Search

We experimentally and theoretically demonstrated that transmission was over 4 km by using InGaAlAs Fabry Perot MQW lasers over a wide temperature range. We also present a new approach for estimating coefficient k of mode partition noise which we confirmed by transmission over 2 km in dark fibers with zero dispersion over the 1.29 to 1.33-/spl mu/m wavelength range.

Summary form only given. The longitudinal modes of a multimode laser diode suffer from intensity fluctuations that, when coupled with spectral filtering, result in mode partition noise (MPN). We have significantly decreased the MPN in a modelocked, multiwavelength semiconductor laser. A gain-guided GaAs-AlGaAs semiconductor optical amplifier (SOA) is inserted in an external cavity which incorporates a spectrometer and binary amplitude spatial filter to define individual wavelength channels. Our method exploits the saturable nonlinearity in a multiple-quantum-well (MQW) saturable absorber, along with high speed gain modulation, to establish hybrid modelocking in the external cavity semiconductor diode laser. The result is an error-free optical pulse train in each of the wavelength channels generated by the multiwavelength laser.

It is obvious that the development of a wavelength tunable source of short optical pulses will be of paramount importance for future Wavelength Division Multiplexing (WDM), optical time division multiplexed (OTDM), and hybrid WDM/OTDM optical communication systems. Self-seeding of a gain-switched Fabry-Perot (FP) laser is one of the most reliable techniques available to generate wavelength tunable optical pulses. An important characteristic of these self-seeded gain-switched (SSGS) sources is the variation in the side-mode-suppression-ratio (SMSR) as the wavelength is tuned, as this may ultimately affect their usefulness in optical communication systems. A quantitative analysis of how mode partition noise and side mode suppression ratio affects a wavelength division multiplexed system is investigated here. Bit Error Rate measurements have been carried out on a 4-channel wavelength division multiplexed set-up using tunable self-seeded gain-switched pulse sources. These measurements demonstrate the degradation in overall system performance, due to mode partition noise, as the side mode suppression ratio of the self-seeded gain-switched optical pulse sources is reduced. The results also show that the constraints on the minimum side mode suppression ratio required increase with the number of channels in the system.

A polarimetric interferometer is used to enhance mode partition noise. A theory is developed on agreement with the experimental results. The experimental setup is implemented using basically the output of a 786 nm wavelength multimode laser diode coupled to a 1.56 m long, 1.2 mm at 820 nm beat length birefringent optical fiber, 2 linear polarizers, and 4 lens. The optical signal with mode partition noise enhanced has a constant spectral density up to 100 MHz. An application is provided with the characterization of three photodetection circuits.

Summary form only given. With a reflective grating, the laser modes can be partitioned and noise level will be enhanced. This generic noise-enhanced feature is named as mode partition noise, which has a quantum mechanical origin, in which the noise level of modal output is larger than that of total output. Here, we report a unique manifestation caused by the interplay between noise and chaos in a free-running multimode Nd:YVO/sub 4/ microchip laser, mainly, as chaos appears, mode partition noise could be smaller than total output noise.

The impact of the 1.3-/spl mu/m Fabry-Perot laser spectrum width on the maximum transmission length is investigated theoretically, and 15-km transmission at 2.5 Gbit/s was demonstrated by using a strained MQW FP laser.

An analysis is carried out to determine the parameters affecting mode partition noise in a laser transmission system operating at a gigabits per second data rate using long dispersive fibers. The analysis is based on a Monte Carlo simulation technique of mode partition noise in systems using single-mode or multimode lasers. Some interesting results were found when the fiber dispersion and/or bit rate were high and raised cosine shaping of the data was used. The results show that, as the dispersion is increased, the bit error rate floor, due to the mode partition noise, increases to a maximum value and then decreases substantially to a minimum and oscillates periodically between those two extremes. The period between maximums is determined by the bit rate, fiber dispersion, and mode spacing. Experimental results substantiating the theoretical predictions are given for a single-mode laser. For the case of a multimode laser, it was found that because of pulse shaping there are certain modes which contribute most to mode partition noise. Specifically, the modes contributing most to mode partition noise in typical 560 Mb systems are typically the fifth to seventh modes, which are displaced from the central mode on the short wavelength side. This result suggests a new criterion for choosing multimode lasers with low mode partition noise based on the intensity of those modes. >

