P. Ian Nicholson, Daniel Mitchard, Carl Forrest and Philip Wallace
Thomas Bantel, Dirk A. Neumayer
Fraunhofer -Institut für Produktionstechnik und Automatisierung
Paper presented at BINDT 2006 Conference, 12-14 September 2006, Stratford-upon-Avon, UK.
The need to increase the component population density in Printed Circuit Board (PCB) assemblies has resulted in evermore-complex surface mount technologies. Typical surface mount technologies such as Ball Grid Arrays (BGAs) andother similar types of assemblies, whereby an array of solder balls are placed beneath the component and the connections made are between the component and the substrate, render the component to board connections visually inaccessible.Non-destructive testing the integrity of these joints and the PCB assembly is a problem because of the lack of access and the increasingly complex fine pitch connections. The situation is additionally complicated by the introduction oflead free soldering, through the potential to create further defects, and its different surface finish and joint profile. Clearly, there is a need for an in-line system capable of detecting these defects automatically during theproduction process. This paper describes the development of an in-line X-ray system for non-destructive inspection of ball grid arrays and other small foot print devices. Algorithms for defect detection are presented and theirintegration into an automated computer controlled in-line X-ray system is described.
A BGA is a surface mount integrated circuit (IC) package with its underside face covered in a grid pattern of solder balls. These solder balls connect to the die and are used to conduct electrical signals from the IC to the PCB.During PCB population the IC package is placed on a PCB that carries copper pads in a matching pattern to the solder balls. Heat is applied by a re-flow oven causing the solder balls to melt and fuse the IC to the PCB. Both thetemperature and solder alloy are parameters that can affect the reliability of the connections and good process control is required. BGAs can be ever increasingly found in PCBs today for a number of reasons: BGAs can offer manyhundreds of connections per footprint and thus occupy less space than leaded packages. Compared to leaded packages, connections to a BGA are separated from the PCB by a shorter distance thus providing potentially better electrical andthermal performance. There are disadvantages to BGAs: They are more difficult to inspect since the solder ball connections are hidden under the IC package, and any defect requires removal and replacement of the entire package.
In summary, the kind of defects that need to be detected are shorts caused by solder bridging between the balls, tilted BGAs, BGA voiding, insufficient/excess solder, BGA missing balls, BGA poor wetting and misalignments.
Already many PCB manufacturers employ off-line X-ray systems for debugging purposes. There is a reluctance to put these systems in-line because X-ray systems can slow throughput and cause a production bottleneck. Europeanmanufacturers can compete with their Asian competitors by manufacturing low-volume, high-value, and high-margin PCBs rather than mass products. It is in these situations that an in-line X-ray inspection system can offer many benefitsby improving board quality and thereby reducing scrap and the expense of warranty returns.
This work presents preliminary findings concerning the development of an in-line X-ray Inspection prototype that has been developed primarily for inspection of BGAs. The development of the prototype has followed in two phases, inthe first phase an X-ray system was designed and constructed and images obtained manually. The resulting images were then used to develop algorithms for automatic defect detection. In the second phase, which is on going at the time ofwriting, the intention is to fully integrate the X-ray equipment defect detection algorithms together with a conveyor system to provide an in-line solution.
2. X-ray prototype development (Phase 1)
2.1 The principle
X-rays are high-energy electromagnetic radiation, and are used in many modern day applications, the most well known of which is in medicine. However, X-rays are becoming increasingly used in industry for the inspection of theinternal structure of components increasing reliability and efficiency and helping to identify defects at an early stage of manufacture and during service. X-ray photons have energies of between 0.1 and 1,000 kV, with wavelengths inthe region of 0.1 to 10 nm.
Typically, during an inspection, a component is placed between an X-ray machine and detection media. The penetration of X-rays through the component depends on the energy of the X-ray, but also depends on the density and nuclearcharge of the material being inspected, and X-ray machines can be tuned to take these factors into account. The detection media usually produces a visual image of the X-rayed component, and can be traditional radiographic film or moremodern digital media such as phosphor screens and direct digital arrays. Digital media is generally better than film due to its flexibility, high contrast and good resolution, and also reduces the time and waste associated with thedevelopment of traditional radiographic film. Most X-ray systems comprise a fixture for holding and manipulating the component to allow views from every angle.
