Jayson Wilkinson
Image acquisition and machine-vision algorithms can be combined with optical- power monitoring and motion-control equipment for the most accurate and efficient solutions in both coarse and fine alignment processes. Use of open standards creates a system that is scalable throughout an organization, flexible, and open to reuse as alignment systems change or grow.Using human operators to perform precision alignment can result in long manufacturing times and poor yields. To achieve better results and streamline processes, machine builders are taking advantage of new, automated technologies. For high-tech production, however, machines must be flexible and adaptable to grow with the number and complexity of components.
Access to the latest and most advanced technologies is available by leveraging computer-based solutions based on open standards. Machine builders can use computer-based systems for precision alignment techniques to shorten development time and create productive machines with the flexibility to adapt to changes in the future. Because these systems use computer-based technologies, they can easily perform statistical process control (SPC) analysis or report to central databases and manufacturing-execution system applications.
One strategy to perform computer-based, optical-component alignments is to use an integrated approach that combines motion, vision, and data-acquisition products. While a motion controller guides the component and/or the stage, data-acquisition products, in combination with sensors, take optical-power measurements while vision devices provide visual feedback for coarse alignment. A motion/vision combination allows several measurements to be taken—movement, position, and alignment—along with visual inspection and laser-beam position. In addition, by using open industry standards like PXI, machine builders can easily adapt and synchronize these measurements to any machine configuration.
COARSE ALIGNMENT AND IMAGE ACQUISITION
Alignment systems typically break the alignment process into two steps—a coarse alignment followed by a fine alignment. In general, the purpose of coarse alignment is to align the optical components close enough for one of the components to detect light transmitted through the other. The initial detection of light through the optical components is referred to as "finding first light." If possible, the fixture for the components should be designed so that the components are already aligned well enough to find first light before the automated alignment process begins.
FIGURE 1. A motion system, in combination with a vision system, can perform coarse alignment quickly. The motion controller aligns the two components using data provided by the vision system through the computer's image processing software.
Vision-guided motion can be achieved by applying a threshold to isolate the components. A simple particle analysis or edge detection can then identify which component is which, and assign a spatial coordinate in the two dimensional field of view. This spatial coordinate consists of pixel values but can also be represented in the computer by engineering units such as microns.
Using good monochromatic lighting, a high-power lens, and a high-resolution camera, machine builders can design systems that capture images and make accurate measurements of features as small as a few microns. For systems requiring only a few microns of accuracy, a vision system may be appropriate. However, many precision alignment applications require resolutions of less than 10 nm. No camera and lens combination viewing visible light can resolve features smaller than the wavelength of visible light. Because of this limitation, vision systems are not used for very high-precision alignment.Another method for performing coarse alignment is to monitor optical power at various positions, measure coupling efficiency, and move to an aligned position. Although not as fast as vision-guided alignment, this method can use the same feedback device used for fine alignment, and therefore is often a lower-cost solution than vision-guided motion. The components can be aligned by searching for the peak transmitted light intensity and thereby determining the physical conditions for optimum coupling.
Optical-power measurements typically characterize insertion loss by measuring light through an optical device under test. Machine builders can use these same power measurement techniques to align a fiber or array of fibers to an optical component. The source (such as a tunable laser, a continuous-wave laser, or an LED light source) is used to emit a signal through one of the components to be read by an optical-power meter or photo detector connected to the other component.
Two of the most widely used coarse search paths are boustrophedon (raster) and rectangular (square) spirals. To perform a coarse scan, the motion controller makes repeated point-to-point moves, measuring the power between each move while the mechanical system is stationary. Motion control occurs along the x- or y-axis, so interpolation or coordinated multi-axes moves are not needed. The rectangular, or square spiral, is easier to use in a blind search, largely because this path has less variance.FIGURE 3. A motion controller coupled with computer-based data acquisition hardware can perform very fast fine alignment. The motion controller moves the first component in a search path around the area of the second component while triggering the data acquisition hardware to get position vs. optical-power data. Software uses the optical-power data as well as the corresponding position information to find the peak location. (Photo courtesy of National Instruments)
One drawback of both the rectangular spiral and the raster scan is that the motion system must change directions by 90° at the end of each line. Because real mechanical systems are unable to accelerate and decelerate infinitely fast, a 90° direction change requires the motion system to decelerate to a stop and accelerate again back to the specified velocity. Stopping and starting in this manner causes undesirable vibration and inertial effects that can reduce the accuracy of the system and cause wear on the mechanical components.
The circular spiral is more efficient than both the raster and rectangular spirals in coarse alignment. For circular motion, a controller that can smoothly coordinate movement between multiple axes can be used. The Archimedes, or equiangular spiral, is a popular choice for evenly distributed sampling. Many other spiral search paths also exist, such as logarithmic spirals for scans with greater density near the center (see Fig. 2). The advantage of using the circular type spiral is that the motion system does not have to move through corners and therefore does not need to stop and start as often. The disadvantage of a circular spiral is that the four corners of the search area are not covered as they are with the raster scan and the square spiral.
The benefit of using commercial, off-the-shelf, digital motion controllers is that machine builders have the choice of using the search pattern best optimized for a system. Several open-standard motion controllers offer smooth movement through arbitrary paths such as spirals. In addition, open-standard motion controllers have a significant advantage over closed architecture systems in that they are flexible enough to meet future needs that will surface in the fast-moving optical-networking industry.
FINE ALIGNMENT
Like the coarse-alignment process, algorithms as well as control and measurement hardware guide the fine-alignment process (see Fig. 3). The most commonly used fine alignment technique is a simple dynamic gradient process.
In this algorithm, point-to-point steps move along a single axis until optical-power measurements decrease; then, the step size is halved, the direction reversed, and so on. This algorithm is reliable and also achieves high final resolution. However, if step sizes are not optimized, it is possible to find local maximums instead of the true spatial maximum, resulting in a poor alignment. This is especially true when dealing with non-Gaussian distributions of light, like those from vertical-cavity surface-emitting lasers (VCSELs).
PRECISION ALIGNMENT ELEMENTS
Assembling the elements of a precision alignment system starts with choosing hardware that works best. In addition to motors to drive the motion control, the mechanical elements that make up a motion system are needed (see Fig. 4).
In general, the optical components must appear clearly enough to perform vision analysis through the initial search-window length. In addition, it is important to have space between the devices and the lens, so that the user can adjust the optical components without interfering with the vision system, and also to allow for other manufacturing tasks such as final verification test and bonding. By identifying these requirements, cameras and lenses that have the appropriate charge-coupled-device (CCD) sensor size, focal length, working distance, and depth of field can be selected.
High-level machine-vision software tools that make precision alignment easier to prototype and develop can also be used. Having the right vision-hardware components also simplifies choosing the correct algorithm.
An alignment system constructed using open standard tools, for instance, integrates the PCI bus for high bandwidth, the real-time systems integration (RTSI) bus for synchronization, and the motion controller for contouring and high-speed position registration. In addition, a high-resolution, 16-bit data-acquisition board is needed. Optical power can be sampled using photodectors from a variety of vendors. The data-acquisition board sends a clock signal through the RTSI to the motion controller to synchronize and collect time-based measurements and position information.
ACKNOWLEDGMENTS
The author would like to thank Joseph Ting, motion control product support engineer, and Jim Balent, business development manager, at National Instruments for their contributions to this article.
Jayson Wilkinson is motion product manager for National Instruments, 11500 N. Mopac Expressway, Austin, TX 78759-3504, 512-794-0100. He can be reached at [email protected].