photonic integrated components applied to secure chaos

FP6-2006-IST-34551

PICASSO

Photonic Integrated Components Applied to Secure chaoS encoded Optical communications systems

Deliverable 6.1: Report on the development and performance of the transmitter/receiver modules

Due date of delivery: Apr. 2009

Actual submission date: May 2009

Start date of project: 01.10.2006 Duration: 3 years Organisation name of the lead contractor for this deliverable: Universite de Franche-Comte

Revision [draft] Project co-funded by the European Commission within the Sixth Framework Programme

(2002-2006) Dissemination Level

PU Public PP Restricted to other programme participants (including the Commission Service) RE Restricted to a group specified by the consortium (including the Commission

Service) X

CO Confidential, only for members of the consortium (including the Commission Service)

IST-2006-34551 PICASSO Deliverable 6.1 (Apr 2009)

Contents 1. Introduction........................................................................................................................... 3 2. Development of integrated transmitters/receivers............................................................. 3

2.1 Hybrid integrated transmitters/receivers. ............................................................................ 3 2.2 Monolithic integrated transmitters/receivers....................................................................... 5

3. Performance analysis ............................................................................................................... 6 3.1 Hybrid integrated transmitter receiver modules.................................................................. 7 3.2 Monolithic integrated devices .............................................................................................. 8

3.2.1 Selection of devices for packaging ................................................................................ 8 3.2.2 Synchronization among monolithic integrated chaos transmitter/receiver ................ 10

4. Stability analysis..................................................................................................................... 11 4.1 Instabilities in hybrid and monolithic integrated devices .................................................. 11 4.2. Properties of packaged monolithic integrated devices ..................................................... 13

5. Electro-optic emitter receiver modules based on discrete devices..................................... 14 5.1 Transmitter / Receiver architecture ................................................................................... 14 5.2 Performance analysis......................................................................................................... 16

5.2.1 Dynamical regimes...................................................................................................... 16 5.2.2 Rest points stabilization .............................................................................................. 16

5.3 Synchronization results ...................................................................................................... 18 5.4 Packaging issues ................................................................................................................ 19

6. Conclusions ............................................................................................................................. 19

-2-


WORKPACKAGE 6: System development and single channel operation

Deliverable 6.1 (M30): Apr. 2009

“Report on the development and performance of the transmitter/receiver modules”

1. Introduction

In this deliverable, the developed transmitter/receiver modules of different chaos generation methods will be thoroughly analyzed in terms of their performance as chaos telecommunication transceivers. NKUA has studied the all-optical feedback chaos generators based on hybrid and monolithic integration, whereas UFC developed emitter / receiver pairs based on commercial discrete devices, required for proper operation the new principle of electro-optic phase chaos generation. We recall that III-V semi-conductor optoelectronic integration was shown to be not yet mature enough for a reliable operation of integrated modules dedicated to the electro-optic approach, due to insufficient separation and isolation between the laser source and the electro-optic modulators (see previous reports and deliverables). However, the potential shown by the electro-optic phase chaos setup in terms of bit rate and synchronization quality over a huge bandwidth, decided the consortium to continue the exploration of the electro-optic approach, but with discrete electro-optic and integrated devices.

The specific study described in this deliverable includes chaos generation characteristics, investigation of the mismatch between transmitter and receiver relying on synchronization experiments, and finally a stability analysis enforced by numerical simulations that explain the experimental findings. The effects that degrade synchronization are detected and methods for their mitigation are proposed. The first packaged monolithic integrated chaotic transmitters/receivers are presented, and a compact box packaging is proposed for the discrete device electro-optic approach approach.

2. Development of integrated transmitters/receivers 2.1 Hybrid integrated transmitters/receivers. In this paragraph, the development of hybrid integrated is presented. Hybrid integrated emitters for chaos generation have been designed, based on the application of a fiber external cavity to a semiconductor laser that emits at 1553nm. Two different lasers have been employed: a distributed-feedback quantum-well (DFB-QW) laser from HHI and a selective-mode Fabry-Perot laser from EBLANA. When integrated with the fiber cavity, no isolation is applied so that the laser is susceptible to external cavity dynamics. The integrated fiber cavity spliced to the above lasers includes a 90/10 coupler, a thermally tunable variable optical attenuator (VOA) and a phase section (PS) fabricated using conventional fiber-technology and a high-reflective fiber end with reflectivity above 95%, as shown in Fig. 1 and already presented in deliverable 3.1. The total length of the device is measured to be approximately 30cm, however, through the dynamics that emerge for specific operating conditions, the exact length of the cavity can be extracted. One of the most crucial parameters that determine the dynamics arising from such devices is the feedback strength from the external cavity. In the specific configuration, the feedback strength is determined by adjusting the VOA voltage in the fiber integrated cavity. At zero voltage the VOA induces losses over 40dB, isolating practically any reflection from the cavity. By increasing the VOA voltage up to 3.3V, the losses are gradually minimized to less