As the X-rays pass through the component, the different densities within the component attenuate the X-rays by different amounts resulting in light and dark areas being produced on the detection media. For example, a crack in asolder ball is easily visible because the density of the solder is greater than that of the air in the crack, and will attenuate the X-rays more. The solder ball will appear as a light area (i.e. where the X-rays have been attenuatedmost) containing a dark crack (i.e. where the X-rays have been attenuated very little).
For this project, the X-ray source is positioned at the bottom of a lead-lined chamber and a digital detector is positioned directly above. A PCB enters the chamber by means of edge conveyer belts and is positioned on a manipulatorin-between the source and detector. This is illustrated in Figure 1 below. The system can then carry out a complete inspection of various components on the PCB by manipulating the PCB's position accordingly. Magnification of a specific area of interest can also be achieved by movingthe board closer to the source.
Fig. 1. The set-up arrangement for the X-ray inspection of PCBs
The X-ray source is fully computer controlled. It is housed within a vacuum continuously pumped by a turbo-pump backed by a two-stage rotary-pump and monitored by a Penning gauge. It contains a microfocus electron gun with amagnetically focussed electron beam intercepting a 15µm air-cooled tungsten foil target behind a 0.5mm thick aluminium window. The X-rays are collimated at the source to a beam with a fan-angle of 14°. The cathode voltage isvariable from 40 to 160 kV, and the maximum current is 500µA, giving 20 watts target power at 40 kV.
The detection media system consists of a single field 6-inch aluminium window with an intensifier coupled via a large aperture lens to a digital camera. The camera produces images of 758 x 580 pixels and 12-bit depth with anadjustable frame rate of 25 fps and less, and is computer controlled. The nominal source to focus distance is 600mm, which allows magnifications of up to 1200 times depending on the sample position.
2.3 Mechanical manipulation of the PCB
A four axis manipulator is used for panning the PCB in X,Y and Z planes of motion and for tilting the PCB. The manipulator can be directly user controlled via a joy-stick for manual inspection. Additionally, the manipulator can moveaccording to a pre-programmed script and throughput of the inspection system can be increased by only looking at components of interest.
The Z motion controls the distance of the PCB with respect to the X-ray source, which changes the magnification. In order for defect detection algorithms to operate reliably in the experimental setup, a PCB sample has to be adistance of not more than 20 mm from the source. This results in a higher magnification, but at the expense of imaging less of a component at the time. The manipulator is then used to position the BGA at different positions so a wholeimage of the component can be built up. Tilting the PCB allows for angled shots of BGAs which can be more revealing for certain defects. However, in the experimental setup a compromise has to be made between tilt angle andmagnification, since tilting the PCB necessitates increasing the distance between the PCB and source, to avoid sample collision with either the source or conveyor.
2.4 Results & discussion
The digital radiographic technique is ideal for determination of flaws and defects that are characterised by local density and/or structure variation. Trials and experimentation with the system have already shown that micro focusradiography can identify the following range of defects: BGA tilting, voiding, shorting, misplaced components, foreign body entrapment, chip capacitor cracking and flip chip under fill.
Figure 2 shows images for a 180 ball lead free soldered BGA imaged at decreasing distance between the PCB and the X-ray source. From the 100 mm height it is already evident that there is some solder bridging between the ballsoccurring. With height just 10 mm above the X-ray source, voiding in the BGA balls is visible. Increased voiding in solder is a typical finding when changing from a lead to a lead free process. It results from an oven profile problemduring solder reflow. Generally, it can be said that each ball should be of the same size and circular in shape. Although an oven profile varies in the same process across different PCB manufacturers a starting point for a defectdetection algorithm would be to accept BGA balls containing voids of a maximum permissible size less than 10% of the minimum ball joint measurement.
