X-ray inspection—Not all black or white
The high I/O densities of semiconductor packages mandate x-ray inspection techniques.
Irene Leszkowicz, Matrox Imaging -- Test & Measurement World, 4/1/2005 2:00:00 AM
|
|
Inspecting printed material
For More Information |
|
Read other articles from this issue:Table of contents, April 2005 FEATURES Combining talents Cellular or WiFi? X-ray inspection—Not all black or white Keep it safe Biological pulses |
Advanced automated optical inspection (AOI) can catch solder-paste defects as well as calculate area and volume, but it's not helpful when inspection targets are buried under component packages. Part 1 of this article (Ref. 1) covered mature optical inspection techniques. This part explores the newer edge-analysis and x-ray techniques that can meet the challenges posed by packaging technologies for applications requiring large numbers of I/O pins.These new packaging technologies emerged in response to the limited perimeter lengths of quad flat packs and other leaded technologies. Greater chip densities have led to extremely fine-pitch lead spacing, with delicate leads that easily bend and lift. Manufacturers can counteract this tendency by using the entire area of the underside of the package—as they do on ball-grid arrays (BGAs), flip chips, and chip-scale packages (CSPs)—to accommodate a larger number of I/O connections while maintaining a larger pitch that potentially allows for greater reliability and easier manufacturing. The major quality-control challenge is that the connections are not accessible for optical inspections or for easy rework.
These factors put additional importance on the solder-paste inspection discussed in part 1. Solder-paste inspection is necessary but not sufficient by itself, however, and automated x-ray inspection (AXI) has emerged to augment AOI for high-density I/O packages.
The use of x-rays also opens up the possibility of other inspection steps that could identify unreliable solder joints that will pass AOI and in-circuit test (ICT) but are at risk of future failure because of excessive voids inside the solder joints. The voids appear as lighter spots in a solder joint's x-ray.
Simple transmissive x-ray images of multilayer boards with components on two sides produce extremely cluttered and ambiguous images, with components and traces from each layer overlapping. To improve the images, vision manufacturers have developed techniques that display one layer in focus at a time while blurring the others.
|
|
| Figure 1. In an x-ray image, voids in solder balls appear as slightly lighter gray-scale rings (left), which can be located and highlighted (right) by contour finder tools. Courtesy of Soldering Technology International. | |
Two principal methods, laminography and tomosynthesis (Ref. 2), employ various combinations of x-ray sources and detectors to generate one or more images per scan. In each image, one focal distance is in focus; objects nearer or farther than the focal plane are blurred.
|
| Figure 2. An isometric x-ray view of wire bonds demonstrates how x-ray techniques can resolve depth relationships. Courtesy of Soldering Technology International. |
The images exhibit a wide range of gray values, with small variations in gray where thin objects overlap. For example, the detection of a properly wetted pad or a well-centered ball depends on inspecting for "slightly darker rings" or "well-formed concentric plateaus." Voids in solder appear as slightly lighter areas. Simple threshold and blob-analysis methods cannot handle the required multilevel gray inspections—what's required are image-analysis tools that extract contour edges from the gray-scale image or that analyze radial intensity distributions. Figure 1 illustrates how contour-finding tools with color augmentation can extract solder-ball void information from x-ray images and highlight them in color. To better resolve defects related to the depth, thickness, and vertical alignment of solder pads, balls, J-leads, and so on, some systems use oblique views while others use stereo views to try to resolve depth relationships (Figure 2). Analysis of 3-D oblique x-ray images remains largely at the research stage.
Towards a quantitative approach
|
| Figure 3. The streak highlighted here does not constitute an actual short, but it does extend more than 50% of the way across the gap separating one trace from its neighbor, making it the source of a potential field fault. |
High-speed AOI systems strive to classify a feature as good or bad based on simple measurements that are "local" to a small inspection window. Typical measurements include average intensity, percentage of pixels that are white, no interruptions in a dark band (that is, no bridges across), and no circular inclusions (voids) in a region. These kinds of inspections are important in answering the question, "Will this board function?" But they provide no information about reliability issues and are unsuited to answer the question, "For how long is this board likely to continue functioning?"
