Field] The present invention is capable of automatically and rapidly performing real-time (real-time) identification of known/unknown parts, pattern recognition, defect identification, etc. during inspection of objects. The present invention relates to improvements in methods and devices.

[Conventional technology]

The most recent and appropriate approach to these problems is 19
U.S. Patent Application No. 4772, filed March 21, 1983, by Yasuhito Motokawa, Bertronics Inc.
No. 64, ``Method and apparatus for rapidly identifying or recognizing known/unknown parts, including defects, etc. of an object during inspection in real time''. The basic steps involved in the method of this application are outlined below, and the method does not address deficiencies in the prior art and others as applied, for example, to printed circuit boards and related applications (as described in the aforementioned U.S. application). This is different from the testing method. Applications to printed circuit boards, etc. may require pre-programming and software programming, including storing images of each image characteristic to be monitored, or the desire to pre-compute or program standards such as minimum conductor width and similar criteria. This is done to eliminate the need for the use of detailed masks, etc.

A preferred and improved application of the method according to the above-mentioned US application, and of the present application, is that the device "learns" the desired shape of a "good" object, such as a standard printed circuit board.
Either physically scan the “good” target or use CAD/CAM,
or by other signal inputs), and later on recognizing the displacement of the object from the "good" object while scanning the object in real time. Briefly, signals representing images of such objects or parts thereof at sufficiently different magnifications are sent to constant field-of-view (liOV) processing units, each of which has an image at each magnification. Each has its own memory. The image format is learned by storing the image signal in each memory. The object to be inspected is scanned and signals displaying the image at the sufficiently different magnifications mentioned above are sent to the corresponding processing device. The stored signal is compared with the most recent signal and a decision is made about the non-previously learned configurations, thus indicating defects.

Supplying a signal to a storage unit corresponding to the image with the lowest magnification but the highest resolution, extracting various parts of the image for different magnifications and quantizing them into sub-elements constituting the constant field of view. This allows display at sufficiently different magnifications.

Compared to the resolution corresponding to the magnification for identifying errors due to unlearned shapes, the disturbance due to magnification changes in the signal image is also preferably small.

For example, an inspection may be performed to locate damage between conductive pads and conductive wires on a printed circuit board or other similar device. First, until you find the damaged part,
Perform a scan of the circuit board at high magnification. Once discovered, lower the magnification and determine if the conductive pad is surrounding the damaged area. To accomplish this task, pattern information obtained at different magnifications for the same area of the object is correlated in order to identify the pattern of the inspection. On the other hand, the FO where the pattern in question is given by the objective lens with the lowest magnification
In other cases, such as larger than V, the test involves observing adjacent fields of view and combining pattern information from each field for object identification.

There are two categories of operations to consider. These include operations for learning images and operations for recognizing images. Items that go into the process of visual acquisition are ■: Obtaining animal body images at a magnification of 1 or 2 or more.

2. Construct the image using the minimum number of required light intensity levels. (In the inspection of printed circuit boards with copper conductors, the inspection is concerned only with the presence or absence of metallization, and small variations in the reflectivity of the copper are out of the question. Therefore, the number of luminous intensity levels involved is limited to facilitate the recognition process. (It is possible to do so.)
and 3. decomposing each video into N2 individual receive elements for each magnification.

In the process of image recognition, in order to locate related patterns,
Perform any number of combinations of the operations below.

1. Examining Nt elements separately for each magnification in each field of view to search for patterns of interest within each field of view.

2. Correlate patterns in the same area of objects generated at different magnifications.

3. combining pattern information generated from adjacent fields of view at a single magnification; and 4. above 3. recognition at various different magnifications and correlate the obtained pattern information.

When applying this type of inspection concept to the simulation or actual operation of mechanical and electronic equipment, the first step (a), as disclosed in the above-mentioned U.S. application, is to perform various magnifications. The purpose of this invention is to generate a high resolution large field of view (HRLF) video memory for simulating a FOV in a field of view. The range of this field of view is comparable to that of a low-magnification objective, but with higher magnification resolution. The size of the field of view is selected so as to memorize the largest single pattern of interest. For printed circuit boards (cards), this size typically corresponds to a field of view that is sufficient to store one or two pads. It is interesting to note that even for a moderately sized HRLF image, an astronomically large amount of memory capacity would be required if all patterns were to be stored. For example, even if an IIRLF video contains only 32 x 32 pixels, and each pixel is represented by a single binary bit, a total of 232.lO'5' memory locations are required. Become. To eliminate this memory requirement and still accomplish the desired task, following the inspection technique described above, (b) select a FOV representative of the desired magnification, and (c) divide each pov into NXN elements. (d
l setting the luminosity value of each element to be equal to the average luminosity value of the pixels in each element; (e) quantizing the luminosity value of each element to the minimum number of luminosity levels needed to display the pattern; , and (nit) Use the quantized element value as the address for the recognition memory device and store information about the pattern at the specified address location. If each element is quantized to b bits and the total number of elements is If there are N2 locations, then the recognition store will have 2bn locations.

