The advent of practical optical gain raises the prospect of lightwave networks with very large capacities, especially when multiple wavelengths are employed. Erbium-doped fiber amplifiers can be arranged in lengthy cascades that deliver strong, well- regulated light levels in single-wavelength systems. However, substantial new challenges arise in amplified multiwavelength networks. We survey means of solving them.

An integrated-optic parallel directional coupler waveguide modulator is the subject of this paper. The device consists of two directional coupler modulators in parallel so that a single laser can be used as a source. Advantages of one laser source are to minimize a relative frequency drift difference and allow for high speed communications on separate channels. Applications include television and videophone where audio and video signals can be transmitted simultaneously. A feedback phase shift noise reduction scheme will be investigated for high speed information flow with high fidelity and high linearity. As the result, the parallel modulation technique approaches transparent transmission which will increase the dynamic range and minimize signal distortions.

Er-doped glass thin films are drawing increasing attention for optical amplifiers suitable for integrated optic configuration. We have investigated deposition of highly Er-doped (mid 10¹⁹ atoms/cm³ silica films using Rf magnetron sputtering. Erbium was doped into the host material by cosputtering techniques. Deposited films (0.5 to 1.2-micrometers thick) were characterized by photoluminescence, secondary ion mass spectroscopy, and fluorescence decay measurements. We have also deposited Er-doped three component silicate glass (SiO₂ + Al₂O₃ + MgO) films to investigate the dependence of Er³⁺ luminescence on the host material's composition. Optimum anneal temperature for the silicate glass films was found to be lower than that for the silica films while the Er³⁺ luminescence intensity was slightly higher for the silicate glass case.

Analog AM photonic links are used in many applications such as cable TV and microwave links in phased arrays. The major issues of these links are linear dynamic range and rf efficiency. We will show in this paper that, while the rf efficiency is always increased by the optical amplifiers, the linear dynamic range may be significantly reduced. The effect of the optical amplifier on the signal to noise ration can always be expressed as a laser source with an equivalent RIN noise. The signal to noise ratio is affected less when the amplifier is placed closer to the source.

This paper reports on an experimental study made of various laser characteristics such as threshold pump power, line width, output power etc. of an Erbium-doped fiber laser using a ring resonator geometry and comparing the results obtained with a linear resonator geometry. The EDF used in this study had a core diameter of 6.8 micrometers and an Erbium concentration of 1200 ppm-wt in a matrix of Al/P doped silica and a N.A. of 0.13. The pump wavelength used was 514 nm obtained from a CW argon ion laser. The resonator configurations studied were a ring geometry with a single plane dichroic mirror at 45 degree(s) to the pump beam. This mirror coupled 80 percent of the pump beam at 514 nm into the EDF and the reflectivity at 1.5 micrometers was 50 percent. The linear geometry resonator used a dichroic mirror with 80 percent transmittance at 514 nm and 60 percent reflectance at 1.5 micrometers . The ring resonator laser had only a single peak in the laser emission spectrum at 1550 nm for longer fiber lengths but at 1530 nm for shorter lengths of Erbium-doped fiber. The linear unbutted resonator configuration spectrum has 1 peak at 1530 nm and the emission depended on the fiber lengths employed. Experimental studies were made of the laser output power and laser spectrum as a function of input pump power, fiber length and cavity losses for both these configurations and the results obtained were compared.

Erbium-doped fiber lasers represent an important class of efficient coherent light sources useful in optical communication systems. Fiber laser have been built with various resonator configurations, (a) with mirrors not in contact with fiber ends and (b) with mirrors butted to the fiber ends. It is difficult to estimate the resonator losses accurately, but they can be measured in erbium-doped fiber lasers under actual operating conditions by studying the relaxation oscillations of the laser. This paper reports on a study made of relaxation oscillations in an erbium-doped fiber laser and the determination of the threshold pump powers, cavity lifetimes and single pass losses of the laser for a linear resonator configuration with (a) butted mirrors (b) mirrors not in contact with the fiber ends and (c) also for a ring resonator configuration. The laser thresholds obtained by this method has been compared with those determined by us from other methods. This technique of determining the resonator characteristics from the study of relaxation oscillations permits in situ loss measurements in erbium-doped fiber laser.

