Package visual inspection
10/01/2000
BY RUSS DUDLEY
Automated PVI supports current and future inspection requirements, reducing production costs and minimizing defects.
 |
For most semiconductor manufacturers, customer demands for product quality dictate a final quality inspection of every outgoing device. In many cases, customers hold manufacturers to a maximum allowable rate of defects per million (DPM) to ensure that customers' products are as reliable and cost-efficient as possible when they reach the market. To this end, more manufacturers are using package visual inspection (PVI) to realize quality improvements approaching zero DPM, while also reducing production costs.
PVI represents the final visual inspection of a semiconductor package immediately before final packaging or shipment. After all assembly and test processes have been completed, packages are inspected for visual defects ranging from cosmetic problems, such as scratches across the surface of the package, to defects that can result in catastrophic failure, including pits, chips and blisters in the surface, incomplete molds and voids in the mold compound (Figure 1). To detect such a wide range of defects reliably across many different package types (including QFPs, BGAs, TSOPs and SOICs), the PVI system must combine flexible camera and illumination technologies with powerful defect recognition algorithms.
Manual Inspection
Today, the majority of facilities rely upon human inspection to ensure the visual integrity of outgoing packages. Traditionally, manufacturers have opted for manual inspection based on the affordability of the technology to support it, as well as low labor costs in the regions where their products were being assembled. However, advancements in automated inspection technology, increased labor costs and the push to minimize errors by eliminating human intervention within the manufacturing process have combined to drive widespread adoption of automated PVI.
Certainly, the considerable visual and analytical capacities of human inspectors give a great deal of power to the process of manual inspection. People can easily and rapidly discern specific defect types, variations in color or depth and other abnormal features, therefore bringing a high degree of accuracy to the inspection process. As new issues arise, human inspectors can be easily trained to look for these variations.
 Figure 1. PVI inspects the top and bottom of the package for a wide range of defects |
However, while human qualities bring many capabilities to the inspection process, they also preclude it from attaining the reliability and precision of automated PVI. Fatigue is a significant factor in the performance of human inspectors, particularly when considering the number of hours they work and the number of packages they inspect per shift. Additionally, the consistency of inspection varies from one inspector to the next, given the subjective nature of the inspection process and the differing capabilities of individuals. As considerable as these performance issues are, the simple act of human intervention in the manufacturing process introduces other significant risks: parts can be damaged by mishandling, and costly material routing mistakes are often made. To protect against the hazards associated with device handling, some automated PVI tools perform the inspection process without removing the devices from their protective carrier trays (Figure 2).
The Advantages of Automating PVI
In addition to reduced labor costs and consistent inspection, upgrading the inspection process to automated PVI offers manufacturers considerable improvements in product quality and the ability to use obtained data to make process and yield improvements. All of these benefits combine to deliver a good return on investment (ROI), along with the competitive advantages and customer confidence that come from delivering top-quality products.
 Figure 2. To eliminate the risks of device handling, PVI can be performed with the devices in-tray. |
Automating the PVI process provides thorough, real-time statistical data regarding the number and types of defects captured. The data may be collected and analyzed to identify immediate or potential process control issues with other equipment in the line, such as marking or molding systems. By using the data to track process trends and determine corrective actions, manufacturers are able to minimize production defects, optimize good product yields and move closer to the DPM rate to which they are committed, with an ultimate goal of zero DPM.
As manufacturers strive to optimize equipment productivity and obtain zero DPM, automated PVI can be deployed at other steps in the manufacturing process. For example, a manufacturer might install separate PVI systems, one within the marking process and the other as part of the molding process, in addition to the end-of-line final inspection system. These dedicated PVI units act as process monitors, able to shut down or automatically correct equipment that deviates from its defined process parameters. This ability to monitor product quality throughout the manufacturing process can result in significant improvements in overall yields.
The cost of shipping failed parts can significantly impact a manufacturer's revenues and profitability because of potential returns, loss of customers and damaged reputation. However, even when those factors are set aside, automated PVI systems offer users substantial ROI when compared to manual inspection. The ROI equation invokes a number of variables, but the most straightforward assessment requires analysis of the system cost, system throughput and the existing cost per unit of the manual inspection being performed.
For example, if the existing inspection cost is $0.01 per device and 120,000 are produced per day, the inspection cost incurred on a daily basis is $1,200. If the inspection system costs $75,000 and provides throughput of 5,000 units per hour with round-the-clock utilization (figured at 100 percent), complete ROI would be realized within three months. The total ROI equation will take additional factors into consideration - some tangible and some intangible - but manufacturers can achieve full ROI from their PVI systems within six months, with minimal recurring expenses to follow. Other benefits include a reduction in floor space, because automated PVI is typically integrated into an existing process.
One of the added benefits of automated PVI is the position of strength it gives manufacturers in their customer relationships. Companies that can ensure and quantify high yields and consistent product quality hold a competitive advantage over those that cannot.