-3-


than 1dB, covering thus the entire desirable region of optical feedback values that could lead to the generation of a broadband chaotic optical signal. The maximum optical power Pf estimated that re-enters in the laser is just above 10% of the initially emitted optical power Pin – defining a feedback strength ratio of kf = Pf / Pin, by considering the losses induced though the optical path. This value is considerably higher than the ones (around 1-2%) for which broadband chaotic dynamics begin to grow. The enclosure of the PS within the fiber cavity and its adjustment to different operating conditions, do not contribute to any alteration of the dynamics determined by the cavity length, the feedback strength and the laser biasing current. It can only affect the dynamics in devices with very short external cavities – of the order of cm – as explicitly shown in [1-3]. Thus, in the specific configuration, the PS is included merely for using these chaos generators in chaos communication systems, where an identical receiver device should be fabricated. The mechanically replicated cavity may usually induce a small length mismatch, which can be compensated by applying the appropriate voltage biasing at the phase section.

Figure 1. Schematic (upper) and photo (lower) of a hybrid integrated emitter with a DFB-QW laser and all-optical

feedback.

An affluent dynamical behavior was experimentally recorded for the hybrid integrated emitter as shown in the spectra of Fig. 2. Given the fixed length of the cavity, the two parameters that determine the spectral distribution of the generated dynamics, are the laser biasing current and the feedback strength. In Fig. 2, two values of laser current are considered, 1.5·Ith and 1.8·Ith, along with two cases of optical feedback: a low kf value (0.5% - 1%) and a moderate kf value (around 3% - 4%). Both the feedback cases are sufficient for leading the laser to operate in the coherence collapse regime and generate a broadband chaotic signal. Slightly lower feedback values trigger only a small number of external cavity modes (ECM) around the relaxation frequency of the laser, while for kf <0.1% the laser operates as a solitary device, practically isolated from the external cavity. The spectra of Fig. 2 are completely indicative for the variety of dynamics these long external cavity emitters generate. For relatively small laser current values and feedback strengths (Fig. 2a, 2c) a broad spectrum rises with suppressed ECMs around the relaxation frequency of the laser, whilst by increasing the feedback strength or the laser current values even further, the external cavity modes are the dominant dynamics competing the broad spectral continuum (Fig. 2b, 2d). Furthermore, from the ECM spacing additional information can be extracted regarding the cavity length, which can now precisely measured to be 29.83cm.

-4-


Figure 2: Experimental microwave spectra of the hybrid device optical output for various laser current values and

optical feedback strengths.

2.2 Monolithic integrated transmitters/receivers. The monolithic integrated device incorporates the fundamental principles of the all-optical feedback technique, using elements that provide the capability to control the most dominant parameters that determine the chaotic properties of the optical emission accurately and reproducibly. The device consists of four successive sections (Fig. 3): a DFB InGaAsP semiconductor laser operating at 1556nm, followed by a Gain/Absorption section (G/As), a phase section (PHs) and a 1cm long passive waveguide (PW) grown by a selective area epitaxial growth. The overall external cavity length is defined by the internal DFB laser facet and the facet of the waveguide end which is highly reflective coated (HRC) (R=95%) and includes the G/As and PHs. Criterion of the selection of this long cavity length is the ability of the device to produce chaotic attractors with high complexity. The necessity of integrating a G/As outcomes from the requirement to control the optical feedback strength. The G/As can be either an amplifying (semiconductor optical amplifier, SOA) or an attenuating element (variable optical attenuator, VOA) once it is forward or reverse biased respectively. Consequently, a very wide range of optical feedback values can be set and thus various types of dynamics can be generated. The phase section in this device, in contradiction to the long cavity hybrid device, plays an important role for controlling the dynamics produced by this short-cavity oscillator. The biasing current of the PHs can be several mA, providing a total phase shift of several multiples of 2π. Another important task for the PHs is the exact matching of the cavity length of emitter-receiver pair devices intended for communication applications.