Strained balanced InAsP/InGaP superlattices for optoelectronic applications were studied with materials grown by low pressure MOVPE. For 20 and 30 period InAsP/InGaP superlattices with a 23 percent As mole fraction in the InAsP layers, sharp x-ray satellite peaks are obtained. The superlattices show an average lattice constant close to that of the InP substrate. Strong photoluminescence with narrow emission linewidth are observed at room temperature around 1.11 micrometers . PIN diodes with an intrinsic region consisting of the InAsP/InGaP superlattice show efficient electroabsorption at wavelengths around 1.15 micrometers with a small residual absorption of 59 cm^-1.

A high-sensitivity, wide-bandwidth, optical preamplifier is demonstrated in the form of a monolithically integrated semiconductor optical amplifier and waveguide photodetector. By fabricating the same device with a shorter photodetector section, part of the signal is passed through to an output fiber, forming an optical tap. Amplifier gains are sufficient to overcome the fiber-coupling and detector loss, and exhibit zero-insertion-loss operation. A lossless tap with an electrical bandwidth of 7 GHz, a responsivity of 26 A/W, and a fiber-to-fiber gain of 3 dB is shown to have a receiver sensitivity of -22 dBm at 3 Gb/s.

In this communication we focus on the potential use of travelling wave semiconductor laser amplifiers as high-speed all-optical inverters. Although semiconductor amplifiers suffer from saturation and crosstalk effects which can degrade the device performances required in some applications, we show here that both phenomena can be induced in order to produce optical transference and inversion of the information between two channels in a direct-detection system.

Recently, intense research on single-mode rare-earth doped fiber lasers and amplifiers[fl has led to the development of a range of active devices. The possibility of ion-doping of standard integrated-optic waveguides has recently allowed the demonstration of Nd- and Er-doped single-mode channel waveguide lasers in single crystal Ti:LiNbO3 waveguides[12] and various glass waveguides[35]. However, there are a number of intrinsic limitations associated with these material systems. First, glass waveguides have no electrooptic(EO) effect therefore, an active device cannot be made using these substrates. Second, LiNbO3 waveguides have significant walk-off between the refractive indices of microwave and optical waves. Consequently, the modulation speed of the EO modulator, which serves as the input signal generator for the waveguide amplifier, is limited to 4O GHz[6]. Third, the waveguide fabrication methods for LiNbO3 and glass substrates are not universal. They are not transferable to other substrates. For instance, Si and GaAs are the most frequently used substrates for optoelectronics. LiNbO3 and glass waveguide lasers cannot be implemented on these substrates as amplifiers without violating the monolithic integration preference. Photo lime gel-based polymer waveguide is an excellent guiding medium due to its wide transmission bandwidth(300nm to 2700nm). The GRIN characteristic of this material[7] allows the formation of high quality(loss <0.1 dB/cm) single-mode passive and active devices on an array of substrate[81. In this paper, we report a graded index(GRIN) polymer waveguide amplifier working at 1 .O6im wavelength using Nd-doped photolime gel as the active medium. Fluorescent spectrums at 1 .32im (Nd3doped) and at 1.56 im (Er3-doped) were also observed and the results are shown in Figs.l and 2, respectively. The polymer introduced is soluble in water. As a result, the chemical compounds containing rare earth ions(REIs) can be mixed with the host polymer as long as they are hydrophilic. But the doping concentrations should be below the level of microscopic clustering[5] which quenches the active ions. There are other quenchers which shorten the lifetime of metastable states of active ions. The most serious quenchers are the admixed 0-H groups whose concentration must be less than 3-5x10'8/cm3 [4] for glass waveguide amplifier. The existence of 0-H groups generates a number of intermediate states between the transition states. Such states, primarily from water molecules, significantly reduce the lifetime of metastable states. The general rule of thumb is that the lifetime of the excited state will be temperature dependent if the gap is less than ten times the effective phonon frequency and completely quenched if less than four times[5I. Experiment results will be presented at the conference.

Today's high-speed electronic devices and networks challenge the capabilities of commercially available test instrumentation. Modern techniques using ultrashort light pulses and electron beams are now routinely used to characterize these fast circuits but are not without their limitations. In this paper I discuss the performance characteristics common to all these measurement techniques and compare some of them in a context that will be applicable to future techniques as they develop.

