E-mail messages, sports scores, stock quotes, even color graphics—the merged functionality of many of today’s electronics products has led to a clear evolution in liquid crystal displays (LCDs) and light-emitting diode (LED) displays: smaller, but displaying more information.
To get more information into a smaller area, the resolution and density of pixilated displays are increasing. Unfortunately, this has made the process of ensuring product quality increasingly difficult for electronics manufacturers.
No matter how small the individual pixels, manufacturers must ensure that each pixel functions properly so information appears correctly on the display. For instance, consider the 3 and 8 characters on a typical cell-phone display. Only a few regions of illuminated pixels differentiate the two. So if you press the 8 key and see a 3 on the screen, it may be because faulty pixels went undetected during the manufacturing process.
Display pixels fail for a variety of reasons. Problems range from solder shorts, in which excess solder paste bridges connector leads, to firmware bugs, where the software governing display behavior generates the wrong output. Whatever the root cause, to remain competitive, electronics manufacturers must find these defects early in the production process, before more value is added to the device.
Limitations of Manual Inspection
Traditionally, manufacturers have inspected displays by in-circuit testing, manual inspection, or some combination of both. In-circuit testing, which still is performed at many manufacturing sites, is conducted during assembly to catch any gross defects before more value is added to the product.
Typically, a display attached to a printed circuit board is fixtured into a testing station in which an overhead bed of spring-loaded pins touches down on each exposed test pad on the board. Then current is applied, turning on the entire set of display pixels at once to indicate general circuit functionality. Unless gross defects are present, such as a whole section of the display malfunctioning, the product passes the testing stage and proceeds to assembly.
Manual display inspection primarily is performed at the end of the production cycle. An operator examines the product, turns it on, and determines whether the display appears to be working normally or whether any noticeable defects are present. In some cases, a command can be punched into the product’s keypad to illuminate display pixels in specific patterns, such as a square or checkerboard, to make faults more visible to the operator.
While these approaches have helped catch display defects, both have significant limitations. Since in-circuit testing primarily is used for gross-defect detection, it typically will not ensure individual pixel integrity. Nor will it detect dependent behavior between individual pixels or segment clusters that can cause ghosting, a situation where turning a single pixel or segment on will turn on several others in its proximity.
On the other hand, the operator may catch this type of defect during final test. However, at this point, a device typically has been fully assembled and enclosed and must undergo costly and time-consuming rework involving reopening the device to investigate and fix the problem. Not only does this require a technician’s time, but, in many cases, also the replacement of new parts damaged by reopening a device.
Manual inspection also suffers from subjectivity. Manufacturing engineers prefer benchmarks to estimates, and specifying acceptance thresholds for contrast, brightness, and uniformity can be a great challenge. An operator’s estimate of these parameters can vary throughout their shift and may conflict with the judgment of other operators. Even when compared with fixed standards such as a photo of a good product or a previously passed product, operator opinions will vary.
Of course, the most common liabilities of manual inspection are repetition and boredom. In an eight-hour shift, an operator may inspect up to 1,500 displays. Even the most conscientious would have difficulty ensuring 100% product quality at this pace.
Advantages of Machine Vision
During the past several years, there has been an increasing trend toward using machine vision technology for automated, 100% display inspection. In a typical system, a video camera above a conveyor captures images of an LCD. While the product is under the camera, it is electronically controlled to display a sequence of images that would reveal independent and dependent behavior defects.
The images are processed on a PC which uses a combination of machine vision hardware and software to determine if any defects are present. All the inspection
parameters such as contrast, brightness, and uniformity can be determined relative to a fixed standard.
The speed of the inspection typically is within fractions of a second. While a complex display can take 1.5 s, simple LCD inspections can take 1/100th s. At the completion of an inspection, the vision system communicates the status of the product to an automated reject spur.
Until recently, most vision systems that inspected LCDs and LED displays have been built from general vision components and adapted to the task of inspecting displays. The success rate usually is proportionate to the R&D that went into the project, and implementations can be very display-specific.
For example, many LCDs come with a protective film coating that can wrinkle, polarize, or distort the displays they protect. Also, bad material combinations can make the lens, protective film, or LCD background reflect or blur the light used to illuminate the product. As a result, process parameters must be selectively opened or tightened to reflect the variation seen from display to display.
With general-purpose vision components, this can be very time-consuming and can intimidate manufacturers that do not have the expertise to tailor vision systems to accommodate the variety of process variations that can exist. Consequently, many manufacturers in the past have shunned away from using vision for automated inspection.
Machine Vision for Display Inspection
Recent advancements in both PC and machine vision technology now address the needs of display inspection. For example, a new vision system from Cognex supplies a variety of display-specific tools that minimizes R&D time by wrapping the tool complexities in GUI interfaces.
One tool applies a virtual probe to each pixel on a display, enabling the system to determine whether the display is good or defective despite inconsistent lighting or creases that have formed in the transparent lens tape. A graphical Windows-based interface allows access to features without writing software.
New machine vision technology also can look for other types of defects. For example, the same vision systems that inspect pixels can measure the alignment of a display within the housing. They can check the consistency of backlighting, verify the print quality of keypads and the color of logos, and examine the product enclosures for scratches.
Perhaps the most important benefit of machine vision technology is the capability to detect errors early in the manufacturing process. Then operators can take care of problems on the spot and make modifications to the process before more value is added to a product.
The cost savings here can take a variety of forms. Manufacturers estimate that catching a defect on a device at the board level, before final assembly, can save up to 90 minutes of a technician’s time per defect in the repair loop. This includes debugging, disassembling, reassembling, retesting, process logging, and reinserting the device on the assembly line.
Many small electronics housings were not meant to be reclosed after opening. As a result, the housings and any other associated parts such as gaskets, rivets, and shock pads become 100% scrap.
Return-on-investment for such systems varies greatly because it depends on so many variables including parts cost, labor rates, defect rates, the manufacturing process, the product design, and the inspection station implementation. Few manufacturers even attempt to quantify this themselves because the cost justification is gross enough to be considered empirical.
Conclusion
Depending on how an inspection station is implemented, cost can vary by more than an order of magnitude. Implementation can be as sparse as a PC, a display inspection package, a camera, and an electrically driven reject mechanism.
As today’s electronics devices continue to increase in functionality and drive greater capability in their displays, more manufacturers are turning to machine vision technology for consistent and reliable inspection. By helping to improve quality and productivity while lowering production costs, the technology is proving essential for highly competitive industries that want to profit within tight margins.
About the Author
Milton Yarberry is a senior vision solutions engineer at Cognex. Before joining the company, he held various manufacturing engineering roles at Motorola and Lucent Technologies. Cognex, One Vision Dr., Natick, MA 01760, (508) 650-3000.
Copyright 2000 Nelson Publishing Inc.
April 2000