Figure 3: Schematic diagram (upper) and photo (lower) of the monolithic PIC that includes: an InGaAsP DFB laser, a 100µm gain/absorption section, a phase section and a 1cm passive waveguide. The end facet of the waveguide is

-5-


highly reflective coated. Several indicative operating regimes of G/As biasing have been identified that provide diverse dynamics in the generated optical carrier. Specifically, for a laser current value of 1.6·Ith, these regimes are:

a) G/As reverse biasing (VOA operation): The total attenuation inside the cavity corresponds to optical power feedback ratio kf less than 1%. Only limit cycle dynamics or stable solutions are observed excluding any chaos dynamics.

b) No G/As biasing: The dynamics of the output signal differs considerably when altering the cavity phase through the PHs current IPHs. The device mainly operates in a coherence collapse regime, with the complex dynamics expanding around the relaxation frequency of the laser (Fig. 4a). However there are some narrow allocated regions of phase values, periodically repeated over 2π, in which the dynamic states include mainly limit cycles with one or two intense peaks in multiples of the external cavity resonance frequency (around 3.3GHz).

c) G/As forward biasing (SOA operation): When the G/As is positively biased the provided gain in the cavity compensates the internal losses, increasing kf to values over 4%. This strong feedback leads to intense broad-spectrum chaotic dynamics (Fig. 4b), providing also enriched low-frequency spectral components and diminishing simultaneously any phase dependence within the cavity. The broadening of the optical spectrum has been measured to extend over 15 GHz, making thus this chaotic carrier an efficient candidate for chaos-encrypted optical communication systems supporting up to 10Gb/s bit streams.

Figure 4: Experimental microwave spectra of the monolithic PIC device optical output for two different levels of

optical feedback strength: ~3.3% (a) and ~5% (b).

3. Performance analysis The chaotic optical generators described above, besides their classification of “smart-structured” emitters for controllable broadband signal emission, are also subsystems of a communication scheme in a transmitter - receiver configuration. Based on such an approach, a receiver should fulfill all the appropriate criteria and locally reproduce the initial carrier. The synchronization of the receiver to the incoming signal (or the initially generated signal by the emitter) follows strict conditions for the optimal performance, principally emerging from the identity of the two subsystems. The performance of both developed integrated configurations will be thus evaluated in terms of synchronization analytically, by considering both approaches that appear in the relevant bibliography: open-loop receivers – in which the receiver laser lacks of an external cavity and complex dynamics emerge from injection locking – and closed-loop receivers – in which the receiver subsystem is a replica of the emitter and has its inherent complex dynamics.

-6-


3.1 Hybrid integrated transmitter receiver modules For the hybrid integrated devices two different pairs of modules have been evaluated in a transmitter – receiver configuration. The first pair employed two DBF-QW lasers from the same production wafer, with almost the same emitting wavelength and same characteristics. The accurate matching of the lasers is achieved through a fine temperature control of the devices. These lasers were integrated with fiber cavities modules as the ones presented in Fig. 1, however the mismatch in the length of these two cavities was measured to be about 5.6mm. More specifically, the cavities PICCAV10 & PICCAV11 are 29.84cm and 30.39cm long respectively. According to PhX fabrication specifications the lengths were predicted to be 29.83cm and 30.39cm respectively verifying the experimentally extracted values. This length mismatch for closed-loop synchronization proved to be large enough, resulting in practically no synchronization at all. When the receiver external fiber cavity was isolated by maximizing the internal losses, synchronization in an open loop configuration was achieved.

Figure 5: Experimental microwave spectra of the hybrid emitter optical output (black) and the subtraction signal

between the hybrid emitter and the synchronized open-loop receiver (red), for two different chaotic emission profiles.

By controlling suitably the laser’s biasing current and the feedback strength, conditions that lead to a relatively uniform spectrum with deteriorated ECMs, such as the spectra presented in Fig. 5a, could be potential carriers for real message encryption, even if this should probably require a subcarrier modulation technique. Conditions that result in intense ECMs, such as in Fig. 5b, can be avoided, since such carriers could not fulfil the scope of broadband real message encryption, but only single frequency tone encryption. In the cases of Fig. 5b, the intense cavity modes can be partially subtracted – with a much better performance around the relaxation frequency of the laser – but such a scheme would not be adequate to result in sufficient chaos cancellation for pseudorandom sequences. However, there are spectral regions that ECMs are cancelled in excess of 15dB. When applying lower feedback, as shown in Fig. 5a the ECMs are significantly deteriorated and chaos cancellation is performed within a wider region – again around the relaxation frequency of the laser – with an improved performance.

