LED advances providing increased benefits for machine vision users
Light-emitting diode (LED) lighting technology has steadily progressed, allowing machine vision customers to take advantage of improved uniformity, better thermal dissipation, increased wavelength availability, and lower costs.
Over the last decade, light-emitting diode (LED) suppliers and chip designs have proliferated. "Not too long ago we were all limited to certain chip vendors for different products, and there wasn’t much differentiation between component manufacturers," said Jason Baechler, president of Moritex North America, Inc. "The sheer volume of options enables more applications because now you can get the required level of output or access certain wavelengths. End users don’t necessarily need to request something custom all the time for more challenging applications."
Machine vision customers have more choices for coupling the LED to optics to better control the light’s output or directionality of the output, as well as having access to more light-controlling accessories. Baechler also cites a large variety of materials, some of which have trickled down from the flat-panel industry, now available to the machine vision market, "so that we can produce lights with different directivity-not just based on the chip but what we couple it with, whether that’s an optical component or film," he said.
Taking some heat
One of the biggest drawbacks of LEDs is the heat they generate. Because heat reduces LED efficiency, less light is produced. But improvements in thermal management of LEDs are addressing these challenges.
"LED manufacturers are creating better thermal bonding pads, and that creates better thermal contacts to help get the heat out. They’re also reducing the amount of wafer bonds inside the LED itself and using chip-scale packaging," said Mark Kolvites, technical sales manager, Metaphase Technologies. "These features increase the lifetime and output of the LED."
The improvements have prompted Metaphase to reevaluate its product line of LED lights to accommodate the increases in efficiency. "What we’re seeing now is that we can put out a whole lot more light compared to what we were doing a year ago just by driving the LED harder because the thermals are so much better," Kolvites said.
Kolvites estimates that Metaphase’s lights are running about 15% to 20% cooler now using the same driver as a year ago. "We’re actually increasing our driver currents, taking more advantage of the light," he said.
Because of the vast enhancements in efficiency, "there’s still a lot of potential intensity that we can unlock in those LEDs just by adding a bit more heat dissipation," said Kevin High, operations and engineering manager at Metaphase. "Customers are always asking for brighter lights so they can run shorter exposures and run their production lines even faster. In our standard products, we are not yet pushing anywhere near what they can do."
Increases in overall optical efficiency have affected the design of coaxial illumination for telecentric lenses. In these products, light goes through a beam splitter and then is projected through the lens optics.
"Even with the best designs this can create reflection, and efficiency has grown so much that we actually have to step down the illuminators so that we can control the light level to ensure there isn’t too much reflection that washes out an image or affects how the lens performs," Baechler said.
Wavelength availability in machine vision
Matt Pinter, head of engineering at Smart Vision Lights, sees the biggest advancements in LED lighting coming not from the die or chip but rather from the availability of wavelengths for nearly any machine vision inspection task.
Research from the agriculture college at Michigan State University demonstrates the importance of wavelengths while offering lessons for the machine vision industry. Smart Vision Lights provided LED lights for a project that studied how wavelengths affect plant growth. Researchers found that certain wavelengths made plants grow while other ones made them bloom. The project also revealed that some types of lettuce only use one wavelength.
"You give the plants all these wavelengths and it’s a waste of time and energy," Pinter said. One wavelength garnering interest from machine vision customers is shortwave infrared (SWIR), primarily in response to increased demand for SWIR cameras. SWIR light penetrates a few millimeters of most materials. "With SWIR, you can solve applications that could have never been solved before by using visible or standard lower-wavelength IR," said James Gardiner, sales engineer at Metaphase. "It’s not just a niche part of machine vision. It can be applied to any market, whether it’s health care, automotive, or technology."
Gardiner offered the example of inspecting the seams and content of medical IV bags. "It’s very difficult to find a clear liquid in a clear material bag," he says. "Especially in the 1450 nm IR spectrum, any water-based product will actually absorb the SWIR and appear black so that you can get very good contrast between a clear background and the water-based product." The same goes for clear glues and lubricants commonly used in automotive applications.
