GB2417072A

GB2417072A - A machine vision inspection system and method

Info

Publication number: GB2417072A
Application number: GB0418094A
Authority: GB
Inventors: Sean Doherty; James Mahon; Peter Conlon; Malachy Rice
Original assignee: MV Research Ltd; MV Res Ltd
Current assignee: MV Research Ltd; MV Res Ltd
Priority date: 2004-08-13
Filing date: 2004-08-13
Publication date: 2006-02-15
Also published as: GB0418094D0

Abstract

For an electronics circuit production line 1 there is a series of nodes 2 - 5 each located after a particular production stage 10 - 13. Each node performs a line scan of the full circuit. It then extracts component images and saves them to local storage indexed on circuit identifier. A controller 6 polls the nodes to retrieve a series of images for a particular component of a circuit if a defect is detected during downstream testing. This series of images gives a defect creation history (DCH) for quick identification of the source of a defect.

Description

24 1 7072 - 1 "A machine vision inspection system and method"

INTRODUCTION

Field of the Invention

The invention relates to machine vision inspection of electronic circuit products.

Prior Art Discussion

At present, machine vision inspection machines (referred to as "automatic optical inspection", or "AOI" machines) are used for this purpose, such as described in US 6,151,407. Many such machines are built to a high standard of precision in terms of their gantry, camera, and illuminator sub-systems. Also, they incorporate sophisticated image processing capabilities so that they provide comprehensive machine vision inspection at the particular point on a production line at which they are located.

However, such machines are expensive, and typically require a significant level of human resources for operation and maintenance. It is for this reason that many manufacturers must make the difficult decision of where to locate the limited number of machines for optimum inspection. If one is placed at the end of a line, defects may be carried through for an unacceptable length of the line before detection. If one is located too close to the beginning of the line a defect may be 2 5 missed.

The invention addresses this problem.

Summary of [invention

According to the invention, there is provided a machine vision inspection system comprising a series of inspection nodes each for location after a production station in an electronics circuit production line, each node acquiring an image, and the system collating a series of images representative of successive stages of production of a particular circuit.

In one embodiment, each node acquires and stores images, and the inspection system further comprises a controller for causing retrieval of images from the nodes for a series.

In another embodiment, each node scans an identifier on a received circuit and indexes the captured image on said identifier.

In a further embodiment, each node performs an in-line image capture without removing the circuit from a conveyor linking the successive production machines.

In one embodiment, each node comprises a barrier for preventing further movement of a circuit beyond a registry position, and acquires an image at said registry position.

In another embodiment, the system further comprises an AOI inspection machine, which performs image processing to identify defects.

In a further embodiment, the controller is connected to the AOI inspection machine; and wherein the controller automatically retrieves a series of; images from the nodes for a circuit for which the AOI machine has detected a defect; and wherein the controller notifies the nodes of fmal positive testing of a circuit, and the nodes free up image storage locations of images of the circuit.

In one embodiment, each node captures a full circuit image, and extracts from said image an image for each of a plurality of components and saves each of said component images.

In another embodiment, the controller tracks progress of circuits progressing through the production line, and generates a user interface indicating this progress.

In a further embodiment, each node performs component presence, component polarity, and optical character recognition of printed data analysis operations.

In one embodiment, a single controller is linked with a plurality of series of nodes, each for a production line.

DETAILED DESCRIPTION OF THE INVENTION

Brief Description of the Invention

The invention will be more clearly understood from the following description of some embodiments thereof, given by way of example only with reference to the accompanying drawings in which: Fig. 1 is a diagram showing an inspection system of the invention; Fig. 2 shows an alternative system of the invention; Fig. 3 is a diagrammatic plan view of layout of a node of the inspection system; Fig. 4 is a sequence of diagrams of elevational views showing operation of a node; and - 4 Figs. 5(a), 5(b) and 5(c) are three series of illustrations to exemplify series of images captured by different nodes.

