X-ray Advances in Support of Advanced Packaging – Today and Tomorrow

IMAPS UK recently celebrated the 50^th anniversary of its foundation and during its annual conference posed the question: “Advanced Packaging – Where Have we Been and Where are we Going?” [1]. In the case of electronic features, devices and packages, the answer is that they have shrunk and will continue to do so. 01005 packages are already only 0. 4 × 0.2 mm in size and yet smaller devices are on their way. Coupled to this shrinkage is the fact that more of the joints are hidden from optical view, with BGAs, flip chips, QFNs and TSVs providing the functional and I/O demands required of the most technologically advanced devices. With these developments, there also comes the need for tighter manufacturing tolerances [2] and an arguable discussion of the perceived value of electronics manufacturing has in today's world.

To support these developments and their growth, X-ray technology must continue to improve. For X-ray, the drivers that need to be accommodated are:

• The need to see ever shrinking features and joint sizes by providing increased magnification.

• The regular use of devices in board manufacture that do not just have optically hidden joints but also contain more of them [3].

• The need to check the joint / package quality non-destructively.

• Ensuring that manufacturing guidelines such as IPC-A-610 & Military Specifications are met [2].

• Replacement of lead containing solders with Pb-free alternatives that has resulted in manufacturing at higher working temperatures that are much closer to the glass transition temperature of the board substrate.

• The continued, and possibly increasing, prevalence of voiding within joints [4].

• The increased use of new X-ray imaging techniques, such as CT (or 3D X-ray), Helical CT and Partial CT (PCT).

The use of X-ray techniques in electronics test and inspection is quick, non-destructive, provides image magnification and is available with in-line and off-line modalities. Therefore, X-ray provides the capabilities that is, and will be, required for the electronics industry. Where X-ray can be of assistance is with identifying the presence and quantity of voiding [4], both within the joints and in the seals of packages. For solder joints and ground planes any voiding may compromise the thermal and / or power performance of the board or device. The level of voiding may be a key identifier that will allow an early estimation, or check, for the subsequent strength, stability, reliability and ultimate performance of the electronics. This is because the voids prevent the solder from touching areas of the joint interfaces and so potentially compromise the efficacy of heat and signal transfer.

2D X-ray imaging goes back to the discovery of X-rays in 1895 by Wilhelm Röntgen. CT, or 3D X-ray as it is also known, in electronics applications, became available in the early 1970s, with Hounsfield and Cormack winning the 1979 Nobel Prize for Physiology or Medicine for its invention. In ‘traditional CT’ analysis, that shall be called ‘full CT’ from now on in this paper, the X-ray tube and detector remain fixed with the sample rotating perpendicularly between them, as shown in Fig. 1.

During sample rotation, many 2D X-ray images are taken from 360° around the Field of View (FOV). From these images a mathematical / computational reconstruction is performed that creates a 3D X-ray model of the FOV. In other words, it converts the pixels of 2D images into 3D voxels (or volume pixels). Visualisation of the model then allows virtual 2D X-ray slices (not the original 2D X-ray images) to be taken at any plane within the FOV. For example, investigating just the interfacial plane between board and package of a large pad under a QFN.

CT is computationally very hungry, and it also takes time to acquire the necessary images. Within reason, the more images that you take from around the FOV then the better the quality of the resulting CT model. However, more images taken requires a longer acquisition time. Up until the last 10 years, only limited and expensive computational capabilities were available, usually through custom hardware. Therefore, reconstruction (CT model generation) and visualization took time to do. It is also arguable that the CT reconstruction algorithms used, such as the Feldkamp (or FDK) algorithm [5], were more optimized to the available computation rather than to providing the best CT models. Recently though, powerful yet relatively inexpensive GPUs in graphics cards used for gaming have provided substantially increased, and continuing to increase, computational performance that can be utilized for CT. This has now allowed CT reconstruction and visualization to be performed, very quickly, on high end home PCs. In addition, it has allowed new CT methodologies / algorithms to be used. For example, helical CT scanning now becomes practical on X-ray systems used for electronics applications. In helical CT, the operational setup is as for ‘full CT’ but the detector and X-ray tube also move together parallel to the axes of sample rotation during image acquisition, see Fig. 2.

