Thesis (Ph. D.)--University of Texas at Dallas, 2006. / Includes vita. Includes bibliographical references (leaves 51-54).

Thesis (Ph.D.)--University of Delaware, 2006. / Principal faculty advisor: Michael W. Haney, Dept. of Electrical and Computer Engineering. Includes bibliographical references.

Electronic signal processing techniques have assumed a prominent role in the design of fibre-optic communication systems. However, state-of-the-art systems operate at per-channel data rates of 100 Gb/s, which constrains the class of communication algorithms that can be practically implemented. Relative to LDPC-like codes, product-like codes with syndrome-based decoding have decoder dataflow requirements that are smaller by more than two orders of magnitude, which strongly motivates the search for powerful product-like codes. This thesis presents a new class of high-rate binary error-correcting codes, staircase codes, whose construction combines ideas from convolutional and block coding. A G.709-compliant staircase code is proposed, and FPGA-based simulation results show that performance within 0.5 dB of the Shannon Limit is attained for bit-error-rates below 1E-15. An error-floor analysis technique is presented, and the G.709-compliant staircase code is shown to have an error floor below 1E-20. Using staircase codes, a pragmatic approach for coded modulation in fibre-optic communication systems is presented that provides reliable communications to within 1 bit/s/Hz of the capacity of a QAM-modulated system modeled via the generalized non-linear Schrodinger equation. A system model for a real-world DQPSK receiver with correlated bit-errors is presented, along with an analysis technique to estimate the resulting error floor for the G.709- compliant staircase code. By applying a time-varying pseudorandom interleaver of size 2040 to the output of the encoder, the error floor of the resulting system is shown to be less than 1E-20.

Electronic signal processing techniques have assumed a prominent role in the design of fibre-optic communication systems. However, state-of-the-art systems operate at per-channel data rates of 100 Gb/s, which constrains the class of communication algorithms that can be practically implemented. Relative to LDPC-like codes, product-like codes with syndrome-based decoding have decoder dataflow requirements that are smaller by more than two orders of magnitude, which strongly motivates the search for powerful product-like codes. This thesis presents a new class of high-rate binary error-correcting codes, staircase codes, whose construction combines ideas from convolutional and block coding. A G.709-compliant staircase code is proposed, and FPGA-based simulation results show that performance within 0.5 dB of the Shannon Limit is attained for bit-error-rates below 1E-15. An error-floor analysis technique is presented, and the G.709-compliant staircase code is shown to have an error floor below 1E-20. Using staircase codes, a pragmatic approach for coded modulation in fibre-optic communication systems is presented that provides reliable communications to within 1 bit/s/Hz of the capacity of a QAM-modulated system modeled via the generalized non-linear Schrodinger equation. A system model for a real-world DQPSK receiver with correlated bit-errors is presented, along with an analysis technique to estimate the resulting error floor for the G.709- compliant staircase code. By applying a time-varying pseudorandom interleaver of size 2040 to the output of the encoder, the error floor of the resulting system is shown to be less than 1E-20.

Fiber-optic networks are continually evolving to accommodate ever-increasing data transport rates demanded by modern applications, devices, and services. Network operators are now beginning to deploy systems with 100 Gb/s per-wavelength data rates while maintaining the 50 GHz dense wavelength division multiplexing grid that is (generally) standard for 10 Gb/s systems. Advanced modulation formats incorporating both amplitude- and phase-based data symbols are necessary to meet the spectral efficiency requirements of fiber-optic data transport. These modulation formats require coherent detection, enabling future networks to take advantage of advances in silicon CMOS via digital signal processing algorithms and techniques. The primary challenge for future networks is the fiber nonlinear response; changes in the intensity of the propagating optical signal induce changes in the optical fiber refractive index. Limiting the allowed propagation intensity will reduce these nonlinear effects and correspondingly limit the total available signal-to-noise ratio (SNR) within the channel. Predicting the nonlinear SNR limits of fiber-optic transport for data rates 100 Gb/s and beyond is a primary purpose of this research. This dissertation expressly matches several novel expressions for nonlinear interference accumulation to experimental data and demonstrates robust theoretical prediction of nonlinear transmission penalties. The experiments were performed to isolate the transmission performance of the fiber medium in the highly dispersive regime -- no dispersion compensation or Raman amplification was employed and all other hardware was kept static. These results are the first experimental validation of the nonlinear interference expressions on a fiber-type basis. Second, this dissertation moves to data transport beyond per-wavelength rates of 100 Gb/s by employing 16QAM at baud rates as high as 32 GHz. It examines signal processing strategies for 16QAM transport and extends the nonlinear interference prediction techniques to 16QAM. The results reveal that the SNR requirements of 16QAM as limited by nonlinear interference will likely limit deployments to high-density regional and metro networks.