The second pair employed two EBLANA DM lasers, also from the same production wafer, with the same emitting wavelength and same characteristics. The specific lasers were provided for a low-cost solution emitter, without any temperature control. These lasers were also integrated with the fiber cavities PICCAV14, PICCAV15 as the ones presented in Fig. 1, but in this case the mismatch in length was measured to be only 0.6mm (Fig. 6). This length mismatch for closed-loop synchronization proved to be sufficient, after adjusting the PS of the fiber cavity, at least for achieving synchronization in the first few GHz of the chaotic spectrum, as shown in Fig. 7.

-7-


Figure 6. Photo of the matched pair of hybrid integrated emitter / receiver with selective mode Fabry-Perot lasers.

Figure 7: Experimental microwave spectra of the hybrid emitter optical output (black) and the subtraction signal between the hybrid emitter and the synchronized closed-loop receiver (red), for four chaotic emission profiles.

Different operating conditions of laser biasing current and feedback strength have been tested also with these hybrid devices providing various chaotic emission profiles, as shown in Fig. 7, similar to those of Fig. 5. Using these devices in a transmitter / receiver scheme, with a closed-loop receiver, synchronization is partially feasible. In cases of low laser current and optical feedback values ECMs are subtracted in the first 3 GHz spectral region (Fig. 7a). By increasing slightly the lasers biasing current (Figs. 7b, 7c) the ECMs are suppressed in favour of a broader chaotic continuum. Cancellation of this continuum is observed in the first 4GHz, with cancellation excess of 10dB. However, any further increase in the feedback will result in intense ECM dynamics, which are proven to be unsuccessfully reproduced at the receiver, as revealed from the poor cancellation efficiency (Fig. 7d). The fact that, in all cases of Fig. 7, no cancellation is observed for frequencies beyond 4 GHz is attributed to the not exactly matched lengths of the cavities. Additionally, the above performance was observed only when the two lasers emitted the same wavelength with a small tolerance in mismatch up to 5pm. Beyond that frequency, synchronization is practically lost. 3.2 Monolithic integrated devices 3.2.1 Selection of devices for packaging The packaging of monolithic integrated devices is a time-consuming and demanding procedure; hence a careful selection of devices that are as matched as possible was carried out in order to achieve packaging for the most identical devices. The basic criterion for the selection of

-8-


matched devices was the emission wavelength and the similarity of the generated dynamics as recorded in the electrical spectra. Indicatively, the emission wavelengths for the series C21-C30 and C41-51 are depicted in Fig. 8.

Figure 8. Emission wavelengths of C21-C30 and C41-C51 monolithic integrated chaotic lasers.

The main conclusions coming from this analysis are the following:

a) Wavelengths of each 10-devices strip spread within 1nm at least b) Large differences in threshold current of the lasers were also observed.

Hence, devices that are closely positioned inside the same strip are the best candidates in order to form matched transmitter/receiver modules. The dynamical analysis is the second criterion for the selection of devices that seem to have matched properties.

Figure 9. Electrical spectra of a monolithic integrated device (F25) at different operating conditions of bias current

and VOA/SOA section.

Many different devices from the C and F series were thoroughly investigated in terms of the electrical spectra they generate at different conditions. The devices providing identical behaviour at a wide range of operating regimes were selected as potential transmitter/receiver pairs. In Fig. 9, we see indicatively the dynamics generated by F25 laser chip. In the next figures, we provide the electrical spectral of F21 and F29 which resemble to each other in terms of their dynamical behaviour.

-9-


Figure 10. Electrical spectra of F21-F29 at different operating conditions of bias current. The similarity of the

spectra the two devices provide is evident

Based on the above methodology, the following groups of devices with similar chaotic dynamics, wavelength and current threshold were selected.

• GROUP A: Devices F18, F26, F27, F28 (∆λ~±0.2nm, ∆Ith~ ±1mA) • GROUP B: Devices F21, F29 (∆λ~0.26nm, ∆Ith~ 1.1mA)

GROUP C: Devices C46, C47, C48, C49, C51 (∆λ~±0.18nm, ∆Ith~ ±1.5mA) 3.2.2 Synchronization among monolithic integrated chaos transmitter/receiver When incorporating the monolithic integrated devices of Fig. 3 in a transmitter / receiver configuration, synchronization using closed-loop receiver architecture is more straightforward. The monolithic fabrication process can guarantee identical cavity lengths for all devices, in contradiction to the mechanically fabricated hybrid fiber cavities. Nevertheless, even if a small fine tuning in the round trip time of the electric field is needed within a cavity for accomplishing the optimal synchronization conditions, the appropriate biasing of the active phase section will provide that.

