PCB Inspection Shifts to AXI

As loss of access grows, it�s time to look into viable alternatives to AOI.

Rapid changes in printed circuit board (PCB) assemblies will force manufacturers to modify their test and inspection strategies. Electronic Trend Publications estimates that, due to the onslaught of area array packaging, half of all solder connections will be invisible to automated optical inspection (AOI) systems by 2007.

The problem worsens as board complexities increase, lead-free solder materials are introduced, and higher-frequency/lower-voltage devices restrict electrical testing. In the coming years, board manufacturers will face greater inspection challenges as loss of visual and electrical access climbs to new proportions.

Automated X-ray inspection (AXI) is gain-ing in popularity because, like its counterpart AOI, it is noninvasive while producing images sufficient for detecting defects and identifying manufacturing flaws. However, X-ray imaging is not hampered by ball grid array (BGA) devices, shields, heat sinks, and high-density double-sided boards. And when integrated with in-circuit test (ICT), AXI equipment can increase fault coverage to nearly 100% of all process and electrical defect classes.

Reduced Electrical Access
New chip packages and board assembly technologies will impact inspection techniques as increasing component and pad densities reduce electrical access to PCBs. The average density on wireless devices is approaching 50 I/Os per square centimeter, the limit at which ICT is effective.

I/O density is not growing uniformly across all market segments. Industry insiders believe that automotive products may become too dense for ICT by 2007 while office systems and computer products should still be electrically testable through 2011.

Higher speeds and lower voltages also will restrict electrical access. ICT signal integrity and accurate measurements tend to become problematic above 500 MHz where clocks for processors, high-speed memory buses, and high-speed serial communications paths often cannot be tested electrically.

Lead-free solder also presents test challenges. Experiments show that flux residues from lead-free solder can build up on test-probe tips and increase contact resistance. As a result, new probe styles, a more aggressive cleaning schedule, and shorter probe replacement cycles may be required.

Lead-free solder is slightly more brittle and susceptible to damage due to board warping and flexing during test-fixture actuation. Flying-probe testers can damage lead-free solder joints, especially during repetitive probing. In response, test-equipment manufacturers will be forced to develop new probe shapes, styles, and techniques.

The transition to lead-free solder can be problematic for many board manufacturers governed by IPC-A-610 standards that mandate the industry�s acceptance criteria for post-assembly inspection. As currently written, IPC-A-610 states that the maximum acceptable percentage of the ball-to-board interface area covered by voids should not exceed 10% and that solder joints with more than 25% voiding are classified as defects. The risk of failure is much higher among those who produce complex, high-density, double-sided boards.

Visual Inspection to the Rescue
Visual or optical inspection is gaining in popularity, but the most commonly used technologies have their own problems. AOI is increasingly challenged by the growing use of technologies that obstruct visual access. They include area array packages and other component technologies with hidden joints as well as heat sinks and RF shields.

The rate at which these technologies are expected to grow is alarming. The percentage of total solder joints that cannot be inspected by AOI systems is projected to increase from 20% in 2003 and 33% in 2005 to nearly 50% by 2007.

The impact of increased complexity on a board manufacturer�s yield also is alarming. At a nominal defect rate of 100 parts per million joints (ppmJ), an increase in the number of solder joints from 5,000 to 10,000 per board can reduce yield from 65% to 38%. Some manufacturers of complex boards with more than 30,000 joints have reported six defects per board, even with their manufacturing process well under control.

Depending upon the type of product, failure rates have a direct impact on end costs and a manufacturer�s reputation. For example, a manufacturer of $20 modems can better absorb shipping products with defects than an automobile brake-sensor manufacturer. If just one in 10,000 boards were found faulty, a vehicle recall could be mandated at a cost of hundreds of millions of dollars.

Mainstreaming X-Ray Inspection
Automobile electronics manufacturers such as Bosch and Siemens were among the first to migrate to AXI. The makers of high-end servers, enterprise systems, telecommunications equipment, and military/aerospace electronics soon followed. Their loss of electrical and visual access�and the increasing importance of defects�has forced the adoption of AXI.