Reliability often depends on imperfections that are not quite faults (see "Inspecting printed material," below). For example, when "voids" exceed 5% of the volume of a solder joint, when a whisker of conductive material extends more than 50% of the way across a gap between traces, or when a spot is more than 13% out of true alignment, there is a high probability of failure after the product is in service and subjected to thermal cycling or mechanical stresses (Figure 3).
Evaluating the state of a component, connector, joint, paste, ball, or board in more precise numeric terms is necessary at several stages:
-
when developing a new process or technique; at this stage, the numeric information will help you determine what the optimal operating parameters will be,
-
when using "forensics" to track down a problem; a thorough evaluation will help you solve the problem and enable you to take steps to prevent similar ones from occurring; and
-
when preparing continuous process-control strategies; here, numerical data about the components will help you determine what parameters are important to monitor on a continuous basis and what their ranges should be.
These more precise evaluations often occur in a QA or research lab setting where the image-analysis tools will probably not be the same as those found on the production line. Therefore, you can consider using a wider set of tools, more adapted to variable conditions and capable of providing detailed numeric results. In a laboratory, you may not be able to set up the extremely controlled lighting conditions that are available in an expensive production inspection system, so simple pixel counting and blob analysis will be less accurate. Instead, you can use contour and crest finders to help determine shapes and boundaries with subpixel precision even with small gray-scale variations.
A lab setup may also offer greater variability in positioning and alignment of the object being imaged. To properly position an object, you can take advantage of new model-finding techniques that quickly align even complex images with large degrees of self-similarity. AOI systems used on a production line require masses of "setup" information for defining the large number of small inspection windows; this setup information is developed from the final product-layout drawings and documentation. In a lab environment, it is better to have tools that require little or no setup and that work more globally on larger sections of an image.
Emerging analysis tools also provide a way to directly measure features such as shape and position, and they measure fit, coverage, and deviation from expected values. In many cases, the quantities of primary interest are often derived results, or those calculated from a combination of direct results. Therefore, the development environment in which you apply the image-analysis tools can help you automate the collection of statistical information on results or can make it extremely simple to transfer results to other applications (such as Excel or Matlab). Easy ways of automating the collection and processing of data are also important if you need to perform the same analyses on a statistically significant number of images.
As electronics packaging and assembly have become more complex in the last decade, image-analysis tools and algorithms have likewise become more sophisticated. The increasing importance of complex gray-scale x-ray and thermal images means that now is the right time to explore beyond the black-and-white pixel-counting tools that power AOI systems.
| REFERENCES |
|
Inspecting
printed material
One approach to inspecting traces, and printed material in general, is to align a captured image with a reference image known to be defect free. Taking the difference between the "master template" and the captured image will result in "blobs" indicating excess material or missing material. The same technique is often used in print verification of logos and part numbers on chips.
Very thin features, whose width is approaching the resolution of the capture device, appear gray even in a high-contrast image. When you perform thresholding for blob analysis, these gray features may be close to the threshold value and may not be reliably selected.
|
|
| (Left) The edge of a defective trace comes too close for comfort to its neighboring edge; (right) an edge-analysis tool presents a view that highlights the problem. | |
Another consideration is that there is an allowable tolerance on the trace widths, so some processing is used to remove very thin chains of pixels in the difference image that correspond to acceptable differences. Unfortunately, this approach will also obliterate any actual faults that are extremely thin. Therefore, to identify thin defects like whiskers and scratches, you need to identify those thin features that do not lie along the edges of traces. An edge-analysis tool, such as the Matrox Imaging Library's Edge Finder module in crest mode, highlights cracks and whiskers.
Print defects in traces can also be tracked down from deviations from straight edges, combined with distance measures to the nearest neighboring trace to see if an incursion is deep enough to be a defect (see figure).
No related content found.
- 0 rated items found.
Datasheets.com Electronic Parts & Inventory Search
185 million searchable parts
- Part Number
- Description
- Inventory
- Products
- Manufacturers
View All Webcasts
