Semiconductor laser amplifiers (SLAs) are multi-functional devices, as they can perform the following functions: amplification, modulation, detection and wavelength conversion. For the optimization of SLAs, a good numerical simulator is required. This gives insight into both the internal behavior of SLAs and the variation of certain system parameters due to changes in the SLA structure or external conditions. This paper describes an extensive numerical model for SLAs. It includes large-signal time dependent analysis, residual mirror reflection, amplified spontaneous emission (ASE) and multiple electrodes. SLAs with active layers made of bulk or MQW material can both be simulated. Important parameters such as the bit-rate, amplification, noise figure, chirp, ASE-spectrum and polarization-dependency can be obtained. It is also possible to use the simulator to investigate the influence of the accumulation of forward and backward propagating ASE in cascaded SLAs. The simulator is used for the design of a fast external modulator and for a wavelength convertor at 1.55 micrometers . Further, a multi-functional two electrode polarization-intensive amplifier-modulator- detector structure is proposed and optimized. This SLA can be used as an insert-drop transparent node for application in LANs and access/distribution networks.

The optical fiber amplifier is a device where signal light near 1.55 micrometers is amplified by stimulated emission of the exited erbium ions as the signal passes through the erbium doped fiber. The critical properties of erbium doped fiber amplifiers are the output power, gain, and noise. These amplifier parameters should be measured as a function of wavelength. For this reason we have constructed a measurement system that consists of a tunable signal source, fast chopper, detectors, and microcomputer. This amplifier measurement system can be used to characterize amplifier properties as a function of wavelength and input or output powers.

Calculation of distortion performances of a given DFB laser for different frequency plans is a well-known problem when working with CATV AM-VSB optical links. One generally takes safety margins or performs measurements on the lasers with the requested frequency plans when it is possible. After investigating the most important theories proposed in the scientific literature, we have come to the conclusion that laser distortion can be described as the weighted contribution of two different distortion sources: leakage current and intrinsic distortion due to the laser oscillation mechanism itself. On the basis of the two preceding ideas we predict the evolution of the second-order two-tone intermodulation (IMD2) response as a function of frequency and operating point by cascading Volterra models of the laser leakage currents and rate equations. The total composite second order (CSO) distortion can then be predicted with a `10.log' adding law and compared to distortion measurements. Regarding the experimental set-up, the optical modulation depth per channel has to be properly chosen in order to minimize the influence of optical reflections and reduce the clipping effects.

In this paper, we present the optical behavior of waveguide PIN photodetectors for millimeter wave applications. The side illumination of these devices allows us to overcome the problem encountered with classical PIN top illuminated photodetectors, which is the compromise between high cut-off frequency and high responsivity. This is the reason why we have modeled PIN waveguide photodetectors grown on InP substrate. Our modeling is based on 3D and 2D FD Beam Propagation Method to describe the propagation of light in the photodetector and on the solution of classical semiconductor equations to describe the electrical behavior of the device. The model is applied to InP P+/GaInAs N-/InP N+ and InP P+/GaInAsP P+/GaInAs N-/GaInAsP N+/InP N+ structures. Cut-off frequencies up to 90 GHz can be obtained for very small devices, typically 12 micrometers for the device length and 5 micrometers for the rib width with 0.3 micrometers thickness of GaInAs absorbing layer, by neglecting parasitic effects due to boundary pads. This is also valid for structures with smaller ribs using a constant surface area. The external quantum efficiency of such a device is strongly dependent on the device structure (GaInAsP thickness, monomode or multimode structure), and also on the conditions of injection of light (width and position of the optical spot, angle of the optical beam with the device). A complete analysis of the quantum efficiency versus the influence of GaInAsP thickness, device length, GaInAs thickness, and optical injection has been performed. It was found that, using lens ended optical fiber, multimode waveguide structures are better devices compared to monomode ones, and can lead to quantum efficiency higher than 90%.

We report on the design, growth, fabrication, and characterization of monolithic wavelength division multiplexed (WDM) modules produced within the Canadian Solid State Optoelectronics Consortium. The transmitter module includes multiple, discrete wavelength, distributed Bragg reflector (DBR) laser diodes monolithically integrated with waveguide combiners fabricated using an InGaAs/GaAs heterostructure. The wavelength demultiplexer unit is based on a Rowland circle grating spectrometer monolithically integrated with a metal- semiconductor-metal (MSM) detector array fabricated on an InGaAs/AlGaAs/GaAs heterostructure. The epitaxial layer wafers for both transmitter and receiver modules were grown in single molecular beam epitaxy (MBE) runs.