If AXI is to meet the demanding standards and lower-cost models of other board manufacturers, X-ray must accomplish three tasks:

� Accurately detect and identify nearly all assembly flaws, solder defects, and other anomalies that reduce manufacturing yields and threaten product reliability.
� Avoid false reporting of defects. It must have a level of technical sophistication, quality, and reproducibility that refrains from making false calls where there are not defects.
� Efficiently operate at the manufacturer�s beat rate, the throughput at which products in high volume are produced.

Whether X-ray can perform these tasks depends, in part, on the resolution and dynamic range of the imaging system. To the test operator and the software designed to locate defects, resolution simply is the capability to distinguish between tiny features that define a fault vs. those that do not. In quality control terms, adequate resolution is the difference between consistently identifying real defect features, such as voids in the body of a solder ball, and shipping unreliable product.

Equally important is the degree of contrast provided by the X-ray imaging system. The distinction between black and white and various shades of gray must be sufficient to accurately report the location and attenuation of various features within the field of view.

A void within a solder ball, for example, should be imaged with sufficient contrast to permit software algorithms to consistently distinguish the void from the surrounding solder. When it comes to resolution and dynamic range, not all X-ray systems are created equal.

2-D X-Ray Imaging
Inside a 2-D X-ray imaging system, the source beam is perpendicular to the board�s surface and the beam and detector plate are stationary. The resulting image includes the entire Z-axis depth of a board and its many traces, solder joints, and components. Edge contrast and resolution usually are more than adequate when inspecting single-sided boards with no BGA devices, flip-chips, fine-pitch components, 0201s, or shields.

When used in high-speed situations, 2-D X-ray systems excel. They typically provide less than 20-ppmJ false fails in high-volume production environments. However, their capability begins to deteriorate when board complexity increases.

2-D X-ray is much less effective for inspecting double-sided boards. The image captures components and features on either side of the board that can overlap and become virtually impossible to distinguish by man or machine.

3-D Laminography
The traditional approach to 3-D AXI is based on laminography, a method widely used to inspect double-sided boards that requires motion of the X-ray source, detector, and board. In Figure 1, the X-ray source is placed above the board and rotates at high speed in synchronization with a detector plate below the board. The board is moved in its Z-axis to achieve the desired focal plane. Laminography uses a mechanical value-averaging approach typically involving 256 gray levels.

The use of a rotating source and mechanical rotating detector requires precise board movement in the Z-axis. Laser mapping is needed to determine the board height. Only points from the focal slice are projected at the same location on the detector and imaged sharply.

Object structures above and below the focal slice are not imaged sharply, leading to low-contrast resolution. In imaging a solder ball, for example, the background level of the board is averaged with the outside of a solder ball. If another component is under the ball, this also is averaged with the ball.

The limited dynamic range caused by averaging can permit the image of a good solder ball to overlap with the image of a defective solder ball. If a third element such as a capacitor is included in the averaging, the distinctions are further reduced. The presence of the capacitor may even cause a darker halo around the solder ball, not the best of conditions for the unambiguous identification of defects or voids within the solder ball.

Laminography cannot achieve the very low levels of false fails that are absolutely necessary in high-reliability production. The combination of mechanical movement and low resolution increases false-call rates to a level typically between 2,000 and 10,000 ppmJ.

On an individual board having several thousand joints, many defects may be reported while only one of these reported defects actually may be present. The false calls simply are artifacts of the blurring, reduced dynamic range, and Z-find instabilities that take place because of the averaging technique.

Off-Center Tomosynthesis
A newer approach called off-center tomosynthesis acquires the angled images needed for 3-D inspection without mechanical movement of the source and detector (Figure 2). It uses a stationary wide-angle X-ray source and a larger flat-panel detector. The source is positioned close to the detector, and the detector is divided into nine equal-size subregions, making it possible to capture nine images simultaneously.

