Optimized PVI
11/01/2000
Process considerations can ensure value and performance from package visual inspection systems.
BY RUSS DUDLEY
 Figure 2. An effective PVI system has to inspect a wide range of package types and configurations. |
Semiconductor manufacturers are faced with increased demand for product quality, often being held to a maximum allowable rate of only a few defects per million (DPM) by their customers. At the same time, product defects are shrinking in size while increasing in variety. While these defects are often cosmetic and the devices that exhibit them will, in many cases, pass electrical test, the defects can lead to premature device failure at the least fortunate time for the customer - after the final product has shipped. In the October 2000 issue of
Advanced Packaging magazine, we outlined the basics of automated package visual inspection (PVI) and its benefits. In this issue, we focus on the process considerations that manufacturers must take into account to attain maximum value and performance from their PVI systems.
To reduce DPM and improve quality, and in so doing, lower production costs, manufacturers can replace manual inspection processes and turn to automated PVI to inspect the final quality of outgoing products. To take full advantage of the PVI process, manufacturers should understand how PVI systems can be adapted to meet demanding or evolving product requirements.
The industry's range of packages, package configurations and materials is significant and continuously increasing, and there are many defect types that a PVI system must be able to reliably detect. To accommodate this, flexibility is crucial to the long-term success of a PVI system.
A Wide Range of Devices
The two major device types in high-volume production today are leaded devices and grid array packages, each having many variations. Leaded devices range from the original dual in-line (DIP) packages for through-hole applications to traditional surface mount configurations, such as single outline (SO) and quad flat pack (QFP) devices. The latter account for the majority of manufacturing volume among leaded devices. When employing PVI with these devices, only the top surface of the device usually needs to be inspected.
 Figure 1. Automated PVI systems often function as part of larger in-line machines. |
Grid array packages pose a greater challenge to the PVI process because many of the packages feature a molded cavity on the top of the package, similar to leaded devices, as well as a circuit substrate on the bottom of the package that must be inspected. Substrate material and color varies from package to package, as do the visible patterns and interconnect points. Newer package types, such as leadless chip carriers, share many of these attributes while raising new issues of their own. To be effective, the PVI system must be capable of detecting defects for any variation of leaded or grid array packages while providing a clear upgrade path for inspection of new package types.
Classes of Product Defects
Aside from common variations within package types, a large number of defects have been recognized by manufacturers. They can be separated into two categories: mark defects (which concern the mark of identification on the package surface) and package defects (which relate to the package itself). Mark defects are more objectively quantified by specific criteria (Table 1).
 Table 1. Criteria for quantifying dark defects. |
With proper illumination, marks are typically easy for a PVI system to image and inspect. While the criteria for mark defects are fairly straightforward, there is some subjectivity with regards to the legibility and completeness of a mark.
 Table 2. Common defects. |
Package defect criteria involve many issues related to the types of defects that can be encountered. Because these defects are normally random in nature, PVI systems must not only be able to recognize each of them, but must do so despite wide variations in size, shape and location (Table 2).
Implementing PVI
Because the variety of packages, materials and defect types can complicate the PVI process, it is important to establish a feasible implementation strategy. Defining defect types, prioritizing defects based on frequency or relative impact on quality, and understanding the capabilities and limitations of the inspection platform are key to successfully implementing PVI within the manufacturing process.
 Figure 3. The PVI camera's imaging area (field of view) will affect pixel size. |
Ideally, the initial qualification of a PVI platform will demonstrate its ability to detect every defect on every package configuration that is presented. However, because of the wide range of package configurations and defects, this is unlikely. It then becomes necessary to determine an optimal implementation strategy that achieves the desired quality and cost improvements as quickly and efficiently as possible.
To do this, the most critical or frequently occurring defects need to be matched with the capabilities of the PVI system. The system is first trained to detect the highest priority defects, which should lead to the reduction of that defect as the PVI system helps root out the cause. As the defects drop accordingly in priority, the system capability expands to detect the next-highest defects on the priority list and so on, continuously moving toward the goal of zero DPM.
Technical Considerations for Optimal PVI
There are two important factors for manufacturers to consider as they implement an accurate, reliable PVI process. The first - system resolution - is determined by the camera and optical system of a PVI platform. System resolution defines the size of the defects that a system is able to detect. The second factor relates to the level of contrast of the defect. The defect must exhibit sufficient contrast with its background area on the package for the inspection system to reliably detect it. The contrast of a particular package or defect is determined by a PVI platform's illumination system.
Pixel Size
A pixel is a single picture element within the sensor array of a camera. The sensor array consists of a matrix of hundreds of thousands, or even millions, of pixels with the exact number differing from camera to camera. Low-cost cameras often have fewer pixels in the array. For purposes of discussion, we will refer to a camera that has a format of 1,000 pixels by 1,000 pixels.
 Figure 4. PVI systems require a minimum number of pixels to correctly identify a defect. In this example, the defect must be at least 4 x 4 pixels (an area of 16 pixels). |
A PVI platform's optical system defines the actual field of view of the sensor. This field of view is a specified size and is projected onto the 1,000 x 1,000 pixel sensor matrix. If the field of view of the optics is 25 x 25 mm and this is projected onto a 1,000 x 1,000 pixel sensor, the pixel size can be determined by dividing the field of view size of each axis by the pixel count of each corresponding axis. In this instance, each pixel would be 0.025 x 0.025 mm.