The decay of higher order solitons in optical fiber, initiated by localized channel perturbations such as a step change in dispersion, a localized loss element, or a bandpass filter, is explored theoretically and experimentally as a means of generating pairs of pulses having wavelengths that are up and down-shifted from the input wavelength. The achievable wavelength separation between the two sub pulses increases with increasing the amount of perturbations. Pulse parameter requirements for achieving useful wavelength shifts while avoiding unwanted nonlinear effects are presented. Experimental studies for N=2 solitons having 1 ps initial width are performed to demonstrate tunable wavelength conversion using a step change in dispersion and using a loss element. Wavelength shifts are tunable by varying the magnitude of a dispersion step or loss element that is used to disrupt the soliton cycle. Competing nonlinear effects, such as cubic dispersion, self-steepening, and stimulated Raman scattering, can be minimized by using input pulsewidths of one picosecond or greater. The separated pulses at two wavelengths can in principle be amplified to form separate higher order solitons. The process repeated to produce multiple wavelength replicas of an input data stream, and may thus be of possible use in multi-casting applications in fiber communication systems. The possibility of soliton recovery is also studied. For soliton recovery, conditions are stringent, in that the precise temporal overlap and phase relationship between sub-pulses that occurred at the point of separation is in principle needed at the reverse perturbation location. Experimental studies on soliton recovery for an N=2 soliton is performed by using a dispersion-compensated intermediate link, and reversing the dispersion step. Detrimental effects on soliton recovery are studied.

No description available.

No description available.

The objective of the proposed research is to develop multiwavelength lasers as cost-efficient sources for broadband optical access networks. Todays telecommunications networks have widely adopted optical fiber as the backbone transmission medium. Optical fiber systems are promising candidates for the broadband access networks to offer high-speed and future-proof services. To harness the available bandwidth in fiber and to meet the ever-growing bandwidth demand, wavelength division multiplexing (WDM) techniques have been investigated. There have been intense research activities for the creation of new low-cost laser sources for such emerging applications. In this context, multiwavelength fiber ring lasers have been significantly investigated as they present many advantages, including simple structure, low-cost, and selectable multiwavelength operation. We propose a new laser system architecture that emits alternate multiwavelength picosecond pulses operating at room temperature. Optical signal generation is based on a single active component, an unbalanced Mach-Zehnder interferometer, inserted in an actively mode-locked erbium-doped fiber ring laser to provide both intensity modulation and wavelength-selective filtering. Time and frequency controls of the light emission are reached by inserting an additional modulator and a periodic filter in the cavity. This approach focuses on the application of multiwavelength lasers as sources for WDM passive optical networks.

Optical communications is expanding into new applications such as infrared wireless communications; therefore, designing high performance circuits has gained considerable importance. In this dissertation a wide dynamic-range variable-gain transimpedance amplifier (TIA) is introduced. It adopts a regulated cascode (RGC) amplifier and an operational transconductance amplifier (OTA) as the feed forward gain element to control gain and improve the overload of the optical receiver. A fully-differential variable-gain TIA in a 0.35µm CMOS technology is realized. It provides a bit error rate (BER) less than 10-12 for an input current from 6µA-3mA at 3.3V power supply. For the transimpedance gain variation, from 0.1kΩ to 3kΩ, -3dB bandwidth is higher than 1.7GHz for a 0.6pF photodiode capacitance. The power dissipations for the highest and the lowest gains are 8.2mW and 24.9mW respectively. A new technique for designing uniform multistage amplifiers (MA) for high frequency applications is introduced. The proposed method uses the multi-peak bandwidth enhancement technique while it employs identical, simple and inductorless stages. It has several advantages, such as tunability of bandwidth and decreased sensitivity of amplifier stages, to process variations. While all stages of the proposed MA topology are identical, the gain-bandwidth product can be extended several times. Two six-stage amplifiers in a TSMC 0.35µm CMOS process were designed using the proposed topology. Measurements show that the gain can be varied for the first one between 16dB and 44dB within the 0.7-3.2GHz bandwidth and for the second one between 13dB and 44dB within a 1.9-3.7GHz bandwidth with less than 5.2nV/√Hz noise. Although the second amplifier has a higher gain bandwidth product, it consumes more power and occupies a wider area. A technique for capacitance multiplication is utilized to design a tunable loop filter. Current and voltage mode techniques are combined to increase the multiplication factor (M). At a high input dynamic range, M is adjustable and the capacitance multiplier performs linearly at high frequencies. Drain-source voltages of paired transistors are equalized to improve matching in the current mirrors. Measurement of a prototype loop filter IC in a 0.5µm CMOS technology shows 50µA current consumption for M=50. Where 80pF capacitance is employed, the capacitance multiplier realizes an effective capacitance varying from 1.22nF up to 8.5nF.