By KEVIN McCARTHY
Chief Technology Officer
Danaher Precision Systems
Edited by Leland Teschler
All moving objects possess six degrees of freedom — three linear and three rotary. The task of a linear-motion guideway is to eliminate, as closely as possible, five of these degrees of freedom, leaving a single rotary or linear axis of motion.
Air bearings are the purest means of defining linear or rotary motion. As many engineers know, an air bearing distributes the load-bearing surface over a large area. The load is borne by a moving carriage that rides on a thin film of pressurized air that separates it from a fixed guideway. Distributing the load over a large area averages out any microdefects in the machined surfaces, creating extremely smooth, straight, frictionless motion. The air-bearing guideways and carriages typically are ground to extremely high precision, often with flatness, parallelism, and perpendicularity being less than 0.5 m. Linear departures from straight-line motion are often less than 200 nm over 100 mm of travel in these devices.
Advantages of air bearings over conventional mechanical guideways become more pronounced as desired resolution increases. It is for that reason that air bearings find use in the assembly of photonic components. Here, misalignments of even a few tens of nanometers between two optical components can degrade performance severely.
In particular, new air bearing stages have become available recently that have been "miniaturized" for use in photonics alignment, fiber fusing, and Bragg writing. They employ direct-drive technology and are compact, in contrast with traditional air-bearing stages which tend to have a bulky cross section and length. This is because the semiconductor industry has been the main user of these devices, and there has been little incentive to design stages much smaller than a semiconductor wafer.
The friction-free nature of these air-bearing stages and their compact form factor makes possible the accurate dither motion usually used to align optical components. This sine and cosine oscillatory movement is at submicron amplitudes and takes place in the plane perpendicular to the optical axis. Friction nonlinearities and lubricant effects in conventional stages prevent them from producing such small movements. Piezo stages can produce the required dither motions, but their limited travel is a decided disadvantage. To get optical components in close proximity, they typically must mount on an additional set of stages.
Absence of friction brings with it a variety of other advantages. For one thing, the tuning process tends to be simpler for air-bearing stages than for electromechanical alternatives. On any given stage, the only free variables are the desired servo-bandwidth and the payload mass.
Stages with the same payload mass are identically tuned; there is no fine-tuning to handle unit-tounit variations. The tuning on day one is the same as at year 10 because there is no part wear.
In contrast, traditional stages have a reputation for needing careful servotuning. The reason is their assortment of mechanical bearings, leadscrews, nuts, and so forth. Attempts to minimize move and settle times often lead to settings that are marginally stable and that are individually hand-tuned for each axis.
The problem is exacerbated with the need to resolve ever-smaller distances. Tuning a conventional stage is difficult because it depends on nonlinear physical parameters such as leadscrew torque, linear bearing preload, and lubricant properties. These vary along travel in any given unit, as well as from unit to unit, and over time.
Frictionless operation also brings with it a settling time that is substantially shorter than that possible with conventional mechanics. There is no need to wait for the fairly slow effect of the servoloop integrator term to overcome friction. Standard stages might make a 5-micron move in 200 msec, settling to under 50 nm. An air-bearing stage can make the same 5-micron move in 20 msec. This order-of-magnitude increase in throughput, most noticeable with the numerous small moves characterizing photonic alignment, is compelling.
Additionally, air-bearing stages can move with an ultraconstant velocity and trace out motion that is extremely straight. Numerous photonic applications require extremely precise motion that has a constant velocity. Examples include fused-fiber couplers and MachZender interferometers, as well as FBG (Fiber Bragg Grating) writing. Suitable component selection can keep tracking errors during motion to levels below 10 nm. Any residual errors are due to (primarily) encoder errors, amplifier nonlinearities, and magnetic field departures from sinusoidality.
Low particle counts are an ancillary benefit of air-bearing technology, thanks to the absence of conventional gearing or screws and their obligatory lubricants. In some cases, pure dry nitrogen is used to operate the air bearings, producing nearly undetectable levels of particulate. If air pressurizes the bearing, conventional filters can effectively make it dust free by filtering out 99.7% of all particles above 0.1 microns.