The feasibility of self-tuning WDM technique has been experimentally evaluated and demonstrated on planar lightwave circuits. This techniques utilizes only one optical window a simple filter operated as not only WDM-MUX/DMUX but also a tuning gauge, and adopts no designation of specific emission wavelength. It is applicable for both DFB-LD and FP-LD to be the optical source of high-bit-rate transmission systems by designating appropriately the free spectrum range of the WDM filter. Proposed tuning control algorithm with a master/slave scheme secures the stable operation under a wide range of environmental temperature variation including various aging effects. A highlight of this paper is the proposed WDM system's capability to settle the defects of conventional WDM systems.

This paper gives a review of the state of the art of the main optical technologies suited to optical add-drop multiplexers (OADM) in a WDM ring network with their respective potential advantages and level of maturity. Experimental results of a four port multilayer OADM are given. Then, experimentations including these filters are described with the corresponding results we have obtained. Further prospects are given for evolution and growth.

At present, most of the proposed architectures for interconnecting nodes in processor networks are based on graph topologies. In this paper we study topologies based on the hypergraph theory, taking advantage of the wavelength division multiplexing (WDM) that allows the huge bandwidth of optical fiber to be divided into several high-speed logical channels. We show that our topologies compare very well against graph-based ones, in terms of maximum hop distance, expandability, and connectivity. Furthermore, they use a constant number of transceivers per node, and a sub-linear number of multiplexed channels. We also give schemes for some global routing operations: the one-to-all and the all-to-all broadcast, analyzing their performance in hops.

A WDM system has been realized, intended for digital broadband (e.g. video) distribution in an optically splitted subscriber access network. Four standard 565 Mbit/s optical transmitter units have been equipped with wavelength-matched 1557 nm DFB laser diodes. The four transmitters are wavelength-locked to a temperature-stabilized fiber Fabry-Perot (FFP). The lasers are FSK modulated with a tone spacing of 6.7 GHz; the 565 Mbit/s input signals are 7B8B coded to avoid problems with the laser's non-linear FM response. Another tunable FFP precedes a standard direct-detection receiver for both wavelength channel selection and FSK demodulation. A wavelength control circuit (WCC) ensures locking of the FFP to the correct tone of the selected channel. The WCC continuously maximizes the average FFP output power, by determining the time-derivative of the mean photodiode current. The main advantage of this approach is the absence of a dithering signal, which interferes with the data signal. BER curves have been measured for each of the four channels and compared with theory, discrepancies are < 0.75 dB. A net system power budget of > 28 dB for each channel has been achieved, and no evidence for BER floors has been found down to a BER of 10^-11.

In this communication we report the design of a clock recovery circuit produced for the 20 Gbit/s demonstrator of the RACE 2011 project `TRAVEL' of the European Community. The clock recovery circuit is based on an open loop structure using a dielectric resonator narrow bandpass filter with a high quality factor. A detailed electrical characterization of the circuit and also its sensitivity to temperature and detuning variations are presented. The experimental results show that the circuit is a very attractive solution for the forthcoming STM-128 optical links.

This paper presents a novel design of a burst-mode transceiver with local real-time intelligence for point-to-multipoint upstream transmission at 155.52 Mb/s. Using embedded intelligence for controlling laser drive parameters, and a trailer canceling technique at a receiver, excellent combined performance of high sensitivity (-37 dBm at BER 10^-9), large optical power budget (36 dB), wide dynamic range (33 dB), and short run-in time (26 ns) was successfully demonstrated. It can connect up to 32 subscribers through a passive optical tree network using a burst TDMA scheme.

This paper describes an eight channel optical fiber interconnection module based on linear array of lasers and photodiodes and multimode fiber ribbon. From the experimental results the skew was estimated to be lower than 5 ns. No error was observed during a long term test. A twenty-two channel optical interconnection module was implemented based on the eight channel module and it was installed in a TROPICO switching system with no degradation in the performance.

The performance of bandpass optical receivers can be improved using a lossless equalizer between the photodiode and the amplifier, as a consequence of the mismatch between the source and the input section of the amplifier. We propose two design approaches based on noise matching and on power matching, as well as a performance evaluation criterion to make comparisons between these different techniques. We have found the limit of noise performance of the approaches at a single frequency, and an upper bound, derived from the Bode-Fano limit, in a wide-band application for the power matching approach.

