Commercial applications of this tri-waveplate technique do exist and have indeed produced respectable results. Unfortunately there are many drawbacks with this technique, most noticeably its high cost, high insertion loss, and high wavelength sensitivity. First, the collimation, alignment, and refocusing process is time consuming, resulting in high labor cost. Second, the wave plates are made of expensive material, and together with the micro lenses, must be anti-reflection coated or angle polished to prevent back reflection, resulting in extra manufacturing cost. Third, because the optical beam has to be coupled out of a fiber and then refocused into another fiber, the insertion loss is high. Finally, the wave plates are inherently wavelength sensitive, making the device sensitive to light's wavelength variations.

In 1980, H. C. Lefevre devised an all fiber polarization controller based on the same tri-waveplate concept (see Figure 1). In this device, the three wave plates were replaced with optic fiber coils. Coiling the fiber induces stress on the fiber, which in turn produces birefringence via the photoelastic effect. The amount of birefringence is inversely proportional to the diameter squared of the coil. By adjusting the diameter of the coil and number of turns in the coil, any desired fiber wave plate can be created. Because the wave plates are made of a fiber, it is not necessary to bring the light out of the fiber, thus eliminating the time consuming process of collimation, alignment, and refocusing. In addition, material cost is greatly reduced because these fiber wave plates are much less expensive than the bulk wave plates. Although numerous names, such as the Micky Mouse Ears, the Dog Ears, the Butterfly, the Three Panels, and the Fiber Coils, have been given to this clever idea, the Lefevre technique represents a major step forward from the tri-waveplate controller because it utilizes the intrinsic properties of an optical fiber.

However, this all-fiber polarization controller also comes with many problems. First, it is bulky. The current commercial version stands up to 10 cm high and 30 cm long. As one can imagine, it is awkward to insert these devices in places where size is of importance, such as in commercial optical receivers and transmitters. Second, because the diameter of the fiber core and its material properties vary from vendor to vendor, the birefringence of the coils may also vary as a result and therefore have to be adjusted accordingly. But a precise adjustment is difficult to obtain in such a device since the diameter of the fiber coils cannot change once created. Furthermore, coiling a fiber may cause micro-bending and result in high insertion loss at long wavelengths (such as in the 1550 nm window). Finally and most disappointingly, just like its free space cousin, the Lefevre controller is also wavelength sensitive due to the inherent wavelength sensitivity of its fiber wave plates. When the optical wavelength changes, the diameter of the fiber coil has to change accordingly. In practice, making such a change is extremely difficult.

Wavelength insensitivity is important in polarization control. For instance, in today's wavelength division multiplexing (WDM) systems, many wavelengths (ranging from 1300 nm to 1600 nm) are simultaneously present in a single optical fiber to carry multiple channels of information. The complexity and cost of these systems can be greatly reduced if a single polarization controller is used to manipulate the multiple optical signals of different wavelengths. The Babinet-Soleil compensator could be a good solution to these multi-wavelength systems; its key component is a wave plate of variable retardation free to rotate about the optical beam. However, this device is very expensive to make and only available for free space applications.

In 1995, X. S. Yao took an entirely new approach to the polarization control problem, by exploiting the intrinsic properties of an optical fiber. His all fiber polarization controller possesses the Babinet-Soleil compensator�s wavelength insensitivity but not the many problems inherent in its free space technique. As illustrated in Figure 2,  the Yao controller consists of a strand of single mode fiber, a center squeezer which can be rotated about the fiber, a left and a right fiber holder. The center portion of the fiber strand is clamped inside the fiber squeezer (made of two special pressure plates). Tightening the knob on the squeezer will apply a pressure to the fiber center portion and produce a linear birefringence in this portion of fiber with its slow axis aligned in the direction of applied pressure. By changing the pressure, the retardation between the slow and fast axes can vary between 0 and 2pi . The induced birefringent axes can vary from 0 to more than pi /2 with the squeezer rotating about the fiber, thus in essence creating an all fiber Babinet-Soleil compensator. The Yao controller is currently licensed and manufactured by the General Photonics Corporation under the trade name of PolaRITE^TM; its derived products are shown in Figure 3.

Compared to the tri-waveplate controller made of bulk phase retarders, the Yao controller has no intrinsic loss, no intrinsic back reflection, simple construction, and low cost due to its simple construction. Compared to the Lefevre polarization controller, it is compact and easy for in-line insertion, insensitive to wavelength variations, insensitive to fiber variations, and has no bending losses. More noticeably, the Yao controller is easy to use; it has only two adjustment procedures - applying the pressure and rotating the knob, to convert an input light of any arbitrary polarization to an output light of any desired polarization.

The above discussion presents two pairs of polarization control techniques. There are also many other types of fiber optic polarization controllers, such as the liquid crystal controller, the fiber crank controller, and the tri-fiber-squeezer controller. However, these controllers have had limited commercial success due to either high cost or complicated control schemes.

The Applications:

Polarization controllers have many uses. They are essential for device testing, such as polarization dependent loss and polarization mode dispersion. They are also critical in interferometeric systems, such as fiber sensors, fiber lasers, and coherent communication systems. Keeping the optical path balance in these systems is extremely important. The Yao controller, commercially known as PolaRITE^TM, is especially fit for these applications. Because its in-line version can easily be inserted into these systems without having to disconnect any fibers, as shown in Figure 4, PolaRITE^TM can preserve the perfect optical path balance.

Polarization controllers are most often placed in front of polarization sensitive components, such as electro-optic modulators. Because it is compact and wavelength insensitive, PolaRITE^TM is particularly suited for insertion in WDM transmitters where the polarization states of a large number of signal channels with various wavelengths need to be precisely aligned, as illustrated in Figure 5.

With an addition of a polarizer, a PolaRITE^TM can also be used to reduce the amplified spontaneous emission (ASE) noise and improve the noise figure of an optical amplifier, as shown Figure 6. This is based on the simple fact that the ASE noise is unpolarized while a signal is generally elliptically polarized. The input signal, once converted into a linear polarization state by a PolaRITE^TM, can pass the polarizer with minimum loss while the unpolarized ASE noise is cut by half. Incidentally, the same combination of a PolaRITE^TM and a polarizer can also function a polarization maintaining variable attenuator since the Yao controller allows the polarization state of the signal entering the polarizer to vary continuously, thus causing the signal level to change consequently. On the other hand, the output polarization state, determined by the polarizer, does not change when the signal level is varied.

 Another important application unique to PolaRITE^TM is that it can be used to join two polarization maintaining (PM) fibers without the expensive PM fiber fusion splicers. One may simply splice a short (~ 8 cm) regular single mode fiber between two PM fibers using regular fusion splicers and then insert the regular fiber segment into an in-line PolaRITE^TM. Adjusting the PolaRITE^TM can couple the optical signal from one PM fiber to the other with a correct polarization state, as illustrated in Figure 7. Because the regular fiber is embedded inside the controller, it is isolated from external disturbances. Similarly, a fully connecterized PolaRITE^TM can be used as an adapter to connect two PM fiber jumpers of regular connectors, eliminating the need for expensive PM connectors.