The only meaningful source of particles is the cable plant because it is the sole component that experiences wear. Even so, the effect is small.
What cable movements there are give rise to forces that are small and nonfrictional. The variations in these forces are vanishingly small over the miniscule moves encountered during photonic alignment. The servointegrator function can produce an equal and opposing force to compensate.
Air-bearing guides are less stiff than rolling-element steel-bearing ways. This could, in principle, give air-bearing guides a first resonance that is lower than in conventional electromechanical stages and hence a lower servo bandwidth. As it happens, the numerous mechanical elements in traditional stages are usually what limit their servo bandwidth. Experience shows that servo bandwidths of air bearing stages are comparable to those of mechanical stages having a comparable size.
Air-bearing stages also function as both precise force generators and force transducers at the milliNewton level. Sensitive load cells are in essence built into each stage by design and provide this function.
The measurement and generation of force is quite useful for tasks such as detecting touch-points at extremely low force levels. The ability to detect part locations at the 100-nm level lets minute physical parameters (epoxy film thickness, for example) serve as process control variables that would otherwise be poorly controlled.
It is interesting to look in detail at the effects of friction and contact in conventional stages. These parameters fundamentally limit the resolution of such equipment and make it difficult to apply them in photonics assembly.
In particular, the stick-and-slip action that takes place as motion commences in conventional stages makes it almost impossible to accurately realize moves on the order of 20 nm. This is one of the reasons leadscrews run out of gas for resolutions at or below such levels. While there are tricks that can make leadscrews cooperate in this regime, they require fairly high-strung and gimmicky tuning and compromise dynamic performance. All in all, they are best avoided.
Similarly, a few mechanical bearing systems can be pushed to resolve distances on the order of tens of nanometers. But bearing friction, preload variations, recirculator cogging, and lubricant issues make this an uphill battle. The usual approach is to depend on the servoloop integrator function to paper over the problem.
In contrast, air bearing, directdrive stages have no intrinsic resolution limit. This allows them to handle tasks such as the production of Bragg writing systems with encoder resolutions of 31 picometers (less than the classical Bohr radius of the hydrogen atom).
There are, of course, a few disadvantages to air-bearing technology. Though lack of friction is generally a benefit, it has drawbacks as well. One positive aspect to friction is that it provides position stability in the presence of external stimuli and does so without the intervention of the servo loop. In a frictionless direct-drive stage, the servo loop is the sole means of suppressing axial vibration.
Vibration sources can be either the environmental background or servoamplifiers, especially those of PWM design. To compensate, makers of air-bearing stages incorporate high-quality servoloops that keep jitter to negligible levels.
Use of a direct-drive linear motor also gives air-bearing stages less sustained axial force capacity than in mechanical stages. The reason is linear motors have no mechanical advantage equivalent to that of ball screws. Typical continuous force levels in these miniature air-bearing stages range from 10 to 30-N peak.
However, photonic assembly operations normally involve small and delicate parts, so the inability to generate more than 30 N is not usually a disadvantage. Where necessary, higher forces are available from larger motors.
Infrastructure requirements must include provisions for an air supply. In that most manufacturing facilities distribute compressed air, the only additional costs for air bearings may be that of adding spot regulation and filtration. In the absence of house air, a compressed-air station next to the stage can make noise and vibration an issue.
Certain "traumas" can damage the stage. Dropping a vise on the air bearing surfaces, for example, will leave a dent that may prevent motion. Pumping oil instead of air into the compressed-air line is another failure mechanism, as is putting 20 A into a 5-A linear motor coil, or a 24-V supply on a 5-V encoder. Most of these issues can be easily prevented with fuses, I2T current limiting, coalescing filters, voltage clamps, and so forth.
Finally, air-bearing stages can handle high centered loads but only limited torques caused by cantilevered loads. There are several reasons (Abbe error, for example) why overhung loads are not a good idea in general. Tasks where they are unavoidable may be candidates for conventional bearings.