Charles Clarkson
The flexibility and versatility of remote fiberoptic illumination has led to widespread adoption of this technology for microscopy and machine-vision applications.
From microscopy through industrial task lighting and machine vision, fiberoptic illumination systems are being implemented in an ever-widening range of lighting applications. In optical microscopy, illumination has long been recognized as a critical aspect of system design. The use of polarized light and spectral filters, for example, dates back more than 100 years. But surprisingly, the need to fully optimize illumination is often overlooked in industrial/production-line applications, such as visual inspection and machine vision, where the cost of failure can be poor product quality or a stalled production line. Indeed, many viable applications have failed for lack of appropriate lighting. Also, there is a common misconception that illumination problems can be compensated for downstream by clever image processing software.
Although these attitudes are slowly changing, illumination design often still is left to the very last during application development. Illumination-systems suppliers have responded to this market reality with a modular approach to custom design, enabling rapid delivery of novel prototype systems.
Illumination optimization
Many parameters can be optimized during illumination system design. The most important include source intensity properties, spectral characteristics, direction/angle of incidence, structure/intensity pattern, polarization, and temporal properties. Each of these factors can have a critical impact on an application.
The light intensity delivered to the workpiece (usually defined by the irradiance) is almost always a significant consideration. Low intensity may result in an unduly long image integration time, thus limiting the ability of the system to keep up with the speed of a production line. On the other hand, too much light can heat or otherwise damage the object being imaged. This is often the case in optical microscopy of biological material.
Source wavelength characteristics can be defined by a spectral description or color temperature. Choosing the appropriate illumination wavelengths can be vital in creating the necessary contrast to image a colored object against a colored background. In microscopy, certain wavelengths can be used to excite fluorescence in specific dyes or markers.
Lighting angle can be used to good effect in several ways. For example, an oblique angle can highlight surface features or defects on a flat surface. Alternatively, dark-field illumination can be used with transparent parts. In dark-field illumination, the source is directed so that none of its output directly enters the camera or eye; only light scattered from cracks and defects will enter the detection apparatus, thus appearing as bright features against a dark background (see Fig. 1).
Light polarization can be another important illumination parameter. Polarized light, for example, has long been used in microscopy to highlight transparent interfaces and is now finding increasing use in industrial applications for this same purpose.
The use of structured light is particularly useful in machine vision. A line projector is one of the most common examples. Line illumination can be used in gauging measurements in which the shape of the image provides information about the three-dimensional shape of the object (such as out-of-plane misalignment of car body panels).
Some applications also can benefit from pulsed or strobed lighting. This lighting is particularly useful for eliminating image blur when examining items on a fast-moving, continuous production line.
Why fiberoptic illumination?
Fiberoptic illumination dominates many vision and lighting applications because it can be easily and rapidly customized in all these areas. More important, it allows the light source (lamp, power supply, and such) to be remote from the task, saving critical space as well as simplifying thermal loading issues. Both of these factors help minimize the overall cost of the application. Remote source placement also removes any vibrations and minor air turbulence due to lamp cooling from the imaging location.
In addition, fiberoptic systems provide mobility and flexibility; the lighting can be readily moved, adjusted, and reoptimized to accommodate changes in a process requirement. Furthermore, the ability to form fiber bundles into novel shapes is a much simpler and more compact way to provide structured illumination than the use of imaging optics.
Recently, there has been some interest in using the new high-brightness LEDs for illumination tasks, particularly in machine vision. While it is true that LEDs are compact and produce little heat, they do not yet provide the performance necessary for most applications. Specifically, LEDs do not offer sufficient intensity and suffer from wavelength limitations. Until recently, only long-wavelength LEDs were available, and the new white-light LEDs are not only expensive but are blue biased, resulting in images with distorted color representations.
Modular components
The basic components of a simple fiberoptic illumination system are the lamp source, power supply, optional filters, fiberoptic bundle, and illumination delivery optics. In a successful implementation, these elements must be carefully matched with each other, as well as the needs of the application.
Four basic types of light sources are typically used in fiberoptic illumination systems. The most common are metal halide and quartz halogen lamps. The applications for these lamps fall into two categorieslow power (25 to 30 W) and high power (approximately 150 W). It is also quite common to overspecify the lamp and then operate at a derated power level in order to extend bulb lifetime to thousands of hours. The two other, less commonly used source types are fluorescent and xenon strobe lamps. All these light sources are offered as modules that typically incorporate a regulated power supply with smart functions, allowing the light intensity to be varied on demand by remote control.