As step (g), an independent recognition storage is used for each FOV. To inspect printed circuit boards,
Each element is quantized into two luminosity levels to indicate the presence or absence of metal in the element. This corresponds to b=1. b-1
In contrast, the following table lists the number of 64 memory chips required to store 2N'' patterns as a function of N2.

Number of elements (N2) Number of possible patterns (2N964 and number of memories 3" 5.12 x 10"
14” 6.55 x 10”
15" 3.35 x 10"
5126” 6.87
10" 1.048.576 [Objective of the Invention] To learn a new object for the first time, perform all the above operations and create a pattern of unique elements (1'OE) each consisting of 6N" bits. Generate every NOV. Each POE
is applied to the address lines of the recognition memory for a particular FOV. An identification code is written for each address accessed, which tells the memory that this pattern has already been seen.

In inspection mode, the object is scanned and operations (a) to (g) above are performed. The contents of each accessed memory address for each FOV is read to determine whether the pattern has been seen before. If a certain region of an object contains a considerable number of patterns that have not been seen before, that region is considered to be a foreign object. Such foreign substances can be detected when inspecting the printed circuit board (card) mentioned above.
Usually a defect.

However, if the shape and size of the conductor on the printed circuit board differs within certain limits from that of the conductor on the standard or standard type, i.e., "good" printed circuit board, etc.
Sometimes there is no problem at all. This means that even if there is a difference in shape or size from what you have learned, you should not signal that the product is defective. For example, a variation (thinner or thicker than the learned shape) may be acceptable if it is within a certain tolerance from artwork, such as CAD/CAM data, discussed below, or a printed circuit board (more commonly However, etching other types of objects is also acceptable.

These improvements provide electronic flexibility to automatically teach the machine safe shape and dimensional tolerances, i.e. allowed variations, allowing for improved detection and registration of defects in artwork. It is the main purpose of the present invention to achieve this.

It is therefore an object of the present invention to set standards for pattern recognition, defect detection, and solving problems associated therewith, without following the limitations hitherto defined in other systems. For automated, real-time, high-speed inspection of objects, providing a universal method with increased flexibility to tolerate variations or displacements within predetermined limits in the shape and/or dimensions of patterns learned from "good" standard objects. An object of the present invention is to provide a new improved method and improved device.

A further object of the present invention is to provide a new high speed automatic recognition and identification method and apparatus operable under the inspection philosophy and techniques described in the above-mentioned US application for printed circuit boards receivable through the holes in the conductive pads. and alignment or alignment tolerances, including but not limited to, along with additional improvements such as object shape B discrimination capabilities and other actuation capabilities.

An additional objective is to provide new methods and devices that are more versatile and have a wider range of uses.

Further objects are described below and more particularly in the claims.

[Means for solving the purpose]

In summary, from an important aspect of its application to printed circuit boards and similar devices, the subject matter of the present invention consists in storing digital signal information indicating the morphology to be monitored during inspection of the circuit board. This is a method of inspecting the shape of conductors on a circuit board. The method comprises storing digital signal information representing images of the desired predetermined object configuration to be learned; conductors, connections, patterns and pads, and acceptable holes and hole locations in the pads; presetting at least one acceptable dimension or size in the form of an object, such as the displacement of the object; generating a fictitious image of the form to be learned; storing modified digital signal information representative of the fictitious memory image; and detecting tolerable defects in the fictitious memory image for later inspection of the circuit board. According to another aspect, it is possible to check the alignment or alignment by comparing the displacements of the coordinate positions. be. Other features and best mode embodiments of the invention, including preferred apparatus and construction details, are described below.

Considering FIG. 1, the system is constructed as described in the above-mentioned U.S. application, where objects pcn < printed circuit board (Pr
The area of interest (abbreviation for C1rcuit Board) is shown to be optically scanned with a CCD imager (charge-coupled device image sensor). Real-time digitization of the scan signal is accomplished by operating an analog-to-digital converter A/D to produce a series of continuous digital signals corresponding to the optical scan line. C
The digitized signal of the CD is then transferred to a two-dimensional buffer memory I, and then to a sliding window memory SWM to create a high-resolution image with a large field of view. This technique makes it possible to increase the resolution to be comparable to that obtained with a high magnification objective, while the FOV is equal to or greater than that obtained with a low magnification objective. Furthermore, the object PCB is moved in a plane perpendicular to the axis of the CCD (displayed horizontally by
By reading the CCD every time the object moves (1/0 inch) and reading it into the buffer memory IIM, it is possible to completely eliminate the blurring of the object. In particular, the analog signal output of the CCD is processed by an A/D converter (
After being converted to digital form by a conventional sampling pixel quantizer (perhaps a flexible version of the sampling pixel quantizer), it is stored in a buffer memory BM (which can be formed into a series of shift registers, each with a length equal to the number of elements of the CCD). be transported. The digital output of each shift register is coupled to the input of the next register and to one column of the sliding window memory SWM, so that the image appearing in the window memory SWH is visible when the magnification lens is moved across the object. Be what you are. This means that the FOVS and FOV seen in the window memory SWM correspond to high and low magnifications, respectively.
Illustrated schematically in block SWM by L (described below).