Fig. 6 shows a still further inspection system, in this case for multiple production processes;

Description of the Embodiments

Referring to Fig. 1 a machine vision inspection 1 system of the invention comprises a series of nodes 2-5, one after each stage of a production line. The nodes 2-5 are connected to a controller 6. The nodes comprise: a node 2 after a paste screen printer 10; a node 3 after a chip pick and place machine 11; a node 4 after a Me pitch pick and place machine 12; and a node 5 after a solder reflow oven 13.

Each node 2-5 is less comprehensive in terms of functionality than a conventional AOI machine. The functionality of each node 2-5 is limited to specific imaging tasks required for the particular production stage which precedes it. Also, the mechanical system of each node does not necessarily have the precision of a conventional AOI machine. The controller 6 operates to sequence output of images from each node and to control their operation according to operator instructions.

Referring to Fig. 2, an alternative inspection system 20 comprises a series of nodes 21, 22, and 23 after, respectively, paste screen printer 30, pick and place 31 and fine pitch pick and place 32 stages. There is also a lull AOI machine 24 alter a - 5 solder reflow oven 22 stage 33. In this embodiment the machine 24 is a "full" AOI inspection machine. All nodes, and the machine 24, are connected to a controller Referring to Fig. 3 the node 22 is located directly on a conveyer line 35 between the placement machines 31 and 32. It does not have a gantry system, but instead a mechanism for stopping the board travel at a particular registry position and performing a full board scan. This is illustrated in more detail in Fig. 4. In a step 1 a board B leaves the placement machine 31 and is conveyed towards the node 22. At the node 22, in a step 2 the board B passes through an imaging head 36. The head 36 comprises a line scan colour camera having full width of the board B. The node 22 also comprises appropriate high intensity lighting and triggering/control electronics.

A stop barrier 37 of the node 22 moves into position to prevent further movement of the board B at a registry position for operation of the imaging head 36. In a step 3 the imaging head 36 moves along the board B. capturing a complete image of the board. Since inspection of the board is not the goal of the imaging node the resolution can be kept low. This may however be unnecessary with the availability of cheap full width line scan camera. Typically the resolution would be similar to inspection and measurement systems at approximately 20 to 25pm.

Finally, in a step 4 the board B is moved from the node 22 to the next placement machine 32. At this stage the board scan image has been saved, and the system will analyse the image to locate the board fducials and extract the component images.

It will be appreciated that the board handling is very simple and very little delay is introduced for the imaging operation. Also, the chances of boarcUcomponent faults being introduced because of board handling are very slim because of the in-line and simple nature of operation of the node 22. - 6

In general, the nodes 2-5 of the inspection system 1 and the nodes 21 to 23 of the system 2() perform the following operations: - identify board fiducials, capture board identifiers by bar code scanning, capture scan images, and extract component images from the scan image, store the component images, indexed on board identifier, and i upon receipt of an asynchronous request from the controller upload a particular image to the controller.

Images are not uploaded until they are asked for, thus saving on network bandwidth. Each node will have local storage capable of holding board image data until instructed to delete them when the board has passed final inspection/repair at which point the manufacturing process has completed.

In the case of the system 20 the AOI machine 24 performs the full range of operations of conventional AOI machines. Where a defect is detected by the machine 24, a "post-mortem" analysis can involve analysis (automated or visual) of a series of images each taken after the individual production stages. This allows identification of where a defect arose and what factors may have lead to the defect arising. For example, if a tombstone defect was detected post-reflow, a defect creation history (DCH) would allow a determination of whether the defect arose from bad paste printing or mix-placement of the component. '['his allows the process engineer to target line maintenance efforts. - 7

The role of the controller 6 or 25 is to activate the nodes by transmitting the activation and sequencing instructions. The controller is configured as the master access node in a distributed database. It retains track of boards that are passing through the production line, and it allows the user to make DCH queries but does not actually store any data. The data itself (in this case image data) is stored at each node and is only transmitted to the controller when requested. The request instructions take the form of a query. The query itself may request a single image from a particular board, identified by barcode; it may be a request for a sequence of images for a particular part designator defined by a barcode sequence or within a certain time frame. Once the node has collated the data it may be uploaded to the controller to be viewed by the user.