The light source can be directly interfaced to the fiber bundle or optical filters can be used to spectrally alter the lamp's output. The most common such filter is the so-called hot mirror, which selectively rejects infrared while efficiently transmitting all visible wavelengths, thus protecting the illuminated object from unnecessary heat.
The fiber bundle itself consists of thousands of individual fibers within a protective sheath of metal, PVC, or even Teflon. Currently, the industry standard diameters for the individual fibers are 25 and 50 µm (including cladding). The standard numerical aperture (NA) for these fibers is 0.55. Recently, demand has been growing for higher (0.66) NA fibers. Higher NA enables collection of more light from the source and results in a wider illumination cone at the output end. Typical lengths range up to 30 ft.
At Dolan-Jenner we draw fibers from several types of glass and quartz to suit specific needs. For instance, a flint glass core with soda lime glass cladding offers low cost but is only rated to 260°C. Conversely, all-quartz fiber bundles are more expensive but can withstand temperatures up to 1095°C, as might be required in a kiln or foundry.
The ends of the fibers are bonded at both the lamp end and the remote end before these bundle ends are polished. Epoxy bonding is most common, but other methods, such as fusing, deliver special performance, for example, high-temperature operation.
Of course, the ability to form the bundle into a termination of virtually any shape is one of the major advantages of fiberoptic illumination. Terminations include square back-lights, microscope interfaces, circular and annular illuminators, line projectors, linear flares, right-angled probes, and a wide range of custom end pieces.
Integration of these system components is the key to a successful illumination application. At the same time, the "design it last" attitude toward illumination systems means that fast turnaround on both custom and standard systems is crucial for success in this market. The only way for a supplier to meet these needs is to become vertically integrated. We not only produce our own lamp modules and power supplies, for example, but even draw our own fibers, giving us complete control over product quality and delivery schedules.
In machine vision, the other important factor in successful illumination is to fully evaluate a prototype system under realistic conditions. Lighting application systems that work on the bench do not necessarily translate to the manufacturing floor where color variations in parts and fluctuating ambient light create distortions. That is why it's important to work with an experienced system vendor who will support the application through both design and implementation.
Applications
The applications for fiberoptic illumination span many industries and technologies; a partial list includes dentistry, biomedicine, web (plastic, paper, foil) production, automotive, electronics, and semiconductors. Examination of a few diverse examples illustrates the value of illumination system optimization.
Manufacturers of cast and machined metal parts increasingly use machine vision to perform metrology on their products, as well as monitor for scratches or other surface damage. Here the camera views the edges and surface at normal incidence and the fiberoptic illuminator is set up to produce oblique lighting, such as an angle of incidence close to 90°. This results in a high-contrast image (see Fig. 2), forming an ideal input for thresholding software, which produces a binary image for edge metrology. In addition, raised features throw sharp shadows with length proportional to their height above the surface. Last, any surface scratches or nicks will scatter the incident light, appearing as bright features.
Several interesting applications have been found for polarized illumination, particularly for highly reflective or transparent objects. Typically, the source is polarized using a polarizing filter; often a polarizer is also placed in front of the camera to remove specular reflections from the part under test. A typical example is the QC/QA of bottle necks by both bottle manufacturers and beverage companies. Conventional illumination from one side reveals some of the vertical cracks, but specular highlights prevent crack detection in some portions of the bottle lip (see Fig. 3). In contrast, polarized illumination eliminates most of the specular light. This same illumination method can be used to visualize stress in transparent objects.
One final application powerfully illustrates the utility of spectral filtering. Colored glass filters are often useful in monochromatic machine vision, a fact which sometimes surprises people. For example, a shampoo manufacturer wishes to verify the print quality on the back of the label on a clear plastic bottle (see Fig. 4). The problem is that the bottle is filled with a red, semitransparent liquid. Using conventional quartz halogen illumination produces poor image contrast. However, when a cut-off filter is used in the illumination system to transmit only red light, the print is clearly visible in the image, with excellent contrast.
CHARLES CLARKSON is president of Dolan-Jenner Fiber Optics, 678 Andover St., Lawrence, MA. 01843; e-mail: [email protected].