We investigate the computational complexity of adaptive equalization in coherent receiver digital signa l processing (DSP) for data center interconnect (DCI) systems. We propose modified DSP procedures with a first-stage static filter and a second-stage short-length adaptive equalizer. Typically, coherent DSP requires 11 and 17 adaptive equalizer taps respectively for 60 Gbaud and 80 Gbaud signals in the 100 km fiber link. The modified DSP procedures combine matched filter, chromatic dispersion compensation (CDC), and non-ideal channel effects and polarization effect s in to a static filter. Since polarization demultiplexing is partially realized by the static filter, the size of the following adaptive equalizer is substantially reduced. Results showed 35% - 87% overall complexity reduction in adaptive equalization a s compared with the reference DSP, depending on link length and symbol rate.
We introduce a new technique for obtaining the frequency offset introduced at the receiver due to the heterodyning of the transmitter laser and the local oscillator. This technique is needed for multi-subcarrier systems, as the offset must be removed without knowledge of the modulation format, making well-known algorithms that correct using it impractical. By detecting the spectral edges for the transmitted and received signals, it is possible to get a coarse estimate of the intermediate frequency by finding the difference between these two values. We then use this approximation as a starting point to fine search for the true value, by taking a window of possible values around the approximation. By taking the cross-correlation between the received data and a subset of the transmitted data for the subcarriers, we can verify whether the correct value is chosen, and if not, move on to the next estimate.
Proc. SPIE. 11309, Next-Generation Optical Communication: Components, Sub-Systems, and Systems IX
KEYWORDS: Signal to noise ratio, Transceivers, Digital signal processing, Quadrature amplitude modulation, Signal attenuation, Digital filtering, Wavelength division multiplexing, Single mode fibers, Monte Carlo methods, Data centers
Exponential growth of Internet traffic demands data center interconnect (DCI) systems to provide 400 Gb/s and higher per wavelength capacity under tight power consumption limitations for optical transceivers. We investigate the potential advantages of applying ultra-low loss and low dispersion fibers in DCI systems. Link optical signal to noise ratio (OSNR) and capacity analysis shows that ultra-low loss fiber (0.16 dB/km) provides significantly higher data capacity as compared with regular single-mode fiber (0.2 dB/km) for 80 km long DCI links. Also, the lower fiber attenuation reduces the required transceiver output power by 10 dB to achieve the same data capacity for 100 km DCI links. This implies substantial simplification in optical transceiver design. Digital chromatic dispersion compensation (CDC) is one of the major power consumers in optical transceivers. Our analysis shows that low dispersion fiber (4 ps/(nm·km)) reduces CDC computational complexity by 20% to 71% for different DCI link lengths versus regular single-mode fiber, indicating significant reduction in power consumption. Moreover, employing the CDC capability of the built-in adaptive filter in coherent receiver digital signal processing (DSP), the digital CDC unit could be completely removed using low dispersion fibers in DCI systems. Finally, we performed Monte-Carlo simulations of DCI links with different fiber types and confirmed the benefits of ultra-low loss and low dispersion fibers.
Nonlinear Schrödinger equation (NLSE) is solved using different split-step Fourier methods (SSFM). The uniform step size distribution is traditionally utilized and the efficiency of the scheme can be improved by using the adaptive step sizes. One scheme of using the adaptive step size is the local error method, which updates the step size based on the local error of the current step. In this paper, a novel scheme which combines the minimum area mismatch (MAM) and the local error method is proposed. The MAM method can be used to find the optimal step size distribution based on minimizing the area mismatch between the ideal effective nonlinear coefficient curve and its stepwise approximation. The local error is a criterion to choose the total number of steps per span. The simulation results show that the proposed scheme outperforms the schemes using uniform step size and adaptive step size.
We present a novel LMS equalizer (FDE) for compensating polarization-mode dispersion (PMD) in polarization-multiplexed coherent fiber-optic systems based on the least mean squares (LMS) method with unstructured channel estimation (USE). It is a low-complexity algorithm compared to the traditional time-domain decision-directed LMS (DD-LMS) algorithm. Also, it converges faster and shows better bit error rate (BER) performance as compared to the frequency-domain block training-directed LMS (BTD-LMS) method.
An analytical cross-phase modulation (XPM) model is developed for coherent fiber-optic transmission systems. The XPM fields are calculated using a first-order perturbation theory. Statistical analysis is then applied to calculate the XPM variance. The analytical XPM model is applicable to dispersion-managed fiber-optic systems with arbitrary pulse shapes. For non-Gaussian pulses, a summation of time-shifted Gaussian pulses is used to fit the target pulse shape, which makes it possible to analytically derive the XPM variance. The analytical XPM model is validated by extensive numerical simulations.
We demonstrate a 160 Gb/s orthogonal frequency division multiplexing (OFDM) system using an all-optical symbol
generator based on planar light circuit (PLC) technology. Excellent bit error rate (BER) is observed after long-distance
transmission. The proposed symbol generator fundamentally eliminates the processing speed limits introduced by electronics
and is suitable for high integration, making it physically realizable to build high-speed all-optical OFDM systems with a
large number of subcarriers.
A novel TE mode transmission waveguide polarizer has been designed based on SiO2/Si waveguide birefringence effect
and based on the coupling mode theory. After numerical simulation by 3D FD-BPM method combined with transparent
boundary condition, the typical polarizer can achieve a high extinction ratio over 50dB with the device length of 8mm at
the wavelength of 1.55 μm. Without increasing the complexity of waveguide manufacturing, this structure can be
directly used as a polarizer, and also can be integrated easily with other waveguide devices.
A novel high speed Optical Frequency-domain Transmission method and System (OFTS) based on two time
lenses is proposed and analyzed in this paper. By using the unchanged spectral profile through light
propagations in fiber, the long-haul optical transmission over 100Gbps per channel can be achieved with
highly tolerance of linear distortion such as PMD, chromatic dispersion, time jitter etc. In the novel method
and system, a time lens is used at the transmitter to convert the initial time domain signal into frequency
domain signal. And at the receiver, another time lens is used to transform the received frequency domain
signal back to time domain. To prove the advantages of OFTS, a specifically designed 10×100Gbps OFTS is
numerically simulated over 1880km transmission without PMD compensation.