Fiber-optic communication permits digital data transmission over longer distances and at higher data rates than conventional electronic communication. The optical fiber consists of a core surrounded by a cladding layer with a refractive index smaller than that of the core. Ideally, the core is circularly symmetric, and the two orthogonally polarized modes of a modulated signal will travel at the same speed. The cross-section of a real fiber core, however, is usually elliptical due to manufacturing defects or external stresses (see Figure 1
). Consequently, the light polarized along one axis will travel faster than that along the orthogonal axis. After propagation through a certain length of fiber, a delay between the two polarization modes will occur (see Figure 1
). This is called a differential group delay (DGD), i.e., a first-order polarization mode dispersion (PMD).
Figure 1. The polarization mode dispersion (PMD) effect introduced by the asymmetric profile of the optical fiber core. PSP: Principal state of polarization. DGD: Differential group delay.
Such an effect can broaden the transmitted signal and introduce errors or outages in optical networks. Moreover, as the PMD effect along the fiber link is totally random, with a statistical distribution, combating the PMD effect is extremely difficult, especially for channel data rates higher than 10Gb/s. Research efforts have focused on emulation and mitigation approaches,1
including tuning DGD values.
Solutions to PMD must include fast response time (typically millisecond scale), superb optical characteristics such as low loss and polarization-dependent loss (PDL), optimized dynamic behavior, and so on. For PMD or DGD emulation, accurate statistical generation is desirable, especially since the DGD of real fiber links follows the Maxwellian distribution (i.e., the tail extends to infinity). Determining effective control algorithms for PMD compensation also poses challenges.
A well-known method of constructing a tunable DGD element for PMD-related applications varies the delay of one polarization arm following a polarization beam splitter and then recombines the two states using a second device. Such a scheme can replicate the DGD effect, but the setup is generally bulky, unstable, and low speed.
We have developed a novel approach for programmable DGD generation using the configuration shown in Figure 2
The DGD element consists of multiple birefringent crystal sections (delay sections) separated by magneto-optic (MO) polarization switches. The arrangement of the crystal lengths follows a binary power series with a factor of 2 (either increasing or decreasing). The MO switches rotate the polarization state between crystals by 0° or 90° with the application of one of two different saturation currents. Thus, at the input to each section, the state can be switched to align with the slow or fast axis of the subsequent birefringent crystals, thereby adding to or subtracting from the total DGD of the device without adding any higher-order PMD. The demonstrated module (commercially available) has several advantages: fast tuning speed (typically <1ms), negligible second-order PMD (comparable to polarization-maintaining fiber) as the generated DGD increases to 45ps, low loss (<1.5dB), and low PDL (<0.2dB).
Figure 2. Programmable DGD element using binary polarization switches and birefringent crystals.
Such a programmable DGD element is useful for more than optimized PMD compensation3
and first-order PMD emulation. It is also key to constructing all-order PMD emulators, which can be built from several tunable DGD elements separated by dynamic polarization controllers: see Figure 3
A big advantage of this kind of configuration is that the final generated statistics (both first-order and higher-order) depend on the individual applied (programmed) statistics of each DGD element. Compared with other all-order approaches, this scheme is much more cost-effective, as it can provide tunable statistical distributions without any hardware modifications: see Figure 3
(b). Moreover, thus far only this approach has demonstrated the capability for importance sampling.4
With just 1000 statistical samples, the emulator can investigate extremely low probability events (down to 10-24
), condensing years of evaluation time to minutes.
Figure 3. An all-order PMD emulator generates tunable statistical distributions using programmable DGD elements. (a) Schematic and prototype. (b) Three typical generated distributions with Maxwellian curve fittings (real fiber case). P-DGD: Programmable DGD. PC: Polarization controller.
As previously noted, DGD programmability is essential for various PMD solutions as well as for optimized PMD compensators. Tunability of statistics significantly benefits network designers in terms of economy and efficiency. In addition, programmable DGD elements may also find application in microwave photonic networks for tuning the relative time delay.
General Photonics Corp.
Lianshan Yan is a chief scientist and manager of engineering at General Photonics Corp. He has about 110 journal and conference publications and holds four US patents. Yan is a senior member of the Institute of Electrical and Electronics Engineers (IEEE) and was a recipient of the IEEE Lasers and Electro-Optics Society graduate fellowship. He serves as a member of the editorial board and is a frequent referee for multiple journals.