An Inside Look: 3D AOI of Electronic Assemblies Seminar
The prestigious Manufacturing Technology Centre in Coventry, UK, was an appropriate venue for SMART Group’s seminar on 3-Dimensional Automated Optical Inspection of electronic assemblies, “3D AOI." The event brought together experts from leading AOI suppliers, to discuss and explain different approaches to three-dimensional inspection and to present the latest in technology to an attentive audience of engineers from the electronics manufacturing industry with a collective interest in yield improvement, process control, and quality assurance. All were welcomed by SMART Group Chairman Keith Bryant, who introduced the speakers and moderated the panel discussions.
“Putting 3D into Perspective” was the scene-setting theme of the first presentation, from Andy Bonner, managing director of NordsonYestech Europe, who demonstrated how AOI technology had evolved from the mid-1990s to the present day. “Test and Inspection--who needs it?” was a rhetorical question as he described how AOI began to take the place of in-circuit testing once confidence in the quality of integrated circuits had improved to the point where their functionality could be consistently expected. The weaknesses of early AOI machines were low data processing power, limited lighting capability and analogue top-down cameras. Systems were used to check for component presence and orientation, and the presence of solder.
Lighting progressed from plain white through added coaxial red light, which could show up tombstoning, to RGB phased lighting with arrays of different-coloured LEDs at different angles which gave some information on topography, for example the shape of a solder joint. The next step into the third dimension was to use a series of angled high-resolution cameras, which could for example look at IC packages from the side and see leads and solder joints out-of-sight of the top-down camera. A typical configuration for the inspection of assemblies had five cameras with resolution from 25 down to 8 microns and simultaneous image capture. To measure the height of components and features accurately, the options were laser or moiré optical techniques, or a combination of both. Now, with clever software, “Real 3D” was possible--3D systems could be integrated with 2D applications and optimised for speed and repeatability, with fast programming and easy updates for component changes during production.
So what could be achieved with AOI other than just funding faults? Bonner listed yield improvement through feedback to the line, historical data, production statistics, and designer and supplier feedback.
Gary Leong, director of business development with ViTrox, remarkably bright-eyed after a 14-hour flight from his base in Malaysia, gave an informative presentation entitled “Capturing lifted lead, coplanarity and missing component defects effectively with 3D AOI.” Reflecting on his early career as a post-reflow inspector, he was concerned for his future when the company invested $300,000 in an AOI machine. In fact his job was just the same--his boss saw him as an insurance policy--because placing the machine post-reflow still left the opportunity for many in-process defects and the machine, with top-down cameras, gave too many false calls and missed defects such as lifted leads, missing components, coplanarity and foreign material in unpopulated areas. Fifteen years on, today’s 2D + 3D combination systems were capable of consistently detecting these defects, the 2D functions revealing missing, offset, skewed, wrong part, wrong polarity, billboard and tombstone components, lifted or bent leads, excess or insufficient solder, and solder bridging, complemented by the 3D functions measuring package height and coplanarity to an accuracy of 2 microns.
Leong described the principles of phase-shift profilometry and explained how phase-shifted structured light was used to re-construct 3D images, with advanced algorithms to identify lifted-lead, coplanarity, and missing-component faults. He showed many examples and case studies, illustrating the differences in capability and inspection speed between 2D, 3D, and 2D + 3D systems, including instances where 3D AOI was a faster and more cost-effective technique than X-ray for detecting lifted leads. However, in spite of the outstanding capabilities of advanced AOI systems, the fact remained that it was not yet possible to eliminate the operator--the dream was for a fully robotic system, the reality was that Leong’s post-reflow inspection job of 15 years ago was still a necessary activity!
In an informative presentation on image capture and modelling technologies used in 3D AOI, Sean Langbridge from CyberOptics gave a more in-depth explanation of phase measurement profilometry, which was widely used for accurate and repeatable measurements in 3D solder paste inspection systems, although component height was a potential limitation in 3D AOI SMT applications. Multi-view 3D sensors gave improved capability; the example he showed consisted of four 3D sensors, with a single projector and a 2D camera. Five images were acquired simultaneously and merged together by parallel processing, using 3D fusing algorithms and reflection-suppression techniques to produce high-precision 3D images with sub-micron resolution at production line speeds.
Langbridge reviewed current image-analysis technologies. Algorithm-based systems were generally complicated to programme. They relied on libraries of pre-defined inspection tasks, and tuning these libraries to understand process variation was very time consuming. Statistical appearance modelling was a knowledge based technology that learned from examples and adopted a single analysis technique for all inspections tasks, giving a simple intuitive approach to SMT inspection. Useful inspection could be achieved within five to 10 PCBs and because the user did not have to anticipate defects, undesirable variation was detected. Autonomous image interpretation was an advanced modelling technique based on appearance modelling, designed to support low volume production and to operate with minimum supervision, with the ability to predict variability from a single sample.