Figure 11: Experimental microwave spectra of the monolithic integrated emitter optical output (black) and the

subtraction signal between the hybrid emitter and the synchronized closed-loop receiver for phase-matched (red) and phase-unmatched (blue) conditions, for a feedback strength kf of ~3.3% (a) and ~5% (b).

In Fig. 11 the dependence of the phase matching condition in the synchronization process of short-cavity devices is presented, after considering two different feedback conditions for both emitter and receiver devices. In the first case (Fig. 11a) the G/As section is positively connected but biased with 0mA, however the feedback strength lays in adequate levels to generate powerful chaotic dynamics gathered around the laser’s relaxation frequency. Cancellation maximum reaches up to 20dB in the most powerful spectral regions, when a phase matched condition is applied (Fig. 11a, red line), while the cancellation performance is indisputably

-10-


much better when compared to the synchronization of the hybrid integrated cavities. On the contrary, as also expected from numerical simulations unmatched phase conditions will result in severe synchronization efficiency deterioration (Fig. 11a, blue line). In the second case (Fig. 11b) when feedback is increased the spectral distribution of the chaotic carrier is changed as a result of the much different dynamics that prevail in the device. Also in this case, under appropriate phase matching, a good synchronization performance is recorded, with a cancellation maximum over 20 dB. An even better performance should be expected in terms of synchronization, if a minimization of back-reflections in the laser-fiber coupling could be achieved. These reflections explain the intense fringes, particularly present in the very good synchronization cases, attributed to the correlated emitter output at the receiver with the same delayed signal reflected at the tapered fiber end used for coupling at the receiver device. 4. Stability analysis 4.1 Instabilities in hybrid and monolithic integrated devices Synchronization between optical generators that emit chaotic signals is based on the identity of the counterparts. Closed-loop receivers can be efficient regenerators of complex carriers as long as external cavities are well matched. For efficient closed-loop receiver operation, the exact cavity length at both transmitter and receiver subsystems is of extreme importance. Assuming that the two cavities have a constant length mismatch equal to ∆L, one could claim that a phase

shift equal to ∆φ= 4π�∆Lvg

�f em,rec is able to anticipate the length mismatch. In this equation, vg

is the group velocity inside the cavity, and fem,rec is the emission frequency of the emitter and receiver laser respectively, which are considered identical in this analysis. Based on the Lang-Kobayashi rate equations model, numerical simulations have been performed in order to designate the synchronization dependence on the mismatch of cavity lengths.

Figure 12: Numerical calculation of chaos cancellation in regards of cavity length mismatch.

The results of this analysis are depicted in Fig. 12. The simulated system ignores any other mismatches either in the lasers’ intrinsic parameters or noise sources that could degrade the synchronization efficiency apart from the length mismatch. This explains the excellent performance, with cancellation of almost 24dB at ∆L=0mm. As the absolute value of ∆L increases, a rapid decrease of the cancellation is observed, showing that ∆L>1mm is enough to cause even 10dB synchronization decay. Considering a fiber cavity, this length mismatch corresponds to a 10ps difference in the round trip time inside the cavities of the emitter and receiver. Hence, even such short time differences have the tendency to de-correlate the two chaotic spectra. In our experimental case that incorporates the hybrid cavities, the measured 0.6mm mismatch is expected to deteriorate the cancellation up to 6dB with respect to the ideal case of perfect length matching according to Fig. 12.

-11-


Additionally, the matching conditions – including the operating conditions of the lasers – should be preserved with high accuracy over time. This is not a straightforward task, since the phase matching between two devices is extremely sensitive to the environmental conditions. For example, for the hybrid integrated devices that did not use a temperature control unit, synchronization was apparent only when the lasers’ wavelength mismatch was within several GHz. Local thermal sources could drift this mismatch to much higher values, making synchronization impossible. For the case of the monolithic integrated devices, temperature control has been utilized. Provided that the temperature dependence of these devices was measured equal to 0.14nm/oC and the temperature stability was ±0.02oC, a ±3.7GHz uncertainty in the wavelength of the emitting signals was expected. Indeed the measured uncertainty was around ±4GHz resulting in a relatively unstable cancellation spectrum, with the duration of the optimal performance being no more than several seconds. Stability of synchronization performance is expected to improve dramatically, if the wavelength shift uncertainty is reduced. The reason for the above downgraded cancellation performance practically emerges from the suspended phase matching condition. The above intuitive explanation for the experimental findings was also verified by numerical simulations. For such a study, the effect of frequency fluctuation was introduced in the traditional Lang-Kobayashi model, considering a random fluctuation around the emission frequency. The phase mismatch attributed to the frequency fluctuations is equal to: ∆φ= 2π�∆f�T , where ∆f is the frequency fluctuation and T the round-trip cavity time. As far as

