Spectral Pre-emphasis Research Articles

A rigorous analysis of digital pre-emphasis and DAC resolution for interleaved DAC Nyquist-WDM signal generation in high-speed coherent optical transmission systems

The Nyquist spectral shaping techniques facilitate a promising solution to enhance spectral efficiency (SE) and further reduce the cost-per-bit in high-speed wavelength-division multiplexing (WDM) transmission systems. Hypothetically, any Nyquist WDM signals with arbitrary shapes can be generated by the use of the digital signal processing (DSP) based electrical filters (E-filter). Nonetheless, in actual 100G/ 200G coherent systems, the performance as well as DSP complexity are increasingly restricted by cost and power consumption. Henceforward it is indispensable to optimize DSP to accomplish the preferred performance at the least complexity. In this paper, we systematically investigated the minimum requirements and challenges of Nyquist WDM signal generation, particularly for higher-order modulation formats, including 16 quadrature amplitude modulation (QAM) or 64QAM. A variety of interrelated parameters, such as channel spacing and roll-off factor, have been evaluated to optimize the requirements of the digital-to-analog converter (DAC) resolution and transmitter E-filter bandwidth. The impact of spectral pre-emphasis has been predominantly enhanced via the proposed interleaved DAC architecture by at least 4%, and hence reducing the required optical signal to noise ratio (OSNR) at a bit error rate (BER) of 10−3 by over 0.45 dB at a channel spacing of 1.05 symbol rate and an optimized roll-off factor of 0.1. Furthermore, the requirements of sampling rate for different types of super-Gaussian E-filters are discussed for 64QAM Nyquist WDM transmission systems. Finally, the impact of the non-50% duty cycle error between sub-DACs upon the quality of the generated signals for the interleaved DAC structure has been analyzed.

Digital signal processing techniques in Nyquist-WDM transmission systems

In single-carrier wavelength-division-multiplexing (WDM) systems, the spectral efficiency can be increased by reducing the channel spacing through digital signal processing (DSP). Two major issues of using tight filtering are crosstalk between channels and inter-symbol interference (ISI) within a channel. By fulfilling the Nyquist criterion, Nyquist spectral shaped WDM systems can achieve a narrow channel spacing close to the symbol rate (Rs) with negligible crosstalk and ISI. In principle, DSP can generate any signals with arbitrary waveforms and spectrum shapes. However, the complexity of DSP is limited by its cost and power consumption. It is necessary to optimize the DSP to achieve the required performance at a minimum complexity. In this paper, we investigated the use of digital finite impulse response (FIR) filters to generate Nyquist-WDM 16-ary quadrature amplitude modulation (16QAM) signals with the raised-cosine (RC) and root-raised-cosine (RRC) shape spectrum. The system performance of both the RC and RRC spectra are also examined. Moreover, we explored the various methods to reduce digital-to-analog-converter (DAC) sampling speed, such as using super Gaussian electrical filters (E-filter), and spectral pre-emphasis.

Generate Nyquist-WDM Signal Using a DAC With Zero-Order Holding at the Symbol Rate

We demonstrate the Nyquist spectral shaping using a two-channel interleaved DAC with zero-order holding at the speed of the symbol rate, and investigate the required spectral preemphasis, bandwidth, and resolution. Compared to a conventional DAC at a speed of two times the symbol rate, we find that an extra spectral preemphasis is required for compensating the extra spectral roll-off in the two-channel interleaved DAC. The spectral preemphasis can facilitate Nyquist spectrally shaped signals generation with negligible OSNR penalty (<;0.1 dB). With proper spectral preemphasis, both the required electrical filter bandwidth and DAC resolution are similar to those with a conventional DAC used. To obtain <;0.1 dB OSNR penalty for root-raised-cosine Nyquist spectrally shaped WDM systems with a roll-off factor of 0.1 and a spectral shaping FIR filter tap number of 37, the two-channel interleaved DAC needs a Gaussian electrical filter with the bandwidth between 0.3 and 0.8 times of the symbol rate for PDM-QPSK signal, and between 0.35 and 0.65 times of the symbol rate for PDM-16QAM signal. Meanwhile, the required DAC resolutions are 6 bits and 7 bits for PDM-QPSK and PMD-16QAM, respectively. We also demonstrate how a polyphase FIR filter structure is applied in the transmitter to reduce the calculation by half and to increase the processing parallelism.