The PCB is indexed across the field of view (FOV) of the X-ray source in a pattern called lawn mowering. Nine images, one taken at each PCB position, come from nine separate fields of view and later are rearranged to combine with the FOVs that belong together. The final step of combining the images to generate a horizontal slice is carried out by shifting the images according to their unique reconstruction vectors. Then the max-value algorithm eliminates out-of-plane objects that do not reside at the same vertical location.

To create a tomosynthetic horizontal slice, the subregions for the same FOV are shifted in computer memory based on their reconstruction vectors and then combined using the max-value algorithm. Each of the subregions has a unique reconstruction vector that corresponds to how much the image shifted to cause a vertical change in the slice.

The reconstruction technique selects the high and low points within the FOV and uses these as the anchors for its dynamic range, which is divided into 4,096 gray levels. The result is a vastly improved dynamic range. A solder ball stands up against the background of the board. If there is a void in the solder ball, the void also stands out.

Magnification and a high-resolution detector help achieve image clarity. One approach puts the part to be inspected close to the X-ray source and locates the detector at a distance equaling the magnification required from the inspection plane. The X-ray source or spot size needs to be very small, which limits the available power output from the X-ray source, and the final FOV size is small. An alternative approach is to use a high-power source, larger FOV, and a very high-resolution detector.

Off-center tomosynthesis is unique in its capability to more accurately detect subtle defects and manufacturing flaws, even as board complexities grow. It ensures that at least one of the images it produces will provide an unobstructed view of every solder joint. For example, a slice can be obtained that removes all bottom-side components so that solder connections on the top side can be easily inspected (Figure 3).

This approach also operates at a much faster speed since the multiple images required to obtain an unobstructed view of every solder connection are now captured simultaneously rather than sequentially. The elimination of moving parts and laser mapping greatly reduces cycle time, making it possible to operate at production beat rates. The elimination of mechanical errors and averaging typically decreases the number of false calls to below 500 ppmJ.

Keeping false calls below 500 ppmJ can substantially decrease the number of operators required for inspection, reduce required operator skill levels and training requirements, and improve operator confidence levels. A study of 100 operators using an inspection system with 4,000 ppmJ false call rates showed that only three in 100 operators were able to detect all defects and that 57% of all defects reported were allowed to pass. Using a $40-per-hour fully loaded operator cost model, an off-center tomosynthesis AXI system with a 500 ppmJ false-call rate will save $1 per board in high-volume automotive products, $2 per board on medium-volume server boards, and $7.50 per board in low- to mid-volume data-communications products.

Rapid changes in manufacturing technology are making the already difficult job of determining the optimum test/inspection strategy even more complex. At the same time, test and inspection technologies are continuing to evolve with the most recent development being off-center tomosynthesis technology that can raise defect coverage, reduce false calls, and increase throughput to beat rate levels. This rapidly changing environment has made it more critical to develop an optimized test strategy during the product development phase and leverage the strengths of each test/inspection platform.

3-D ClearVue

Instead of averaging the values from angled images, ClearVue selects the high and low points within the FOV and uses these as the anchors for its dynamic range, which is divided into 4,096 gray levels. As a result, all the features from the object being inspected are retained.

The XStation MX provides resolution on the order of 40 lp/mm and throughput rates of more than 5 sq in./s. The detectors used in the XStation MX range from 0.75″ to more than 2″ FOV with resolution retained at ~10 to 25 �m per pixel. This is accurate enough to detect fine-pitch components and all commonly used area array packages.

About the Author
Paul Groome is the director of automated X-ray inspection products at Teradyne. Educated at St Albans College in England, Mr. Groome has 26 years of experience in the test and inspection industry. He also worked in the telecommunications and military industries and developed the modular scan techniques used in military ASICs. Teradyne, Assembly Test Division, 600 Riverpark Dr., North Reading, MA 01864, 978-370-2700, e-mail: [email protected]

FOR MORE INFORMATION
on off-center tomosynthesis at Teradyne
www.rsleads.com/510ee-178

October 2005