Faults Can’t Hide From X-Ray Inspection

An engineer becomes very frustrated when told that his or her design is breathtakingly elegant, futuristic, and extremely powerful; but, unfortunately, it can’t be produced as a commercial product. Of course it can be built, but testing it economically may be a different matter altogether.

There’s no doubt that circuit density—the number of components in a given area—has increased greatly as ball grid array (BGA) and chip scale packages (CSPs) have gained acceptance. These types of IC packages are only slightly larger than the encapsulated die, but nevertheless they provide a large number of connections because their contact locations are not limited to the edges of the devices. Instead, an array of contacts covers most of the device-mounting surface.

In addition to BGAs and CSPs with hundreds of contacts on one device, minute new packages have been developed for less complex ICs requiring from four to eight connections, for example. Together with shrinking passive components such as 0201-size resistors, approximately 0.02² × 0.01², and increased use of fine-line PCB etching and multilayer technology, printed circuit boards (PCBs) have become remarkable structures.

According to a recent article by three test professionals from Lucent Technologies, “[PCB] assemblies that exceed a 5,000-node count are a concern to us as they approach the resource limit of our current in-circuit test (ICT) equipment….For example, one design currently on the drawing board has approximately 11,600 nodes, over 5,100 components, and over 37,800 solder joints that require testing or verification. This unit also has BGAs on both the top- and bottom-side of the board with the BGAs located back-to-back.”1

Lucent previously had determined that there were approximately 30 missed contacts typically per million probe attempts. On boards with as many as 5,000 nodes, this represents a significant portion of ICT detected faults. This meant that, in spite of the increased test coverage required by more complex assemblies, adding ICT capabilities wasn’t the answer. Instead, the company determined that a mix of fewer ICT contacts together with X-ray inspection would be a viable solution.

Actually, X-ray inspection has been used for many years on so-called mission-critical assemblies as a way to gauge solder-joint quality. However, other than the medical, space, and defense industries, few manufacturers adopted it for routine PCB inspection. That’s changing because there’s no other practical way to inspect assemblies with BGA packages and CSPs. The solder balls that attach them to the PCB are underneath the packages, so the solder joints can’t be visually inspected.

Any Old X-Ray System Won’t Do

If all you need is a two-dimensional (2-D) view of the overall PCB, maybe any old X-ray system really will do the job. But if you want to inspect solder joints having defects a few microns in size, you need a modern microfocus system, perhaps one with three-dimensional (3-D) capabilities as well.

It all starts with the X-ray tube. The density of material that X-rays can penetrate is proportional to the tube voltage. But the higher the operating voltage, the more likely arcs or discharges are to occur inside the tube, making the tube gassy and reducing its lifetime.

One way to avoid this problem is to include a vacuum pump with the X-ray system. Then, each time it is used any gas residue from discharges is removed as the new vacuum is drawn. Tubes that operate in this manner are called open tubes because they can be disassembled and their constituent parts replaced as necessary, extending their life indefinitely. They usually are constructed of steel, provide a very sharp focus capable of detecting 1.0-µm details, and can operate from voltages as high as 225 kV.

Closed tubes typically are made of glass and, as the name implies, cannot be disassembled because they are permanently sealed when made. The highest operating voltage for a closed tube is about 130 kV, and their focus is limited to detecting 5-µm details. A reasonable lifetime for a closed tube is 5,000 to 10,000 h.

An important consequence of the constructional differences between these types of tubes is the magnification available. Sure, you can expand an image after it is formed, but here we’re considering only the geometric magnification defined as the fraction: focus-to-detector distance/focus-to-object distance.

Because the focus-to-detector distance typically is 500 mm to 600 mm for almost all X-ray systems, magnification depends on the focus-to-object distance. For closed tubes, this dimension varies from 9 to 19 mm; for open tubes, it is 0.4 to 6 mm. Consequently, the range of magnification for closed-tube systems is about ×26 to ×60 and varies from ×100 to ×1,400 for open-tube systems.

In the comparison chart that accompanies this article, magnifications greater than 1,400 are achieved by combining geometric magnification, optical magnification after the detector and ahead of a camera, and pixel zooming after the image has been digitized. The field of view is inversely proportional to magnification. At a given magnification, the field of view will be larger for a larger detector.

A conventional image intensifier provides about 4.2 line pairs (lp)/mm at a geometric magnification of 1.0. Higher values may be quoted, but generally include magnification >1.0. Glenbrook Technologies is an exception because its scintillator actually has >22-lp/mm resolution.

Seeing the Light

X-rays are invisible but scintillating materials, such as cesium iodide, emit light pulses when struck by X-rays. By using a detector made from such a material, a visible image can be formed. For many PCB inspection systems, the image is magnified optically and viewed by a TV camera. In medical applications and single-image fault-detection systems used to inspect welds, for example, the image may be transferred to photographic film.