Direct observation of nonuniform operation in GaAs microwave high-power FETs has been realized by introducing a new electro- optic (E-O) probing system. In the system, ZnTe is used as a longitudinal external E-O crystal in order to make high sensitive measurement. The spatial resolution of the E-O probing system with an electrically synchronized laser diode is as small as 5 micrometers . We apply this system to the measurement of the electric field at the microscopic region of drain electrodes of an X-band 1W FET consisting of 10 FET unit cells (73 micrometers separation). The electric field concentration to the center (1.6 times) and the most outer cells (1.1 times) has directly been measured. The electric fields on the unbalanced FET cells which were damaged artificially by the focused ion beam are also reported.

The characteristics of a family of coplanar transmission lines have been studied at frequencies extending to the terahertz range. Traditional wide-ground coplanar waveguides and coplanar strip lines were investigated together with a coplanar waveguide with narrow ground planes. The technique of nonuniform gap illumination was used to excite subpicosecond electrical pulses as a testing tool of transmission lines for the first time. It is shown that this method is versatile and convenient for testing ultrafast devices and circuits. The experimental results, extracted by both time- and frequency-domain analyses, indicate several interesting features. In the subterahertz frequency range, the 50-micrometers transmission lines are dominated by dispersion, while the narrower 10-micrometers lines are dominated by loss. The characteristics of traditional (wide-ground) coplanar waveguides and coplanar strips are in agreement with theory and comparable to each other up to very high frequencies. The implementation of narrow ground planes can considerably reduce attenuation and dispersion in coplanar waveguides. In some geometries, radiation loss can be eliminated completely. The reduction in radiation is attributed to the change of field patterns at the dielectric interface, which leads to reduced coupling between the coplanar waveguide mode and radiative substrate modes.

An ultra-broadband hybrid integrated pulse shaping circuit for the direct and efficient modulation of semiconductor injection lasers has been developed. The circuitry is optimized for the use in an electrooptic sampling setup that utilizes a 1.3 micrometers wavelength Fabry-Perot laser diode. The realization of arbitrary repetition rates allows the operation of the measurement setup similar to a conventional sampling oscilloscope with a trigger input. The generation of the electrical pulses and the obtained optical pulses as well as the simulation of the large signal modulation is discussed.

The ever and rapidly increasing utilization of single-mode optical fiber bandwidth coupled with rapid improvements in integrated circuit technology, have made ICs for 10+Gb/s lightwave systems a reality. However, even with today's advanced IC technologies, stringent system requirements still impose severe limitations on the design of lightwave ICs.

Device modeling of the heterojunction bipolar transistor (HBT) has been examined. The Gummel-Poon model derived from silicon bipolar transistor is able to predict the base and collector currents of the HBT with good accuracy for negligible self- heating. The present model extends the Gummel-Poon model equations to account for self-heating using a nonlinear thermal resistance. The present model also accounts for base current reversal due to impact ionization in the collector-base depletion region. Effects of self-heating on the performance of analog circuits such as current mirrors and small-signal amplifiers are evaluated.

Digital GaAs FET technology, with its high level of integration, can be employed in high-speed optical systems for optical detector amplifiers as well as for baseband processing. However, the WSi gate of the digital GaAs FETs results in high gate resistance, which increases the noise figure as compared to the Ti/Pt/Au `T' shaped gates used in microwave FETs. In high-speed optical receiver applications, the noise in the FET based amplifier dominates the entire system noise. In this work, we use the 1st and 2nd layer metal to reduce the gate resistance in order to improve the minimum noise figure through layout variations compatible with production digital technology.

In this communication we consider the use of a phase-locked loop with signal injection (PLLI) in clock synchronization circuits. We show that by exploring the correlation between the I (in- phase) and Q (quadrature) components of the disturbance, the PLLI allows significant improvements in the output jitter. Numerical results are given for optical communications with RZ signaling which show considerable improvements relatively to a conventional PLL. Experimental measurements obtained with a Colpitts VCO designed to allow current injection are presented that demonstrate the effectiveness of the PLLI in exploring the I-Q correlation.

Recent progress in the integration of lasers and detectors with transistor amplifiers to make optoelectronic integrated circuit (OEIC) versions of transmitter and receiver circuits is reviewed. Comparison of OEIC receivers based on the electronic devices used is presented along with other novel and notable recent achievements in this field.