Results are presented from the first reported 1.2 Gb/s transmission system trial operating with a 1550 nm multilongitudinal mode laser over 67 km of installed dispersion shifted fibre. Temperature tuning the laser allowed measurement of system penalties arising from mode partition noise, giving operational limits for laser/fibre wavelength matching. It is shown that using multilongitudinal mode lasers with a linewidth of approximately=5 nm, the centre wavelength must be matched to within approximately=2 nm of the fibre dispersion zero to maintain an acceptable error performance. >

A simulation program for the analysis of single-mode fiber optic systems is described. The characteristics of a fiber optic system are more accurately determined by specifying 15 parameters and, in an interactive fashion, determining system requirements for a specific performance level. The program is based on specifying an acceptable dispersion penalty and finds a corresponding link length. The output format also lists the other data important to the system designer, such as major losses and a breakdown of the dispersion penalty into its two components, i.e., intersymbol and mode partition noise. >

A gain-flattened multiwavelength modelocked semiconductor laser transmitter generates 6 Gb/s pulses into 168 discrete wavelength channels. Hybrid modelocking suppresses mode partition noise and produces error-free (pulse Q > 13) pulses simultaneously in each channel of a multiwavelength semiconductor laser.

Chirp or wavelength shifts in distributed feedback (DFB) lasers have been the limiting factor in high bit rate transmission systems. Furthermore, dynamic chirp in direct modulation schemes results in pulse broadening because of the chromatic dispersion in the fiber leading to higher power penalties for the link. It is thus crucial to be able to unambiguously quantify the chirp component at the chip level with a quick and inexpensive method. Towards this, we present results of a set of measurements done to understand the extent of correlation between the observed chirp and the system's dispersion penalty. Strained MQW DFB lasers fabricated in the AlInGaAs/InP system were used. The ratio of the conduction to the valence band offsets in the AlInGaAs/InP system is larger than in the InGaAsP/InP system. This is expected to provide better electron confinement minimizing leakage. Coupled with a MQW structure, uniform hole distribution across wells is possible thus enabling higher bandwidth. Different facet coating combinations were also included to provide a rich subset of the device dynamic properties for this study. The tested devices had a lasing wavelength between 1520-1550 nm. We have experimentally shown the existence of a correlation between the transient chirp in the leading edge of the light pulse and the observed dispersion penalty under 2.5 Gbit/s PRBS modulation. This offers a cost-effective solution to screen chips at the chip level without incurring costs associated with module level components. It is important to note, however, that this is a necessary but not sufficient screen due to issues like mode partition noise and timing jitter, which affect the overall power penalty.

A simple and accurate theory proposed by Jiang et al. (see Optical and Quantum Electronics, vol.22, p.23, 1990) to calculate the mode partition noise in single-mode fiber systems is shown to be in good agreement with the transmission experiment presented. The 560 Mb/s transmission experiment using the multilongitudinal mode laser diode is described and the result indicates that the transmission distance is limited to 10 km by the mode partition noise when 1-dB power penalty takes place. A statistical method is proposed to determine the k-parameter of laser diodes modulated by quasi-random RZ binary signals. The k-parameter is the most important parameter to describe the mode partition characteristics of laser diodes. >

We present, for the first time, a study of polarization instability induced RIN (relative intensity noise) enhancement in VCSELs. Devices under study are nominally single-mode top-emitting GaAs/AlGaAs VCSELs with an operating wavelength of 850 nm. Current confinement and gain guiding are obtained by double proton implantation. Detailed results of bit error rate measurement on free-space optical links with polarization-unstable VCSELs and a rate-equation based model of the polarization instability are presented.

In vertical cavity surface emitting lasers (VCSELs), a single longitudinal mode is supported in a cavity that can allow multiple transverse modes. In contrast to edge emitting lasers where multiple modes are longitudinal with nearly identical transverse profile, transverse modes can have significant differences in their coupling efficiency. Hence, mode selective loss (MSL) is easily introduced and mode partition noise (MPN) is more significant to the intensity noise of VCSELs. In this paper, we discuss the relevant time scales of the mode partition dynamics of VCSELs and explain the low frequency dependence of the relative intensity noise (RIN) spectrum.