Major advances have been made in detector technology. The story begins with the efforts of flat-panel display manufacturers to develop large active matrix display panels. Amorphous silicon (a-Si) has become the technology of choice for most companies because fabricating thin-film transistors requires only a few process steps, and their performance is adequate. Processes developed for solar cell and CD production already are well-proven and have been scaled up to meet the needs of the larger display panels.

If a-Si technology is used to produce a large photo diode array and the array is excited by light from a layer of cesium iodide, a very large-area X-ray detector results. For example, a 12² × 16² model available from Feinfocus has 2,240 × 3,200 pixels (diodes) on a 127-µm grid. Each element is connected to readout circuitry so that an image can be scanned out of the device. The basic resolution is similar to that of the image intensifier—about 4 to 5 lp/mm.

Commenting on the applicability of a flat-panel detector (FPD) to PCB inspection, Vikram Butani, manager of product development and sales at All these approaches present near-photographic quality images of solder joints that can be used to check for faults. Of course, other than for off-line quality-control investigation, large-scale solder-joint inspection must be automated if it is to be routinely used as part of the production process.

Generating 1,000 Words From a Picture

OK. You’ve got a clear, detailed image of the solder joints on one device. What do you do next? If your X-ray inspection system is automated, it will reposition and manipulate the PCB as required to gather the necessary images. Then, it will process the acquired information and compare it against limits that you have supplied or that the system vendor has recommended.

The key to repeatible inspection is the software. For example, the Agilent 5DX System uses program modules organized according to device type and joint type. The algorithms carry out quantitative measurements and feature detection and recognize joint types such as J-lead, leadless chip carrier (LCC), gullwing, and BGA. Quantities measured include average solder thickness, detailed solder thickness distribution, pin/pad/fillet positional relationships, and void volume. Shorts, opens, misalignment, insufficient solder, missing components, solder balls, and excess solder are among the types of defects recognized.

As impressive as these capabilities are for all of the 3-D systems, processing vast amounts of information does take time. According to Joseph Pascente, president of Lixi, “The current PCB systems can do a fraction of the total PCB output when programmed to look at every solder joint.”

In many cases, you won’t want to inspect every joint. Some areas of every PCB are more susceptible to soldering faults than others, and these components are the ones to inspect first. Measuring every joint on every device will reduce throughput for very little quality gain.

For example, the Agilent 5DX System processes about 100 joints/s. It only takes a quick calculation to determine that the very large Lucent PCB design with 37,800 joints would take 378 s or just over six minutes to inspect. If you only make a few very large PCBs per day, that time may be acceptable. But how about a consumer product with 2,000 to 3,000 joints? Perhaps even a span of 20 to 30 s is too long if you are making thousands of PCBs each day.

Until very recently, in-line 100% X-ray inspection has been a contradiction of terms. But, partly thanks to faster computers and better X-ray detectors, throughput really has improved. Agilent Technologies considers its Series 3 version of the 5DX 3-D system to have about twice the throughput of the previous model. No one change made this possible. Instead, several elements of the process run significantly faster, including board alignment, surface mapping, load and unload, and the actual viewing and testing.

CR Technology takes a different approach to high throughput by splitting the inspection process between optical and X-ray. Don Miller, vice president of sales, said, “We realize that both techniques have their place in populated PCB inspection. It generally is recognized that optical inspection is a more effective means for component inspection and X-ray is the preferred method for solder-joint inspection.” See Figure 2 (right).

So, another way to improve throughput is to limit X-ray imaging to the things it does best—solder joints. Lixi’s in-line system illustrates this approach to manufacturing process control. Instead of concentrating on the analysis of every solder joint, this system captures X-ray images of successive PCBs at a rate up to 20¢/min and provides a macro view of the assembly. The in-line system determines the solder density of the assembly and linearizes the display gray scale to represent density.

Helping to put the broad capabilities of X-ray systems into perspective are comments from Tony Melton, western U.S. sales director at FeinFocus: “Some individual defects or special process weaknesses can be detected using alternative technologies, but only microfocus X-ray inspection can review multiple variables. For example, it is not possible to read a printed number by ultrasound or detect delamination with an optical camera.

“In most cases, you can determine inspection technology and identify production errors with a well-designed, highly accelerated stress-screening program,” he continued. “However, such a system usually will not identify the reason for the problem. Manual or semi-automated X-ray systems analyze and determine the causes of defects, help eliminate the process weakness, and define criteria for 100% inspection.

“Currently, only manual systems take advantage of high-resolution X-ray technology,” he concluded. “In the future, look for fully automated inspection systems that have overcome the speed, cost, and inspection quality issues to better meet the market’s needs.”

References

  1. Crane, E., et al, “Tackling Advanced Technology Boards: Combining X-Ray and ICT,” Circuits Assembly, September 1999.
  2. Lecklider, T., “X-Ray Inspection Underpins Assembly Process Quality,” EE-Evaluation Engineering, November 1998, pp. 111-115.

Return to EE Home Page

Published by EE-Evaluation Engineering
All contents © 2000 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publishers.

November 2000

Sponsored Recommendations

Comments

To join the conversation, and become an exclusive member of Electronic Design, create an account today!