To implement high-bandwidth optical analog or digital communication links based on optical modulators, ultrahigh performance modulator driver amplifiers are required. Design considerations for such amplifiers are discussed here. Designs are significantly different than for laser driver amplifiers. For modulator drivers, an emitter follower output stage is appropriate. Active pull-down circuitry is beneficial to reduce power dissipation in a digital driver. Inductive tuning is beneficial to extend frequency response. Flip-chip mounting is beneficial to reduce bonding parasitic capacitance. Examples are given for driver designs for 10Gbit/s digital applications, and for 20GHz analog links, employing GaAs/AlGaAs HBT technology.

A novel punch-through heterojunction phototransistor (PTHPT) integrated with a compatible modulation doped FET (MODFET) was fabricated. The two terminal operating PTHPT can provide the superior properties of high optical gain and high speed. This is done without introducing an amplified shot noise associated with a base bias current as found in a three terminal operating heterojunction phototransistor (HPT). The PTHPT fabricated in a 0.8-micrometers GaAs/AlGaAs material system exhibits an optical conversion gain as high as 1250 at an incident optical power as low as 0.5 (mu) W. The gain changes less than 15 percent over a 20 dB range of the incident optical power. The transient measurements show that the PTHPT has a higher response speed than that of a conventional HPT. The compatible MODFETs with 1-micrometers gate lengths exhibit transconductances over 200 ms/mm and a current density of 160mA/mm. The principle presented here can also be applied to other material systems such as GaSb/AlSb and InGaAs/InP for long wavelength optical communications.

Future communication and computer systems will require major advances in the integration of optical, optoelectronic and high- speed electronic componenets if packaging costs are to be reduced to acceptable levels. A hybrid integration approach called optoelectroic glass microwave integrated circuit (opto-GMIC) has been developed. Opto-GMIC comprises a glass substrate on a silicon carrier which can accommodate discrete and monolithic microwave and optoelectronic devices. In addition, fibre alignment grooves that simplify pigtailing can be defined. Hence a wide variety of high-speed optoelectronic multichip module designs can ultimately be produced from the same production line. In contrast to rival monolithic optoelectronic integrated circuit (OEIC) technologies, opto-GMIC can interconnect devices fabricated from different materials. This feature has been demonstrated by the fabrication of prototype fibre-optic repeaters, transmitters and receivers. These modules are designed for synchronous optical network (SONET) applications at a bit rate of 622.08 Mb/s and at a wavelength of 1.3 micrometers . All electronic functions on the modules are performed by GaAs microwave monolithic integrated circuits (MMICs) while optoelectronic functions are provided by InGaAs photodiodes and InGaAsP laser diodes. By off-loading electronic passive elements onto glass, the number of parts in assembly is reduced drastically.

Negative differential velocity in nonintentionally doped GaAs/A1As superlattices has been evidenced by an optoelectronic picosecond time-of-flight experiment in the temperature range from 10 K up to 300 K. In the same structures bursts of high frequency oscillations up to 61 GHz are optically induced in the same temperature range. Experimental records are correctly reproduced by numerical simulations in the classical drift- diffusion approximation. A particular design of the superlattice structure gives high yield of mm-wave/optical excitation conversion.

Integral Gummel charge-control relation for both linearly graded- and non-graded-base and for single or double-heterojunction bipolar transistors has been derived. In addition to modeling the collector current density for different heterojunction biopolar transistor with and without graded energy gap in the base. Comparisons with numerical and experimental data show excellent agreement.

We have fabricated photoconductive and photovoltaic ultraviolet sensors from GaN single layers and pn-junctions. These sensors exhibit a sharp long wavelength cut-off in responsivity at the bandgap (365 nm). The active layers (GaN) were deposited using low pressure MOCVD. The p-type doping was accomplished using Mg as the dopant. Photoconductive, and schottky barrier detectors were then fabricated using photolithography, reactive ion etching and contact metallizations. These processing techniques were developed specific to the A1_xGa_1-xN material system. We will discuss growth, fabrication and characterization details for these various device types. The measured values of device parameters will be contrasted with those estimated from active layer material characterization.

High-speed electronics was introduced into the subpicosecond regime by the applicaiton of short-pulse lasers to semiconductor systems. Photoconductive sampling, electrooptic sampling, as well as other ultrafast phenomena, have provided tools for the investigation of the behavior of solid-state materials on the ultrafast time-scale. Low-temperature-grown GaAs (LT GaAs) holds unique advantages for high-speed optoelectronics when employed with metal-semiconductor-metal photodetector (MSM) technology. To properly utilize this material we must consider its properties and the nature of the circuits and devices in which it will be used. The use of the MSM on LT GaAs as a small-signal detector and as a high-speed sampling gate has been demonstrated. Work is under way to develop materials with picosecond response and high resistivity, but with longer-wavelength sensitivity.