the monolithic integrated devices are concerned, the round trip time is in the order of 240ps, hence for 1GHz frequency fluctuations, the calculated phase mismatch of π/2 is expected to degrade the cancellation a few dBs. For higher frequency deviations, i.e. 4GHz, the phase will vary over two times π, expecting much higher effect on the phase-sensitive closed loop synchronization. Concerning the hybrid devices, similar frequency instability could force the phase mismatch to evolve over multiples of π, due to the much longer cavity incorporated in this scheme. The effect of random frequency fluctuations was numerically estimated for the short-cavity monolithic devices as a function of the receiver laser current (Fig. 13).

Figure 13: Numerical calculation of chaos cancellation in regards of random frequency fluctuations between the

emitter and the receiver laser emission frequencies.

The numerical results confirm the simple theoretical analysis described above. It is evident that 1GHz frequency fluctuations reduce the cancellation value up to 3dB, while for 4GHz the achieved cancellation is in the order of 10dB, close to the experimentally measured cancellation values. The above designate that the appropriate packaging of such devices with efficient thermal properties will stabilize thermally the devices with accuracy below ±0.01oC, thus a wavelength uncertainty of less than 1GHz is expected leading to much more promising results in terms of synchronization.

-12-


4.2. Properties of packaged monolithic integrated devices Two of the selected pairs are already packaged by HHI. The specific devices are the F18 and F26 respectively. A picture of the packaged device is shown in Fig. 14.

Figure 14: Picture of the packaged monolithic integrated chaos transmitter

The packaging process reassures coupling to fiber up to 40% and accurate temperature control of the device utilizing a heatsink and a Peltier element underneath the laser chip. Moreover, the three SMA connectors are the electrical interfaces for the laser current supply, VOA/SOA voltage/current supply and the phase section current supply. High frequency connectors were finally utilized in order to test the dynamics and the synchronization in regimes where optical feedback strength and phase are modulated at high rates. The RF facility for the laser current provides the possibility to test direct modulation as the codification method (chaos shift keying), hence providing all the transmitter functionalities in a chip. The electrical spectra of the generated chaos from the two devices are depicted in Fig. 15.

Figure 15. Electrical spectra of transmitter (red) and receiver (blue) and subtraction signal at the receiver (black).

The two devices provide similar dynamics and the result of their subtraction is stable in contrast to what was observed for the bare chips. The synchronization efficiency as extracted by the power of the signal coming from subtraction is strictly dependent on the current feeding the phase section. Precise adjustment of the latter was realized in order to obtain the lowest possible power for the subtracted signal. In the next figure the dependence of synchronization on the phase section current is clearly depicted.

-13-


Figure 16. Electrical spectra of transmitter (black) and receiver (red) and the result of the subtraction at the receiver

for precise phase matching (blue) and arbitrary relative phase (green).

The reduction of subtraction power (synchronization efficiency) up to 7dB is recorded. 5. Electro-optic emitter receiver modules based on discrete devices The electro-optic phase chaos setup will be described in this section, considering its performances in terms of emitter-receiver modules, chaos generation capability, and synchronization performances. Stability issues will be also addressed, which is related to stable long term operation of the emitter-receiver synchronization state. As already stated, this electro-optic approach was first implemented with commercial standard devices, before customized parts are available in the frame of the project (essentially phase to intensity nonlinear converters, based on a tunable multiple wave imbalanced interferometers, fabricated with fiber optics technology by PHX). 5.1 Transmitter / Receiver architecture The emitter / receiver setup is depicted in Fig.17, showing the basic devices and their arrangement in emitter / receiver modules. It allows for chaos generation at the emitter output, and chaos cancellation at the receiver output, both being uni-directionally coupled by the transmission channel.

Figure 17. Electro-optic phase chaos setup.

The actual devices used for the reported results are the following: − a DFB high power laser (EM4 from EMCore) providing up to 50mW monomode laser light,

at any wavelength of the ITU grid (the other devices are indeed specified to operate on the

-14-


whole C-band, thus giving the flexibility to choose the ITU channel with which the phase chaos has to be generated). The laser has a Butterfly package (built-in elements for temperature and optical power control), and is pigtailed with a polarization maintaining (PM) fiber.

