Automated optical inspection (AOI) has gone through a number of evolutionary steps and, with each increment, improved the value delivered to electronics manufacturers as an element of an overall test and inspection strategy. From its humble origins as a replacement for inconsistent, human-based inspection to modern state-of-the-art systems that minimize defects, boost yields, and cut repair costs, today’s AOI systems have won a deserved place in the modern high-volume surface-mount (SM) production lines.
The value of AOI was established when electronics manufacturers recognized successive improvements in labor costs, scrap rates, and rework rates. Product and process quality improvements realized along the way also were significant.
Over time, AOI became viewed as an adjunct to in-circuit test (ICT). In some cases, the delivered coverage of AOI has enabled manufacturers to reduce or eliminate ICT coverage in certain products, with proportional reduction in ICT fixture, programmer, and test costs.
Since the salad days of AOI, the technical inspection requirements for machine vision have intensified substantially. The improvements in AOI capabilities have not emerged as rapidly.
Each generation of incrementally more powerful portable electronic products is useful as a barometer of the sophistication and capability of that era of manufacturing. By way of example, today’s mobile-phone technology yields products that are far smaller and lighter and have a greater range of functionality, incorporating applications unrelated to telephony such as still cameras, live video, personal organizers, and high-speed wireless web access. To accommodate this expanded functionality within the ever-smaller size demanded by the end users, manufacturers must design and produce denser, high-complexity printed circuit boards (PCBs).
The changes that make these portable electronics possible are several: increasing semiconductor content, reduction in size and power of silicon die, innovative types of packaging, increased connectivity, and the use of more complex, multilayer PCBs. Designs for small, dense PCBs often restrict or eliminate the traditional probe access for electrical process or ICT, with the close proximity of RF, video, audio, and digital functions that mandate increasing amounts of shielding.
The progress of the commodity personal-portable electronic product segment is a harbinger of things to come for the general PCB assembly industry. The experience and breakthroughs in packaging, shielding, and component technology used in personal electronics eventually will find their way into other areas and products where increased functional density, size, and cost drive innovation in different dimensions.
The Need for AOI
The technological innovations that enable manufacturers to deliver smaller, highly complex products to consumers create application and use headaches for traditional ICT. In particular, the resulting loss of access for bed-of-nails ICT has driven manufacturers to implement automated optical inspection technology as a noninvasive countermeasure to ICT loss of access. While this is a significant opportunity for AOI suppliers, it highlights a number of challenges that remain in AOI: improvements in programming efficiency, reliability, and flexibility.
Maintaining a high yield environment is more difficult for high-density boards. Large numbers of small components result in a higher probability of defects, and higher interconnect and assembly complexities create more opportunities for defects.
Although the defect production rates (measured in defects per million operations or DPMO) in modern processes appear mathematically low, growing board complexity represents larger numbers of opportunities for errors. The number of defects produced equals the probability multiplied by the opportunities, so a low error rate can yield unacceptable numbers of defects for sufficiently complex boards.
AOI systems primarily are successful in high-volume applications and, in some cases, have replaced electrical test applications where board volumes and process stability are established. However, the Achilles heel for AOI is deployment in lower volume or moderate mix applications.
Where increased process and induced variations such as many lot changeovers are common, a conventional AOI shortcoming is apparent: needing sufficient board volume and programmer skill to set nominal operating parameters. These systems use human judgment to establish the programmatic standards for good and bad and what range should exist around each decision. How good is good enough depends on judgment and initial results.
Because conventional AOI systems depend on empirical programming methods, their competency is established over time and volume, using statistically measured values and volumes as proxies for good and bad. Where unforeseen process variation exists, the conventional programming paradigm of AOI is challenged to accommodate it. Where time and volume are unavailable to establish these stable test norms, it is difficult to deploy and run AOI effectively.
Conventional AOI faces another challenge in process variation (Figure 1). Once set, conventional AOI thresholds are fixed. As machinery, environment, and operators constantly change, such normal variations expose another limitation of AOI—programming thresholds do not vary.
As a consequence, when operating or process conditions diverge from the expected thresholds, the generation of false fails, good results being judged bad, occurs. Changes in process that are not already established in the program give rise to false fails.
With production changeovers, different component suppliers introduce sources of substitutional variances that cause false flags. Such variances are episodic and can’t be easily accommodated in an inspection program a priori.
Where volumes are small or the mix of PCBs increases, variation exposes the limits of conventional AOI. AOI requires users to compensate for variation, but the variation cannot be compensated for unless first seen by the programmer. A vicious circle results.
Today’s lean manufacturers require a new form of advanced machine vision systems to adequately compensate for loss of physical access. Smarter machine vision solutions would allow manufacturers to succeed with smarter, noninvasive alternatives to defect detection while providing process quality checking to detect error conditions in the first place.
