How to Select an AOI System

With the needs to reduce cost, increase output, maintain quality, and follow technology trends of smaller components and higher-complexity printed circuit boards (PCBs), it is becoming ever more difficult for manual visual inspection to keep up with today’s pace of electronic manufacturing. Help, though, is available in the form of automated optical inspection (AOI) systems.

Successful use of AOI systems results in reduced manufacturing costs and improved product quality. So why are AOI systems not more prevalent on manufacturing floors?

AOI systems, available for more than 10 years, are a branch of machine vision systems that use image-detection hardware and image-analysis software to detect assembly defects in PCBs. Early generations of AOI systems suffered high false-call rates and disappointing inspection results, leaving many early adopters with AOI systems sitting off-line and unused.

Technological advances have brought vast improvements to AOI systems. But because of the extreme expectations for AOI systems, only a few have been implemented satisfactorily. To ensure success, you must be fully aware of what the newer AOI technologies can deliver and how they can reduce your manufacturing costs.

Setting the Correct Expectations

If one recommendation could be made, it is know the details about what faults you need to detect. All shorts and opens are not equal. Some faults are detected exceedingly well by AOI systems, and some are detected poorly or not at all.

If you can match the faults you need to find with the strengths of AOI systems, you will not be disappointed. In concert with other test and inspection systems, such as in-circuit (ICT) or X-ray inspection, a complete spectrum of faults can be found (Figure 1).

For example, if lifted leads are faults you wish to detect, any of the three technologies could be a viable solution. Before making a decision on which technology to implement, look into the reason for the faults. Are the leads lifted due to insufficient

solder? Are they bent upward during the placement process? Did the solder not flow properly in the soldering process?

AOI will detect leads that are physically separated from the solder. ICT will discover the lifted lead if electrical contact is not made. X-ray technology will find leads that are lifted due to bending, insufficient solder, or cold solder joints. Knowing the details of your needs and the inspection competency of the technologies available will help you have realistic expectations for AOI as a tool to use alone or together with ICT and X-ray.

Location, Location, Location

Post-paste, pre-reflow, or post-reflow? Many AOI systems can be placed in all of these locations. The correct strategy depends upon your inspection needs. Know the trade-offs involved in selecting any of these strategies (Table 1).

Benefits

With its strengths in mind, AOI can be correctly incorporated into your manufacturing-line inspection strategy for maximum results. But how do you go about selecting a particular system among the many choices available? Here are some key areas to evaluate.

False Calls and Escapes—False calls are defect indictments (calls) that are actually not defective. Escapes are defects that are not indicted. In a good AOI system, both of these should be low.

An AOI system that attempts to inspect for marginal solder joints (or any other defect outside its natural capabilities) will have a higher false-call rate. This adds time

and cost through rework and reinspection. High escape rates run the risk of sending suspicious products to your customers, only to have them return later—defective.

Fault Coverage—The investment in AOI will have a higher return on investment (ROI) when used in the manufacturing line where fault coverage can be maximized. If placement faults and solder faults are a high percentage of your fault spectrum, then a post-reflow AOI system may be the right solution for the highest ROI. The choice will depend upon the application, product life cycle, and production processes already in place.

A product’s test and inspection strategy using AOI may be mixed with a combination of ICT, X-ray inspection, and functional test to further increase fault coverage and ROI. As the assembly process matures, fewer assembly defects may be anticipated, lengthening the break-even time.

Throughput—Keeping up with the output of the manufacturing line will be a continual challenge faced by all AOI systems. Speed through the AOI system often results in a trade-off in fault coverage that, in turn, leads to higher levels of false calls and escapes.

There is no doubt that throughput is a key decision parameter. But keep your priority on quality (low false calls and escapes with high fault coverage), and make sure that continual throughput advancements are viable without sacrificing the initial AOI system investment.

Choices

In AOI, there are a myriad of solution implementations from which to choose: flashing strobes, moving cameras, moving boards, line-scan, rings of color, domes of

light, multiple-camera systems. None of them has yet to demonstrate its superiority over the others.

Until a standard implementation emerges, investing in a platform that uses technology based on industry standards will help guarantee that advances made in these industries are leveraged by your AOI system. Faster processors, better cameras, and stronger lights probably will be needed, so allow for their evolution.

For the hardware, keep it simple. A simple hardware implementation with technology based on industry standards will make installation, operation, training, and maintenance easier and overall cost of ownership less expensive.

The core competency of an AOI system is in the image chain and image analysis. Good data into good algorithms means good data out.

Some other choices to consider are:

Source of Programming Data—CAD data-based inspection allows for the manufactured product to be inspected according to the design. Learning or self-programming systems have higher false calls and escapes. This is analogous to learning a foreign language from non-native speakers. There is acceptance as long as you only speak with those nonnative speakers. But once exposed to native speakers, the faults are obvious.

The learned systems rely on the manufacturing process and the condition of the known-good board. There is no native data to expose faults; consequently, no actionable information to help control the process. Look for an AOI system that bases its calls on the original CAD design.

Image Capture—Clear images correctly represent the product and capture a large number of component and manufacturing features. Compromises made here will be reflected directly in the quality of the results. System complexity at this stage will not necessarily guarantee a higher-quality image and may only add to the knowledge needed to program, operate, and maintain the system.

Algorithm Theory—The operator uses the results delivered by the AOI system to make an informed decision. With more information, a better analysis can produce better results.

The level of confidence the operator has in the results is based directly on the amount of information used by the AOI system to produce the result. Low false calls and escapes with high fault coverage also help increase the confidence of the operator in the AOI system.

Use Model—AOI systems should use a common-use model and operating system for programming and operating the system. This makes training, system management, and capability expansion much easier.

Vendor Experience—Work with AOI vendors who can provide the depth of knowledge about AOI as well as the breadth of knowledge regarding manufacturing test and inspection needed for various applications.

Conclusion

Setting the correct expectations and understanding the capabilities and limitations of AOI will help you to stop investigating and begin implementing AOI as a valuable step in the manufacturing process. Implement your AOI system within its

capability to detect faults. With the successful implementation of AOI, the bottom line will be a reduction in the cost of manufacturing.

About the Author

Joanna Forbes is a product marketing engineer for the Manufacturing Test Division at Hewlett-Packard. Before joining HP, she lived and worked in Japan. Ms. Forbes received a B.S. in mineral engineering mathematics from the Colorado School of Mines and a master’s from the American Graduate School of International Management. Hewlett-Packard, Manufacturing Test Division, 815 14th St. S. W., Loveland, CO 80537, (970) 679-5358.

Table 1.

Copyright 1998 Nelson Publishing Inc.

September 1998