The height experiments are a useful consideration when taking the X-ray System throughput into account. It has already been mentioned that the height affects the magnification of the X-ray images. As can be seen from Figure 2 a certain magnification will result in a different number of balls in the image.
Fig. 2. X-ray images of 180 ball BGA with different separation distance between PCB and X-ray source (95kV, 60.0µA)
100mm separation distance
When applying defect detection algorithms, the accuracy of the results will decrease as the magnification of the balls become smaller because the number of pixels representing each ball and each void become smaller. For an analysisof the balls the computation time per image increases with the number of balls. However, if only a certain number of balls are to be inspected then the total computation time will be least when inspecting all of them using a singleimage and greatest if a different image is used for each ball because of the extra image acquisitions and multiple manipulator movements required. Therefore increasing throughput to meet the in-line requirement by using fewer images isonly possible if the images visualising multiple balls are of sufficient magnification.
3. In-line X-ray prototype development (Phase 2)
3.1 In-line modification
An edge belt conveyor has been designed to enable the system to be placed in-line with existing in-line transfer systems. PCB equipment suppliers use a standard called SMEMA  to facilitate the interface of their equipment at a mechanical and electrical level. This standard has been implemented for the prototype and includes standardisation of conveyor height, PCB edge clearance, and in-line system tosystem electrical communication. As such the X-ray system behaves as a pure 'pull' type system, where the PCB is not released or moved to the next process step until authorised or pulled by a vacancy downstream.  The advantage of using the SMEMA standard is that a separate cell controller is not needed to manage the flow of the material. Figure 3 shows the X-ray system prototype integrated with the conveyor system. This is not a finalised solution as the conveyor entrance and exit holes and the conveyor need to be lead shielded to be radiation safe.
Fig. 3. X-ray prototype integrated with conveyor system
3.2 X-ray safety
The X-ray inspection system is to be incorporated into a production line that will only allow less than 30 seconds for the inspection of a BGA component on a PCB. However, the system is housed inside a lead-lined chamber andallowing automatic access for the boards without leaking any radiation into the surrounding areas is a major issue. The source is required to be on constantly to prevent issues with 'warming up', and open access is also required forPCBs. Lead lined motorised trap doors were originally considered. However, the fact that the X-ray source would have to be switched on and off according to when the trap doors were open or closed would reduce system throughput.
6mm lead shielded ply tunnels are to be constructed, that are up to 1m long, on either side of the X-ray chamber to house the internal conveyer belts. The PCBs will enter and exit the chamber through a small gap, approximately 30mmhigh and as wide as the PCBs to be inspected, at the end of each tunnel. The external conveyor belts will pass the board through the gap and onto the internal conveyor belts.
Using a known X-ray beam angle of 14° (i.e. 7° from the normal) reported by the manufacturer, the dimensions of the X-ray chamber, and the absorption of lead, simulations of the scattered and secondary radiation within thechamber can be calculated  as shown in Figure 4. With this information the safe optimum length of the tunnels and size of the entrance gap can be determined to ensure that X-ray safety regulations are met.
Fig. 4. X-ray scatter inside the X-ray chamber
4. Automated defect detection
In order to provide for fully automated analysis of BGAs, software is being developed, that identifies, analyses and classifies the balls of BGAs. This section describes the applied algorithms for pre-processing of the images andoutlines proceedings for the classification of the balls. Since algorithms for ball counting, void detection are already known  emphasis in this paper will be on the analysis of the wetting of the balls.
Due to short exposure times, the X-ray images are noisy. In a first step this noise is reduced by a rank filter using a moving 5x5 square window. In the next step the objects are segmented from the background. The goal is a binaryvisualisation, where black pixels denote objects like BGA-balls and white denotes background. However, simple segmentation using a global threshold for binary visualisation fails due to variations in the background as can be seen in Figure 5. Here, a global threshold was set to separate the balls on the right side. The balls on the left side are still mixed with the background. If a global threshold for the balls on the left side is chosen, the balls onthe right side would get smaller and change shape.
