The rapid adoption of wavelength-division multiplexers, laser diodes and optical switches, modulators, splitters, and filters has raised the economic stakes for optoelectronics manufacturers. Price pressures and volume demands require truly industrial approaches to assembly and packaging. Yet one crucial step in the production process remains a manual one for all but a handful of manufacturers: the exacting alignment and bonding of optical fibers. Clearly, making automation of this step affordable for a wider segment of optoelectronics manufacturers is key to ongoing industry profitability as the market continues its expansion.
In single-mode applications, such automation requires transverse position accuracies better than 25 to 50 nm, plus real-time compensation of drift processes. Transverse alignment accuracy requirements can be eased if a collimating lens is used, but this can impose additional difficulties for angular alignments. Cumulative dimensional tolerances—such as core centration, diode placement and orientation, cleave angle, mounting surface uniformity, and a host of other minuscule variables—rule out use of a dedicated tool to achieve alignment accuracies of the required order. Such fixturing still requires a physical alignment step. Often, the alignment must be repeated during or after the bonding process, due to misalignments introduced by the application or action of the bonding agent.
A skilled worker rapidly develops a feel for the alignment procedure and can usually perform an alignment in less than two minutes using a coupled power meter and possibly a video microscope. Pick-and-place and other procedures in the packaging process easily double this time, so a worker can handle perhaps 10 to 15 units per hour for the simplest devices. The inevitable breakage in manual processes slows the output even further, and such a production rate is not economical. To make matters worse, many optoelectronic devices have more than one fiber or other element requiring alignment: optical switches can have three or more fibers sprouting from the package and high-power pigtailed laser diodes can require submicron/submilliradian six-axis orientation of confocal optical trains.
Challenge of automation
Alignment automation is one of the most challenging processes in motion control. Elements that start out arbitrarily oriented to each other must be brought to optimum alignment quickly and with submicron accuracy, blind except for feedback from the coupled power meter.
Automated alignment systems fall into two categories. The first covers laboratory-duty instruments with analog phase demodulation or hill-climbing decision paths. These open-loop mechanisms are usually unsuited for multimode applications and generally offer only transverse alignment automation.
The other category includes industrial process tools configured to perform pigtailing, bonding, or other specified procedure. These tend to be costly and require custom application engineering. Angular alignment automation is often unavailable, or restricted to very constrained packaging and fixturing situations. The automation often uses video feedback with unimpressive angular resolution and is incapable of correcting the undesirable transverse movement of the device arising from fixturing offsets. As a rule for custom systems of this sort, any change in the physical design or configuration of the devices being processed requires substantial redesign of fixturing tools, which is costly and time-consuming.
Flexible alignment-automation subsystems capable of automating angular, transverse, and longitudinal alignment are therefore needed for production applications. Such subsystems should be suitable for general application by the customer, original-equipment manufacturer, or systems integrator, eliminating the need for costly engineering and software.
In many cases, all that is necessary to satisfy manufacturing needs is a reliable, stable, responsive, and high-resolution positioning capability. Piezoelectric positioners are certainly capable of the necessary resolutions, and conveniently configured compact positioners are now available with integrated multiaxis capabilities. With travels of 100 µm and nanometer-scale resolution, these mechanisms offer up to five axes of integrated motion. They are frequently used by manufacturers and systems integrators as the heart of a manual or computer-controlled alignment workstation.
A desire to make fixturing easy, along with a continued need for high throughput, drove the development of a specialized transverse nanopositioner with piezo-class high-speed responsiveness and a novel freespace mounting platform, which makes access easier and eliminates the need for intervening fixturing between the stage and the optoelectronic load. It integrates closed-loop positioning technologies into a subassembly suitable for a wide variety of industrial production and test applications. As with all piezo devices, its speed and resolution allow it to do a wide-field scan of the devices being aligned in a practical time frame.
Despite its unusual configuration, this device is a general-purpose x-y positioner compatible with our analog and digital closed-loop controls. Four fieldless low-voltage piezoelectric actuators operate in push-pull mode for high-speed actuation, yielding micron-scale motions in less than 3 ms. A patented coprocessor eliminates the resonant reaction of loads and neighboring components for high-speed actuation. Integral, noncontacting linear variable differential transformer sensors monitor the actual position of the rigid freespace actuation platform, eliminating hysteresis and backlash and yielding closed-loop resolutions to less than 25 nm and less than ~0.1 µm maximum DZ runout error.
Most flexure devices are based on a monolithic flexure design and use no bearings. Bearings are susceptible to contamination and damage and are a significant source of particulate generation. Besides offering motion that avoids static friction—the key to high resolution and repeatability—flexure designs use no lubricants and require no scheduled maintenance or mechanical adjustments. There are no parts like leadscrews that can slide or wear down.
While motion devices such as those discussed above are compatible with any alignment methodology, an object-oriented software package with high-throughput alignment automation and two-dimensional profiling capabilities can also be used. Our example package, based on LabView 5.1 from National Instruments (Austin, TX), performs rapid nanoscale positioning synchronized with metrology.
The systems described so far address high-resolution transverse (x-y) alignment automation. Equally important for today's advanced packages—which can incorporate isolators, lenses, and confocal optical train subassemblies—is the process of angular alignment in qx and qy. This can be addressed using rotation stages, tilt stages, and goniometers, but such mechanical solutions are unsatisfactory because these devices have fixed rotation points that can be counted on not to coincide with the fiber tip, lens focal point, or other desired rotation point. Thus, each angular adjustment causes a transverse misalignment, leading to the time-consuming process of making a transverse alignment following each angular step.
Applications such as these drove the development of six-axis hexapod microrobots. Ours is equipped with fully integrated transverse and angular automated alignment algorithms. Though such motion devices are still exotic to most engineers, they have been produced for five years, and their basic elements are the familiar dc servomotors, encoders, and PC-card motion controllers. The key to these parallel kinematic positioners lies in the real-time coordinate-transformation algorithms that simultaneously move the six linear motion struts to achieve a desired position and orientation. The principle is similar to that of flight simulators and "virtual reality" rides: the six linear actuators are coordinated by computer to allow arbitrary linear and angular motions in six degrees of freedom. Rather than constraining their trajectory by bearings, such systems are constrained by the algorithms that define the path planning of the system in real time while making the systems easy to use.
In particular, this means that the center of rotation can be mathematically placed anywhere rather than residing at the fixed center of a rotary bearing. For example, the center of rotation can be placed at a fiber tip, lens focal point, or any other desirable point. A software subroutine can rapidly sequence transverse and angular alignments, automatically determining the ideal rotation point to correct for geometrical and fixturing offsets and tolerances.
This system works for packages containing several elements that all need to be aligned to the best mutual orientation. For example, a laser diode package containing a confocal optical train (COT) can be assembled according to the following sequence.
The package is inserted into a power socket mounted on the microrobot's platform. The COT is brought into position with a handling stick mounted on a simple motorized or pneumatic actuator and held stationary. Software sets the rotation point of the microrobot to coincide with the location of the back focal point of the COT. The package is oriented in qx and qy and bonded into place. The handling stick is then retracted.
The fiber is then brought into position by another motorized or pneumatic actuator. The rotation point is reset to coincide with the front focal point of the COT. Transverse and angular autoalignments are performed, usually in 10 to 15 seconds. For applications requiring better than 0.2-µm transverse alignment resolution, a high-resolution piezoelectric transverse positioner is used along with the microrobot and activated to perform the final transverse alignment.
By taking a modular approach to alignment automation subsystems, manufacturers and systems integrators can build workstations that work fast enough to make the construction of optoelectronic devices more affordable.