The following technique compares GaAs monolithic processes regarding the requirement of transimpedance amplifiers to be used in optical front-ends. This technique uses a lumped element model of the forward amplifier to establish relationships between the small signal model of the devices which form the forward amplifier and the closed loop performance, as well as the root locus representation of the loop gain transfer function to set clear constraints to the frequency response. The comparison is performed by establishing a set of constraints to ensure the effectiveness of solution found and to reduce the number of degrees of freedom in this problem. This comparison has been carried out by taking into consideration the presence of the parasitic capacitance of the photodiode which plays an important role in the bandwidth performance.

In this paper, a balanced optical front-end and IF module for a coherent heterodyne optical receiver are presented. The balanced optical front-end is realized with two commercially available GaInAs/InP p-i-n photodiodes and a full custom designed GaAs MMIC transimpedance amplifier. In this front-end a transimpedance, 3 dB bandwidth, and bandwidth- average input-referred noise current density of 58 dB(Omega) , at 1.7 GHz through 6.3 GHz, and 10 pA(root)Hz, have been measured. The IF module consists of three amplifier stages, a variable attenuator, a pass-band filter, and a broadband coupler. The amplifier stages and the variable attenuator are realized with commercially available MMICs chips. The filter and coupler have been built in hybrid technology. The measured characteristics are 47 dB gain with +/- 1.5 dB ripple, and more than 12 dB input/output port return losses. A transimpedance of 104 dB(Omega) has been measured in this complete optical receiver.

Nonlinear optical fiber loop mirrors (NOLM) are attractive devices for multiplexing optical signals with bitrates of up to 100 Gbit/s and beyond. So far, less attention has been paid to the relation between design of the demultiplexer and the resulting system performance characterized by parameters like input power penalty and wavelength tolerance. Based on our own measurements of a polarization-independent 40 Gbit/s demultiplexer and in conjunction with a new theoretical approach modeling NOLM transmission and worst-case input power penalty of the demultiplexer, a device capable of 100 Gbit/s operation is analyzed. The survey illustrates the potential and trade-offs of the NOLM concept assuming a limited cw power of the switching wave. E.g., the calculations show that a 14.5 dBm booster EDFA together with a loop fiber length between 1 and 2 km allow low penalty drop rates (< 6 dB worst-case penalty) of the demultiplexed channel between 1.3 and 8.8 Gbit/s. Then 100 Gbit/s operation should be possible within a wavelength range of at least 5 nm.

Reflectively tapped fiber-optic delay-line matched filters with metal film reflectors evaporated on cleaved fiber ends have been fabricated. This novel device can provide ultra-high speed signal processing capability, small size, and low loss, and be used as an optical packet address detector for all-optical wavelength division multiplexing switching networks. Using the filters, detection of 2.5 Gb/s, 6-bit packet addresses for optical ring networks has been successfully demonstrated. Experimental and calculation results on correlation outputs of the filters are reported.

The alignment of laser diodes to monomode fibers has to meet extremely close tolerances for a low coupling loss. Typically < 0.5 micrometers in lateral and vertical direction and less than two degrees in angle deviation are allowed for a coupling loss below 2 dB. Presently such close tolerances can only be met by gluing or soldering both components on separate base plates and combining them via piezoactivated alignment monitoring the output of the circuit and then gluing them using UV-hardening epoxies. Such a procedure is not very economical and not useful for mass applications. This paper presents the principle and realization of a silicon micro-optical bench for laser-fiber-coupling, which avoids the above mentioned disadvantages. The micro-optical bench is realized using well controlled plasma etching processes to transfer the guiding patterns for the laser and the fiber into the silicon substrate, keeping geometry tolerances below +/- 0.5 micrometers in lateral and vertical direction. Mounting the laser diode by means of a self-aligned soldering process, an additional contribution to the precise alignment of the laser is further improved.

Polarization fluctuations are one of the obstacles in coherent fiber-optical transmission systems. Active polarization control is the most favorable method of matching the polarization of both the signal from the transmitter and the local oscillator. The author presents a new endless control scheme which makes reset procedures of finite range polarization controlling devices obsolete. The principal idea is to maximize the mean-square distance between the retarder drive signals and their physical limits in each operation step. Consequently, the drive signals are smoothly decreased to their center retardation when they approach the outer boundaries of their physical range. Calculations proved the worst-case receiver sensitivity degradation to be less than 0.2 dB, owing to SOP-mismatch. In most cases, the resulting degradation is less than 0.1 dB, corresponding to an error angle of about 3 degree(s) in Poincare- sphere representation.

This paper reports how to realize a new type of tunable filter using the Faraday effect. The properties of a Fabry-Perot etalon whose cavity is made with a Faraday rotator are recalled. The analysis of the polarization state of the emergent light permits us to find the configuration of the device to realize such a tunable filter. This latter finds application in wavelength demultiplexing.

Acousto optic filters (AOF) with optical size approximately 8 - 10 mm for visible and ultraviolet spectrum on the base of TeO₂ and KH₂ PO₄ crystals correspondingly are developed and manufactured. The spectral resolution is better than 1 nm. The frequency bandwidth of both filters is 55 - 90 MHz, that allows us to use the only high frequency generator with two different power amplifiers. Acoustic power level for the AOFs on the base of TeO₂ and KH₂PO₄ was about 1 W and 5 W, respectively. The optical tuning is carried out by means of a computer program.

An approach to the design of all-optical universal integral-optical logical elements (UIOLEs) for digital optical systems of data processing and transmission based on effects of changing of reference radiation spreading direction under the influence of control radiation is examined in this article. The article presents the possibility of construction of multichannel universal logical elements with high-speed programmed tuning on a realized Boolean function. Their throughput may be up to 100 billions oper./second and higher.

Variable delays in fiber optic optical transmission have been achieved by using photochromic- doped sol-gel silica glasses. Gaps between two commercial fibers in V-groove connectors, when filled with these materials, produce variable propagation times. Variations of the optical propagation time depending on the input power, wavelength, and dopant concentration are shown.

In many amplification regimes, the EDFA gain is saturated by noise. The accurate computation of noise spectra is then necessary to predict the EDFA exact gain and noise characteristics. This is done by solving the rates equations (ordinary differential equations, ODE) describing the noise spectra. The first way to optimize the EDFA modeling tool is the noise spectrum discretization. Such a method is detailed in this paper. Another interesting approach has been given by E. Desurvire. One ordinary differential equation (ODE) describes the whole noise spectrum in one direction. This is the usual concept of noise effective bandwidth (NEB). The NEB definition we use has been modified in order to take into account the spatial distribution of the light power in the active fiber section and the population inversion. The NEB method we propose uses 1, 2, or 4 NEBs, leading to small computation times (approximately equals 4 s). The experimental set-up allowing the determination of the input parameters is also detailed. The predictions provided by one modified NEB are at least better than the results obtained with 4 standard NEBs, as defined by E. Desurvire. The computations fairly agree with experimental results: the larger discrepancy is lower than 0.7 dB. Prototype performances are discussed.

A theoretical model based on propagation-rate equations has been developed for modeling the continuous wave operations of Er³⁺/Yb³⁺ co-doped fiber lasers. It includes uniform upconversion mechanisms from the ⁴I_13/2 and ⁴I_11/2 erbium levels and describes the coupling between Er³⁺ and Yb³⁺ ions through a pair-induced energy transfer process. The nonlinear differential equations of the model are numerically solved by means of a Runge-Kutta based iterative procedure. Numerical results, based on realistic input fiber parameters, show that slope efficiencies close to the quantum limit with threshold pump powers of few mW, can be achieved in short Er³⁺/Yb³⁺ co-doped fibers equipped with input dielectric mirrors and output grating Bragg reflectors. Furthermore, we show that single frequency cw operations can also be performed, with lower efficiency, in Er³⁺/Yb³⁺ co-doped Fabry-Perot fiber microlasers

At the present time, standards are not yet available to characterize EDFA performances. Several methods have to be tested to provide an accurate and reproducible measurement technique. This paper presents such an investigation for optical and electrical measurement methods. In the first part, the interpolation and polarization nulling optical techniques are presented. The experimental set-up is described in detail. These methods have been applied to measure the gain and the noise figure of EDFA. These measurements have been carried on a commercial EDFA. Meaningful discrepancies have been highlighted. These discrepancies have been analyzed. More especially, the insertion of a narrow band optical filter and the polarization dependence of noise have been investigated. Conclusions have been drawn explaining the measurement results differences. In the second part, the electrical method is detailed and analyzed. The influence of the laser RIN on the noise figure measurement is put into light. The signal spontaneous noise becomes a major contribution and care has to be taken to avoid it. Optical filtering of the signal is then highly recommended. Finally both electrical and optical methods are compared. The errors originating from respective methods are highlighted.

The erbium doped fiber amplifier (EDFA) is presently the key element of a long distance transmission. Experimental results of noise characteristics (1 Hz <EQ f <EQ 500 kHz) of an Er³⁺ erbium doped fiber amplifier pumped by 1.48 micrometers laser diode are reported in this paper. Several kinds of noise generated in EDFAs (1/f^(alpha ) noise, 1 <EQ (alpha) <EQ 3; hopping noise, quantum noise) are very troublesome in the various applications of the EDFA. These noise levels impede to improve the optical coherence and the signal to noise ratio. Several studies have been made in order to introduce the low and medium frequency noise for transmitters but nothing yet concerning the EDFA. This communication deals with optical noise simulations of an EDFA. The output noise depends on the amplified spontaneous emission (ASE) power because no input signal is applied. Thanks to a noise simulation, we have been able to deduce the internal gain G of the doped fiber and the spontaneous emission factor n_sp. The correlation between the monitoring photocurrent noise, the output noise and the input noise of the amplifier are given in order to show the influence of the pump laser on the EDFA output and input.

The vectorial beam propagation method has successfully been applied to a passive polarization converting waveguide structure. A complete polarization conversion has been simulated. The propagating fields are calculated and the power attenuation is evaluated. The influence of structural changes of the device on the polarization conversion is investigated.

The control of the group velocity dispersion (GVD) in the waveguiding structures may be of crucial interest in optical communications. It may be achieved controlling one of the components of GVD, the waveguide dispersion, which depends on the geometry and the refractive index profiles of the guiding structures. In this paper the waveguide dispersion in practical guiding structures is evaluated by means of a general numerical procedure, based on a FEM formulation and a second order perturbation of the corresponding eigenvalue equation. The evaluation is performed for various values of geometrical and physical parameters of the guiding structures, in order to seek for a best structures configuration, in which waveguide dispersion may compensate the dispersive effects of the optical materials in a given frequency range. Particular attention has been devoted to the possibility to control the sign of GVD in the single mode region which is of great concern for the design of all optical components.

In this paper, the fabrication and characterization of polymeric thermo-optic (T/O) single switching elements and (4 X 4)-switching matrices are described. The first loop experiment with cascaded transparent OFDM crossconnects including a polymeric (4 X 4)-switching matrix is reported. In order to reduce insertion loss, power consumption and crosstalk, the optimization strategy for the polymeric T/O switches is discussed.

Asymmetric Fabry-Perot (AFP) reflection modulators containing thin film electro-optic polymeric materials are described and designed for operation at a low drive voltage. Prototype device structures have been fabricated using a crosslinked electro-optic polymer, optical modulation of 43% has been obtained at an operation wavelength of 633 nm and at a rms drive voltage of 88 V. A 4 X 4 array of devices has also been fabricated and characterized. The uniformity of the array was evaluated and a contrast ratio exceeding 5:1 can be obtained simultaneously at each of the 16 device positions.

Using a grating coupler technique, we report on accurate determinations of bulk and mode refractive indexes in the 1.5 micrometers wavelength region. On the one hand, InP bulk refractive index is measured within a relative accuracy of 10^-3 as a function of wavelength range, temperature, and for two doping levels. On the other hand, quantum effects are evidenced on index measurements using a GaInAs multiquantum well waveguide. As a result, it is shown that the guided mode birefringence is strongly enhanced as predicted from energy levels and selection rules of fundamental electronic transitions.

The guided and leaky mode propagation in a multilayered c-rotated anisotropic waveguide having an equatorial dielectric tensor configuration is examined. In particular, the electromagnetic behavior of titanium diffused lithium niobate (Ti:LiNbO₃) and proton exchanged lithium niobate (PE:LiNbO₃) waveguides is compared. The dependence of modal characteristics on the equatorial angle (gamma) is shown and the effects of different metal layers are investigated. The transition from leaky to guided modes occurs only for the PE:LiNbO₃ waveguide while the modes do not change their nature for the Ti:LiNbO₃ waveguide. The number of guided modes, for the PE:LiNbO₃ waveguide, increases by increasing the angle (gamma) . The first guided mode, having the real part of the refractive effective index n_r equals 2.2701, appears at (gamma) equals 25 degree(s), while, for (gamma) > 73 degree(s) nine different guided modes exist.

Polymeric materials such as poly[diethylene glycol bis(allyl carbonate)], simply called CR-39, subjected to (gamma) -rays or ion beams (H, He, Li...) undergo structural, physical and chemical modifications. In the case of ion beam, these modifications lead to a sufficient and localized increase of the polymer refractive index to achieve optical waveguides. The first CR-39 polymer studied at the laboratory was a commercial one, we recently fabricated our own material dedicated to our work, thanks to a collaboration with the Polymer Group (University of Bristol, U.K.). Consequently, its composition is mastered, notably by the concentration of the initiator used during the polymerization process. Then we showed that the refractive index of the substrate can be controlled by the curing conditions. The losses of the waveguide fabricated by irradiation have been systematically measured in order to apply this technique to integrated optics (microstructure fabrication). The device used has been recently improved and it is based on a CCD camera related to an image analysis system and an exponential regression calculation program. The values are obtained with an accuracy of +/- 0.5 dB/cm and they remain too high (2 to 4 dB/cm) for a lot of applications. That is the reason why we studied the behavior of the waveguide under thermal treatments with temperature around the glass transition temperature (Tg) of the unirradiated CR-39 polymer. These experiments showed that, on the range of Tg - 10 degree(s)C, Tg + 10 degree(s)C, the attenuation after the annealing treatment remains equal to its original value. They also pointed out the thermal stability of the refractive index, then we can conclude that the radiation induced modifications have stable structural consequences.

By time domain measurements on a carrier injection switch using Ge-indiffused waveguides in silicon the plasma dispersion and the thermo-optical effect were separated and quantified. Optimized thermo-optical switches in silicon-on-insulator with low switching powers are also presented.

In the course of the growing complexity of computer and communication networks and the corresponding high data rates which have to be distributed within isolated and between communicating systems, optical fibers and the corresponding optoelectrical receivers are widely used. Until now these components are usually added to the IC-network. With the increasing number of such OE-internets an integration of the fiber-detector coupling into the integrated circuits would be advantageous. This paper presents the realization and the results of a fiber-detector coupling structure fabricated in a technology which is compatible with C- MOS or BIC-MOS technology. The deep (approximately equals 65 micrometers ) trench for the fiber is fabricated by an anisotropic plasma etching process using a photo resist mask. This process has an accuracy of +/- 0.5 micrometers in lateral and vertical direction, which is necessary for the coupling of monomode fibers, and an etching rate of up to 15 micrometers /h. The PIN-photo diode can be fabricated both laterally and vertically orientated as part of standard C-MOS, BIC-MOS or bipolar processes. The diode exhibits a cut-off frequency exceeding 400 MHz at a capacitance of less than 1 pF and a high coupling efficiency of more than 70% for wavelength below 1.1 micrometers due to a lateral injection of the photons into the active area of the detector.

An adiabatic 3 dB-coupler is realized on InGaAsP/InP for the first time. The device consists of an asymmetric and a symmetric Y-junction, connected by a bimodal waveguide. The operating principle of such a device is not based on interference effects, but on evolution (or adiabatic propagation) of modes in the Y-junctions. This allows a tolerant fabrication and a low sensitivity to operating conditions, wavelength, and polarization. In general mode evolution components however tend to be rather long. The length of our coupler is reduced with respect to previous adiabatic couplers by using an optimized design for the asymmetric Y-junction, which contains two sections with different branching angles. This design permits a total length of the coupler of 3 mm. The device is realized using a double masking technique in order to obtain high quality splitting points in the Y-junctions. Experimentally a coupling ratio of 3 +/- 0.3 dB was found, independent of the length of the bimodal waveguide and the polarization, at a wavelength of 1.5 micrometers . The excess loss is below 1 dB.

We present in this paper recent results obtained in the field of bidirectional transmissions, coherence multiplexing, and integrated optics. We report the description of a new kind of Mach-Zehnder interferometer integrated in lithium niobate (LiNbO₃), which is the key device of a coherence multiplexed bidirectional link. The orientation of the waveguides, parallel to the Z-axis of LiNbO₃, yields a polarization independent behavior. The working conditions of the modulators and the results obtained as they are used in a full-duplex bidirectional link are presented.

A Ti:LiNbO₃ acousto-optical TE-TM converter followed by a 45 degree(s) linear polarizer allows us to measure the Stokes parameters. A wavelength-selective demonstration unit with 9 MHz electrical measurement bandwidth, roughly 0.04 rad accuracy and 250 GHz optical bandwidth has been developed. An acousto-optical TE-TM converter whose transducer is placed at 1/3 of the interaction length forms a polarization-independent optical depolarizer, suitable to suppress polarization hole burning in long haul EDFA links. We present a double- stage depolarizer with