− an integrated optics LiNbO3 phase modulator (PhM), Z-cut, from EOspace, specified for 20Gb/s (16GHz analogue bandwidth), has 1.2dB insertion loss, and a half wave voltage Vπ=4V. The device is polarization sensitive, hence pigtails are making use of PM fibers. The consequence is that polarization control is required at receiver PhM input, but this functionality is now achievable with commercial modules.

− A 50x50 PM fiber coupler is used to split the chaotic phase modulated light beam between the transmission channel (to the receiver), and the local optoelectronic nonlinear delayed feedback path.

− a phase to intensity nonlinear converter, actually performed by a commercial (itf) DPSK demodulator, with 2.5GHz free spectral range (FSR). It was shown in previous deliverables that the FSR needs to be smaller than the PhM bandwidth, which condition is fulfilled by a factor of 6. This commercial device consists in a fiber based Mach-Zehnder imbalanced interferometer, mounted in an athermal package which provides a heating wire for fine tuning of the offset phase between the two arms of the imbalanced interferometer. In a future version, the device will be replaced by a recently delivered matched pair from PHX, of two 3GHz-FSR imbalanced Michelson-like fiber interferometer, also tunable through a heating wire (see previous deliverables). Another future option within the project, is to use a currently developed customized matched pair of 3-wave interferometers (also fabricated by PHX), with independently tunable relative phases. This experiment is programed in order to demonstrate the customization capability of this key-element in the chaotic oscillator, which is performing the nonlinear phase to intensity conversion.

− a variable optical attenuator (M420 Eigenlight) to set the normalized feedback gain of the chaos generation oscillator, without changing the wavelength (optical power level can be varied through the DFB laser injection current, but this also slightly changes the operating wavelength, thus deviating the operating rest point of the DPSK demodulator). Insertion loss is typically 1dB, and attenuation can tuned over a 40dB range. This attenuator can also be integrated in the future, within the phase-to-intensity converter, since PHX has a compact technology to induce electrically tunable attenuation within a fiber (see VOA used for the hybrid integration).

− A broadband amplified photodiode (Miteq DR125), with 2V/mW conversion efficiency, and a 30kHz-13GHz analogue bandwidth. The maximum output voltage swing is limited to 4Volts peak-to-peak, thus limiting the input optical power to ca. 2mW (the total loss from the laser to the photodiode is ca. 10dB, so that a laser source with ca. 20mW could be enough to achieve a maximum gain for highest chaos complexity).

− A broadband telecom driver. Two types were successfully tested, Photline/ADLightec DLM12111 (50kHz-22GHz bandwidth, 28dB gain 12Vpp output voltage swing), and SHF100CP (20kHz-25GHz bandwidth, 18dB gain, 12Vpp output voltage swing). With 12Vpp voltage amplitude driving the PhM, we can achieve a 3Vπ scan, which corresponds to a highly nonlinear motion of the chaotic optical phase oscillator. Notice that ultra-low Vπ LiNbO3 phase modulators appeared recently on the market: with 2V for the Vπ, the maximum normalized feedback gain could be multiplied by a factor of 2, thus increasing even further the maximum achievable complexity.

− The different patch-cords and pigtails are defining the total delayed feedback time, which is a very sensitive parameter in the chaos cancellation process (the most important secret physical key). A variable delay line is required, at the emitter, or at the receiver, in order to match within a fraction of ps the delays at the emitter and receiver.

-15-


The receiver intended for chaos cancellation, is made of the same devices, with matched pairs when available. Different matching procedures were described in the previous deliverables, in order to obtain the best chaos cancellation situation. 5.2 Performance analysis 5.2.1 Dynamical regimes Figure 18 shows a quick overview of the bifurcation diagrams, both in the amplitude and spectral distribution (over different spectral ranges). The chaotic regimes are typically starting at β=2, and then, nearly Gaussian amplitude distribution, as well as very flat and broadband spectra over more than 13GHz, can be achieved with the maximum value of β=5. More details about the dynamical regimes and the bifurcation scenarii of the electro-optic phase delay dynamics can be found in [4].

Electrical amplitude PDF High freq. spectra

Middle freq. spectra Low freq. spectra

Figure 18. Characteristic experimental bifurcation diagrams.

