Bill Holtkamp and Tom Stakelon
Optomechanical switches have been the mainstay of switching when testing multiple components during manufacture or operation. The authors argue that a switch based on a Risley prism pair offers the accuracy, stability, and speed required for test equipment.
Testing is becoming more intensive and frequent during component manufacturing and packaging, and in implementation and operation of the final system. A key to managing costs in all this testing is parallel switching.
Specifically, the use of M x N parallel switches allows multiple components and systems to be accessed and tested on demand. This type of switch can also be used to rapidly test one or more components with multiple instruments sequentially. A single-channel, parallel switching mechanism based on an optical technique long used in aerospace applications satisfies the most important needs of this application—repeatable performance, random channel access, bidirectionality, low insertion loss, long-term stability and a short settling time.
A fiber switch directs the light from an input fiber to different output fibers on demand. The techniques by which this light can be redirected can be summarized as electro-optic and optomechanical. Electro-optical devices, such as addressable waveguides, have the advantage of very high speed, although for testing applications typically a few hundred milliseconds is fast enough, and is much shorter than the testing/data acquisition times.
The drawback of waveguides, however, is that they introduce insertion loss and polarization-dependent loss (PDL) effects, and can be unstable, compromising the ability to perform accurate quantitative tests and limiting the ability to carry out high power testing. As a result, all testing switches at this time are based on optomechanical methods to deflect a beam to the desired output fiber.
While microelectromechanical systems (MEMS) technology appears attractive for this type of programmable beam deflection, practical issues prevent its use in demanding applications. Specifically, the lightweight mirrored membranes tend to distort during actuation, which affects the beam profile and in turn affects the coupling efficiency into single-mode fibers.
Several macroscopic approaches have been implemented. Two examples are the use of a rotating ring supporting multiple butt-matched output fibers and the use of a rotating prism. The butt-matched rotor approach can have accuracy limitations, because consistent, efficient coupling of butt-matched fibers requires very precise alignment, which in turn depends on very accurate rotor motion.
The rotating prism approach is less sensitive but still is highly dependent on repeatable, accurate motion in the switch, which affects manufacturing cost as well as long-term performance. The same is true for switches based on tilting mirrors. Also, these switches are susceptible to variable coupling losses due to out-of-axis vibration or drift.
Furthermore, all these switching technologies can only be used to produce 1 x N switches. The only way to achieve true (M x N) parallel switching is to gang multiple switches together to form a switching tree, which multiples cost and can lead to cumulative performance errors.
RISLEY PRISM PAIR
The Risley prism pair has long been used in demanding aerospace applications that require the ability to perform small, repeatable beam deflections. Typical examples include testing guidance systems in "smart" weapons, satellite alignment systems, and various optical tracking devices.
The Risley prism pair is often the preferred choice in these applications because of the high degree of accuracy and stability it offers. The major limitation of the Risley prism pair is that it can only be used to perform deflections of a few degrees at most. However, this is not an issue in a fiberoptic switch, which does not require a high degree of beam motion.
A Risley prism pair is a pair of wedged circular windows mounted in independent rotation stages (see Fig. 1). Each window acts as a thin prism, and rotation of these wedges causes tilting of the beam due to refraction. Because the wedged prisms can be independently rotated, an input beam can be deflected in both Θx and Θy (for example, anywhere within a cone). Specifically, rotation of the prisms with respect to each other sets the magnitude of deflection (2a), and overall rotation of the pair sets the polar angle (for example, the direction of the deflection).
The deflection angle for each prism is given by:
a = (n - 1) aw
where n is the refractive index and aw is the wedge angle.
The total beam deflection is given by where β is the relative rotation of the two wedges, and ad is the deflection due to each single wedge.
The simplest way to construct a commercially manufacturable switch based on this technology is to use collimated light (see Fig. 2). The input end of the switch consists of multiple individual single-mode fiber connectors. Short fibers from these connectors are formed into a tight bundle. The output of this bundle is collimated by an achromatic collimating lens placed at one focal length from the fiber ends.
This lens is chosen to have a larger numerical aperture than the single-mode fiber so that the Risley prisms are underfilled and there is no possibility of vignetting the beam and thereby increasing insertion loss. After the prisms, a second low-aberration lens is used to focus the deflected light; the focal plane is set to coincide with the face of the output fiber bundle array.
These short fibers are stacked in a close-packed array; each individual fiber then feeds into a single-mode output connector. The focusing lens not only serves to efficiently couple the light into the output fiber, but also converts the angular deflection caused by the prisms into a change in the position of the focused beam. In this way, the prisms can be rotated so that light from a given input fiber can be precisely directed into any chosen output fiber.
In practice the most stable and rugged way of rotating the prisms is to use hollow shaft stepper motors. Each prism is mounted directly onto the motor shaft and the light actually passes through the hollow core of the motor. Direct mounting on the center line of the motor delivers the maximum short term accuracy as well as long-term switch stability.
The light is always efficiently coupled between the designated input and output fibers with a repeatability of ±0.012 dB or better, no matter whether fibers are sampled sequentially or at random. This repeatability is a direct result of the optomechanical reduction that occurs in Risley prisms.
As can be seen from the second equation, a large rotation of the prisms only produces a small beam deflection. In fact, beam motion at the output fiber can be as small as 1/60 of the prism motion, depending on the absolute angle. Thus, small errors in prism rotation have no effect on switch performance. This is in contrast to mirror-based systems, where any error in mirror deflection angle results in twice that error in beam deviation.
As a result, conventional stepper motors deliver more than sufficient resolution and accuracy for this application, eliminating the complexity and expense of precision encoders. Moreover, the optomechanical arrangement is very stable in terms of both vibration and temperature drifts. In fact, the temperature variation in insertion loss is so small that it cannot be measured with conventional test equipment (for example, ±10 mdB over a 6°C temperature range). Other optomechanical switches are typically ±0.3 dB or worse over this same range.
Another advantage is switch flexibility. Switches can be customized in many M x N configurations with widely different values of M and N. For example, a 5 x 50 configuration allows several test setups or instruments to automatically test up to 50 components or subsystems via SMF28 fiber connectors. In addition, these switches are completely bidirectional, further enhancing their utility in systems testing applications.
Because the optics are directly mounted on stepper motors, this type of switch also delivers short switching and settling times (see Fig. 3). Typically the total is 200 ms or less, whereas other commercial switch technologies typically deliver >300 ms of switching time, plus 200 ms of settling time. And in certain switches, this time is additive; for example 16 x 300 ms, plus settling, to go from output channel 1 to channel 16.
This switching technology platform can conceivably be extended as demands change—specifically in the areas of higher power and all-optical network technology. At present, a power rating of up to 300 mW per input channel is available, which meets the needs of current applications. But as channel count and the power of testing sources increases, a higher power rating may be necessary.
It would be straightforward to integrate an optical shutter within the switching assembly, so that a high power input beam is only "on" when pointed directly at the desired output fiber. The result of this integration would be an even higher power rating.
In addition, it might be possible to integrate a diffraction grating into this type of switch. This would mean that the angular deflection and hence output channel number would become wavelength dependent—a major advantage for testing wavelength-selective components and subsystems in an all-optical network.
Bill Holtkamp is director of marketing and Tom Stakelon is director of engineering at Spectra Physics Active Telecom Systems, 1335 Terra Bella Avenue, Mountain View, CA 94043. Bill Holtkamp can be reached at firstname.lastname@example.org.