Celestino S. Martins

Time-Domain Volterra-Based Digital Backpropagation for Coherent Optical Systems

We propose a novel closed-form time-domain (TD) Volterra series nonlinear equalizer (VSNE) for the mitigation of Kerr-related distortions in polarization-multiplexed (PM) coherent optical transmission systems. The proposed TD-VSNE is obtained from the inverse Fourier analysis of a frequency-domain VSNE based on a frequency-flat approximation. Employing novel TD approximations, we demonstrate the equivalency between the VSNE algorithms formulated in time and frequency domains. In order to enhance the computational efficiency, we insert a power weighting time window in the TD-VSNE, yielding the weighted VSNE (W-VSNE) algorithm. We demonstrate that the convergence of the W-VSNE to its maximum performance is much faster than that of the TD-VSNE, thus requiring fewer parallel filters. Through numerical simulation of a 224-Gb/s PM-16QAM optical channel, we compare the performance/complexity tradeoff of the W-VSNE with the well-known split-step Fourier method (SSFM) and with the computationally optimized weighted SSFM (W-SSFM). Enabled by the use of fewer iterations and only two parallel W-VSNE filters, we demonstrate a reduction of up to ~45% on computational effort and ~70% on latency, in comparison with the W-SSFM.

Nonlinear compensation with DBP aided by a memory polynomial

Using digital backpropagation (DBP) based on the split step Fourier method (SSFM) aided by a memory polynomial (MP) model, we demonstrate an improved DBP approach for fiber nonlinearity compensation. The proposed technique (DBP-SSFM&MP) is numerically validated and its performance and complexity are compared against the benchmark DBP-SSFM, considering a single-channel 336 Gb/s PM-64QAM transmission system. We demonstrate that the proposed technique allows to maintain the performance achieved by DBP-SSFM, while decreasing the required number of iterations, by over 60%. For a transmission length of 1600 km we obtain a complexity reduction gain of 50.7% in terms of real multiplications in comparison with the standard DBP-SSFM.

Distributive FIR-Based Chromatic Dispersion Equalization for Coherent Receivers

We propose a novel and efficient multiplierless finite-impulse response (FIR)-based filter architecture for chromatic dispersion equalization (CDE) in coherent optical communication systems. After quantizing the FIR coefficients, we take advantage of the high multiplicity of their real and imaginary parts, employing the distributive property of multiplication over addition to sharply reduce the number of multiplication operations, obtaining the distributive FIR-CDE (D-FIR-CDE). Furthermore, the implementation of multiplication operations with shifts and additions allows us to obtain a multiplierless D-FIR-CDE (MD-FIR-CDE). The proposed equalizers are experimentally validated in a 100G polarization-multiplexed (PM)-QPSK long-haul optical link and compared against benchmark FIR-CDE and frequency-domain (FD)-CDE implementations. We demonstrate computational resources savings of over 99% in number of multiplication operations and 40% in number of additions, relatively to the FIR-CDE implementation. In addition, the D-FIR-CDE is also shown to compare favorably relatively to the most widely used FD-CDE, achieving significant gains both in terms of required chip area and latency: more than 99% and 30% fewer multipliers and additions, respectively, and a latency reduction of over 90%. We have also experimentally demonstrated that the performance penalty imposed by the coefficient quantization tends to decrease with increasing propagation length, rendering it as an attractive solution for efficient and high-performance chromatic dispersion compensation in long-haul optical fiber links.

Real-time demonstration of low-complexity time-domain chromatic dispersion equalization

We demonstrate real-time CD equalization (CDE) for coherent optical transmission systems using a low complexity time-domain (TD) multiplierless finite-impulse response (FIR)-based equalizer, based on a field-programmable gate array (FPGA) implementation. The real-time operation is performed for a single-channel 2.5 Gb/s QPSK optical signal with a performance penalty of only ∼0.15 dB with respect to the maximum performance. The hardware complexity is also evaluated in terms of occupation in a Virtex-6 FPGA-XC6VLX240T, revealing the high efficiency of the proposed CDE algorithm.

Multi-carrier high-speed optical communication systems supported by digital signal processing

We report our recent advances in the implementation of very high-speed optical channels (targeting 400 Gbps and 1 Tbit) based on multi-carrier super-channels supported by intense digital signal processing. We present results on modulation format agnostic polarization demultiplexing, computationally efficient chromatic dispersion equalization, and digital compensation of nonlinear impairments.

Real-time digital signal processing for coherent optical systems

Digital signal processing (DSP) is nowadays a key enabling technology in coherent optical transmission systems, triggering an increasing research effort on the development and optimization of DSP subsystems. The first DSP development step is often based on the use of simulation tools and offline processing of experimental data. Complementarily, real-time implementation is a critical implementation step to assess the performance and feasibility of advanced DSP subsystems. In this work, we address both the offline and real-time stages of development, supported by an experimental optical testbed. Real-time demonstration is enabled by an FPGA processing platform and 1.25 Gsample/s analog-to-digital conversion.