Also, the controller 6 or 25, can poll each node in turn to retrieve an image of a particular board. The controller displays single component images or sequences of component images. Saving and transmitting an image of the entire board would be an expensive operation and unnecessary. This reveals very important process information for the operator as it reveals where exactly a particular defect arose.

This allows prompt corrective action to be taken.

Referring to Fig. 5, three examples are given. Fig. 5(a) shows an image 40 captured by the node 2. This reveals offset (to the LHS) printing of deposits 41 onto pads 42.

The node 3 captures an image 43 after placement, revealing the fact that the left side of a component 44 is not on a paste deposit. An image 45 captured by the node 5 after reflow reveals a "tombstone" defect in which the component 44 is upright.

Referring to Fig. 5(b), the node captures an image 48 revealing correct printing of solder deposits 49 onto pads 50. However, an image 51 captured by the node 3 reveals skewed placement of a component 52. This results in the component 52 being offset, as shown by an image 53 captured by the node 5. - 8

Referring to Fig. 5(c), an image 54 captured by the node 2 reveals a blocked-stencil fault resulting in a pad 55 being free of solder. The node 3 acquires an image 56 revealing good placement of a component 57. The node 5 captures an image 58 showing an open joint 59.

It will then be appreciated that the inspection system of the invention captures a huge amount of relevant information, which can quickly reveal the source of a defect. This is achieved without need for expensive equipment or for significant processing delays. The nodes are of very simple construction. While they require a good deal of image storage capacity, this is inexpensive. Also, there is no need for the controller to be complex or expensive. It merely needs to activate the nodes for a batch and to subsequently poll them for images.

In another embodiment, shown in Fig. 6, an inspection system 60 has three series of nodes 61-64 for a first production line 65, a series of nodes 71-74 for a second line 75, and a series of nodes 81-84 for a line 85. This may be repeated for all of N lines. A controller 70 controls all nodes and captures the images from them. The controller 70 also performs central storage of databases, and central computer aided design (CAD) processing. The controller 70 also allows portability of programs across multiple lines.

In an alternative embodiment there is more extensive local image storage and processing at the nodes. More specifically: Fiducial Location - to locate the boarding of the image Barcode reading - to read the barcode for tagging. This may incorporate a hardware barcode reader, or one residing in software. - 9 -

Knowing the fiducial locations, the system can locate every component on the board, and store their locations in a file tagged with the barcode. Images ol each component would be retained until such time as the controller directs the node to delete the images (such as after that particular board has exited repair).

In one embodiment, the system nodes can automatically analyse the images for the following attributes: Presence of component Polarity of Component Location and orientation of component - measurement could be achieved using local features rather that global flducials. For instance using pad data, via data or local fdueials may allow measurement of sufficient accuracy to be usable in closed loop. Components could be measured with respect to the paste deposit (lead to paste) and paste deposits could be measured with respect to the pad (paste to pad).

Read text on the component (OCR) Verify that labels are on the component.

Joint defects on post reflow solder pads. This may be able to perform gross analyses of solder fillets to ensure the creation of a correct structure. This would allow the monitoring of the reflow process by this invention.

This analysis will depend on image quality, primarily due to image registration and smooth movement. -

In another embodiment the nodes and the controller can perform the following: A: Detect defects on components, and record them or stop the production line, or inform machines further up the line of' the defects.

B: Return an image history of a given component on a given board as it progresses down the production line.

C: Return an image history of a given component on a run of boards on a single machine down the line.

A major advantage is ease of programming the nodes. This is mainly because there is relatively little local analysis.