Experimental Demonstration of a 10 Gb/s Subcarrier Multiplexed WDM PON

We experimentally demonstrate a 10 Gb/s bidirectional subcarrier multiplexed (SCM) wavelength-division multiplexed (WDM) passive optical network (PON). We use digital signal processing and square-root raised cosine pulse shaping techniques to generate M-ary quadrature amplitude modulation electrical signals for the downlink and uplink transmissions using simple intensity modulation and direct detection optoelectronics, and 10 GHz transceivers. The design includes spectral preemphasis and offset optical filtering to increase the effective bandwidth of a commercial 2.2 GHz reflective semiconductor optical amplifier based optical network unit. We realize symmetric 10 Gb/s transmission over a 20 km single feeder PON with optical line terminal launch powers from 1 to 9 dBm. We also demonstrate the ability of these SCM WDM PONs to withstand the effects of upstream impairments.

Optimization of DSP to Generate Spectrally Efficient 16QAM Nyquist-WDM Signals

We investigate transmitter-side digital signal processing to generate 16-ary quadrature-amplitude-modulation (16QAM) Nyquist wavelength-division-multiplexing (WDM) signals by simulations. We first study required digital finite-impulse response (FIR) filter tap numbers and optimum roll-off factors to generate Nyquist 16QAM WDM signals. We show that a root-raised cosine spectrum with a roll-off factor of 0.1 is optimal for Nyquist 16QAM signals considering the tradeoff between the system performance and complexities, and a digital FIR filter with 37 taps is sufficient to generate such signals. Then we look into the effects of digital-to-analog converter (DAC) sampling speeds on the performance of Nyquist-WDM systems. We show that the electrical filter shape and spectral pre-emphasis have a big impact on the system performance. With a fourth-order super-Gaussian electrical filter and spectral pre-emphasis, DACs with 1.2 samples/symbol sampling speed can generate 16QAM Nyquist-WDM signals at 1.05-symbol-rate channel spacing, with an OSNR penalty below 0.2 dB compared with 2 samples/symbol DACs.

Combining Auditory Preprocessing and Bayesian Estimation for Robust Formant Tracking

We present a framework for estimating formant trajectories. Its focus is to achieve high robustness in noisy environments. Our approach combines a preprocessing based on functional principles of the human auditory system and a probabilistic tracking scheme. For enhancing the formant structure in spectrograms we use a Gammatone filterbank, a spectral preemphasis, as well as a spectral filtering using difference-of-Gaussians (DoG) operators. Finally, a contrast enhancement mimicking a competition between filter responses is applied. The probabilistic tracking scheme adopts the mixture modeling technique for estimating the joint distribution of formants. In conjunction with an algorithm for adaptive frequency range segmentation as well as Bayesian smoothing an efficient framework for estimating formant trajectories is derived. Comprehensive evaluations of our method on the VTR-formant database emphasize its high precision and robustness. We obtained superior performance compared to existing approaches for clean as well as echoic noisy speech. Finally, an implementation of the framework within the scope of an online system using instantaneous feature-based resynthesis demonstrates its applicability to real-world scenarios.

Optimum pre-emphasis in ultra-long-haul networks

We investigate experimentally and theoretically the shape of the optimum spectral pre-emphasis for ultra-long-haul networks and systems in the presence of optical nonlinearities. In a 40×10 Gbits/s×2240 km system, we experimentally show the advantages of a pre-emphasis scheme that ensures minimum variations of path-averaged power among the wavelength-division-multiplexing channels between the cleanups performed by dynamic gain-flattening filters. This is achieved by enforcing a straight-line spectrum near the middle of the link between two consecutive gain-flattening nodes. We explain these results by a simple model, in which we derive the conditions for optimum pre-emphasis in the presence of either distortionlike (e.g., self-phase modulation) or noiselike (e.g., four-wave mixing) nonlinearities. We show that, for relatively small gain ripple, both of these optimum pre-emphasis techniques are equivalent to equalizing the optical signal-to-noise ratio for all channels and producing a flat power spectrum near the middle of the link. Hence, enforcing the straight-line spectrum in the middle of the link is the closest practical approximation to the optimum pre-emphasis.

Some Issues in Vibrator Data Processing

Recent improvements in vibrator control, such as the use of phase and amplitude control of ground force, require that some of the issues relating to vibrator data processing be reviewed. One central issue concerns the polarity of processed vibrator data as compared to processed dynamite data. At the field processing stage, alternate methods of correlation that include conventional zero-phase correlation, zero-phase inverse filtering, and minimum-phase inverse filtering are compared to determine their compatability with subsequent steps of wavelet estimation and inverse filtering or deconvolution. This issue is shown to be of particular importance when nonlinear sweeps are used. Conventional correlation is shown to be responsible for causing deconvolution to induce different phase shifts on the data when different spectral pre-emphasis sweeps or methods are used. Techniques for avoiding this problem are reviewed.