Doug Jones from TRI Innovation reported the results of an assessment of the capability of a 2D + 3D AOI system to inspect 0201 chip components based on 03015 chip analysis. The system evaluated used a top camera with multi-angle red, green, blue, plus white lighting for clear illumination of component features, with four separate lighting phases for improved detection of component bodies, marking and solder, and dual laser modules for shadow-free 3D inspection and true 3D profile measurement including board warpage. Blue lasers enabled a high dynamic range whilst maintaining high inspection speed. Feature analysis and comparison were by algorithms analysing visible reflected light, enabling shape, colour, gradient, and location inspection, combined with image-based techniques.
For very small components, camera resolution became a significant issue. For example, for pixels of 8 microns, the dimensions of an 0201 component corresponded to 31 pixels x 16 pixels and sub-pixel resolution enhancement was necessary to calculate the relative position of the edge of a component within a pixel. This was achieved by measuring differences in shades of grey in adjacent pixels. Jones discussed detection algorithms for component shift, component missing, insufficient or excess solder and foreign matter, with a series of examples of detected defects, and the potential benefits of future developments in coaxial lighting in improving the detection of dry soldering pads and the position of non-tilted silicon-based bottom-terminated components.
This capability assessment had concluded that the 3D AOI system as described was capable of inspecting 03015 and 0201 micro-components and of detecting defective symptoms such as component shift, component lifted, component missing, insufficient or excess solder or bridging, solder balls, and solder splash.
Pierre Williams from Yamaha discussed new technology available in SMT equipment and software to help raise production yield and reduce waste in a presentation entitled “Boosting Quality Control and Productivity through Closed-Loop AOI.” The goals were to produce more good units and achieve a higher end-of-line inspection pass rate, with fewer line stoppages and faster fault resolution. To achieve these goals, the requirements were high-quality inspection data, real-time information and feedback of defect data, and access to detailed reports, either at the machine or on the move.
In a typical SMT line there was only one AOI and this was located post reflow so that defects were not found until after board has been soldered, meaning that several defective boards could be in reflow by the time the first defect was found, and significant human intervention was needed to monitor reports and stop the line. On a short production run, the run could be finished before the defect was found and there was no benefit from feedback. Until recently, there was no facility for the AOI to communicate with the pick-and-place machine, so quality information was not shared between the inspection area and the production area and there was no scope to eliminate a problem whilst it was occurring.
In the knowledge that reflow could correct minor placement errors, Williams suggested locating the AOI machine after placement and before reflow as the first stage in the transition to closed-loop feedback so that defects could be detected before boards were reflowed. Large amounts of manual rework and repair could be avoided and high-value components could be retrieved and reused. With suitable software, this gave the option to automatically feed-back data from the AOI to the pick-and-place and pinpoint the origin of the defect, and conversely, close the loop to feed forward data from the mounter to optimise AOI performance.
With increasing trends to automation, and improvements in the up-time of printing and placement machines, it was likely that just one person would have responsibility for running the line, and might not always be present. Advanced software could take “human” decisions and the operator could have full control of the line through a mobile phone app.
In his alternative capacity as Technology Editor of SMT Today Keith Bryant returned to the rostrum with a wrap-up presentation designed to provoke controversy in the final panel discussion, entitled “Is (was) 2D really so bad?” After reviewing scanning methods and extra or floating component detection, he explained the technology of the telecentric lens and the relative benefits of coaxial lighting and multi-light technology, somehow managing to include a photograph of Bob Willis clad in a Hawaiian shirt and carrying an inflatable parrot! He concluded that the “real” answer was for all 3D systems to contain some 2D technology. Although the third dimension was just Z, a measurement, it represented “a whole new dimension.” He commented that component and board warpage become bigger issues each year. Because an increasing number of components came in bottom-termination packages, AOI was of limited value in determining solder joint integrity and in-line X-ray was not fast enough for many applications. He suggested a link between AOI and X-ray such that that the X-ray went to any questionable areas and the operator only needed to inspect suspect components, which would save a lot of time and improve yields.
AOI systems have undergone rapid development in recent years, and have evolved from very basic 2D image recognition machines to sophisticated 3D systems that can reliably detect the full range of assembly defects at production-line speeds. The ability to look for trouble has truly taken on a new dimension, and this well-organised SMART Group seminar delivered a wealth of up-to-the-minute information and provided an opportunity for existing and would-be 3D AOI users to understand and discuss the relative merits of the different approaches of major players in the industry.