To further illustrate the impact that the camera and field of view can have on pixel size, Tables 3 and 4 show the difference in pixel sizes between a standard PVI camera that has a pixel matrix of 640 x 480 and a high-resolution camera that has a pixel matrix of 1,300 x 1,030 for different lens configurations.
 Table 3. Pixel sizes for a standard PVI camera that has a pixel matrix of 640 x 480. |
Ideally, an inspection should be configured so that the entire package fits within the field of view. Therefore, a 15 x 15-mm package would use a 35-mm lens with a field of view of 26 x 21 mm, providing a pixel size of 0.020 x 0.020 mm. If a particularly small defect demands a higher resolution, a lens with a smaller field of view can be used to capture multiple images, inspecting different sections of a single package. Conversely, if the defects being sought can be captured with a lower resolution, the system can be set to inspect multiple devices in a single image, increasing the throughput of the system. Typically, the higher the resolution (and the smaller the field of view), the lower the throughput.
 Table 4. Pixel sizes for a high-resolution PVI camera that has a pixel matrix of 1,300 x 1,030. |
For a PVI system to properly identify a defect, the area of the defect must contain a minimum number of pixels (a number that may vary from system to system). This minimum number of pixels is specified for both the X and Y axes, creating a minimum area of pixels within which a defect may be recognized. While PVI systems normally have the ability to detect smaller defect areas than the specified minimum, they may not be able to do so at the required level of inspection reliability, causing over-rejection of good parts and under-rejection of bad ones.
Given that pixel size is determined by a combination of sensor array size and system field of view, and that a minimum number of pixels are required to reliably locate a defect, the concept of minimum defect size can be understood. For example, if the system resolution provides a 0.025 x 0.025-mm pixel size and the system specifies a minimum of 4 x 4 pixels to reliably detect a defect, the minimum detectable defect size would be four times the pixel size for each axis: a 0.1 x 0.1-mm minimum detectable defect.
 Figure 5. LED-based lighting, shown here in a ringlight configuration, illuminates a wide range of defects within a PVI system. |
To detect a smaller defect, the resolution must be increased, which means a smaller pixel size must be used. This can be done by either increasing the number of pixels in the array while maintaining the same field of view or by decreasing the field of view while maintaining the same pixel array. Either method may be employed, but there are technical and financial limits on the number of pixels. Costs increase significantly for high-resolution equipment. On the other hand, reducing the field of view may not allow the complete device to be viewed within a single image, requiring multiple images to inspect the entire device and therefore decreasing throughput.
Once a manufacturer understands the relationship between defect size and pixel size, it is simple to select an appropriate configuration or inspection technique to achieve reliable inspection results. Field of view and system resolution must be optimized based on the minimum detectable defect size that the manufacturer requires.
Image Contrast
In addition to pixel size and area requirements, the other key to establishing reliable PVI is contrast, which is primarily determined by the illumination system. Contrast represents the difference in gray level between the surface of the object under inspection and a defect that may exist on that surface. A PVI system is typically capable of imaging 256 different levels of gray, with the two extremes being black and white. To reliably identify a defect, the difference between the gray level of the defect and the background area where the defect resides must be a minimum defined number of gray levels as imaged by the system. As with the amount of pixels required to reliably locate a defect, the system may be set up to detect defects that are less than the minimum amount of gray levels specified, but it is best to maximize the contrast of the defect to ensure reliable inspection results.
Expanding PVI Technology
To accommodate the wide variety of package types, materials and defects, a PVI system must be flexible with regards to image processing and illumination. The ability to interchange sensor technologies, such as moving from a standard camera to a high-resolution camera, is beneficial when the inspection process demands improvements in resolution or reliability.
Image Processing
The evolution of an imaging platform should be toward higher processing speeds and new imaging algorithms, enabling manufacturers to leverage the process expertise developed on the original platform for more demanding applications. Once an algorithm has been proven and optimized, it is ideal to embed it into the system firmware, enhancing speed and ease-of-use.
As the image processing capability of the PVI system is upgraded with new inspection algorithms, it is important to increase processing speed to ensure that overall system throughput is not impacted. To do this, the system needs to accommodate additional processor boards or offer some other means to increase processing capability without impacting system stability. This allows processing speed to be increased without changing the platform, enabling the user to keep the same proven inspection algorithms to maintain performance and overall platform stability.
Illumination
A PVI platform's illumination system is key to achieving reliable inspection results. In most cases, an illumination source can be identified that will properly highlight a particular defect on a particular type of package. The challenge is to find a flexible illumination scheme that can illuminate any defect on any background surface that the system may encounter. In an effort to meet this goal, a flexible illumination system will provide programmable intensity, angle, and even a particular segment or quadrant of the lighting system that may be controlled independently. Different types of lighting, such as light-emitting diode arrays (LED), incandescent sources and flash lamps may be considered singly or in combination.
Illumination is often difficult to optimize across a broad range of defects because the characteristics of the defects and the background surfaces where they reside vary significantly. To accommodate large variations in lighting requirements between different inspection tasks on the same device, manufacturers may employ multiple sources to capture multiple images. However, this may reduce system throughput.
Cost of Ownership
Manufacturers who know the roles that device types, defect sizes, contrast, image processing and illumination play within the PVI process can understand the importance of a PVI system that has the flexibility to be optimized and expanded as product requirements evolve. This flexibility enables manufacturers to maintain reliable inspection over the long term while achieving maximum efficiency, throughput and value from their PVI systems.
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; E-mail: [email protected].