This combination of movements results in the 2D X-ray images being acquired 360° around the FOV but over a helical trajectory. When these images are reconstructed, it provides a better final CT model as this approach removes artefacts which do exist in the FDK method. These days, the rate determining step for CT analysis is effectively the time to acquire the quantity of suitable images that will provide an adequate CT model from which analysis can be explicitly determined.

As all CT methodologies can take a relatively long time to obtain the highest quality model, and electronics requires high magnification of the FOV to see the smallest features in sufficient detail, full CT approaches may be limited to working on only a small section of a board, rather than the whole object. In other words, the sample may have to be cut down. Obviously, cutting down a sample then renders it unusable. Therefore, this is why the Partial CT (PCT) technique, which is also known as Inclined CT (ICT), Limited Angle CT or Laminography, has become recently an important part of the X-ray inspection and test arsenal for electronics. In PCT, the methodology requires that the sample remains horizontal, the X-ray tube is fixed in place and the detector is set at a specific oblique angle relative to the tube, see Fig. 3. Then either the sample is rotated 360° relative to the fixed detector, or the detector rotates 360° around the fixed sample. In both modes, 2D X-ray images are taken around the sample and from which a 3D model can be reconstructed.

The advantages of using PCT is that it allows the sample to remain intact, as different FOV of the sample can be separately analyzed without any need to cut the sample. PCT can also be quick, with a 3D model generated from only a relatively few images taken from around the FOV. This is why the PCT technique is used within in-line X-ray systems as it provides reasonable slice detail in the z-plane of the sample. In other words, it can be used to virtually slice down the board and see the layers and devices, especially at the interfaces.

However, the disadvantages of PCT is that the quality of the CT model is much poorer than ‘full CT’. The information in the z-direction of the 3D model is adequate but it is very poor in the × and y directions. This is because there is ‘missing information’ in the 2D image set that is used to create the model. In PCT, you do not have information from all around the sample, only in a cone volume. Full CT, meanwhile, has excellent model detail in all orthogonal directions. It must also be remembered that, as with all CT techniques, the time to acquire the quantity of suitable images determines analytical throughput. Therefore, and especially for in-line systems, there is always a trade-off between the throughput and the analysis quality achievable from the models generated. So, an in-line system may be fast but only for larger features being analyzed. As the feature sizes reduce, the model detail needs to be increased to adequately find the desired faults to a high confidence level. Acquisition times are then likely to dramatically increase and so throughput times to drop.

By using the PCT method, it is possible to look independently at different levels within the board or device. For example, the die attach layer in an IGBT can be checked for voiding as well as die alignment. Within the same 3D model, the slice at the substrate attach can also be selected to check its voiding. Underlying all this capability to check the level of voiding, there must also be a consideration of what is an acceptable level of voiding. Does a device manufacturer define what percentage is acceptable? Is there an agreed specification between the manufacturer and their client? For example, is the total voiding in a pad or interface less than a certain percentage? And / or does a single void exceed a specified area? Both these values may affect the performance of the device and / or the board that they are used on.

Although relatively simple and low-cost devices, LEDs are a good example of the remarkable developments in packaging and technology over the last 20 years, let alone the last 50. From the earliest NIR LEDs, commercially available in 1960s, they have also moved towards shorter wavelengths ever since and dramatically increased in volumes produced. High power, blue LEDs were not commercially available until the late 1990s. Only with the availability of blue / UV light does white light become possible, either through the blending of red, green and blue light, or using the phosphor method where blue / UV light excites a yellow phosphor. These methods now provide lighting with long operational lifetimes and lower total power usage compared with incandescent bulbs. As such, they have revolutionized domestic and commercial lighting, automotive lighting and LED TVs, amongst other applications.

