• Font Size:
  • A
  • A
  • A

Feature Articles

X-Ray-Based Machine Vision - Part 1

by Nello Zuech, Contributing Editor - AIA

In the electronic industry there are compelling reasons to adopt X-Ray-based machine vision systems. The cost of failure becomes significant when faulty product is shipped to a consumer. Detecting rejects immediately after a value-adding step such as soldering can significantly reduce warranty costs, field repairs, etc.

Furthermore, the shrinking sizes of passive components (resistors as small as 0.02' X 0.01') and finer line widths of printed circuit boards all reflect the trend to more dense board designs as designers attempt to get more functionality on less real estate. Consequently, solder joints are ever more critical as they in turn are smaller and closer together, making them more difficult to inspect by people.

Another reason for the adoption of X-Ray-based machine vision systems is that even if people can inspect solder joints, studies have demonstrated repeatedly that people are very subjective and the results of inspection can not be duplicated from inspector to inspector or even by the same inspector. The result of using human inspectors to inspect solder joints, especially where the population density of a board is high, is inevitably a high incidence of poor solder joints escaping detection while at the same time a high incidence of solder joints being falsely perceived as rejects. This brings up the third reason for using X-Ray-based machine vision systems to inspect solder joints, That is, a high incidence of false rejects leads to excessive rework and inevitably that rework results in damage to the board creating scrap.

Another rationale in the case of high density boards where test probe points are difficult to establish is that X-Ray-based machine vision systems can eliminate the need for in-circuit testing or at the very least increase yields at in-circuit and functional testing.

Studies have also suggested that 25% of all post-solder defects cannot be detected electrically. These include fault types such as insufficient solder, cold solder joints, misaligned components, voids and excess solder. All of these can be detected by X-Ray-based systems.

Applications in Electronics

In the electronics industry, X-Ray-based machine vision systems are used to inspect the inside of integrated circuit packages to assure package and interconnect integrity (perhaps an application more identified with the semiconductor Industry), to align board layers in multilayer manufacturing, to inspect multilayered boards before hole drilling to qualify hole-pad alignment and to inspect solder joints. By far the largest market is the last application. There are a number of reasons why the solder joint inspection application is being more widely addressed by X-Ray-based machine vision systems. With the increased adoption of chip scale technology, such as ball grid arrays, and the result that solder joint interconnects are hidden from view there is no way that people really perform a comprehensive inspection. Similarly, conventional machine vision systems can not see the solder joints.

Technology Readiness

Significantly, as the soldered assembled board application becomes more demanding, the technology that serves as the basis for responding has also emerged. Microfocus X-Ray systems are now available compatible with the requirements to detect artifacts a few microns in size. Solid-state cameras including those made of amorphous silicon are available with the properties required to handle the dynamic ranges of X-Ray images. Microprocessor power has evolved to the point where three-dimensional data can be processed, modeled and analyzed fast enough to keep up with many production lines.

Types of Systems

Not all X-Ray systems used in the electronic industry are machine vision-based. It turns out there are various degrees of 'machine vision-ness.' Many in the industry still rely on real-time X-Ray systems. These are no more than closed-circuit TV systems where the image is derived from an X-Ray of the scene. In these cases, the operator makes the decision regarding solder joint quality, so there is still substantial subjectivity in the decision process. These systems are often used to perform line set-up analysis - first piece approval, sample inspection, or as an engineering tool to debug the manufacturing of a new board design.

The next iteration is one that incorporates machine vision-based software to enhance the X-Ray image to reduce noise, increase the contrast separation between acceptable and unacceptable conditions and/or enhance edges. Some of these packages come with additional tools such as 'calipers' to aid an operator in making dimensional checks on artifacts such as solder voids. Some come with specific tool sets for specific components; e.g. BGA, flip chips, etc. Some come with sufficient memory to store the images of 30 or more sectors of an assembled board giving them the ability to make comparisons against the stored references to detect changes that could signify a solder concern.  Some of these systems are further  'automated' in that there stages can be programmed to follow a specific path or sequence to assure all solder joints are inspected systematically by the operator.

