Automatic PMD compensation uses feed-forward approach


Researchers transmitted 80-Gbit/s optical time-division multiplexed (OTDM) data error-free over 45 km of installed single-mode fiber using a system that allows automatic compensation for polarization-mode dispersion (PMD). Harald Rosenfeldt and others from Technische Universität Hamburg-Harburg (Hamburg, Germany), Heinrich-Hertz-Institüt Berlin GmbH (Germany), and T-Nova GmbH (Berlin, Germany) avoided the need for dithering by adjusting the differential group delay (DGD) of the PMD compensator using a feed-forward controller (see figure).

Polarization-mode dispersion is a major problem for high-bit-rate optical transmission systems. Telecommunication companies avoid the problem for now by using low-PMD fibers and short transmission distances. In OTDM systems with bit rates much higher than 40 Gbit/s, however, PMD reduces signal quality even when using low-PMD fibers.

One approach to PMD compensation uses a closed-loop approach that measures the degree of polarization (DOP) at the fiber output. The DOP increases in the presence of PMD, but DOP alone is not enough information to derive the DGD. The DOP of a received signal depends both on the DGD and on the power-splitting ratio between the two principal states of polarization. Therefore, these systems dither all the parameters of the equalizing unit as well as measuring the DOP to find the best settings for the equalizing unit.

This solution, however, creates a tough timing problem for an adaptive control system: if the parameters of the equalizing unit have to be dithered to find the optimum settings, then the response time of the equalizing elements must be much faster—about an order of magnitude faster, or in the millisecond range.

Rosenfeldt and coworkers eased the timing requirement for the DGD element of the PMD compensator by taking a different approach to measuring the DGD. The researchers used a polarization scrambler at the fiber input (which eliminates dependence on input polarization) and measured both the DOP and the state of polarization (SOP) at the fiber output. Because the scrambler causes the input polarization to cover the whole Poincaré sphere within 18 ms, one can assume that the worst-case DOP occurs during that time interval, and this value can be used to calculate the largest possible DGD for the link. The principal states of polarization can be determined by monitoring the SOP with the largest DOP within that time interval. They used a single polarimeter and a fast MEMS switch to measure the signal at the compensator input and output alternatively.

The group applied feed-forward control in an open loop. The link DGD was continuously measured and fed to a variable DGD element, which set the DGD of the compensator. They demonstrated error-free transmission of an 80-Gbit/s RZ data signal over a PMD-compensated 45-km installed fiber link in Berlin, with an average DGD of 6 ps.

Without PMD compensation, no error-free transmission on this link is possible. With this PMD-compensation scheme, the researchers obtained error-free transmission with a penalty of less than 1 dB. They discovered that the variation of the minimum DOP with DGD was strongly dependent on the pulse shape.

The feed-forward approach for PMD compensation allows the compensator DGD to be adjusted in a single step. Degree-of-polarization evaluation combined with polarization scrambling provides access to link parameters such as DGD and PSP. In future applications, the knowledge about the principal states of polarization also could be exploited to apply feed-forward control for the polarization controller.

For more information contact Harald Rosenfeldt at

Yvonne Carts-Powell


  1. H. Rosenfeldt et al., ECOC 2001, postdeadline paper 28-1.
More in Transmission