The optical configuration that can be incorporated into the system shown in FIG. 1 is extremely flexible. The image size of the target reproduced on the CCD is determined by the CCD lens L1 and the object PCB.
Increase or decrease the magnification (e.g. from 173x to 3x using an optical lens Ll, from 2x to lθx using a video lens or from (from 1/2 times to 1/10 times).

The CCD is relatively very long (e.g. reticon (Re
Ticon's CCD device has 1728 pixels, while Fairchild's device has 200 pixels.
0 pixel), and each pixel is very small (0.64 mil for Reticon and 0.5 mil for Fairchild), allowing for a wide variety of high-resolution, large-field images.

The operation of the improved system of FIG. 1 according to the present invention will now be described using a CCD imager or
The input signals generated by the /CAM device will now be described. In the following example, a superimposed image of both sides of a printed circuit board, the object PCB, is provided to the apparatus.

The recognition code is calculated directly from the signals provided by the CAD/CAM system that designed the circuit board. These codes are then compared to the codes generated from the CCD video signals as the CCD scans the PCB and performs front and back pattern alignment.

To teach the machine a "good" or reference printed circuit board pattern, a digital signal from the CAD/CAM system A of FIG. 1 is fed to a shape modification module B, which will be described below. According to the invention, this module B:
Adding and/or removing pixels along the boundaries of an object within a predetermined tolerance. Adding pixels helps make lines and pads thicker and thicker, while removing pixels helps make lines and pads thinner and thinner. Depending on the number of pixels that are added or removed, the final PC
Set tolerances for line widths and pad dimensions that may exist on B. These modified digital video signals are learned by the device as will be described later.

For example, if a tolerance of up to a±2 pixels is required, the images to be trained are ±O1±1 and ±2 pixels. These correction signals occur in a continuous linear fashion;
Display a continuous line of video. The number of lines "L" is then stored in the puffer memory BM (Fig. 1) mentioned above, and 1
A two-dimensional image is generated that is equal to the width of the CCD scan line of the book.

As mentioned above, the row of pixels has a dimensional dimension W from the buffer distance.
It is supplied to the sliding window memory SWM of x-pixels (where ->l). In one preferred mode of implementation, the SWM contains, for example, 32 x 32 pixels. 3rd a
Shown is a typical image of a pad contained within an SWM of such dimensions. Here, if only 32
To store the total number of possible patterns seen in an SWM of x 32 pixels, assuming each pixel is represented by one binary pixel, the required memory locations would be 2 x . It would be a good idea to emphasize that it will be a number such as IQIS4. To eliminate this need for memory, yet accomplish the necessary task of recognizing all patterns, the following functions are performed in accordance with the present invention.

Function 1: Sliding window memory SW? I reduced field of view (FOV)
and then subdivide each FOV into NXN partitions.

Name each section an element.

Figure 3(a) shows the small field of view FOVS and the large field of view FOVL, which were previously compared to the block SWM in Figure 1.
It was named in relation to Figure 3(b) and Figure 3 (C1 are NXN with these two FOVs as N4.
This shows the case where it is divided into elements.

The next functions performed according to the invention are: Function 2: Set the luminosity value of each element equal to the average luminosity value of the pixels contained in each element, and represent each average element value in V bits.

Functions 1 and 2 are both performed in block (el) of FIG. 1, but in FIG.
C” bit) and N2 elements (E1141+++[j
nl+ E++, , ) - each element contains (C/N) 2 pixels - F with 02 pixels divided into
It shows Ov. The average luminous intensity for each element is then calculated and obtained at output (e) as shown in FIG.

Next, in the present invention, the following functions are performed by utilizing element quantization in FIG.

Function 3: Quantizes the average luminous intensity value of each element expressed in V bits to the minimum number of bits V required for pattern display.

For example, in inspecting an object such as a PCB, only the presence or absence of metallized conductors on the circuit board is of interest. Therefore, if 50% or more of the pixels included in the region of one element indicate the presence of metal, it is possible to consider that all the elements contain metal, and the element is quantized to a theoretical value of 1. If this condition is not met, the element is considered metal-free and quantized to the theoretical value O. In this example, V is naturally equal to 1 bit.

In the present invention, this quantized luminous intensity value is used for the address input of the recognition storage device (J) in FIG. 1 for the execution of the next step.

Function 4: Use the YN'' quantized element values representing the unique pattern of the element named PO[i (top right in Figure 1) as an address to the recognition storage (J), and Store this information in the address.YN”
If there are 1 bits, there will be 2y1'1 memory addresses in the storage device (J).

As previously mentioned, in accordance with the present invention, separate processing is employed for each different FOV, as shown in the following features.