Technologically, LEDs can now use much higher powers within the package. They also have smaller sized features and need increased reliability. This intensifies the need for higher quality, more consistent production. Flaws cannot be accepted; especially as higher usage powers mean higher operating temperatures. Therefore, this requires very good thermal conductivity in the package to move heat away from key areas. Without good heat dissipation then heat stresses at the interfaces can cause delamination or die fractures, so reducing LED lifetimes. The presence of voids, particularly at the die to package interface, creates air gaps that reduces heat transfer efficiency and therefore potentially increases package temperatures. Not only can this impact on the performance, or failure, of the package but can add additional stress to the rest of the board. As many LEDs are potted, or encapsulated, the only non-destructive test option to check for voiding and other faults within them is by using 2D and CT X-ray analysis.

Despite the simplicity and low cost of single LEDs, failures can still occur which, if not corrected during production, can cause substantial wastage and yield loss. Fig. 4 shows an X-ray images of a good, single LED with the wire bond connecting the package to the die. Fig 5, however, shows a wire break at the wedge bond (highlighted).

A 3D representation, created by CT, of a good ball bond on a single LED is shown in Fig. 6.

High brightness LEDs are also still relatively simple devices but typically use higher power and operate at much higher temperatures. However, they may also be part of a larger and much more expensive assembly and / or part of a safety critical component. Therefore, their test and inspection becomes increasingly more important. Fig. 7 shows a top down view of a board containg 3 high brightness LEDs together with an oblique angled view of one of those LEDs. Voiding can be seen in the joint interfaces as the white patches in the grey solder. This is particularly clear in the oblique angle view. This level of voiding is small and may be very hard to eradicate completley. However, it is unlikley to affect device performance. This may not be the case for the high brightness LED sample seen in Fig. 8 – 9. Fig. 8 shows an oblique view X-ray image where substantial voiding is seen under one of the LEDs compared to the others. This is more clearly seen in Fig. 9 at a higher magnification, top-down view. This level of voiding may affect the thermal performace of the device. The indication from the X-ray image of a voiding issue is very helpful for the production process. However, some X-ray systems are also able to provide a calculation of the voiding level, as seen in Fig 10.

Other potential sources for LED failure include the movement of the wires during / after the molding process. An example is shown in Fig. 11. Here, a short between wires has almost been created. This problem is often referred to as a wire sweep issue. Automated calculation of the level of wire sweep can also be provided by some X-ray systems. This is achieved by indicating the percentage deviation of the wire from a straight line when viewed from above.

It should be noted that whilst the X-ray inspection techniques are able to see many hidden flaws, there are limitations. Consideration must be given to the density and quantity of the materials within the objects being inspected, as well as the environment in which they are located. This is because the absorption of the X-rays, which cast the ‘shadow’ on the detector to produce the image, is proportional to the cube of a material's atomic number and that material may have to be imaged through the rest of, or other components of, the whole sample. As a result, less dense and thinner features, such as aluminium wires, are more difficult to image than, say, similarly sized gold wires. Furthermore, wires will be easier to see as individual devices compared to being on, or as part of, a board, where the wire's absorption has to be imaged ‘on top’ of the absorption of the board material and possible second side components (see Fig. 8). This also means that the intermetallic compounds created in joint formation are unable to be seen in the commercially available X-ray systems. Finally, it must also be noted that not all commercially available X-ray systems have the same capabilities. Therefore, the user must ensure that they have access to an X-ray system that is fit for their purpose.

The development and complexity of electronic devices and packages has radically evolved over the last 50 years. The use of smaller features, the ability to handle higher powers and temperatures and the demand for reduced costs yet with increased reliability pushes the demands on materials and manufacturing performance. This requires an enhanced test and inspection regime during manufacture to ensure product quality. X-ray inspection provides one part of this regime. However, X-ray technology itself needs to continue to evolve to provide the increased magnification, the better image resolution and high quality automated analysis to satisfy the market and technological demands. The use of CT and PCT X-ray techniques allows novel analytical methods to be considered in addition to the standard 2D images. The ‘simple’ LED, or similarly the IGBT, provides a great example of the challenges that X-ray must support.

The author would like to gratefully acknowledge the Applications Team at YXLON International GmbH for the supply and use of their X-ray images in this paper.