A highly sensitive photodiode was fabricated with a metal-porous silicon-silicon-metal structure. The porous silicon film was found to be an excellent light trap and antireflection coating for the photodiode. It was demonstrated that close to unity quantum efficiency could be obtained in the wavelength range of 630 to 900 nm. The detector response time is about 2 ns for a 9 volts reverse bias. It was also demonstrated that a relative sensitivity of higher than 90 percent could be obtained with incident angles of 40 degree(s), 80 degree(s)2, and 58 degree(s) for s-,p- and randomly-polarized light. The uniformity and stability of the photodiode were also studied. Porous silicon films can also be used as solar cells to improve the acceptance angle of the sunlight. Possible machanisms to this photodiode and roles of the porous silicon film are discussed.

A subpicosecond electro-optic sampling system was used to measure the picosecond characteristics of silicon-based, metal- semiconductor-metal photodiodes made on both bulk-silicon and silicon-on-sapphire (SOS) substrates with submicrometer finger spacings and widths. The temporal response of bulk-silicon diodes was strongly dependent on the wavelength of excitation light because of the effect of different penetration depths. On the other hand, for the SOS diodes, the device speed is nearly independent of wavelength since the thickness of the silicon layer limits the depth of photogenerated carriers. The external quantum efficiency of SOS diodes was measured at several selected wavelengths and shown to be dominated by the photon absorption coefficient.

Optical fiber communication systems operating in the wavelength range of 1.3 to 1.55 micrometers still demand the improvement of optoelectronic (O/E) transducers at the receiver site with respect to sensitivity, speed, and cost. Merging the O/E transducer function with electrical preamplification on one chip helps to improve these parameters. The scope of this paper is to review the status of such 'amplifying optoelectronic transducers', as there are avalanche photodiodes and optoelectronic-integrated circuits.

Avalanche photodiodes (APDs) are required for long distance fiber optical systems at wavelengths of 1.3 and 1.55 micrometers . Presently deployed systems operate at speeds up to 2.4 Gb/s, with increasing volumes. It is becoming increasingly necessary to produce low-cost, high-performance planar APDs. Most commercial InP/InGaAs APDs employ a separate absorption, grading, and multiplication (SAGM) design. However, this approach has severe process limitations for obtaining functioning APDs. Greater growth and processing flexibility is obtained from the separate absorption, grading, charge and multiplication (SAGCM) APD. In this work the influence of the layer thicknesses, junction depth and doping on the performance of SAGCM APDs is described both theoretically and experimentally. Theoretically, an analytical model of the (delta) -doped (ideal SAGCM) APD is presented, both neglecting and taking into account ionization in the InGaAs. In addition to the performance predictions with respect to fabrication parameters, this model also predicts that the gain depends on the wavelength of light (e.g. 1.3 or 1.55 micrometers ). Experimentally, whole 2-inch wafer performance results are interpreted with respect to variations in growth and processing parameters. The results are in good agreement with the predictions of the model.

InP/In_0.53Ga_0.47As (InGaAs) avalanche photodiodes (APDs) are often used in long distance, high bit rate fiber optical transmission systems operating in the 0.92-1.65 micrometers wavelength region. Lately attention has focused on the separate absorption, grading, charge and multiplication (SAGCM) structure. Recent studies have demonstrated that the room temperature device operation is less affected by changes in fabrication parameters than in the more conventional separate absorption, grading and multiplication (SAGM) structure, emphasising the potential of the SAGCM APD for manufacture. However, for coolerless applications, which are increasingly desired, the device temperature is not directly controlled but rather is determined from environmental influences. Therefore it is necessary to evaluate the APD performance as a function of temperature. In this paper temperature dependent measurements of gain, breakdown voltage, leakage current, bandwidth and noise on SAGCM APDs in the range -50 to 85 degree(s)c are presented. It is found that the bandwidth vs. gain dependence, gain and breakdown voltage are strongly correlated with the temperature. The gain-bandwidth product and minimum gain for useful bandwidth both increase with decreasing temperature or decreasing breakdown voltage. In addition, temperature studies of these devices has been used to evaluate the quality of the grading layer in a more stringent manner than from room temperature measurements alone.