Function 5: Using a separate rFOV processing unit consisting of an element calculation unit (e), an element quantization unit (F) and a recognition storage unit (J) in FIG. F
Processing the OV (note the "pOv processor for field of view S" and the block to the left "FOV processor for field of view").According to the technique of the present invention, the machine first generates the shape of a new object. To learn from function 1 to function 5 above,
is performed for each NOV extracted from the SWM in FIG. A unique pattern of elements (POE) consisting of YN" bits is generated for each FOV and applied to the address line ("Address input" in Figure 1) of the recognition storage (J) in each FOV processing unit. .

An identification code is written for each accessed memory address, indicating that the pattern has been learned. This process can also be called function 6.

In this example, the CAD/CAM system provides signals to the inspection device in a learning mode to draw a superimposed shape of the front and back sides of the PCN inner layer for subsequent alignment testing. The field of view (FOV) up to the WXW pixel is extracted from the sliding window memory SWM and the last described operation (function 6) is performed in each 150v processing unit.

CAD/CAM supplies the learning signal or CCD
Once the object is actually scanned, the device is ready for inspection...Function 7: The object being inspected is scanned and functions 1 to 5 above are performed. . The contents of each accessed memory address in the recognition memory of each FOV processing unit is read to determine whether the pattern has been previously learned and written to the recognition memory (J). If within a predetermined area of an object,
If a significant number of patterns that have not been seen or learned are present, the area is identified as foreign, ie, for example, a defect.

In order to carry out this foreign object area detection function, a two-dimensional error code memory shown on the left side of block (K) in FIG. 1 is used. This memory consists of cognitive storage (
J) stores a code indicating whether the pattern has been seen and learned. When we add up the number of unseen patterns in the two-dimensional area covered by memory, this is shown in Figure 1 (K
), and if the number is larger than the minimum predetermined number, the image of the foreign object area is displayed on the computer (L).
and further identification processing is performed as necessary.

In this illustrative example, front-to-back pattern alignment is tested by placing the inner PCB layer on a glass-topped X-Y table (on the left in Figure 1) and illuminating it from below. be able to.

Illumination light passes through the inner layer and creates an image on the CCD, which is a superimposition of both sides of the circuit board. X-
When the Y-table is scanned across the CCD and the image is scanned, functions 1 through 5 and function 7 above are performed. When a foreign object area is discovered, the video of this area is displayed on the surveillance TV T shown in Figure 1.
V and the X-Y coordinates are recorded. If a considerable number of foreign object patterns are present, both sides of the inner layer of the PCB are determined to be misaligned.

Another application of the invention is to produce a punching tape for automatic punching of finished PCB pads. In this application, the negative of the artwork with the positions of all pads serves as the object. The artwork is scanned on an X-Y table with a light placed directly below it. The image of the pad is learned. This artwork is scanned once more, this time with the device in inspection mode. In each case the pad is recognized and its center coordinates are recorded (eg at L) on a punching tape (not shown) which is later used to punch holes in the finished PCB.

- As previously mentioned, one of the important improvements of the present invention is that it "learns" all tolerance deviations in the shape and/or dimensions of "good" objects such as PCB leads, pads, etc.
and to eliminate the need for storage in memory. This is the morphology modification module shown in Figure 1 that adds or removes pixels along all boundaries or peripheries to create a hypothetical modified image that is thicker or thinner than the actual image within acceptable limits. This is done by using B. If the thickness of a particular object had been learned, hypothetical thin or thick variants (within tolerance) would be recognized as if they were standard, as shown in Figure 4. It may be appropriate to describe the useful form of the shown shape modification module B in the order of nine one-unit delays ^ to I, two lines whose length is equal to the length of the CCO. Delays LDI and LD2 and one read-only memory ROM constitute the module. The signal input comes from the CCO of FIG. 1, but this signal is then delayed at points A, B, and C by three one-unit delays. The output of the CCD is also delayed by one line in the LDI and then fed into one unit delays D, E, P.

Furthermore, the output of line delay LDI is line delay L
The second line is delayed by D2, and its output is added to a one-unit delay G, 11.1 to create a 3x3 matrix or block data.

Each delay (delay unit) A, B, C, mouth, E, F
, G and I are connected to the corresponding address lines of the read-only memory ROM (with the same reference numerals).
is applied to The contents of the read-only memory ROM are 3.
This is a newly calculated value using the central element of the ×3 data block. An appropriate output from this read-only memory ROM can be selected to indicate what the new calculated value is. For example, one output line OO is the data when you want to obtain an image of a thin conductor, and the data when you want to make the image of a conductor thick (output line O1).Thicker or thinner conductor If you want to get an image of
A method is used in which circuits of the type shown in FIG. 4 are connected in a cascade, and the output of the first circuit is used as the input for the next circuit. The final circuit is coupled to the CCD buffer memory l of FIG.

Thus, by presetting a desired amount of dimensional variation of an image within an acceptable range, the present invention allows for the electronic removal or addition of vixel signals from a stored image. This provides alternatively stored hypothetical but acceptable video information that displays different dimensions of the object form.

