Researches On Optical Fiber Communication System With High-speed And High Spectrum Efficiency

As the single channel capacity of optical fiber communication system evolves from 100-Gb/s to terabits, the bottleneck of electrical components has become an obstacle. In order to satisfy the increasing demand on single channel capacity, an optical parallelization, called superchannel, has been proposed as a promising solution. In the superchannel architecture, the multi-sub-channel are tightly packaged in frequency domain, generating a"seamless" spectrum, and propagate as a whole channel along the fiber,providing a much higher spectrum efficiency and data rate compared to traditional wavelength division multiplexing (WDM) systems. One of the supechannel implementations has manifested itself to be a good candidate which consists of a frequency-locked multi-carrier source with high OSNR,pulse shaped high order QAMs and coherent detection based powerful DSP equalization. On the other hand, the future optical network will have another feature of elastic and flexible, and can be well supported by Nyquist-WDM and all optical wavelength conversion. The Nyquist pulse shaped channels with different bandwidth can be easily packaged seamlessly due to their square spectrum shape. The wavelength converter can provide great flexibility for optical routine and wavelength allocation,which can maximize the ultilization of the spectrum resource. The research work of this article focuses on several key issues of such a superchannel system, including 1) low noise multi-carrier generation based on frequency shifting loop; 2) Digital-to-analog converter (DAC) based Nyquist-16-QAM generation and optimization; 3) Nyquist-WDM based superchannel long haul transmission, and 4) wavelength conversion of Nyquist-WDM signal in a flex-grid network. Details of each part are detailed in the following.We theoretically study the noise accumulation process of Single-side-band modulation based frequency shifter (SSB-RFS), and propose three improved SSB-RFS schemes: 1) linear IQ modulator (LIQM) based SSB-RFS, and with which the numerical simulation study shows that about 4 dB power increment of shifted signal could be achieved while maintaining 35 dB sidelobe suppression ratio (SSR). This value can be equally transmitted into OSNR improvement of SSB-RFS. 2) The optical finite impulse response (FIR) filter based in-loop noise suppression scheme is studied and shows a significant OSNR improvement. The simulation results show that with 8-tap filter deployed in the loop, all the 100 tones can provide adequate OSNR to carry 16QAM. With a 2-tap optical FIR filter, we experimentally generated 69 carrier tones with about 8.7dB OSNR improvement compared to an ordinary source. 3) The stimulated Brillouin amplification based SSB-RFS is studied experimentally. By replacing the EDFA with Brillouin gain media, the induced ASE noise can be significantly reduced. We experimentally generate 8 sub-carriers with ultra low noise using Brillouin gain based method.For the Nyquist-16QAM generation and optimization, we experimentally study the trade-offs for the DAC based pulse shaping considering hardware restrictions, such as clipping ratio and modulation index. We also study the impairments caused by the transmitter (Tx) IQ skew and receiver-end channel skew. Two digital blind post-equalization schemes are proposed to overcome the Tx IQ skew, one is complex conjugated signal based adaptive multi-input and multi-outout (MIMO)algorithm, and the other is real valued one. The two post-EQ schemes are compared experimentally in a 25Gbaud Nyquist-16QAM back-to-back(b2b) test, and the results show that the real-valued MIMO outperforms,providing ?1.5 dB required OSNR improvement at BER threshold of 7%FEC overhead. The optimized Nyquist-16QAM signal is launched into fiber loop for distance test, and with the optimum launch power, the signal spans ?800km with BER below the 7% FEC threshold.The superchannl transmission is studied by generating 26.43GHz spaced 20-carriers with noise suppressed SSB-RFS to deliver 20×25Gbaud dual-polarization (DP) Nyquist-16QAM. First a back-to-back (B2B) test is carried out to verify the effectiveness by deploying noise suppression scheme in the SSB-RFS source. The B2B BER results of the 4-Tb/s superchannel show that, compared to an ordinary SSB-RFS based superchannel, the noise suppressed 20 carriers contributes significant BER improvement. With noise suppressed SSB-RFS source, all the 20 sub-channels have a similar BER around le-3 while only first several channels can reach that value for an ordinary SSB-RFS based system. Then the low noise SSB-RFS based 4 Tb/s superchannel signal is launched into SSMF fiber circulating loop. Assume 20% FEC overhead, the 4 Tb/s superchannel signal can span 1120 km with a net capacity of 3.2 Tb/s and a net SE of 6.03 bit/s/Hz.With 3x25Gbaud Nyquist-16QAM signal, we experimentally study the semiconductor optical amplifier (SOA) based wavelength conversion(W/C) of 3×25Gbaud Nyquist-16QAM signal, measuring the performance in terms of BER and conversion efficiency for different channel spacing.The results show that all the three converted channels with 27GHz,37.5GHz and 50GHz grids can reach the BER below 7% FEC threshold.The W/C induced OSNR penalty indicates that the closer spaced grid strategy for the three channels gives a more consistent BER performance after W/C. The Nyquist pulse shaping is usually used to balance the intra-channel crosstalk (ICI) and SE, while a higher SE could be expected in the presence of SOA based W/C in the link, as a more tightly spaced stratedgy would be considerable to balance the ICI and W/C penalty. These results can be used to optimize routine, grid and spectrum allocation strategy when SOA based W/Cs are present in the optical link.