For local analysis at the nodes the order of operations is Presence, then Polarity, then offset, then OCR. If a node has AXI (X ray system) or ICT (electrical test), capability, it should do Polarity, OCR and Offset tests. If there is a post reflow AOI machine, the nodes have no need to perform Polarity and OCR, as the AOI machine will do these. In all cases, offsets should be measured pre reflow as this information is very useful and will be lost once the board goes through the reflow oven.

The defect component histories (DCHs) need only be retained until the board under test has been repaired or until such time that the process engineer has analysed the DCH and attributes a cause to the defect and store it in a line defect analysis database (LDAD).

The LDAD could be use to guide line operators to fix causes of defects without having to consult the line engineer. Typical defects would have an assigned cause and so an assigned remedy. If an atypical defect occurs then the resolution of this defect would require the intervention of the process engineer.

Defect component histories would generally require board/panel-based barcodes in order to track the component through the process. Barcodes may not always be necessary. A time-based tracking approach could be used where the line production rate is known and the time taken at each stage is known. In this instance it could be possible to still create viable DCH's allowing the process engineer to view a sequence of candidate images. This could strengthen the case for manual closed loop.

It will be appreciated that the invention allows inexpensive monitoring of every stage of a production line. When a defect is found by AOI (or AXI, ICT) the user can better determine the process where the error was introduced. This analysis process can be manual (users looks at a series of images) or automated.

For post renow defect prevention - the expensive AOI equipment may be put after the oven with an inspection node installed post paste printing or between the chip device placement machine and the fine pitch placement machine. In this mode the pre-reflow inspection node provides limited component inspection measurement and full post reflow joint inspection. Here, the pre-reflow node could capture placement defects before they reach the oven, reducing end of line defects and improving yield.

For pre-reflow defect prevention, the AOI machine may be installed before the oven and possibly between the chip and fine pitch placement machines. Here the AOI machine provides inspection and measurement data with the aim of allowing the user to implement closed loop control on the SMT line. An inspection node would be placed after the oven, which would then provide data on defect that may be produced during the SMT boards run through the oven. - 12

The invention is not limited to the embodiments described but may be varied in construction and detail. - 13

Claims

Claims 1. A machine vision inspection system comprising a series of

inspection nodes each for location after a production station in an electronics circuit production line, each node acquiring an image, and the system collating a series of images representative of successive stages of production of a particular circuit.
2. A machine vision inspection system as claimed in claim 1, wherein each node acquires and stores images, and the inspection system further comprises a controller for causing retrieval of images from the nodes for a series.
3. A machine vision inspection system as claimed in claim 2, wherein each node scans an identifier on a received circuit and indexes the captured image on said identifier.
4. A machine vision inspection system as claimed in any preceding claim, wherein each node performs an in-line image capture without removing the circuit from a conveyor linking the successive production machines.
5. A machine vision inspection system as claimed in claim 4, wherein each node comprises a barrier for preventing further movement of a circuit beyond a registry position, and acquires an image at said registry 2 5 position.
6. A machine vision inspection system as claimed in any preceding claim, further comprising an AOI inspection machine which performs image processing to identify defects. - 14
7. A machine vision inspection system as claimed in any of claims 2 to 6, wherein the controller is connected to the AOI inspection machine; and wherein the controller automatically retrieves a series of images from the nodes for a circuit for which the AOI machine has detected a defect; and wherein the controller notifies the nodes of final positive testing of a circuit, and the nodes free up image storage locations of images of the circuit.
8. A machine vision inspection system as claimed in any preceding claim, wherein each node captures a full circuit image, and extracts from said image an image for each of a plurality of components and saves each of said component images.
9. A machine vision inspection system as claimed in any of claims 2 to 8, wherein the controller tracks progress of circuits progressing through the production line, and generates a user interface indicating this progress.
10. A machine vision inspection system as claimed in any preceding claim, wherein each node performs component presence, component polarity, and optical character recognition of printed data analysis operations.
11. A machine vision inspection system as claimed in any of claims 2 to 10, wherein a single controller is linked with a plurality of series of nodes, each for a production line.