In the previously introduced method for illumination, it was explained that the object should be illuminated from below the x-y table.
It is clear that there are many effective ways to apply the reflected light from the top of an object, and this is also discussed in the above-mentioned US application.

The concept of providing a range of acceptable conductors and shapes that differ in size is useful when an object is illuminated with reflected light, negative of circuit board artwork by a camera, or if it has a similar transparency. Of course, it is also useful to illuminate by the transmission of light from.

Returning to the subject of artwork, we will now explain why it is most useful to place the artwork in the position of the object in Figure 1 and to illuminate it from behind.

Combining front and rear lighting is also possible.
Useful for inspection and alignment applications of CBs and other objects. If the lines, pads, or other items on the front and back sides are not aligned with each other, and the machine has learned to identify a "good" inner layer when it is illuminated from behind, it may not be possible to inspect the circuit board. If in practice one would see a misaligned pattern due to misalignment of the pattern on the front and back sides, this previously unseen misaligned pattern would be signaled as a defective product.

When processing PCNs or inner layers without transmission holes, reflected and diffused illumination can be utilized to generate optimal effects. Also, a combination of transmitted and reflected illumination can be incorporated to scan the inner layer with transmission holes, and the transmitted illumination can also be used for artwork processing.

By allowing light to pass through the artwork, defects such as pinholes and hairline cranks can be reliably detected, which could not be detected using traditional techniques that reflect light from a mirror placed behind the artwork. When the artwork is placed on a mirror surface, the light passing through the hole will be as shown in Figure 6A (1).
The mirror surface reflects the light in the direction away from the hole, keeping the angle of incidence equal to the angle of reflection. If the artwork is not perfectly flat, the reflected light beam may be blocked by the dark areas around the holes, making it impossible to detect defects.

For example, if the artwork has a protrusion that lifts it slightly above the mirror surface by 10 mils (h=10 mils in Figure 6A(2)), and the light entrance angle is 70 degrees. , holes or tears smaller than 7.28 mils will not be detected.

According to the backlighting method of the present invention shown in FIG. 6B, all defects are reliably detected without being affected by small protrusions or surface disturbances. This is because any defects, such as holes oriented in different directions along the irregularities of the artwork, will be detected by the light beam passing through the holes and directly towards the camera.

Therefore, the optimal form of main body for scanning the artwork has a transparent synthetic resin (trademark Lucite) upper surface and an internal illumination source at a stable position under the table, as shown in Figure 1. A hollow stage (X-Y), which generates a beam of light perpendicular to the stage surface. Place the artwork on the top of this table and fix it by suction in at least two places.

To inspect the inner layer using the transmission hole, as shown in Figure 7, the inner layer is illuminated from above and below and the hole is removed from the image.
Eliminate false errors due to perturbations in the positioning of the transmission holes. The reflected illumination also allows for checking the alignment of the front and back sides. To accomplish this, increase the intensity of the transmitted illumination source so that the light passes through the inner layers but not the PCB.
The learning and inspection of overlapping front-back patterns is made possible in the manner described above.

Sensing of dimensional tolerances, features, and positions presents problems in conductive pads, PCBs, or similar applications, particularly in connection with through holes, vias, and the like. artwork and C
The AD/CAM database does not indicate the location of the communication hole and does not have information on such location. Even if the communication hole appears perfectly centered on the pad in the final finished PCB, in reality it will be offset in shape and location compared to the center hole of a "good" circuit board. There is no problem with the presence of the communication hole anywhere within the band, and therefore it should not be flagged as a "defective product".

In order to recognize the acceptable thinness and thickness of conductive wires, fictitious "learning" videos have been generated that are thicker or thinner than the conductive wires of "excellent" objects that have actually been "learned." A somewhat similar idea turned out to be an excellent solution to the via hole problem. That is, a pad having an acceptable level of communication holes is imaginaryly formed so that it appears as a hollow pad during the inspection process.

In order to enable the inspection of circuit boards with pads with communicating holes (i.e. annular rings) without compromising the ability to find pinholes, defective holes and internal defects, the method according to the invention first comprises: A unique method of identifying the annular ring, checking its proper shape, and making it appear as if it were a solid pad in the original artwork during the inspection process. To do this, the PCB is illuminated from both the front and back sides simultaneously, a constant beam of light is reflected from the top surface as shown in Figure 1, and a beam whose appearance has been changed by coding or other methods is emitted from the bottom of the circuit board. It is sufficient to pass through the annular ring formed by the communicating hole in the pad and the hole for the component in the board. This process is shown in more detail in Figure 5. The bottom beam is clearly the reflected beam at the top. The upper luminous flux is coded (1l
r1 encoding) is performed by a rotary chipper RC for light modulation, and the resulting pulsed light flux is sent by an optical fiber bundle to a linear illumination head located at the bottom of the object PCB (the PCB is also substantially Helps reduce heating effects by removing infrared energy, or
(filtering of other spectral bands may also be performed).