Fig. 5 shows:
a) radiograph of BGAs with
b) corresponding segmented image using a global threshold
In order to facilitate accurate segmentation, thresholds are computed locally. In the newly developed algorithm a mask is moved across the image in steps. At each step, the histogram of the grey values inside the mask is analysedand a threshold is determined that applies for the whole mask. According to this threshold the image is binary visualised.
Fig. 6 Shows the segmented image of (5a) using a variable threshold
In the histogram analysis the threshold is determined by finding a value that on the one hand separates the dark and light values evenly and on the other hand, defines a local minimum. Figure 6 shows the segmented image of Figure 5a using this method. All balls are separated properly from the background.
Depending on the viewing angle, adjacent balls may not appear separated sufficiently well in the image or even overlap. Up to a certain degree, it is possible to separate those balls by image processing. First, small holes and gapsare filled by applying a closing operation.  Then the balls are separated by applying a distance transformation followed by a watershed transformation.  Figure 7 shows a magnified part of an X-ray image before and after the separation step.
Fig. 7. Overlapping balls
For each ball, a number of features are extracted. To name a few, the area, angle, height and width are measured. Also the curvature and roughness of the contour of each ball is analysed. The features of well wetted balls are usedto create a model of 'good' balls. Then the individual objects are classified according to their features.
Fig. 8. Classification of the wetting of BGA balls
Objects that are not recognised as balls are removed. The balls are highlighted in the image and colour coded according to their wetting (see Figure 8). Green means that the wetting is acceptable; yellow warns that one condition for acceptable wetting has not been met, and red indicates that the wetting of the ball is not acceptable.
The developed software allows accurate segmentation of the objects from the background. This is a prerequisite for many defect detection algorithms like ball counting or void detection. An algorithm was developed that analyses thewetting of the balls by classifying the balls according to several features like area, angle and contour-based features. The software was tested with X-Ray images of BGAs with artificially introduced poor wetting. The balls with poorwetting were reliably detected. As a next step, it would be important to test the software with many real-world images of BGAs with poor wetted balls to verify if slight adjustments of the classification and possibly of the extractedfeatures are necessary.
This paper describes the development and first results of a prototype in-line X-ray inspection system for detection of defects related to BGAs. The technique can be used after reflow; however the technique also lends itself afterwave soldering, goods in and during failure analysis.
An in-line X-ray prototype has been developed which, when operated manually by the user, is capable of detecting a number of different defects that can be found in BGAs. These first results show that the system is a capable nondestructive technique that can help improve the process control during the BGA component population of the PCB.
Basic algorithms are being developed to allow automatic detection of these defects. In addition, modifications have been made to facilitate a conveyor add in and automatic control of the manipulator to allow in-line detection ofdefects in BGAs.
The results from this work are part of the CRAFT programme: Development of comprehensive in-line quality control system for printed circuit board assemblies, Acronym MICROSCAN, COOP-CT-2003-508613. MICROSCAN is a collaborationbetween the following organisations: TWI Ltd, X-TEK Systems Ltd, Lot Oriel GmbH, Machine Vision Products Inc, Microtel Technologie Elettroniche SpA, Beta Electronics Ltd, Ultrasonic Sciences Ltd, Goodrich Control Systems Ltd,Fraunhofer-Gesellschaft zur Foerderung der Angewandten Forschung E.V. and Kaunus University of Technology. The project is co-ordinated and managed by TWI Ltd and is partly funded by the European Commission under the Framework SixProgramme.
- SMEMA Interface Standard 1.2, Surface Mount Equipment Manufacturers Association
- Burda, Richard, 'Controlling Material Flow and WIP on Existing Semiconductor Assembly equipment', SEMI TAP Conference, February 22-23, 1999
- 'Recommendations for Data on Shielding from Ionizing Radiation. Part 2. Shielding from X-Radiation.', British Standards Institution, British Standard 4094: Part 2: 1971
- Inspect-X User Manual, February 2006, X-Tek Systems Ltd.
- MIL 8 User Guide, Matrox Electronic Systems Ltd., 2005
- Jähne, Bernd, 'Digitale Bildverarbeitung', Springer Verlag, 2005