The side mode suppression ratio of self-seeded, gain-switched optical pulses is shown to be a vital parameter in wavelength division multiplexed communications systems. Experiments carried out on a 2-channel wavelength multiplexed set-up using tunable self-seeded gain-switched pulse sources at 2.5 GHz, have demonstrated the cross-channel interference effects which may be encountered if the side mode suppression ratio of one of the sources becomes degraded.

A magneto-optical data storage system utilizing single-mode fiber is capable of providing high signal-to-noise ratio (SNR) recording if laser noise sources are properly managed. In particular, mode partition noise (MPN) associated with use of a Fabry-Perot laser diode can be a significant problem in a fiber-based system. The various mechanisms leading to MPN as well as to laser phase noise are discussed in the context of a system constructed with polarization-maintaining fiber. The primary noise mechanisms include spurious fiber-endface reflections and errors in the quarter-wave plate on the recording head. An understanding of these effects is essential for fabrication of a fiber-based recording system with suitable SNR performance.

We demonstrate a hybridly modelocked multiwavelength semiconductor laser using an intracavity saturable absorber. Intensity autocorrelations show improved interchannel phase coherence versus active modelocking. Eye diagrams illustrate the suppression of mode partition noise and the production of error-free optical pulse trains in each of the multiwavelength channels.

The sidemode suppression ratio of self-seeded, gain-switched optical pulses is shown to be a vital parameter in determining the usefulness of these pulses in wavelength-division multiplexed communications systems. Experiments carried out on a two-channel wavelength multiplexed setup using tunable self-seeded gain-switched pulse sources at 10 GHz, have demonstrated the cross-channel interference effects that may be encountered if the sidemode suppression ratio of one of the sources becomes degraded. The sidemode suppression ratio of self-seeded, gainswitched optical pulses is shown to be a vital parameter in determining the usefulness of these pulses in wavelength-division multiplexed communications systems. Experiments carried out on a two-channel wavelength multiplexed setup using tunable self-seeded gain-switched pulse sources at 10 GHz, have demonstrated the cross-channel interference effects that may be encountered if the sidemode suppression ratio of one of the sources becomes degraded.

In the above published paper (ibid., vol. 31, no. 15, pp. 2629-2638, Aug. 2013), equations (10) and the paragraph after equation (9) have typographical errors. Corrections are presented here.

Relative intensity noise spectra of weakly index guided VCSELs in a multi—transverse mode regime are analyzedby using a model that takes into account all the transverse modes supported by the waveguide. Several resonancepeaks are obtained in the noise spectra that correspond to the relaxation oscillation frequencies of the transverse modes. In is shown that for low spatially overlapping transverse modes, low RIN operation can be maintained. However, the excitation of transverse modes with a significant mode overlap leads to a clear enhancement of theRIN at low frequencies. Keywords: Semiconductor Lasers, Vertical cavity lasers, Transverse modes, Relative Intensity Noise 1. INTRODUCTION Vertical-cavity surface-emitting lasers (VCSELs) are highly attractive light sources for applications in intercon- nections and optical communications, because of their useful characteristics such as single longitudinal modeoperation, low threshold current, circular output beam, and wafer scale integration.'2 An important consider-ation for such applications of VCSELs is the understanding of their noise characteristics. Several experimentalstudies of the relative intensity noise (RIN) in VCSELs have been performed.39 In general the noise perfor-inance of VCSELs operating in the fundamental mode is comparable to that of edge emitting lasers. However

The performance of a lightwave communication system can be affected by the laser's relative intensity noise (RIN) and mode partition noise (MPN). In addition, measurements of RIN after propagation in fiber can be used to obtain several important intrinsic laser parameters. Therefore an accurate and simple model of RIN after dispersive propagation is necessary. We report here an effect of MPN that can affect RIN in low-noise distributed feedback (DFB) lasers even with side-mode suppression ratios higher than 40 dB and at frequencies up to 5 GHz, and which cannot be explained by previous single-mode models. We present a multimode theory of RIN after propagation in fiber in which the effect of FM to AM conversion in the fiber has been included via a generalization of the transfer function approach to the multimode case. The additional parameters that are needed in this multimode model are experimentally obtained from a measurement of the optical spectrum. This allows us to get theoretical fits that agree well with experimental measurements of RIN.