Detection of a coding pattern at the video output section of the CAD indicates the presence of a communication hole (annular ring) or a component hole (bypass switch S1 is open in FIG. 5). The ring of metal film surrounding this encrypted pattern is then inspected for its continuity, and only if continuity exists is an imaginary pixel effectively added to the center during the inspection process. The entire hole is then filled, making the ring look like a solid pad. If continuity is not found to exist, the ring holes will not be filled and will be signaled as defective.

Further, regarding hole detection and alignment issues, when PCNs are manufactured, they are manufactured using tool holes around the circuit board to hold the circuit board during processing. The holes are used to align the circuit board for drilling, to form a conductor pattern on the circuit board, and once this is completed, to align the multiple layers of circuit boards with one another. It is interesting to examine the relationship between the various conductors on the printed circuit pattern as well as the positional relationship between the tool hole and the printed circuit pattern. If the pattern is misaligned relative to the hole, or if one part of the pattern is misaligned relative to another part by more than a predetermined tolerance, this will result in an error.

To test a circuit board according to the present invention for these purposes, one begins with a known "good" reference circuit board or artwork. The pcn is placed on the X-Y scan table of FIG. At a predetermined location, the reference image is picked up and stored.

This is, for example, the reference hole at the 1st coordinate position in Figure 8 AI, the 2nd
The pad shown in Fig. 8 1 is shown in Fig. 8, and the diagonal line shown in Fig. 8 A3 is shown in the third place.
The fourth line is the two intersecting lines in A5 of Figure 8, and so on.

Inspect the other circuit board and check that the circuit board has circuit conductors,
To determine if interconnects, holes, or especially fiducial images (typically located at the terminal boundaries of the board) are in the proper position, use the tool holes to X- The circuit board is placed on a Y-table and the edges of the table and the circuit board are carefully aligned.The circuit board is scanned, and the image 85 from Figure 8Bl is located at the coordinates of the X-Y table previously used as the reference image. It is shadowed by the position.

Holes drilled through the circuit board (81 to B4 in Figure 8)
matches the standards Al to A4, so the holes in circuit board B are correctly positioned. However, the pad and conductor image B2. B3 and B5 are shifted 8 units to the right, indicating the error in the position of these conductors.

Regardless of whether there is a correspondence between the reference drilling image and the X-Y position,
If lines intersect or if rings or the like shift beyond a predetermined tolerance, an error indication will occur at the location of these conductors.

It will be appreciated by those skilled in the art that technical modifications can be made in a manner similar to the system described in the above-referenced U.S. application and referenced in that description, including the addition and assembly of the components of FIG. It will be good. It is contemplated that the exemplary description of a PCB is merely an example, and that the invented technology is readily applicable to other tests, and that such fact is within the spirit of the invention and the scope of the claims. obtain.

[Brief explanation of drawings]

FIG. 1 is a combination block and circuit diagram illustrating an embodiment of an inspection system operative employing the improved features of the present invention. 2 and 3 (a), (b), (C1 are plan views showing the conductors of the circuit board and their dimensional tolerances. FIG. 4 can be used with particular advantage in the system of FIG. 1. FIG. 2 is a block circuit diagram of an embodiment of a subsystem that enables automatic learning of dimensional variation ranges to pass inspection, and automatic learning of tolerances of learned morphological images, such as manufacturing modification type (PC [ lOther targets, etc.) can be changed from artwork, etc. Figure 5 is a diagram for modification to detect communication holes such as PCB pads and their predetermined tolerances. 6A, 6B, and 7 are explanatory diagrams showing how to use reflected and inverted light or transmitted light, respectively, which is particularly useful for inspecting through-holes in PCBs, etc. is a schematic diagram of a reference circuit board and a circuit board under test at specific X, Y locations.

Claims (1)