PVI Technology
As devices and defects become more complex and continue to shrink in size, automated PVI technology continues to make significant advances and is well-positioned to support both current and future inspection requirements.
PVI imaging platforms consist of specialized hardware and software, including dedicated processors and/or customized components, which enable large amounts of detailed visual data to be processed quickly. These imaging platforms employ a variety of complex inspection algorithms to reliably locate the full range of possible defects while maintaining production-level inspection rates (Figure 3). These cutting-edge subsystems will continue to evolve as future devices demand greater power to process and manipulate inspection data.
To further enhance their data-processing capability, PVI systems offer a variety of available acquisition modules to support different sensor cameras. These modules come in the form of daughter cards that are plugged into the imaging platform's circuit board and provide specialized acquisition capabilities for different sensor technologies and resolutions. This flexibility enables a PVI system to be configured with the ideal sensor for its defined application, which can optimize performance. Furthermore, by employing only as much sensor capability as the application demands, this flexibility can translate into cost reduction.
 Figure 3. Software screen captures demonstrate precise, thorough inspection of the package surface. |
Currently, array sensors are used for a majority of PVI because they offer low cost and high performance for most inspection applications, and different array sensor resolutions are available depending upon application. Typically, higher resolution array sensors are more costly. As package sizes continue to decrease, color- and line-scan sensors will be more widely used because they are able to acquire greater levels of detail than array sensors while maintaining production-level throughput.
Given the availability of dedicated processing and sensor technologies, illumination is critical to successful automated PVI. In order to highlight the wide range of package defects on a variety of surfaces, a PVI illumination system must be able to change the angle and intensity of illumination automatically. An effective PVI system will typically employ multiple light sources that may include both on-axis (bright-field) illumination, which is ideal for defects that exhibit variations in reflectivity, and off-axis (dark-field) illumination, which is well-suited for defects that differ in height.
Implementing PVI
When the decision is made to implement PVI, there are several strategic considerations that should be taken into account to optimize PVI performance.
A good implementation strategy will be jointly defined by an equipment supplier and the user to focus on performance objectives. These objectives should include the size of the defect that the system should detect, as well as the over- and under-rejection percentages that can be expected based on the measurement tolerances required by an application. As with any measurement system, the inspection tolerances for PVI can be adjusted to locate smaller defects than are necessary for the application, but this will result in a small percentage of over-rejection (good parts declared bad). Conversely, measurement tolerances can be calibrated to be more tolerant of variations in the package surface, which should minimize over-rejection while introducing the risk of under-rejection (bad parts declared good). Ideally, inspection tolerances should be defined to eliminate under-rejection while holding the potential for over-rejection to an absolute minimum (typically less than one percent).
The quickest route to reducing a manufacturer's DPM is to focus the PVI process on the defects that occur most frequently. For example, if a manufacturer currently sees voids in the package mold at a rate of 2 DPM, while surface cracks are occurring at a rate of 100 DPM, the PVI system should first be focused on detecting surface cracks to ensure the greatest impact on overall product quality. The best approach is to perform a Pareto analysis of all defect criteria. A Pareto chart will quantify the various defect types that the manufacturer encounters and the rate at which each defect occurs. A PVI system can then be trained on those defects that make up the largest percentage of all defects. Because a PVI system supports the manufacturer in identifying the cause of defects, these commonly occurring defects can soon be resolved. As that happens, other less-common defects will rise to the top of the Pareto chart and can be similarly addressed. As defects are continuously detected, identified and corrected, the ultimate result of this iterative approach to defect management is a DPM approaching zero.
When implementing a PVI system, the manufacturing environment becomes an important consideration. While human inspectors easily can distinguish between defects and debris, dust or other debris on a package can appear as a defect to an automated system. In facilities where debris is prevalent, an air knife or vacuum chamber may be used before inspection to clear debris from packages under inspection and ensure the reliability of the PVI process.
The flexibility of a PVI system is crucial to its long-term usefulness as a production tool. It must be ready to identify new defects as they arise while accommodating the full range of packages found on the market. To accomplish this, a system must maintain a level of inspection capability that is sufficient for the majority of packages and defects, as well as offer adjustment and programmability to enable users to refine and optimize the PVI process for new or unique challenges. An easy-to-use interface can give PVI operators the ability to quickly change inspection tolerances and differentiate between good and bad devices as new packages come on-line.
Conclusion
In response to increased customer demand for reliable assurances of product quality, semiconductor manufacturers are turning to automated PVI to inspect the final quality of every device before shipping. While PVI is often implemented in an effort to reduce DPM and improve product yields, the ROI and cost savings that a package inspection system affords make PVI an efficient tool for reducing production costs and improving the manufacturer's margins on every product. AP
RUSS DUDLEY, vice president of strategic programs, can be contacted at RVSI Electronics, 425 Rabro Drive East, Hauppauge, NY 11788; 631-273-9700; Fax: 631-273-1167; [email protected].