Tunable Diode-Laser Absorption Spectroscopy (TDLAS) is increasingly being used to measure trace-gas concentrations down to low part-per-billion levels (1 ppbv = 10(exp -9) volume mixing ratio). Semiconductor lead-salt diode lasers give access to the mid-infrared spectral region and the application of high-Frequency Modulation (FM) schemes can further improve the sensitivity and detection speed of modern instrumentation. Several factors influence or even limit spectrometer performance. The central elements in such spectrometers are lead-salt diode lasers. Experimental data will be presented, which demonstrate that high-frequency excess-noise contributions above several MHz can be attributed to mode hopping and mode partition noise during multimode laser operation. Additionally it will be discussed how a FM-TDLAS spectrometer can be interpreted as an optimized Michelson interferometer for absolute distance measurements and, therefore, is extremely sensitive towards drift effects. The hig her the modulation frequency, the higher is the drift sensitivity of the spectrometer due to interferometric effects. These drift effects are a second factor affecting ultrasensitive measurements. While wideband-laser noise characteristics call for high modulation frequencies, the aforementioned interferometric effects in the spectrometer require low modulation frequencies.

Mode-partition noise has been shown to impose significant transmission distance limitations on microwave subcarrier transmission systems. While mode-partition noise has been analyzed for pulsed fiber-optic systems,2 these results do not apply to subcarrier based systems. This work presents and analysis of mode-partition noise in subcarrier systems based on a simple model of laser mode partitioning. Experimental data are shown which verifies predicted changes in the SNR due to mode-partition noise in conditions of changing dispersion, modulating frequency, and modulation depth.

At typical read powers, laser intensity fluctuations can be the dominant source of noise in the read channel of an optical drive. Even with the isolation present in most write-once optical heads, optical feedback (OFB) can be a significant contributor to the noise in the system read channel. The dominant laser noise sources are mode hopping and mode partition noise.' The former noise is particularly troublesome as it results from an alternation of the dominant laser mode, which causes output power fluctuations and irreproducible playback performance. At certain combinations of drive current, temperature, and optical feedback, this unstable behavior is observed. The second contributor, mode partition noise, is the broad-band noise associated with the transient variation in the partition of energy per mode.

Free-running gain-switched Fabry-Perot laser diode is an appropriate incoherent broadband optical source for incoherent 2-dimensional optical code division multiple access. However, the mode partition noise (MPN) in the laser seriously degrades performance. We derived a bit error rate (BER) expression in the presence of MPN using the power spectra of the laser. The theory agreed with the experimental results. There was a power penalty and BER floor due to the MPN in the laser. Therefore, this scheme should be operated with a sufficiently large number of modes. At least 9 modes should be used for error-free transmission at 1 Gbit/s for the laser we investigated in this work.

This paper is primarily an overview of data link design efforts in IBM pertinent to local area networks (LANs) using both multimode fiber (MMF) and single-mode (SMF) links, with emphasis on MMF links operating at short wavelengths. Device models (laser and receiver) and multimode fiber models are discussed, as well as noise aspects (modal and mode partition noise). In addition, new simulation and measurement results for a 20-Gb/s 1-km-long link are presented.

We demonstrate the suppression of intensity fluctuations, which are known as mode partition noise, in a multiwavelength semiconductor laser by using a hybrid mode-locking scheme. The laser design incorporates a saturable absorber and a gain-modulated semiconductor optical amplifier, along with spectral filtering, in an external cavity to achieve multiwavelength hybrid mode locking. The mode-locked laser produces an error-free (pulse Q>13) 300-MHz optical pulse train in each of three wavelength channels.

The presence of mode-partition noise in an erbium-doped fibre laser has been experimentally demonstrated. A numerical model has been developed using the Langevin rate equations. Its predictions are in qualitative agreement with the experimental data.

The authors investigate dispersion penalties in an analogue link using an uncoded Fabry-Perot laser operating near 1.3 /spl mu/m. Mode-partition noise caused by mode fluctuations combined with dispersion can rise significantly as the laser temperature rises and therefore its wavelength drifts. It is found that in a link transmitting signals near 1 GHz the noise may rise by over 20 dB even though the laser wavelength is close to the dispersion zero of the fibre.