Waveguide p-i-n photodiodes are theoretically revealed to have a great advantage over conventional surface-illuminated p-i-n photodiodes and metal-semiconductor-metal photodiodes in the performance limit, the product of the bandwidth and the external quantum efficiency. Experimental results show a bandwidth of over 75 GHz and a high efficiency with a mushroom-mesa multimode waveguide p-i-n photodiode.

A monolithically integrated planar dual InGaAs pin photodiode with a 3 dB bandwidth of 18 GHz has been fabricated for balanced optical coherent receiver application. The photodiodes are designed for front and backside butt-jointed fiber coupling. They exhibit a very low dark current of 10 pA at -10V, a capacitance of 85 fF and a responsivity of 1 A/W. After 4500 h operation no chip failure has occurred. High impedance front-end modules have been assembled with a 3 dB bandwidth of 8.9 GHz and a coupling efficiency of more than 90 percent (-0.5 dB).

The design of high-speed, high responsivity PIN optical detectors for fiber optic telecommunication test equipment requires a very careful trade-off between lateral and vertical active area dimensions. A thin absorption layer (i-layer) leads to high bandwidth at the expense of responsivity. A large-area device improves light coupling efficiency from a fiber, but reduces the bandwidth. In addition, packaging constraints must be considered. Our approach attempts to optimize the end-system performance by incorporating an integral micro-lens which increases the effective light collection area; a double-pass illumination strategy that increases the bandwidth without sacrificing responsivity; and a micro-flip-chip die attachment technology that minimizes parasitic capacitance and inductance. An optical receiver composed of a GaAs MODFET-based transimpedance amplifier and a flip-chip PIN photodector exhibited a bandwidth of 7.2 GHz with an optical-to-electrical conversion gain of 560 V/W.

Three systems have been developed to measure the frequency response of high-speed photo-receivers to 50 GHz. I will describe the systems and their calibration, and show measurements made on the three systems which are in good agreement. (1) A fiber and diffraction grating pair compresses the pulses from a mode-locked 1060 nm YAG laser to 2.5 pS FWHM. Peak optical power is adjustable to more than one Watt. We measure a photoreceiver's impulse response and the autocorrelation function of the optical impulse. From these we calculate the photoreceiver's frequency response via Fourier transform. We can observe the broadening of a photoreceiver's impulse response with increasing peak optical power. (2) Light from a pair of diode-pumped 1320nm Nd:YAG ring lasers is combined. The photoreceiver responds to the optical power envelope which varies at the difference frequency. We adjust the difference frequency by changing the lasers' temperature, and record the photoreceiver's frequency response. (3) The frequency response of a LiNbO₃ waveguide optical modulator is measured using a two-tone technique. It is then used in both fundamental and second-harmonic generation modes as a calibrated source to measure a photoreceiver's frequency response.

We report the first comparison of high speed photodiode frequency response measurements up to 40 GHz between NIST and NPL in the 1.3 and 1.5 micrometers wavelength regions. This comparison is an important step in establishing international agreement on photodiode response measurements, with traceability to international microwave power and DC current standards. Measurements at NIST used a Nd:YAG heterodyne system. NPL used DFB heterodyne and integrated-optical modulator-based techniques. Measurements of a photodiode with nominally 20 GHz optical bandwidth show good agreement with average scatter of +/- 0.15 dB (2(sigma) ) below 20 GHz, and +/- 0.30 dB from 20 to 33 GHz. The results diverge systematically above 33 GHz, due to calibration of the RF power sensors. Scatter in the data is well represented by the combined uncertainties of the measurement systems up to 33 GHz.

Physical operation of p4-n photodetector has been studied. Photodiode current in the on and off states are presented. The small-signal conductance and junction capacitance versus light intensity are shown. The conductance decreases with intensity due to an increase of photocurrent. The junction capacitance is insensitive to illumination at low intensity and increases significantly at high intensity due to modulation of photogenerated mobile carriers.

We propose a model for III-V compound PIN and MSM detectors which is both physically realistic and easily modifiable. The complete model consists of two components: the physical and circuit model. In the physical model an intrinsic device is simulated by solving Poisson's equation, the current continuity equations, and a rate equation for charged traps using MATHEMATICA. Carrier injection, charge-storage effects and fringing electric fields are also taken into account. The results can be presented either in the time or the frequency domain. These results are used to determine the parameters of a circuit model for an intrinsic device. Our circuit model was specifically designed to include all the above mentioned effects. The model was tested by simulating the performance of PIN detectors with various geometries in both SPICE and MDS. Agreement between the measured data and the simulated results was remarkable good.