The ultimate class of these X-Ray systems are those that incorporate machine vision-based automated defect detection capabilities as well as machine vision-based image processing and enhancement. These do not require any operator intervention to determine if a solder joint is acceptable or not. These systems have typically been used in an offline scenario but some are emerging that are fast enough to be used for many online applications as well.

Fundamental Technical Approaches

There are two fundamental approaches to these fully automated X-Ray-based machine vision systems: transmission and laminography. In the transmission-based system defects can be correlated to the optical density map of the scene, which, in turn, corresponds to the attenuation map stemming from the actual object density variables in the X-Ray image. While these systems can tell whether there is sufficient, insufficient or too much solder at a joint they can not tell much about the shape of the solder joint itself. Solder purists suggest that the quality of a solder joint is as much a function of the amount of solder as it is the shape of solder fillet. Another potential problem with transmission-based systems is their ability to reliably inspect double-sided boards because board and component properties may hide or alter solder joint images.

The laminography approach essentially takes cross-sectional views along the z-axis of the solder joint so it examines not only the two-dimensional optical density profile but it does this for each slice or cross-sectional view. In other words, a three-dimensional picture is captured of the solder joint and that is the profile that is essentially compared to the 'golden' profile for that specific solder joint. Double-sided assemblies can be inspected in one pass.

Which to Buy

The type of system that one buys will be determined by how it will be used. If it were to be an engineering tool, a real-time X-Ray system with image analysis software would probably be sufficient. If it is to be a process improvement tool then a fully automated system including machine vision-based automated defect recognition tools will be required. In the latter case, considerations include type of facility - low volume/high mix, medium volume/medium mix or high volume/low mix. In the latter case systems may struggle to perform 100% inspection - that is inspect 100% of all the solder joints on 100% of the boards produced. In this case it may make sense to combine an automatic optical inspection system to inspect the more conventional components and use the X-Ray-based system to inspect those components whose designs tend to hide the solder joints. Certainly on those lines where high value boards are produced or boards are produced whose failure would lead to security or safety issues (e.g. many medical devices, automotive, military), then a fully automated system should be considered.

Ken Pelzel of Agilent Technologies made the following observation regarding applications. 'A better way to look at the problem is board complexity, repair strategies, and quality goals. Low mix, high volume which are simple, single-sided consumer products which are not repaired and there isn't high end user expectations on quality, then X-Ray is not a good fit. Conversely, low mix, high volume, high complexity, double-sided consumer products such as a laptop is a reasonable fit.'


As suggested in our earlier article on machine vision-based automatic optical inspection systems for assembled board applications, the purchase of an automated, machine vision-based X-Ray system can be easily justified. The analysis suggested in that article applies as well to these systems.  Factors in justification include predictions of yield improvement, scrap reduction, rework costs, etc. The greater the board component density, higher throughput required, more inaccessible test nodes and higher the logic complexity the easier these systems are to justify as the first-pass yield will be proportionately greater where such systems are used as process monitors and immediate corrective action is taken as required to avoid the production of defects.

Questions to Ask Prospective Vendors of X-Ray-Based Machine Vision Systems

In choosing an X-Ray-based machine vision system many of the issues associated with the selection of a machine vision-based AOI system apply. As you examine the products from different vendors you will find most make the same claims. It is clearly important to get them to put their claims in writing. The following questions are meant to provide the framework for a systematic analysis of the competitive landscape. The answers given should be consistent with the application requirements anticipated. This list is not meant to be complete. Because of different quality management philosophies within the board assembly industry, the set of questions used should be consistent with your own espoused quality strategy. 