Meanwhile, large tech companies are using SWIR lights for inspecting defects in silicon wafers, which become clear like glass beyond 1100 nm. "A good silicon wafer would appear invisible to the camera, so SWIR helps you create great contrast between it and the defect," Gardiner said.
Perhaps the biggest challenge with UV and IR lighting comes from the optics, as few options exist for diffusers that are effective on non-visible wavelengths. Without the proper optical diffusing material, the light can pass right through the object under test and create "hot spots," which show up on the images where the light is too intense, therefore reducing the image’s overall contrast. Conversely, if the optical diffusing material absorbs too much light, none of it makes it to the object under test.
While the main type of material for LED optics is plastic, UV light will cause that to break down because of solarization. Materials such as fused quartz and fused silica offer high UV light transmission. Choices in optical materials for SWIR also are limited but include fused silica, germanium, calcium fluoride, and sapphire.
For several years, Smart Vision Lights has used silicone for the functions of housing, sealing, and lensing direction of its LED lights. According to Pinter, silicone optics have many advantages over their plastic and glass counterparts. Optical-grade silicone features a robust design that offers high transmission across a broad spectrum, withstands a large temperature range, and doesn’t react to UV light. Additionally, silicone molding is available at a fraction of the cost of plastic injection molding.
Doing more with less
Machine vision designers are on a continual quest to create systems that are easier, cheaper, more flexible, and more compact. The machine vision lighting industry is responding with multispectral LED lights that support inspection at a single station with one camera and one light source.
"The more information you can get about a product, the better analysis you can do on it," Gardiner said. "When you can get that information not only in the visible spectrum but in the IR range as well, you have a more robust inspection."
One of the first applications to use RGB + IR LED was ground pork inspection. The customer used the RGB spectrum to ensure that the pork was the correct color. Meanwhile, the IR portion of the light accomplished something that visible light could not: It penetrated the surface of the meat to look for embedded contaminants such as a rubber glove.
While demand for multispectral lighting continues to rise, hyperspectral LEDs could represent the next frontier in machine vision lighting. Customers interested in hyperspectral LED lighting "want a true broadband, solid-state flat light from 600 nm to 1000 nm with every wavelength in there, rather than a multispectral light that has an 850 [nm] LED, a 940, and a 1300 with peaks and valleys in between," said Pinter.
Because hyperspectral LED lighting is in its infancy, customers still have to rely on broadband bulbs such as halogen, tungsten, and xenon. "These bulbs throw away a lot of wavelength, so the LED would be custom tailored to a certain area of wavelength," Pinter said.
The correct lighting is critical to another upward trend in machine vision: 3-D imaging. For metrology applications, advanced light control product—which extends into the UV and IR spectra—uses programmable structured light patterns to produce 3-D data in real time.
"The technology has been used in lithography for years, but we’re bringing it to the mainstream machine vision market for automated optic inspection applications, like measuring PCB components, microelectronics, and other high-end components," Baechler said.
Although LED lighting development continues along a slow and steady path to greater efficiency, by combining the strength of more, and better-controlled lighting, with improvements in optics, sensors, and software, machine vision capabilities are managing to keep pace with customer expectations.
Winn Hardin is contributing editor for AIA. This article originally appeared on Vision Online. AIA is a part of the Association for Advancing Automation (A3). A3 is a CFE Media content partner. Edited by Chris Vavra, production editor, CFE Media, email@example.com.
Machine vision customers have more choices light-emitting diode (LED) products to control the light’s output as well as having access to more light-controlling accessories.
Machine vision users are constantly looking to create systems that are easier, cheaper, more flexible, and more compact.
Hyperspectral imaging’s potential could be the next revolution for machine vision applications.
What other potential advances in LED technology could benefit machine vision applications and users?