The data are useful, but at times overwhelming. Here’s how to cut them to size.

Each AOI or SPI model offers different advantages such as inspection coverage, speed, programming time, vendor support or cost. However, general use models are consistent: Each performs user-defined inspections of a board and provides lists of suspected defects and data for each component inspected. Depending on the implementation strategy – SPI (3-D or 2-D), pre-reflow, mixed mode, post-reflow, or post-wave/final assembly – there is ample information available to the user. Again, depending on the implementation, there will be inherent differences in the data and their usefulness. Certain data will be more applicable for different users, but determining what is available, where and for how long is important to inspection success.

AOI and SPI generally provide measurement data, defect calls and defect images. All this is important, but how can a user quantify and use the information efficiently? How long should data be stored? What information should be readily available and what should be archived or removed? A post-reflow AOI system can generate offset, presence scores, joint measurements and additional component data for every device on the board. On a laptop assembly, this could result in 40,000 data points for every board inspected. For manufacturers running three shifts, that means nearly 200 million data points a day! Should all these data need to be archived infinitely? Should operators be shown every SPC chart known to man? How long are images of the defects useful?

When implementing AOI, certain key steps need to take place.

First, there should be internal meetings with all levels of users to determine the facility needs. Operators and technicians should be involved in the brainstorming sessions, along with engineers and managers, to discuss all variations of needs and practical implementation. Important points to be considered include data storage capacity, customer requirements, liability, factory software databases, and how each group would like the data presented. A prioritized list of the needs is important for the manufacturer and vendor, as this will help determine what is required before machines are moved inline and what can occur after the fact.

Next, sit down with your vendor and determine available data, how long they are stored, and available format(s). The vendor will probably have standard formats available but may have several other options if they understand your goals and needs. Images may be stored locally on the system, but scripts or configurations may need to be deleted or images moved to defined locations.

Third, ensure everyone follows through on and executes the plans. If the goal is to use AOI to improve the process, the relevant information must be provided to manufacturing in real-time so that potential issues can be corrected quickly. If the goal is for liability purposes and customer reports, it’s important to determine what data need be kept and for how long. Backup strategies and accessing the information should also be at the top of the list.

Finally, consider other test and measurement equipment in your production facility. How can the data be used to improve processes or better mold your overall manufacturing and test strategies (Table 1)?


Some of the pitfalls users should try to avoid include not involving operators/technicians in the discussion. These individuals are the front line in defect prevention and process improvement. If not involved in the discussion, they may resist the equipment or strategy. Also, they have firsthand experience with what works well in production and what information would be valuable for their jobs.

Also, AOI strategies at one company may not work for others, and all the experience in the world cannot account for simple human error. Create a simple test strategy, implement it and remain flexible to new ideas and approaches. Users often get hung up on certain charting types or data presentation. Remain open to new ideas and take suggestions from all users, as some of the best ideas come after the systems are installed and users gain experience.

Start out simple and review data in manageable segments (Figure 1). Inspection systems generate ample data, and it can be exciting to think of the Six-Sigma style projects that can be implemented to make the process perfect, yet too much information will slow down – and possibly halt – even the best of plans. Start by analyzing one variable or process at a time. For example, analyze the top five defects caught by AOI and determine their causes; then move to the next five. After that, perhaps analyze issues that cause false failures or look downstream and see what part types are escaping the system. The point is to keep the implementation simple and build the test strategy so that it’s not too overwhelming.


AOI and SPI data can provide a lot of useful information; however, the volume of the data can be overwhelming. Manufacturers should assess their assembly and test environment. An implementation alignment team should be formed to determine the goal of the equipment and use model. Vendors should then be engaged to review available information and possible implementation. The value of AOI is not only as a defect detection system, but also as a tool to improve process and test strategies. ROI can be greatly improved through good implementation plans that include the best ways to use available measurement and defect data.

Jeff Bishop is product marketing engineer at Agilent Technologies (agilent.com)