  1. For which applications does your company offer products: post reflow, post wave solder? Other?
  2. Regarding your system:
    1. Is your system an on-line or offline system? NOTE: If both, please answer the following questions for each style.
    2. Does it perform 100% inspection or sample inspection? In either case, please provide some measure of board density versus throughput? E.g. For board with 4000 components on an 8' board, system can handle 2 sq. in/second or whatever.
    3. Is your system 2D or 3D?
    4. What does the system do and what are the specs?  In italics are the likely responses to anticipate.
      1. Component presence - most likely do
      2. Component missing - most likely do
      3. Correct component - unlikely
      4. Polarity - for some components
      5. Orientation - for some components
      6. Misplaced/offset - most likely do
      7. Skewed - most likely do
      8. Tombstones - most likely do
      9. Solder presence - yes
      10. Insufficient solder - most likely do
      11. Solder bridges both between leads and between components - most likely do
      12. Solder wick - most likely do
      13. Cold solder joint - depends
      14. Dewetting - depends
      15. Solder voids - yes
      16. Bent leads - most likely do
      17. Lifted leads/chips - most likely do
      18. Solder balls - most likely do
    5. What is the finest pad pitch that can be handled?
    6. If 3D what does the system do and what are the specs? Accuracy, repeatability of measurements, etc.?
    7. Do you have a recommended calibration procedure to demonstrate accuracy of the system and, if so, what is it?
    8. If 3D, how is the height of the solder joint measured - based on a local reference plane? A global reference plane? How does the system handle board warpage issues?
    9. Can your systems handle lead-thru-hole components?
    10. If 2D is system based on area camera or line scan camera?
    11. If 3D can you describe the fundamental underlying principles for capturing 3D data?
    12. How is the system trained to handle a new board design? CAD compatibility? Gerber file compatibility? Train-by-showing? Other? Combination?
    13. How long does it take to train on a new board?
    14. What is the changeover time where boards have been previously trained?
    15. What is the throughput at what specific pixel size?
    16. What is your false reject rate? Escape rate? How have these been demonstrated?
    17. Is there an action that takes place if there are 'x' number of consecutive rejects at the same location? Or 'Y' over the entire board?
    18. Is the system design based on your own proprietary hardware or commercially available products such as frame grabbers or vision processors or is it a host-based processing system?
    19. Do you offer an upgrade patch for future generation products?
    20. How many cameras does your system have and why?
    21. Does your system have the ability to adapt field-of-view/resolution as a function of the board design for a specific board design?
    22. Does your system have Internet trouble-shooting compatibility?
    23. Can you comment on your system's suitability for: low mix/high volume operations, medium mix/medium volume and high mix/low volume?
    24. Does the system interface to a rework station? Do you offer a rework station?
    25. What is the price range of your systems?
    26. What options, if any, are offered for your system?
    27. Vendors

This table depicts some of the vendors known to be selling X-Ray-based systems into the North American electronic market. It is understood that there are over 12 companies aggressively pursuing these machine vision-based X-Ray system applications in the worldwide market. The companies in the table are understood to have delivered machine vision-based X-Ray systems into the North American electronic market. An attempt has been made to identify the specific applications in electronic manufacturing the respective companies suggest their products address. This data was largely obtained from information on their website.

When posed this question, Ken Pelzel of Agilent Technologies suggested the following:

"The buyer needs to commit resources to the machine for programming and maintenance. He should also make a commitment to track the data in order to realize the most substantial payback." Insofar as the vendors, Ken suggested, "the vendors need to provide a suite of support products that meet the buyer's needs."   



There are currently no comments for this article.

Leave a Comment:

All fields are required, but only your name and comment will be visible (email addresses are kept confidential). Comments are moderated and will not appear immediately. Please no link dropping, no keywords or domains as names; do not spam, and please do not advertise.

First Name: *
Last Name: *
Your Email: *
Your Comment:
Please check the box below and respond as instructed.

Search AIA:

Browse by Products:

Browse by Company Type: