FR2830088A1 - Electro-optical modulation method for e.g. telecommunication signals applies indexed control signal to electro-optic modulator and samples light at output to update reference value - Google Patents

Abstract

An indexed control signal is applied to an electro-optical modulator (110). Part of the modulator output is fed back through a sampler (400) and compared (210) to a reference value, which is modified. The method involves in converting the sampled light into a digital signal, and in digitally analyzing the signal (220) for updating the reference value (Vdc).

Description

<Desc / Clms Page number 1>

 The invention relates to electro-optical modulation in the telecommunications field.

- de nombreux notamment les formats numériques et le codage des données ; de nombreux progrès, tels ceux obtenus sur les paires de cuivre, permettent de transporter davantage d'informations sur les supports classiques et accroissent ainsi la capacité des réseaux. The development of telecommunications and computer networks brings with it an increasing need in terms of speed of information exchanged between users by means of computers. The most significant example on this subject is the extraordinary development of the Internet which accompanies the increase in the use of personal computers as tools of daily life. These flow requirements also correspond to the number of interconnections, since the number of users is constantly increasing. Technologies are evolving rapidly,

- many including digital formats and data coding; numerous advances, such as those obtained on copper pairs, make it possible to transport more information on conventional supports and thus increase the capacity of networks.

 The challenge that arises is then to determine which moderate cost technology can continue to meet the increasing demand in terms of throughput of an increasing number of users and continuously replace the existing technology. A variable remains to be evaluated, that of the real need of the user in terms of throughput. Will a speed of 100 Mbps remain sufficient for everyday applications? In fact, it should not be forgotten that the generalization of digital signal processing techniques standardizes and popularizes data compression methods, whether multimedia or file type. It is recognized that the use of optical fiber takes all its quintessence for the network of speed higher than 1000 Mbps. Thus, the advisability of the generalization of optical fiber as a physical medium will flow directly from the speed actually necessary.

 The use of multimode optical fibers (glass or polymer) with an index gradient offers interesting prospects in terms of ease of

<Desc / Clms Page number 2>

 connection inside buildings, but their use will probably be limited to a few hundred meters, either because of attenuation or because of the bandwidth limitation due to chromatic or modal dispersion effects. Such a multimode optical network should moreover be connected to an optical network based on single-mode optical fibers such as those used for long-distance optical transmissions at very high speed (> 2.5 GHz).

 This growth in throughput will however come up against the physical limit constituted by the bandwidth of copper devices which, associated with components based on silicon technology, ensure the moderate cost of current installations and lead to their expansion. The development of digital signal processing techniques (implemented in silicon) allows, in part, to push this limit. This is what we see daily when we use our modem to connect, via the PSTN telephone network (switched telephone network), to our Internet service provider. Historically, the first consumer modems used frequency hopping modulation (FSK modulation - frequency hopping modulation) according to the recommendations V21 & V23 of the UTI (international organization which manages telecommunications standards). This technique allows digital data to be transmitted at a rate ranging from 75 to 1200 bits / second. The use of m-state phase modulation (BPSK modulation-binary phase jump modulation or QPSK-quadratic phase jump modulation) increased the bit rate to 2400 bits / second. As these techniques were not sufficient, quadratic amplitude modulation (MAQ-Quadratic Alumina Modulation 16.34 or 256 states) was used in order to push back the speed limitation to 56,000 bits / second. These developments demonstrate that for the same physical medium, intrinsically limited in bandwidth, the implementation of digital signal processing algorithms makes it possible, in a non-negligible ratio, to increase the bit rate. The latest example is that of ADSL (Asymetric Data Suscribe Line) which allows

<Desc / Clms Page number 3>

 a speed of 2Mbit / s without changing the physical medium of the subscriber's line, that is to say that the information is transported by copper. At the same time, the so-called local access networks (LAN-Local Access Network) saw their throughput greatly increase, and in much higher proportions than those of modems. Thus, the physical Ethernet network has moved in a few years from a speed of 10Mb / s to that of 1 Gigabit / s. In local broadband networks, the physical transport of information over long distances (> 100 meters) is mainly done by optical fiber. Thus for the LAN network, it is to this day acquired that broadband and optical fiber go hand in hand because, the threshold of 100 meters being very quickly reached in a configuration of local business networks, optical fiber has immunity to noise unmatched.

 For these local networks, another option is available to the user to date: WAN networks (Wireless Area Network). Local loops are indeed developing in order to offer an alternative to ADSL; the proposed digital bit rate remains around 2 Mbit / s. The second WAN network to develop is the Bluetooth network intended for indoor local micro networks)>. With a limited range (a few tens of meters) and a flow rate of less than 2 Mbit / s, this solution will, for broadband, be ineffective indoors as it will require the use of frequencies above 20 GHz (2.4 GHz today) emission powers prohibited by law. Other solutions exist and are the subject of very active studies.

The use of radio frequencies for end communications allows the mobility of the terminals within a reception cell whose variable size depends on the type of user, but can be reduced to the scale of a single element in the pico-cell case. The signal bandwidth remains limited at present for this type of communications, but the evolution of standards and formats leads to a multiplication of uses which amplifies the demand for speed elsewhere (third generation mobiles to UMTS standard by example).

<Desc / Clms Page number 4>

 The extension of optical networks based on single-mode optical fiber is currently limited, despite its rapid growth, by the costs of connection terminals: they require the use of components and fast electronics based on semiconductor materials III- V.

 The devices currently used for transmission in the field of telecommunications over optical fibers are semiconductor lasers. Light-emitting diodes, taking into account their spectral spreading and the resulting multimode characteristics, are used little or very little. There are two types of laser modulation: direct and external modulation.

 Direct modulation is the easiest to implement, however the variation in the concentration of carriers in the laser introduces variations in optical index leading to a widening of the unit digital bit and a degradation of the error rate of the optical system.

 One of the ways to get around this dysfunction is to use external modulation. In this case, the laser is continuously polarized and a modulator performs a light chopping function at the rate of the binary pulses which are transmitted to it. There are several types of modulators, with different associated operating principles. However and in all cases the choices of operators and equipment manufacturers are guided by the performance / cost ratio. The parameters entering the cost structure are essentially due to the manufacturing processes used for the optical transmission module, ie its level of integration. The monolithic integration on an InP substrate of the laser and the modulator is the most complex to achieve, it is however essential for very high bit rates (from 10 Gbits / s). However, at this level, the additional cost is not a real obstacle to the deployment of a transport network. For lower speeds, in particular covering metropolitan networks and local loops where cost is a key element in the decision, less sophisticated solutions are required, based on simplified processes and assemblies.

Hybrid modules are then considered. Hybridization means association

<Desc / Clms Page number 5>

 the laser, the modulator and the latter's control electronics on a host substrate (there are several varieties: alumina, teflon, duroid, etc.). Each component is manufactured in different ways, with the equipment supplier ensuring supply and final assembly. In addition, the possibility of repairing the module is guaranteed at the level of the unitary component.

 It is at this level that the introduction of the modulator based on electro-optical polymers takes all its significance. The synthesis of the material, the low number of manufacturing steps and their scheduling make it a low-cost component capable of driving the speeds of 100 Mbits / s estimated at the local subscriber over the horizon of 10-20 years.

 A first object of this invention is to provide an interface for optical networks, in particular single-mode networks, which provides a compromise both in terms of speed and costs, and which is fully compatible with the data coding formats used on networks in copper or in computer networks.

 Time stability of performance should also be addressed. It is a second object of the invention.

 These objects are achieved according to the invention by means of an electro-optical modulation method in which a control signal indexed to a reference value is applied to an electro-optical modulator, a process in which light is taken from the output of '' a modulator subjected to this reference value and a signal is deduced from this sampled light which is analyzed to update the reference value, characterized in that the sampled light is converted into a digital signal, and digitally produced on this signal the update analysis of the reference value.

 An electro-optical modulation device is also proposed according to the invention, comprising at least one light modulator, means for supplying a control signal from the modulator indexed to a reference value, means for taking the light at the output of a

<Desc / Clms Page number 6>

 modulator subject to this reference value, means for analyzing a signal deduced from the light sampled and deducing from this analysis an updated reference value, characterized in that the sampling means and the analysis means are provided for converting the light taken from a digital signal, and provide the updated reference value using a digital analysis of this converted signal.

 Other characteristics, objects and advantages of the invention will appear on reading the detailed description which follows, made with reference to the appended figures in which: - Figure 1 shows a coding and stabilization system of the modulator according to a first variant of the invention, feedback by analysis of the coded signal; - Figure 2 shows a coding and stabilization system according to another variant of the invention, feedback by analysis of a test signal on an additional modulator, placed in parallel; - Figure 3 shows another variant with additional modulator this time placed in series; - Figure 4 is a functional block diagram of the additional modulator system of Figure 2; - Figure 5 is a functional block diagram of a calculation in a feedback loop, according to the invention; - Figures 6 and 7 are functional diagrams of a particular calculation of reference voltage correction, according to a variant of the invention; - Figures 8a, 8b and 8c show different implementations of an optical modulator part according to the invention, by means of mirror modulators; - Figures 9a and 9b show two alternative implementation of external modules, according to the invention; - Figure 10a shows an embodiment from a Mach-Zehnder interferometer with a single output;

<Desc / Clms Page number 7>

 - Figure 10b shows an embodiment of a modulator from a Mach-Zehnder interferometer with two complementary outputs; - Figure 10c shows an embodiment from an optical coupler.

 - Figure 11 shows an example of assembly for the test of electrooptical modulators. In the box are shown the modulation signal (bottom) and the detected optical signal (top). Ld: laser diode, PD: photodiode; FIG. 12 represents an example of drift of the operating point measured by the assembly of FIG. 11.

 - Figure 13 shows a block diagram of optical transmission using the FSK modulation format; - Figures 14a and 14b show examples of digital signal frames before and after transmission via the electrooptic modulator whose operating point is digitally controlled; - Figures 15a and 15b show examples of FSK coded signal frames before and after transmission via the electrooptical modulator whose operating point is digitally controlled.

 We will now describe a servo system, control, coding and stabilization of electro-optical modulators according to the invention allowing high speed coding of optical signals from low cost electronics.

 The applications targeted by this system relate to the field of telecommunications and computer networks. This interface is intended to be placed in the interconnection terminals of optical networks. It offers the possibility of using modem formats (modulator-demodulator) for telecommunications applications; it can especially be used as a low-cost modulation system in high speed local area network cards (> 100 Mb / s).

 The proposed solution is at the interface between two technologies:

<Desc / Clms Page number 8>

 - silicon technology with transmission via copper pairs, a low-cost solution with speeds limited to 2Mb / s as part of the most sophisticated solution currently, ADSL (Asymetric Data Suscribe Line).

 - technology using the transport of high-speed optical signals over single-mode optical fiber, the installation of which remains expensive for users by the use of more expensive fast electronics and laser sources with stabilized frequency (in the context of mutliplexing wavelength).

 The embodiment described here comprises a stabilized optoelectronic interface comprising at least one electrooptical modulation component whose operating point is stabilized by a low cost digital circuit.

 The advantages provided are behavior insensitive to the optical wavelength due to the use of electrooptical modulation.

This interface can be particularized for different optical frequencies of use in the case of application in a WDM (Wavelength Division Multiplex) network by the addition of suitable optical fibers.

 The use of a digital circuit to stabilize the operating point of the modulator ensures that the feedback circuit will not be the source of drifts.

 The algorithm used here for the stabilization of components allows real-time feedback on the component. It is particularly advantageous.

 Different implementation schemes of the invention are proposed and make it possible to adapt it to a wide variety of architecture of optical telecommunications networks.

 In the various variants, the optical interface stabilized against drifts comprises the association: of a block 100 for coding an optical wave (modulation block by linear electrooptical effect);

<Desc / Clms Page number 9>

 - digital electronics 200 for servo-control of the modulators' operating point in order to optimize their operation; - possibly optical sources 300 and detectors 400 integrated in variable number according to the implementations.

The optical coding block can have different operating modes:
In a first variant shown in FIG. 1, the interface comprises a single modulator 110. In this case, the control is ensured on the basis of the signal itself or of a low frequency signal superimposed on the useful signal.

More specifically, the signal to be encoded is mixed by a digital signal processor (DSP) 210 with an error correction signal, transformed into analog control and applied to an electrooptical component 110 which encodes an optical beam. Part of the optical signal is detected by the photodiode (PIN) 400, converted into a transmitted digital signal (feedback) and analyzed by the DSP 210 in order to generate the error correction signal;
In this single-element configuration, the feedback and the data signal are managed simultaneously by the DSP, which extracts signal distortions to stabilize the component. In this configuration, the bandwidth of the signal is preferably less than the bandwidth of the DSP, since it is the signal itself which makes it possible to generate the feedback controlling the drifts.

 In a second variant (FIG. 2), the interface comprises modulators 110, 120 associated in pairs (twin modulators): in this case the servoing is done from a test signal applied to a single 120 of the modulators. Only the continuous signal is applied to the second modulator 120, there is therefore very little disturbance of the ideal operation of this (no superposed modulation signal can induce additional coding error). This approach has the advantage of simply allowing low-frequency servoing of a high-frequency modulator.

<Desc / Clms Page number 10>

 Thus, a distinction is made here between two configurations in which the signal coding functions are separated and the drifts are controlled by the use, on the same chip, of two identical modulators, one of which is reserved for coding the data signal while the second is used for the control of daggerboards (Figures 2 and 3).

 In the first case (FIG. 2), the modulators are placed in parallel: an optical source 310 (LED) is coupled in the control modulator 120, while the detection is ensured by a photodiode 400 (PIN). The feedback electrode is common to the two modulators 110 and 120. The possible asymmetry of the two modulators 110 and 120 can be calibrated during manufacture and taken into account by the processor 210 (DSP). The twin modulators 110 and 120 are here placed in parallel for the control of the drifts and the coding of data. The upper modulator 110 codes the data optically. Its operating point is stabilized by the lower modulator 120 controlled by a local oscillator 140 and the DSP 210. The optical input of the lower modulator 120 is provided by a luminescent diode.

 In FIG. 3, the two modulators 110 and 120 are on the other hand placed in series, the first 110 has two outputs which makes it possible to extract the optical data signal with the minimum of losses, while the second 120 has its optical input on the second output of the first modulator 110, which makes it possible to save on an integrated optical source and limits the number of couplings.

 In FIG. 3, the twin modulators 110 and 120 are therefore placed in series for the control of the drifts and the coding of data. The upper modulator 110 codes the data optically. Its operating point is stabilized by a modulator 120 controlled by the local oscillator 140 and the DSP 210. The optical input of the lower modulator 120 is provided by the source of the first modulator 110 (additional signal).

 This interface can find applications in access points of optical networks, in particular for single-mode optical networks

<Desc / Clms Page number 11>

 using wavelength multiplexing (the same interface can be used for each wavelength when these have been separated) and therefore at the interface between these optical networks and other networks such as mobile networks, multimode fiber networks, or local computer networks.

 The aim is, in this embodiment, to modify in real time a DC reference voltage Vdc in order to compensate for the time drift of a pinching voltage of the modulators. If one manages to vary Vdc in correlation with the variation of the pinch voltage Vn, the intrinsic defects of the electro-optical modulator become minimal. The diagram of the servo system connected to the electrooptical modulator is recalled by FIG. 4.

 In order to carry out this function of regulating the control voltage, this device relies on one of the non-linearity properties of the modulator: in fact, when the latter is attacked on one of its arms by a pure sinusoidal signal Vac = A. sin (2.:. fo. t), it can be seen that over time the output signal has non-zero harmonic distortion, that is, harmonics of the excitation signal appear. It has been shown that this distortion is directly linked to the drift of the pinch voltage. The preferred servo system takes advantage of this finding.

 More specifically, the analysis of the signal by the DSP 210 is here as follows: the signal being transmitted in an analog manner, it comprises one or more characteristic frequencies. The DSP 210 detects whether the harmonic two of a characteristic frequency is present in the detected signal, this harmonic two being the signature of a distortion and therefore of an operating drift. The DSP then generates an error correction signal which is a function of the error measured and which makes it possible to reduce it.

 In FIG. 2 (embodiment with two twin modulators), the lower modulator 120 receives as input a constant light signal emitted by a light-emitting diode (LED) 310 Infra Red (1300m). On his

<Desc / Clms Page number 12>

 lower arm 122, a sinusoidal signal Vac is applied either from a local oscillator 140 or from the control system. On its upper arm 124, the control voltage Vdc is applied. It will be noted that, in this configuration, the control voltage Vdc is common to the two modulators.

 At the output of the lower modulator 120, a photodetector 100 is placed (PIN diode for example) behind which the electrical signal is digitized after having amplified it. The series of samples representing the light signal at the output of the modulator 120 will be processed digitally in order to determine the value Vdc corrected for the variation of the pinch voltage.

 Figure 4 shows in more detail the control system. The latter consists of three channels 260, 270 and 280, one of which is optional. The input channel 260 consists of a transconductor which converts the output intensity of the photo-detector into a voltage. This signal is then digitized by an analog / digital converter 220 and transmitted to the processing system 210 which can be a DSP or an FPGA.

 According to a procedure which will be detailed later, the voltage Vdc to be applied by the output channel 270 to the electrooptical modulators 110,120 is determined. This channel 270 consists of a digital / analog converter 230 and an amplification chain.

 The second output channel 280 will be integrated, if a local analog oscillator is not used. Its function is to modulate the control voltage of the lower arm by a sinusoidal signal for Vac.

 As explained above, this modulation is applied in order to quantify the harmonic distortion due to the variation of pinch voltage.

 Indeed, the theory makes it possible to make the link between the amplitude of the line 2fo at the residual voltage Vo. More precisely, the enslavement conditions which we will detail later encourage us preferentially to consider both fo and 2. fo. These two variables

<Desc / Clms Page number 13>

 allow to continuously calculate the control voltage Vdc to be applied to the modulator.

 To do this, we proceed as follows (see Figure 5).

 The sample sequence {x (k)} resulting from the sampling is put in a minimal spectral form, that is to say that we only consider the spectral components fo and its first harmonic.

 This spectral analysis will be done by the implementation of the Goertzel algorithm. By this procedure, the amplitude of the harmonic distortion can be accurately assessed. We will then be able to determine the value AV which must be added or subtracted (according to the sign of AV) to Vdc in order to minimize this harmonic distortion and to compensate for the drift of the pinch voltage. The procedure for calculating the correction is in accordance with the procedure for a PI controller, which we will detail below.

In parallel with this procedure, in the case of a non-integration of an analog local oscillator, the processor 210 (DSP) will have to transmit to the analog digital converter the samples coming from a table (LUT) in order to generate the control signal Vac.

We have seen that, during the enslavement, the variable to minimize is the harmonic distortion. Consequently, we must first evaluate this variable quantitatively in order to tackle the desired correction on Vdc. At this stage, several solutions are available to us: the first would be the use of the fast Fourier transform in order to perform in real time a spectral analysis of the signal x (n):

This solution requires calculating N frequency components in order to use one or two, the speed inherent in FFT (Transformed

<Desc / Clms Page number 14>

 of Fast Fourier) being therefore masked by the amount of information calculated and not used.

 Also, in the particular case of a computation of frequency component of frequency previously known in limited number, the computation by discrete Fourier transform (TFD) seems more opportune.

pour les N composantes fréquentielles se résume à :

Le gain de temps de calcul est proportionnel au nombre d'opérations évitées. The general form of TFD:

for the N frequency components comes down to:

The saving in calculation time is proportional to the number of operations avoided.

 The need to use a discrete convolution product greatly restricts the choice of the digital system in which such an application can be implemented. In order to develop a low-cost hardware solution, we were led to seek an implementation which is, in a real-time context, implementable in a program logic circuit (FPGA-Field Programmable Gate Array), a DSP core (low cost), or even a fast microcontroller.

 In order to allow a low cost realization, we have been brought to optimize the computation time of our algorithm. In this approach, we referred to an algorithm widely used in telephony for DTMF decoding: the Goerzel algorithm. It is known that this approach allows the implementation of the TFD (for m predefined frequency) in the form of a filter with infinite impulse response.

 Figure 6 is a block diagram of such a filter.

 In accordance with the implementation in recursive digital filter recommended by Goertzel, the algorithm can be split into two phases. The

<Desc / CRUD Page number 15>

 first, performed N-1 times (we will specify the value of N below), makes it possible to calculate two variables Qn and Qn-1 from the feedback terms x (k) and -x (k-1) and the coefficient ak. The second, performed once every N samples, calculates its value of the frequency component sought using the variables Qn & Qn-1 and the phase term WNk.

 In FIG. 7, the repeated phase (N-1) times is shown to the left of a vertical dashed line, and that performed every N times is shown to the right of this dashed line.

 Therefore, the difference equation will be of the form: Qk (n) = ak. Qk (n-1) -Qk (n-2) + x (n) (A. 1) With: ak = 2. cos (2. 7i. k / N) Qk (-1) = 0 Qk (-2) = 0 For n = 0, 1, ..., N-1.

 The term y (N) is obtained by the following expression: Yk (N) = Qk (N-1) - WNkQk (N-2) = X (k).

Since the servo rests on the effective value of the frequency component 2. fo, we will have to calculate its energy. However, if the values x (k) are real quantities resulting from the sampling of the signal, the phase term WNk is a complex quantity. We then demonstrate that:

The formulas A. 1 and A. 2 have the advantage of being implementable either in a DSP processor or in an FPGA.

The determination of the value N is defined by the following formula:
N = k. Fe / fO since we are interested in the frequencies fO and 2. fO, the value k is equal to 2.

 N = 2. Fe / f

<Desc / Clms Page number 16>

If, for example, the signal is sampled at a sampling frequency Fe of 180 kHz and if the analysis frequency fo = 1 kHz is used, then N = 360 samples are obtained.

The algorithm implemented, we recover, every 360 samples, a state of harmonic distortion. We then define a variable Tdh (k) Tdh (k) = Vfo (k) - V 2fO (k) NfO (k)
For reasons of stability and noise immunity, Tdh (k) is averaged over 256 values. Consequently, every 500 ms, one obtains a quantity directly correlated to the drift of VO. This data will allow us to define a control law in order to control the voltage Vdc. Due to the numerical nature of the calculation, this data is insensitive to any drifts in the servo electronics.

The difference equation is, as for a P1 controller, Vdc (n) = Vdc (n-1) + A1. T (n) + A2. T (n-1) Or A1 = Kp + K ,. T; A2 = -Kp; T = 1 / Fe
Many variations of the proposed interface are possible. They correspond to the architecture of the chosen network:
In a first variant (FIGS. 8a to 8c) the laser or lasers constituting the optical sources are deported to a central belonging to the network manager. In this case, the interface card does not contain an optical source, but only detectors. The modulator (s) operate in mirror mode: the light they receive is reflected and returned to the network after modulation.

 FIGS. 8a to 8c represent different implementations of the optical part by means of mirror modulators. FIG. 8a represents an embodiment from a Mach-Zehnder half-interferometer, FIG. 8b from a Mach-Zehnder half-interferometer with two complementary outputs, FIG. 8c from a half -optical coupler.

 It can be noted that there are configurations which are insensitive to the polarization of the received optical wave, a priori random polarization.

<Desc / Clms Page number 17>

 In a second variant, the laser or lasers are placed on the interface card. It is for example a low cost laser which cannot be modulated at high frequency. Modulation is then ensured by the external electrooptical modulator. The typical diagram of this interface (transceiver) is shown in Figure 9a. The optical modulation block 100 is described in more detail below.

 FIGS. 9a and 9b show two examples of the implementation of an external modulator with a laser source and a detector for a transmission / reception module (transceiver). The optical modulation block, which can be more complex, has been represented here by a Mach-Zehnder interferometer. In FIG. 9a, it is a full-duplex operation. In FIG. 9b, this is a half duplex operation.

 Other implementations of the interface can take advantage of the wavelength insensitivity of the modulator by the use of different wavelengths for the uplink signal and the downlink signal.

 The electrooptical modulation block can consist of a Mach Zehnder interferometer having an input and an output. The signal component necessary for the feedback is then extracted by means of an optical coupler. It may be advantageous to use a component having one input and two outputs (directional coupler, interferometer comprising a directional coupler or an X-junction). In this case, one of the outputs corresponds to the useful signal while the other output corresponds to the inverse of the useful signal. This inverted signal has the same sensitivity to drifts as the useful signal and can therefore be used in feedback without the use of a coupler (and the introduction of excess losses) being necessary.

 Figures 10a to 10c show three examples of electro-optical modulators.

 In FIG. 10a, using a Mach-Zehnder interferometer, in FIG. 10b, using a Mach-Zehnder interferometer with two outputs (various modulations from one another), in FIG. 10c, using a directional coupler (modulations opposite to each other).

<Desc / Clms Page number 18>

 Tests and verifications have been carried out on such devices.

 To characterize the drift of an electrooptical modulator, the following experiment can be carried out: a modulator is placed on a coupling bench. Light is injected into the modulator through a microscope objective. The output light is collected by another microscope objective. By applying a modulation voltage to the modulator electrodes, a modulation of the optical signal detected by a photodiode is observed (Fig. 11).

In FIG. 12 representing the measured drift, VpAC corresponds to the dynamic extinction voltage (1 kHz), VpDC corresponds to the static extinction voltage (linked to the drift of the static phase of the modulator);
This experiment thus makes it possible to characterize the drift of the modulators when applying a voltage niche. The characteristic times of this drift are of the order of a few minutes.

 The invention has also been implemented with another type of modulation.

 Thus, FSK (Frequency Shift Keying) modulation is a basic modulation technique. This basic protocol is used by communication devices during their interconnection phase. In this experiment, the digital signal (continuation of 0 and 1) containing the information is first coded in frequency by the control unit: to code 0 corresponds to a frequency fO, to code 1 corresponds to a frequency fl.

 A continuous signal (bias voltage defining the operating point of the modulator) is superimposed on this frequency modulated signal. This signal is applied to the modulator electrodes and encodes the optical signal.

 After detection by a photodiode, the signal is demodulated and compared to the initial signal. The interface, object of the invention, allows the modulation-demodulation functions and adjusts the operating point of the modulator by analyzing the drift of the operating point.

<Desc / Clms Page number 19>

 For this, the distortion at the frequency 2fo is detected in the photodiode signal, which is minimized by the servo loop. Thus, one of the elementary functions of the interface is to implement elementary modulation formats such as FSK modulation. This type of function can be generalized to other forms of phase or frequency modulation, since coding and decoding are carried out by the DSP. This experience shows the compatibility of this interface with all existing electrical modem formats.

 The signals transmitted to the interface and received are shown in Figures 14a and 14b. Digital processing allows the signal to be recovered without error. The stabilization scheme of the modulator by digital servoing thus makes it possible to transmit a digital signal using analog coding like the coding formats usually used by modems.

[1] Bellamy, John. 1982. Digital Telephony. New York: John Wiley &
Sounds [2] Blahut, Richard E. 1985. Fast Algorithms for Digital Signal Processing.

Reading, MA: Addison-Wesley [3] Burrus, CS and TW Parks. 1985. DFT / FFT and CorìVolution
Algorithms. New York: John Wiley & Sons.

[4] Oppenheim, Alan V. and Ronald W. Schafer. 1975. Digital Signal
Processing. New York: Prentice-Hall [5] ~. 1982. Reference Data for Radio Engineers. New York: Howard
Sams.

[6] ~. 1986. IEEE Standard for Functional Requirements for Methods and
Equipment for Measuring the Performance of Tone Adress Signaling
Systems. (ANSI / IEEE Standard 752-1986). New York: IEEE Press.

 [7J Fabrication of low refractive index low loss fIuorinated self-cross linking polymer waveguides for optical devices, A. V. OCHS, A.

 ROUSSEAU, B. BOUTEVIN, E. TOUSSAERE, A. DONVAL, R.

<Desc / Clms Page number 20>

 HIERLE, J. ZYSS, Proc. 7th International Plastic Optical Fibers conference'98, Berlin.

[8] Insensitive polarization electrooptic polymer modulator, A. DONVAL,
E. TOUSSAERE, R. HIERLE, J. ZYSS, J. Appl. Phys., 87 (7) (2000), pp3258-3262 [9J Drift correction of polymer based Mach-Zehnder modulators, P.

LABBE, A. CLOUQUEUR, R. HIERLE, E. TOUSSAERE, J. ZYSS,
Nonlinear Optics (Icono'5, Davos, 2000), vol 25, pp393-398.

Claims (14)

1. Electro-optical modulation method in which a control signal indexed to a reference value (Vdc) is applied to an electro-optical modulator (110), method in which light is taken (400) from the output d '' a modulator (110,120) subject to this reference value and a signal is deduced from this sampled light which is analyzed (210) to update the reference value (Vdc), characterized in that the sampled light is converted into a digital signal, and the update of the reference value (Vdc) is performed digitally on this signal (220).  2. Method according to claim 1, characterized in that two electro-optical modulators (110, 120) are adopted, the sampling (400) and analysis (210) steps are carried out at the output of only one of the two modulators (120), and the updated reference value (Vdc) is applied to the two modulators (110,120).  3. Method according to claim 2, characterized in that one provides on one of the modulators a useful modulation signal and a laser light and on the other of the modulators a disturbance signal and a light generated by a light emitting diode .  4. Method according to any one of the preceding claims, characterized in that a selected disturbance signal (Vac) is applied at the input of the modulator (120) on which the light sampling is carried out and disturbances are analyzed numerically. generated at the output of this modulator (120) by this disturbance signal.  5. Method according to any one of the preceding claims, characterized in that the updating analysis comprises the analysis of a disturbance generated by the modulator at a frequency 2fo, with respect to a given frequency component fo at the input of this modulator. <Desc / Clms Page number 22>  6. Method according to the combination of claims 4 and 5, and characterized in that one applies, as disturbance signal, a sinusoidal signal (Vac) having a frequency fo and one analyzes digitally (210) at least the component frequency at the frequency 2fo present at the output of the modulator (120).  7. Method according to any one of the preceding claims, in combination with claim 5, characterized in that the frequency component at 2fo is calculated by recursive digital filtering.  8. Method according to claim 7, characterized in that the recursive filtering is carried out according to a difference equation of the type Qk (n) = ak. Qk (n-

 1) -Qk (n-2) + x (n) with ak = 2. cos (2. Pl. k / N), Qk (-1) = 0 and Qk (-2) = 0, x (n) designating the samples obtained from the modulator output signal, and in that this filtering is carried out in at least two phases, one carried out in N-1 stages, which makes it possible to calculate QN and ON-1, and the other carried out once after these N-1 stages, and providing a state of the frequency component at the frequency 2fo, N being given by the equation N = 2fe / fo where fe is the sampling frequency of the digital signal obtained from the sampled light.  9. Method according to any one of the preceding claims, in combination with claim 2, characterized in that at least two modulators are adopted, the input of one receiving signals from an output of the other , in that the light coming from the downstream modulator is taken for updating analysis, and in that the reference value updated by this analysis is applied to the two modulators.  10. Method according to any one of the preceding claims, characterized in that a digital control signal is supplied (210), and this digital signal is converted into an analog control signal for the modulator, and in that one realizes further correcting the digital control signal (210) taking into account the updated reference value (Vdc) before converting this corrected digital signal into the analog control signal. <Desc / Clms Page number 23>  11. Method according to the preceding claim, characterized in that the updating analysis of the reference value (Vdc) is carried out as well as the correction of the digital control signal using the same processor (210) .  12. Electro-optical modulation device, comprising at least one light modulator (110), means for supplying a control signal from the modulator (110) indexed to a reference value (Vdc), means for taking the light ( 400) at the output of a modulator (110, 120) subject to this reference value, means for analyzing a signal deduced from the light picked up and deducing from this analysis an updated reference value (Vdc), characterized in that the sampling means and the analysis means (400,210) are provided for converting the sampled light into a digital signal (220), and supplying the updated reference value using a digital analysis of this converted signal.  13. Device according to claim 12, characterized in that it includes a mirror modulator.  14. Device according to claim 12 or claim 13, characterized in that it includes an interface card and a laser source, said at least one modulator and the laser source being placed on the interface card.