The performance of random-wavelength switching of a wavelength-tunable three-electrode DBR laser is demonstrated. Packet header information is used to randomly switch the currents of the laser and route 155Mbit/s data packets onto four different wavelength modes. BERs of 10−9 are achieved over a 4.8nm tuning range with a packet guard time 3 μs. Modepartition noise and mode hopping are not degrading factors under random-wavelength-switching conditions.

The performance of optical communication systems employing multilongitudinal mode laser and semiconductor laser preamplified receiver combinations is investigated theoretically, emphasis being placed on the influence of mode partition noise (MPN). Two types of MPN are considered in the analysis: one owing to wavelength-dependent fibre dispersion and the other owing to the wavelength-dependent attenuation mechanism associated with both the amplifier gain and optical filter transmission characteristics. To reduce the influence of amplifier gain ripple and to ensure high receiver performance the mode spacing and the centre mode wavelength of the laser source may need to match closely those of the amplifier, especially for high internal amplifier gains and substantial facet reflectivity. For example, for 1% amplifier facet reflectivity, 30 dB internal optical gain, and 3 nm half-spectral-width laser the receiver sensitivity degrades by 5.1 dB when the lengths of the laser and the amplifier differ by 5%. In contrast with 0.1 and 0.01% reflectivity this reduces to 1.95 dB and 0.4 dB, respectively. We also show that for multimode laser operation there is an optimum bandwidth for the bandpass optical filter at which the required detectable optical power for a given bit error rate (BER) is minimum. Operating with filter bandwidth less than this optimum value will degrade dramatically the sensitivity and may, even in the absence of fibre dispersion, lead to a BER floor in the presence of MPN. The analysis presented here may be used to provide guidelines for the design of optical systems based on laser amplifiers and multimode or nearly single mode laser sources.

A system model for analyzing the effects of dispersion when multilongitudinal-mode laser diodes are used is presented. The analysis considers in a general manner the digital symbol statistics, the system pulse shaping, the laser diode spectrum, the laser diode extinction ratio, and mode partition noise. For mode partition noise, separate consideration is given to the statistics of the longitudinal-mode powers within a baud period and to the statistics between different baud periods. It is demonstrated that failure to separate the mode partition statistics in this manner can lead to underestimation of the dispersion penalty. By incorporating statistical models for the important system parameters, a statistical approach to route design is demonstrated. >

Mode partition noise may become the dominant performance limitation for single-mode fiber systems. We define the mode partition coefficient k for laser diodes which can be used as a universal index, like threshold current and spectral width, to characterize laser diodes. The k value varies between 1 and 0, and for larger k , the mode partition noise will be higher. We also demonstrate two simple techniques for measuring k : the low-pass filter method and the sampling method. We discuss the k values measured under different operating conditions (pulse width, dc bias) for several types of laser diodes, k values are around 0.4-0.6 with short pulse width (1.5-5ns), and decreases to 0.14 pulse width of 50 ns. The largest k values were obtained for a laser exhibiting strong relaxation oscillation. For that laser, k did not depend on the pulse width. The k value is an important factor for system and fiber design as well as laser diode design.

Mode partition noise can be the dominant limitation on a system error rate for single-mode-fiber laser transmission systems. The distribution of power among the longitudinal modes of the laser fluctuates and, as a consequence of chromatic dispersion in the fiber, causes an amplitude fluctuation of the signal at the decision circuit in the receiver. The effect is in essence a pulse-delay fluctuation, and the error rate cannot be reduced by increasing the received signal power. We derive a simple formula describing this mode partition noise, give useful design relations, and verify the analysis by comparison with system experiments.

Design relations are given for semiconductor lasers having both an active and a passive (low-loss) region within a single resonant cavity. Langevin noise sources are included to make possible the calculation of line width and power fluctuations using Monte Carlo techniques. Explicit relations based on equilibrium conditions are given for line width, power fluctuations, transient-response time constant and side mode level. It is shown by example that when the passive section is several times the length of the active section, the statistics of the power fluctuations become more nearly Gaussian and mode partition noise should become far less noticeable.

The injection laser can initially turn on in a side mode rather than in the main mode when the risetime of the driving current is short compared to the electron lifetime, causing mode partition noise in digital transmission systems which have wavelength dispersion between the laser and the receiver. It is shown that increased differential loss between the main mode and side modes reduces both the incidence and duration of side-mode turn-ons and is the most desirable approach to reducing mode partition noise. The needed differential loss is a strong inverse function of the signal pulse width. Gain compression has a lesser influence on the rate of occurrence of side-mode turn-ons. Quantitative results from a computer model indicate that significant numbers of such turn-ons occur even when the side-mode power during quiescent bias conditions never exceeds the main-mode power. Biasing slightly above threshold will always increase the pulse rate at which a given laser can be operated and provides an attractive goal for designers of laser transmitter packages. >

The intensity-noise characteristics of a nearly single-longitudinal-mode semiconductor laser are analyzed by adopting a two-mode model. An approximate solution of the resulting Langevin rate equations is used to obtain an analytic expression for the enhancement of the main-mode intensity noise caused by mode partition. We also obtain analytic expressions for the autocorrelation and cross-correlation functions which are used to discuss noise variances in the two modes. In particular, the cross-correlation function is related to the mode-partition coefficient, a measure of anticorrelation between the two modes that is useful to analyze the system performance. Our results show explicitly how the intensity correlation is affected by the mode-suppression ratio and other device parameters.

The power penalty and bit error rate degradation encountered when using wavelength division multiplexing devices is analyzed. Analysis is applied to the birefringent wavelength division multiplexer (BWDM) with a sinusoidal transfer function.1 The degradation in bit error rate is attributed to laser mode partition noise and crosstalk noise.

We report first measurements of the effect of optical feedback on the bit error rate (BER) as a function of received power of single longitudinal mode cleaved-coupled cavity (C3) and distributed-feedback (DFB) semiconductor lasers. Mode-partition noise, induced by the optical feedback, can cause a bit error-rate floor for large optical feedbacks, but it is shown that high-performance C3lasers and especially DFB lasers can withstand very large optical feedbacks without system degradation.

Theoretical and experimental results are presented for the signal-to-noise (S/N) ratio caused by mode partition noise, intensity noise, and reflection-induced noise in optical data links. Under given conditions an additional noise source with a S/N ratio of 20 dB will cause a power penalty of 1 dB in order to maintain a 10/sup -9/ bit error rate. From numerical simulations the authors predict the maximum allowable dispersion in the presence of mode partition noise to be approximately 40% of a clock period. This figure is almost independent of bit rate and laser structure and agrees well with the measurements and with results of other workers. Numerical simulations of a buried-heterostructure and a TJS laser were carried out at four bit rates from 565 Mbit/s to 4.5 Gbit/s and the measurements were done at 2.2 Gbit/s using a TJS laser. >

A simple model of mode partition noise is developed, which is valid for semiconductor lasers operating CW, having a single active region and oscillating primarily in one dominant mode. It is shown that the intensity fluctuations in the nonlasing modes occur with characteristic times of a few nanoseconds and have exponential probability distributions. We show that the lasing mode is able to follow adiabatically fluctuations in the nonlasing modes in such a way that the total intensity remains constant. Our model quantitatively accounts for the observations of Linke et al. that the frequency of "dropouts" in the lasing-mode intensity decreases exponentially with the ratio of average intensities of the lasing and nonlasing modes, and that in digital communication applications a ratio of about 50 is required to reduce the resulting bit-error probability to \lsim 10^{-9} . We also account for the probability distribution of the lasing-mode intensity, which was observed by Liu et al to be exponential at low intensity and Gaussian at high intensity. The exponential fluctuations constitute partition noise induced in the lasing mode by large Gaussian fluctuations in the field amplitude of the nonlasing mode. Additional noise sources associated with the carrier density and the lasing-mode field produce Gaussian behavior at high intensities plus a negligible broadening of the low-intensity exponential tail.

The dominant penalties are wavelength chirp, line broadening due to optical back reflection, and mode partition noise due to insufficient side-mode suppression. Experimental results on pulse compression at 1550 nm and on penalty reduction by dispersion compensation are presented. The dispersion compensation is achieved by adding a negative dispersive fiber link to a standard fiber link with positive dispersion at 1550 nm. In this way it was possible to transmit 5 Gb/s over a 111-km fiber link. >

The mode partition noise of semiconductor laser diodes in high-speed PCM direct modulation systems is reviewed from the standpoint of communication applications. First, the progress in this field is summarized briefly, and an appropriate noise model to predict the transmission performance is explained. Second, the parameters to evaluate the mode partition noisein optical communication systems are explained. Third, methods to measure the noise parameters are de scribed.

Quasi-linearized multimode quantum-mechanical rate equations with shot-noise sources for nearly single-longitudinal-mode semiconductor lasers are solved analytically by means of a flow graph technique. It is found that mode partition noise is the major intensity noise at low frequencies. At the resonance frequency, the "single-mode" noise dominates. Expressions for the relative intensity noise at low frequencies and the resonance frequency are derived. The correlation of the main mode and a side mode is negative at low frequencies and positive at the resonance frequency. The analysis also shows that intensity fluctuations of a side mode have a time constant of a few nanoseconds and a weak resonance peak caused by the fluctuations of the carrier number.

An evaluation of the influence of mode partition noise on error rate performance in a high bit rate optical fiber transmission system is presented. First, it is experimentally clarified that the intensity in each longitudinal mode of a laser diode fluctuates, although the intensity for the total mode is constant. It is also established that this fluctuation causes degradation of the error rate performance after transmission through a long optical fiber. The fluctuation is named "mode partition noise." Next a simple model for the fluctuation is proposed. The characteristics of the fluctuation are discussed on the basis of this model. Optical waveform fluctuation is found to be introduced by mode partition noise in the course of transmission through a long despersive transmission medium. This optical waveform fluctuation and its frequency spectrum are calculated. Finally, the error rate performance is evaluated and specifications required for a laser spectrum to attain a given repeater spacing are clarified.

A rate 0.964 forward error correcting (FEC) code is integrated into the low-speed tributaries of a 565-Mb/s lightwave system as an exploratory system design approach toward relaxing requirements on laser sources in dispersion-limited operation. By virtually removing error rate floors, regardless of their cause, FEC is shown to provide an increasing advantage in conditions of greater degradation and to be effective against mode partition noise (MPN), mode jumping, and reflection impairments.The experimental FEC code is implemented in a standard gate array. The FEC code is described and its performance is analyzed. A new system design strategy is suggested for low-cost gigabit lightwave systems using FEC. >

This paper describes an attainable repeater spacing for a high bit-rate single-mode fiber-optic transmission in the 1.55 μm wavelength region where laser mode partition noise comes to be significant. An expression for evaluating mode partition noise is given as the form involving the influence of laser spectral fluctuations under high bit-rate modulation, together with the intersymbol interference and the equalized pulse shape in tile optical receiver. After the validity of its numerical results is confirmed experimentally, the resulting evaluation of laser mode partition noise is connected to a systematic discussion on the attainable repeater spacing of a 280 Mbit/s fiber-optic transmission system operating at 1.55 μm, along with fiber loss versus dispersion tradeoffs. This discussion permits the attainable repeater spacing to be 60-70 km for the combination of a laser diode with 1.5-2.0 nm spectrum broadening and a fiber with the loss of 0.5 dB/km and the dispersion of 4-6 ps/km - nm.

This paper presents the theoretical and experimental results of a study on laser diode mode partition noise. Laser diode mode partition noise is evaluated by mode partition constants which indicate the degree of variance in longitudinal mode power. The relationship between the mode partition constants and laser diode characteristic parameters is clarified using rate equations. The mode partition constants for AlGaAs/GaAs and InGaAsP/InP double heterostructure laser diodes are measured experimentally and found to be independent of the structure, wavelength ( 0.8-1.6 \mu m), and bit-rate (400 Mbits/s-l.6 Gbits/s). The experimental and theoretical results show excellent agreement.

Mode partition noise is unavoidable with semiconductor lasers, even when the laser is oscillating in a nearly single longitudinal mode.1-4 This noise shows up in the form of power dropouts in the main mode, and power increases in the side modes. Such power dropouts were first observed by Linke et al.5 Main-mode power dropouts can affect the performance of a coherent optical communication system in the form of signal fading. Knowledge of the statistics of dropout pulse interarrival times, depths, and widths is important if one wants to estimate the error rate of the system. We have studied these statistics using commercial single-mode laser diodes.