[Scope of Claims] 1. A method for inspecting the form and position of conductors on a circuit board, in which digital signal information indicating the preferred form to be monitored during circuit board inspection is stored, which method should be learned first. Storing digital signal information representing an image of a desired, predetermined object configuration; predetermining at least one of the possible dimensions or sizes; and storing a digital signal for generating a hypothetical image of the form of the object to be learned, incorporating acceptable variations. modifying the information; storing modified digital signal information representative of said fictitious memory image; and recognizing acceptable defects in said fictitious memory image upon subsequent inspection of the circuit board; in response to ignoring the above acceptable displacement as an acceptable defect. 2. Modifying the originally stored digital signal information according to the above-mentioned changes in dimensions or magnitude is carried out by electronically implementing the removal or addition of pixel signals from the stored digital signal information. 2. A method according to claim 1, for representing different dimensions or sizes of a form of an object. 3. The above modifications of the initially stored digital signal information regarding the holes and hole locations in the conductor pads are carried out by monitoring the annularity of the pads to ensure acceptable communication holes; electronically adding a pixel signal to the stored digital signal information to substantially fill the hole;
Claim 1 providing a fictitious image of a solid pad
The method described in section. 4. The method of claim 3, wherein said monitoring includes sensing both light rays reflected from the top surface of the circuit board and light rays transmitted through the annular pad from the underside of the circuit board. 5. The method of claim 4, wherein the light rays transmitted from the underside of the circuit board are coded to distinguish them from the rays reflected from the top surface of the circuit board. 6. A method for inspecting conductive features and locations on circuit boards in which digital signal information is stored indicating preferred features to be monitored during subsequent inspection of the circuit board, the method first including the desired features to be learned. storing digital signal information representing an image of the object form at a predetermined coordinate position; determining in advance an allowable positional change in the coordinates of the learned object form; and determining the existence of the learned object form. The circuit board is subsequently inspected to determine its presence, and upon determining its presence, its coordinate position is determined, and any change in the final coordinate position that exceeds the above allowable position change is considered a defect. A method consisting of showing. 7 A device for inspecting the configuration and position of a conductor on a circuit board, in which digital signal information indicating the desired configuration to be monitored during circuit board inspection is stored, the device means for storing digital signal information indicative of an image of the form of the object; at least one change in size
means for modifying the stored digital signal information to produce a fictitious image of the form of the object to be learned, incorporating an acceptable amount of variation; means for storing modified digital signal information indicative of the fictitious stored image; A device combining means for ignoring acceptable displacements as acceptable defects. 8. Said means for modifying originally stored digital signal information according to said change in dimension or size includes electronically implementing the removal or addition of pixel signals from the stored digital signal information. equipped with the means of
8. Apparatus according to claim 7 for representing different dimensions or dimensions of a form of an object. 9. The above means for modifying the initially stored digital signal information regarding the hole and the position of the hole in the conductive pad comprises:
means for monitoring the annularity of the pad to ensure an acceptable communication hole; and electronically adding a pixel signal to the stored digital signal information to substantially fill the hole; 8. A storage device as claimed in claim 7, including means for providing a fictitious image of a solid pad. 10. The monitoring device according to claim 9, wherein the monitoring device includes means for detecting both the light beam reflected from the top surface of the circuit board and the light beam transmitted through the annular pad from the bottom side of the circuit board. . 11. The apparatus of claim 10, further comprising means for encoding light rays transmitted from the underside of the circuit board to distinguish them from rays reflected from the top surface of the circuit board. 12 A device for inspecting conductive features and positions on a circuit board, in which digital signal information is stored indicating the desired features to be monitored during subsequent inspection of the circuit board, the device means for storing at predetermined coordinate positions digital signal information representing an image of the desired object form to be acquired; means for predetermining allowable position changes in the coordinates of the learned object form; means for inspecting to determine the presence of said learned object form; and means for being activated for determining the coordinate position thereof upon determination of said presence; and said permissible position at the final coordinate position. A device that combines: a means for indicating all changes that exceed the change as defects; 13. The apparatus of claim 12, further comprising means responsive to said defect indicating means for recording the coordinate position of said circuit board and producing an image of the defect. 14. Apparatus according to claim 13, wherein the means provided include an X-Y coordinate table for employing said image indicating the location of the actual defect. 15. A method for inspecting the form of a conductor on a circuit board, in which digital signal information indicating the form to be monitored during subsequent inspection of the circuit board is stored, the digital signal information representing the form of a predetermined object to be studied. storing signal information; predetermining inspection-acceptable dimensional changes in the above configuration; and electronically removing or adding pixel signals from the stored video information; providing alternative memory image information representing dimensions different from the form of the image. 16 A device that performs real-time high-speed inspection of an object with the assistance of a video sensor, CAD/CAM video signal generator, etc., which generates a video signal of the object expressed in the form of pixels at a magnification of 1x or higher. means for quantizing the pixels of the image to the minimum number of luminous intensity levels or colors necessary to characterize the object; and storing the quantized pixels to create a two-dimensional storage image equal to the desired maximum field of view. means for selecting a desired predetermined field of view and magnification from said two-dimensional stored image; means for dividing each field of view thus selected into N^2 elements; a luminous intensity of each element; means for setting a value approximately equal to the average luminous intensity value of the pixels contained in that element; and quantizing the luminous intensity values of each such element to a minimum number that is a threshold necessary to recognize the associated pattern. means for applying each group of N^2 quantized pixels at each magnification to a corresponding recognition memory as a pattern of elements;
Each pattern of elements is an address (address) for the corresponding memory.
