LTS PMD Experiments

An RF Tone Sensor for Real-Time PMD Monitoring

There have been a number of different sensor technologies demonstrated for PMD monitoring. For this project, we selected a sensor based on the detection of one of the RF tones contained in the optical signal [1]. A photo of the sensor is shown in Figure 1. The sensor sub-system begins with an optical photodiode, followed by an electrical amplifier to boost the signal power. This is then filtered by a bandpass filter, to select the desired RF tone. In this case, we are using 5 GHz, which is half of the 10 GHz clock rate of the data stream. The filtered power in this tone is converted to a DC voltage by a microwave power detector, whose output is digitized by a data acquistion card which is interfaced to one of the network control computers.

The sensor voltage output is proportional to the DGD impairment experienced by the optical signal during transmission, with a lower sensor voltage indicating a more degraded signal. We tested our sub-system in the laboratory using a commercial DGD emulator and a programmable polarization controller. For different DGD settings of the emulator, we varied the state of polarization (SOP) of the light at the emulator input through 100 different states. The results of this are shown in Figure 2 [2]. The individual dots correspond to different input SOPs, with the worst case for each highlighted in blue. At low DGD settings, we observe very little dependence on input SOP, but at higher settings there are large fluctuations. This is because if the signal SOP is aligned along one of the emulator's birefringent axes, then the signal will not suffer impairment even for very high DGD settings, leading to a sensor value which is similar to that of lower emulator settings. On the other hand, a signal which contains equal components along each axis will suffer a larger penalty, resulting in a low sensor voltage. Thus for high DGD levels in a transmission system, there will the potential for a great deal of variability in system performance, but we see that our sensor is able to accurately track these fluctuations, which will be important when it is installed in the transmission link.

Installed Fiber Tests of the Demonstration System

For the field trials, we are utilizing an installed 52 km fiber link which connects our laboratories in Baltimore, MD (UMBC) and College Park, MD (Mid-Atlantic Crossroads). Each end of the link terminates at a virtual link state router (VLSR) composed of a computer, an ethernet switch (Raptor ER-1010), and an optical add-drop multiplexer (Movaz Ray-Express). We transmit two wavelength channels over the link, one at 10 Gb/s and one at 1 Gb/s. In order to have a more controlled DGD impairment level in our experiments, we place the DGD emulator at the Baltimore node, just after the OADM output. The sensor is located at the College Park VLSR, where a small part of the received 10 Gb/s is tapped out for the sensor, just before detection at the transponder. The sensor is polled every 2 seconds by the VLSR computer, which compares the current value to a pair of pre-programmed threshold value. When the sensor crosses the lower threshold value, the control plane will take the 10 Gb/s channel out of service, and begin routing traffic over the 1 Gb/s channel. The VLSR continues monitoring the sensor output, so that if the sensor crosses the upper threshold value, the 10 Gb/s channel will be returned to service.

The testing consists of first setting a DGD emulator level, and setting the input signal polarization state to the emulator. We then create an end to end circuit (a label switched path or LSP) using the DRAGON user interface. The packet loss over this path is then measured, along with a number of diagnostic values such as the sensor output, the available bandwidth for the 1 and 10 Gb/s channels, the explicit route object (ERO) associated with the LSP, and various power levels. With this, we are able to quantify the system performance, and identify which channel (1 or 10 Gb/s) was used. The LSP is then deleted, and the process repeats for a new DGD impairment setting. The goal is to show that without monitoring and switching, as the DGD level increases the 10 Gb/s channel will become severely impaired, resulting in end-to-end packet loss. However, when the sensor is monitored by the VLSR, and this information is used to select which channel the signal should be routed over, the end-to-end packet loss is negligible, even for DGD values approaching 90 ps for our system [2]. While this simple test uses only a single link, one can imagine placing these sensors at many different nodes within the network, and routing decisions could made with consideration for the current link health and the desired data rate.

References

[1] G. Ishikawa and H. Ooi, "Polarization-mode dispersion sensitivity and monitoring in 40-Gbit/s OTDM and l0- Gbit/s NRZ transmission experiments", In Proc. Opt. Fiber Commun. Conf. (OFC1998), San Jose, CA, Paper WC5.

[2] A. S. Lenihan, W. A. Babson, H. Jiao, J. Sobieski, and G. M. Carter, "An experimental demonstration of a soft-failure approach to PMD mitigation in an installed optical link," Optics Express 15 (1), pp. 24-32 (2007).