We present in this paper the results of two different modelings of photodetectors. The first is based on FD-BPM and presents the study of a waveguide PIN photodetector structure grown on InP substrate. It gives the possible cut-off frequency for such a photodetector. The second one is a modeling of a conventional PIN photodiode including the effects of the external circuit. The goal of this modeling is to analyze the behavior of the photodetector under very high optical power. We considered the case of an optical signal sinusoidally modulated at 20GHz and analyzed the output photocurrent in order to give the limitations of the photodetector.

We report on theoretical and experimental studies on the microwave passive reactive matching of high speed photodetectors. Such matching networks can substantially improve the power transfer of a microwave optical link. As an example, we relate two demonstrations which have been made, at a central frequency of 2 GHz, using both commercially available pigtailed PIN photodiode and laboratory made MSM detector. In both cases, matching networks are fabricated using transmission lines and comparison with unmatched devices is made in order to determine the improvement of the power transfer. Trade-offs between working frequency, bandwidth and available improvement in power transfer are presented taking into account parasitics value and the intrinsic capacitance of the detector. The extent of such a matching scheme to monolithic integration in order to increase either working frequency or power improvement is also reported as conclusion.

This paper presents a theoretical analysis of a series tuned optical receiver front-end design based on a PIN diode and MESFETs/HEMTs for wide-band multichannel lightwave transmission systems. A simple circuit model of this series tuned optical receiver front-end is presented and the method followed in this paper uses basic circuit theory to derive analytical expressions in terms of the elements of an equivalent circuit for characterizing the front-end. This paper computes the transimpedance, equivalent input noise current density and peaking frequency of the front-end from known physical parameters of the commercially available photodiodes and GaAs MESFET/HEMT, InP HEMT. Low noise is achieved for an appropriate choice of device parameters. The results obtained from the analysis presented in this paper are used in calculating power penalty and number of channels at various bitrates to optimize the performance of a multichannel excess shot noise limited coherent optical communication system.

The current trend of increasing data rates for fiber optics communications systems has created a demand for high-speed lightwave measurement instrumentation. For time domain measurements, a key element is a lightwave receiver for converting the optical signal to an electrical signal which can then be analyzed by conventional methods. The requirements for a high-speed lightwave receiver include DC-coupling, sufficient bandwidth to accurately reproduce the optical signal, frequency response flatness, sensitivity for measuring low signal levels, and linearity to avoid signal distortion. A receiver has been designed within Hewlett-Packard to meet these requirements for data rates up to 2.5 Gb/s. The receiver design consists of a high-bandwidth InP/InGaAs/InP p-i-n photodiode and a GaAs transimpedance amplifier. The photodiode has a bandwidth of 32 GHz with a responsivity greater than 0.5 A/W. The transimpedance amplifier has a gain of 600 ohms, flat frequency response, and a bandwidth of over 7 GHz. The combination results in a DC-coupled receiver with a bandwidth of over 4 GHz and a conversion gain of 330 V/W. The receiver provides accurate measurement capability for optical transmitters for both SONET and fibre channel communications systems.

We have investigated monolithic pin-FET photoreceivers in both the InP and GaAs systems. The GaAs-based circuits consisted of a single growth step in which the p-i-n diode was grown on top of the MESFET. The circuits exhibited flatband gains as high as 17 dB and bandwidths of 2.0 GHz. The InP circuits featured regrown MODFETs integrated with p-i-n diodes. These devices exhibited a gain of 17 dB and a bandwidth of 10 GHz.

A 6.6 Gb/s, 1:8 optoelectronic (OE) time demultiplexer was implemented with InP metal-semiconductor-metal switches. These switches were activated sequentially by <25 ps laser diode clock pulses distributed via an optical fiber splitter and delay lines. The demultiplexer capacitive feedthrough was much larger than the switch photoresponse. Therefore we applied a dc bias to obtain a detected clock pulse comparable to the capacitive feedthrough pulse, each alone being an output 0 bit and both in coincidence being a 1 bit with twice the amplitude of a 0 bit. The S/N was 28.5 dB.