High-speed optical transmission system using stokes vector receiver

Abstract

Due to data traffic explosion in datacenter applications, there has recently been a great deal of interest in high-capacity short-reach (e.g., <2 km) optical transmission systems. These systems typically utilize the direct-detection (DD) scheme since it can be implemented cost-effectively even though the DD systems are inferior to the coherent systems in terms of capacity and receiver sensitivity. The transmission capacity of DD systems can be doubled by using the polarization-division-multiplexing (PDM) technique, which exploits two orthogonally polarized light channels. Due to the random polarization rotation during transmission over fiber, however, it is necessary to undo this polarization rotation at the receiver side to demultiplex the signals. Recently, the Stokes vector (SV) DD receiver has drawn much attention since it is capable of demultiplexing the PDM signals with the aid of digital signal processing (DSP) without cumbersome dynamic polarization control in DD systems. The SV-DD receiver can be implemented by using a $1\times4$ optical coupler, three polarizers, a quarter-wave plate and 4 photo-detectors. This 4-branch SV-DD receiver detects the Stokes vector component in a direct-detection fashion at each branch. It has a lower insertion loss than the SV-DD receiver implemented by using a $90^\circ$ optical hybrid. One practical issue related to the implementation of the 4-branch SV-DD receiver is the calibration of polarizers located at the branches. In this thesis, I present the calibration process of adjusting the polarization angles of polarizers. The calibration process of measuring the timing difference between the branches is also presented. However, real implementation of 4-branch SV-DD receiver always accompanies some misalignment of polarization angles. Thus, I analyze the impact of polarization misalignment on the performance of SV-DD receiver. I evaluate the signal-to-noise ratio (SNR) penalty induced by the polarization misalignment of polarizers through numerical simulation. The results show that the 4-branch SV-DD receiver is very robust against the polarization misalignment. For example, an SNR penalty of 0.6 dB is estimated when the quarter-wave plate has a retardance as large as $\pm0.3$ radian. Next, I propose a new 16-ary modulation format for SV-DD receiver. Due to the insertion loss of optical front-end in SV-DD receiver, it requires a relatively high optical power into the receiver. To improve the receiver sensitivity of SV-DD receiver, it is necessary to employ the modulation format tolerant to the noise of the receiver. For this purpose, the 16-ary modulation format which provides us with the largest Euclidean distance for a given average signal power was introduced. This optimum format was found by solving the sphere packing problem in the Stokes space. However, this 16-ary optimum constellation requires numerous number of levels to drive a dual-polarization IQ modulator (IQM). This implies that high-resolution digital-to-analog converters (DACs) are required for the generation of the signal. To avoid this problem, a suboptimum constellation has been proposed very recently. Nevertheless, this 16-ary suboptimum constellation still needs unequally-spaced 9-level electrical signals to drive IQMs, yet requiring high-resolution DACs. Moreover, since the reflected binary Gray coding is not applicable to this format, it suffers from multiple errors when a symbol error occurs. In this thesis, I propose a 16-ary coaxial dual cubic lattice (CDCL) constellation. I compare the transmitter complexity and the receiver sensitivity among five 16-ary modulation formats: dual-polarization 4-ary pulse amplitude modulation, 16-quadrature amplitude modulation polarization-multiplexed with continuous wave, 16-ary suboptimum constellation, 16-ary optimum constellation, and the proposed CDCL. I show that the proposed CDCL format exhibits a better receiver sensitivity than the other modulation formats (except for the 16-ary optimum constellation), but has considerably reduced implementation complexity. For example, the CDCL format incurs only <0.7 dB sensitivity penalty with respect to the 16-ary optimum constellation, but the transmitter can be implemented by using a fewer number of lower-resolution DACs. We believe that the proposed CDCL format can be used to lower the cost and complexity of SV-DD systems.