The next generation of optical machine vision must yield solutions that automatically comprehend and compensate for the practical, day-to-day operational variation in SM processes. This generation of machine vision systems must embody built-in awareness of the process and its variation and of the objects being viewed.
These machine vision systems must replace conventional exhaustive programming efforts, eliminating heuristic methods and judgment of the human programmer. Machine vision systems need to be equipped with truly automatic solutions that know the structure of the objects they inspect and the context of the objects relative to their backgrounds.
Said differently, these systems would replace the domain knowledge today supplied by human programmers with internal, standardized data. Such features would reduce or substantially eliminate the requirements for user skill consistency and experience as proxies for process data.
AOI systems have evolved substantially since their infancy. They are very sophisticated, have a high degree of user flexibility in imaging conditions, and possess multiple lighting solutions, multiple angle lights, multiple cameras, and fancy timing and control systems to bring it all together. Knowing how to deploy these hardware assets is the sine qua non of the next generation of optical test systems whose performance will challenge current electrical test methods for efficiency and ease of use.
New Approaches to AOI
A new generation of robust machine vision systems must be developed that works in low-volume/high-mix environments, using far more sophisticated back-end analysis of images to compensate for part and process variation. Machine vision suppliers are beginning to listen and respond to these needs. Recent developments in machine vision improvements have come in two basic forms that promise to change what we call inspection.
Statistical Modeling
One promising method is statistical or appearance modeling, an automated means to learn and define library entries from good and bad examples distilled from the initial production. Statistical methodology automates the process of defining good and bad.
Statistical appearance methods allow users to input image data collected during prototype or initial production and then automatically compile the examples into a composite good or bad database. The stored samples are normalized for rotation, size, and shape and collapsed into eigenvectors, a set of internal representations of the collective basis appearances for subsequent use. As new examples of good and bad are collected, the library entries are further annealed to eigenvector entries, then adjusted for the incremental variations.
With statistical modeling, AOI does not make manual iterative adjustments to account for variation. Statistical modeling automates this key step, removing the manual tasks of evaluating the impact and making the manual library adjustments. Statistical modeling also automates the process of comparing new images to stored eigenvectors, simplifying runtime operations.
A potential downside to statistical modeling is that it’s still statistical. It operates based on samples culled from a larger population, and it still requires sufficient examples that mimic the larger population to achieve workable library results. As a result, statistical appearance-based techniques for PCB manufacturing need substantial sampling to achieve workable (~100 ppm) false-flag rates.
Statistical appearance techniques depend on initial samples of good and bad assessed by human programmers. For that reason, there are opportunities where good and bad judgments can be wrong, leading to errors in eigenvectors and later coverage or defect detection issues that may go undetected.
Configural Recognition
A second promising method uses combined scene and object recognition. One such solution, configural recognition, is based on detecting objects and scenes made of objects contained in an otherwise ordinary image (Figure 2). Such recognition techniques are based on previously defined spatial relationship and variance allowed in key object regions and the relationship and gradient of those regions to other key regions in the image.
Configural recognition has been integrated into visual search engines and shown to successfully select graphic objects as in a web search. With configural recognition, the knowledge of what good and bad pre-exists in the data coded into the application and is not based on statistical or large numbers of learned or taught examples.
Configural recognition operates with comparatively few practical examples and, in theory, still can be effective with as few as one example. As a result, configural recognition holds great promise for new product introduction and low-volume, high-mix applications. Also, the hardware required for configural recognition is simplified so only a static, fixed light source and a camera are needed.
On the downside, configural recognition is comptationally very intense and, until very recently, outside the realm of possibility for low-cost computing. With the advances in clock speed, multiprocessing, and parallel computing price/performance, configural recognition is a near-term reality. Configural recognition also requires that precise process knowledge be delivered in the application.
Summary
Optical inspection currently is popular in high-volume applications, and future machine vision developments promise extraordinary improvements upon current possibilities. AOI has demonstrated that machine vision can work. Now the challenge for suppliers of equipment is to expand the technological solutions to deliver the true promises of machine vision: ease of use, comprehensive performance, and reliable diagnosis.
About the Author
John Arena is the marketing manager for inspection and test products at Teradyne. He has been employed by the company for 23 years, most recently in new business development and marketing roles. Mr. Arena holds a B.S.E.E. from Rensselaer Polytechnic Institute and an M.B.A. from Boston University. Teradyne, Assembly Test Division, 600 Riverpark Dr., North Reading, PA 01864, 978-370-1012, e-mail: [email protected]
FOR MORE INFORMATION
on AOI enter this rsleads URL
www.rsleads.com/305ee-221
Return to EE Home Page
Published by EE-Evaluation Engineering
All contents © 2003 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.
May 2003