means for entering information into address locations in the memory for teaching properties, including the shape and other properties of a known object, and A device combining means for subsequently recognizing the taught information by reading the information in each memory location accessed by the pattern of elements, and means for monitoring such information characterizing the object. , and pixel signals are electronically removed from the memory location to store video information indicative of one or more different dimensions of the feature of the object that are predetermined to be acceptable for characterizing the object. Or a device with improved means for adding. 17. Said means for electronically removing or adding a pixel signal comprises delay line means connected to receive said signal image and said delay line means connected to said delay line means and applied to an address line input of a read-only memory. Claims: a one-unit delay for generating a matrix of pixel data; and means for outputting from a read-only memory a new calculated value indicative of the removal or addition of pixels at the edges of the stored image. Apparatus according to paragraph 16. 18 A device that performs real-time high-speed inspection of an object with the aid of a video sensor, CAD/CAM video signal generator, etc., which generates a video signal of the object expressed in the form of pixels at a magnification of 1x or higher. means for quantizing the pixels of the image to the minimum number of luminous intensity levels or colors necessary to characterize the object; and storing the quantized pixels to create a two-dimensional storage image equal to the desired maximum field of view. means for selecting a desired predetermined field of view and magnification from the two-dimensional stored image; means for dividing each field of view thus selected into N^2 elements; and means for determining the luminous intensity value of each element. means for setting approximately equal to the average luminous intensity value of the pixels contained in that element; and means for quantizing the luminous intensity values of each such element to a minimum number that is a threshold necessary to recognize the associated pattern. and means for applying a group of N^2 quantized pixels at each magnification as a pattern of elements for the purpose of converting digital signal information. 19 CA of digital signal information in circuit board artwork
16. The method of claim 15 obtained from a D/CAM description. 20. A method for inspecting the form of a conductor on a circuit board, in which digital signal information representing the form to be monitored during subsequent inspection of the circuit board is stored, the method comprising: a digital signal representing an image of the form of a predetermined object to be studied; storing signal information; predetermining inspection-acceptable dimensional changes in the above configuration; and electronically removing or adding pixel signals from the stored video information; providing alternative memory image information representing dimensions different from the form of the image. 21 scanning the further circuit board to be inspected, detecting on the circuit board signal information corresponding to the shape of the object on the circuit board, and detecting the predetermined shape and other shapes within the limits of the different dimensions stored above; 21. The method of claim 20, wherein the method of claim 20 recognizes. 22. Claim 20, wherein said object form comprises a linear strip, and said electronic pixel removal step results in stored image information of a thin linear strip that passes the inspection process. the method of. 23. The shape of said object is curved, and said electronic pixel addition step results in curved stored video information of increased dimension that passes the inspection process. The method described in section. 24. The object has disks or pads placed in near vertical alignment on either side of a translucent or optically transparent surface, and as a next step directly below said surface.
A strip of light of substantially uniform intensity is generated to remove infrared or other spectral energy from the light, and the light is then passed through the surface from a position near the backside of the surface to electronically illuminate the pixel. The step of adding a disk or pad arranged on the opposite side of the surface obtains stored image information of a disk or pad having a large dimension corresponding to an allowable change due to vertical alignment error of the disk or pad located on the opposite side of the surface. Range 2nd
The method described in Section 3. 25. The method of claim 24, wherein said disk or pad is a conductive pad and said translucent surface is a circuit board to which said pad is attached. 26 A method for inspecting the form of an object performed on a light-transmissive surface, etc., in which digital signal information indicating the form to be monitored is stored in subsequent circuit board inspections, the method comprising: generating a strip of light of uniform intensity and directing the light from a position close to the back side of the surface in a direction passing through the surface after removing the infrared part from the light; scanning the surface with the light strip; Detecting signal information on the upper surface of said surface corresponding to the shape of the object above, and digital signal information representing an image of the shape of the predetermined object to be learned as recognition in a subsequent scanning inspection of the surface. A method consisting of remembering. 27. According to claim 26, the strip of light is directed from above the surface to an adjacent area on the surface and is capable of illuminating the surface during scanning along the surface. the method of. 28. The method of claim 26, wherein the object is in the form of a light-opaque artwork and the light-transmissive surface is a substantially transparent surface carrying the artwork. 29. The method of claim 26, wherein the object is in the form of a light-opaque conductive element and the light-transmissive surface is a transparent circuit board to which the element is mounted. 30. Claims in which said conductive element is a disk or pad mounted on the opposite side of said circuit board, the vertical alignment of which can be detected by illuminating said circuit board from the back The method according to paragraph 29. 31 Inspection of the shape of an opaque conductive object on a circuit board, etc., in which digital signal information indicating the form to be monitored is stored during subsequent inspection of the circuit board, directly below said surface. generating a strip of light of substantially uniform intensity and directing the light, after removing infrared or other spectral energy from the light, through the circuit board from a position proximate to the back side of the surface; scanning with said optical strip to detect signal information corresponding to the shape of the object on the circuit board, and digital signal information representing an image of the shape of the predetermined object to be studied, subsequent scanning inspection of the board; A method consisting of memorizing it for recognition. 32. An apparatus for inspecting conductive patterns and locations on a circuit board, wherein digital signal information indicative of a desired pattern to be monitored during subsequent inspection of the circuit board is stored, the digital signal information indicative of the desired pattern; means for storing the pattern, means for later inspecting the circuit board and determining that there is a difference from the desired pattern, and determining the X-Y coordinate position of the pattern in response to the determination that there is a difference; and means operative to indicate the same as a defect;
and means responsive to the indicating means for producing an image of the indicated defect along with recording the X-Y coordinate position of the circuit board;
A device that combines 33. Apparatus according to claim 32, wherein the means provided include an X-Y coordinate table for employing said image indicating the location of the actual defect.