Cascaded ROADM Filter Replication and the Effects of Statistical Variations in the Response of Cascaded ROADMs on 100 Gb/s DP-QPSK System Performance
This thesis focuses on two distinct investigations: filter replications and the testing of a real-time system. The filter replication section used a Finisar WaveShaper programmable filter, a Yenista adjustable bandpass filter and a Luna Optical Vector Analyzer to adjust the input to the WaveShaper in a feedback method. There were two different scenarios for the filter replication: bandwidth narrowing/ center frequency shift and passband impairments. Both were meant to model the effect of concatenating filters from a signal being transmitted through cascaded ROADMs. The real-time testing section was based around using a WaveLogic 3 real-time transceiver in QPSK mode. In this setting, the transceiver sends and receives a QPSK signal, performing all of its digital signal processing in real-time. Similar to the filter replication section, the real-time performance of the WL3 transceiver was tested with bandwidth narrowing/center frequency shift as well as passband impairments. For the bandwidth narrowing, only the Yenista bandpass filter was used between the transmitter and receiver to provide the effect of bandwidth narrowing/center frequency shift. For the passband impairments, the WaveShaper and Yenista filter were used together to provide passband impairments with an adjustable bandwidth setting. Unlike the filter replication case, the WaveShaper was used in a feed-forward capacity. For the filter replication, the metric for measuring success was the reduction of the mean squared error compared to the feedforward case. The bandwidth narrowing/center frequency shift case had limited success, with a mean MSE reduction of 16.08% and one scenario of 10 resulting in an increase in the MSE. The passband impairment filter replication was much more effective, resulting in a mean reduction in MSE of 73.53% and an improvement in all 10 scenarios. With real-time testing, the penalty associated with narrow bandwidth filtering / center frequency shift was explored by determining the required OSNR for a BER of 1x10-3 at different bandwidths and amounts of center frequency shift. The overall range of ROSNR was 2.2 dB, after which there was breakdown in the Forward Error Correction coding. It was found that for the passband impairments, passband ripple had a much larger effect on the performance of the system than passband slope, due to the interaction with the Erbium-Doped Fiber Amplifier. In the ranges tested, the slope was not correlated with the mean SNR, whereas the ripple amplitude was negatively correlated with the mean SNR. The ripple period was also found to influence the performance penalty.