This dynamical regime overview mainly highlights the fact that the chaotic waveform is capable of encoding binary message which spectrum extends up to 13GHz (which limits is fixed by the slowest electronic feedback device i.e., the amplified photodiode). Of course, smaller bandwidth are also possible. The chaotic spectrum needs then to be matched with that of the binary message, so that the influence of the channel noise, as well as that of the receiver noise, can be reduced to the minimum by suppression the useless frequency range. This chaotic spectrum can indeed be tailored using calibrated RF filters in-loop, thus also improving the matching between the emitter and the receiver dynamics. Bessel-Thompson low pass filters with a 7,73GHz cut-off matched with 10Gb/s bit rate, were indeed used to improve the BER (see deliverable 6.2). 5.2.2 Rest points stabilization It was noticed previously that the setup performance (chaos cancellation, or BER measurements when data are applied) degraded with time. This was observed for example, when recording bifurcation diagrams over long time slots, with slowly and finely scanned bifurcation parameter values. The very accurate diagrams in Fig.18 could be obtained only with the implementation of an active stabilization of the operating rest point of the phase-to-intensity nonlinear converter (the DPSK demodulator). This also holds when mismatch between the emitter and the receiver DPSK demodulators rest points has to be kept as small as possible over long operating periods. The rest point deviation is explained by environmental temperature variations resulting in fibers lengths change. In order to deal with this problem an active control system was designed. The demodulators have built-in heating wires allowing to tune the rest point. Stabilization could be performed by measuring the output power at the second (unused) output of the DPSK demodulator, and

-16-


compensating it by changing the current in the heating wire (thus forcing a small drift of the relative phase ruling the static interference condition). However, this simple solution does not work when chaotic phase modulation is applied: the power is dynamically split almost equally between both outputs of the demodulator, practically independently of its rest point. For this reason a part of non-modulated light was split (see Fig.19) and passed through the demodulator in the contra-propagating direction with respect to the chaotically phase modulated light. This backward propagating light was then separated using an optical circulator, and used as the input to the active control circuit.

Figure 19: DPSK rest point control

The control circuit scheme is shown in the Fig.20, and its electronic implementation is shown on the photograph in Fig. 21. It consists of a standard proportional-integral (P-I) control circuit. The error is generated by comparing (OP Amp. subtracter) a (tunable) reference voltage with the measured interference intensity level. After applying a P-I control to this error, the output signal is used to drive the heating wire changing the interference condition. Before activating the control circuit (before closing the control loop), the operator needs to bring the heating wire voltage so that the interference point is close to the desired one (with the stable slope). This is an issue related to the non linearity of the DPSK demodulation transfer function.

Figure 20: control circuit scheme.

The receiver demodulator was stabilized using the same approach. The non-modulated light had to be transmitted to the receiver by an additional fiber. This drawback of a second channel, can be avoided, for example, by injecting modulated and non-modulated light in the same fiber with orthogonal polarizations or by implementing another approach for stabilizing the demodulators (however more complex: this needs a second reference wavelength with an optical frequency discriminator and comparator, with WDM DMUX, and the whole control setup has to be implemented both at the emitter and receiver. This, should avoid the use of a second channel,

-17-


most probably with comparable performances than the one actually implemented with the DPSK stabilization through contra-propagating non modulated light beam).

Figure 21: photograph of P-I control electronic circuit.

With the implemented active control, the performance of the system (in terms of chaos cancellation quality) was very stable over several hours of operation. Notice that due to the nonlinear demodulation transfer function of the DPSK, not all the rest point can be stabilized with this circuit. The position around the extrema can not be achieved for the moment. Modified circuits should allow for the control around the extrema as well, but this would imply an additional slow modulation of the interference condition. 5.3 Synchronization results We reproduce in Fig. 22 the chaos cancellation demonstration, observed in the optical spectrum, with the light exiting the receiver in the scheme depicted in Fig.17. The black curve illustrates the spectral spreading around the laser line due to the chaotic phase modulation (this chaotic spreading was obtained with the maximum feedback gain, β=5); this spectrum is obtained when the chaos cancellation is deactivated, i.e. when the delayed feedback path at the receiver is open (in that way, the measured light beam is the same than the one available on the transmission line, and corresponding to the chaotically phase-modulated light beam).

Figure 22: Synchronization with a message free phase chaos unidirectional coupling

The red curves shows the chaos cancellation efficiency in the side bands of the laser line. A chaotic phase modulation suppression of more than 10-15dB is observed over more than +/-10GHz. This performance was checked to be very stable with the control circuit (stabilizing the rest point with an offset phase of π/4), over several hours of operation.

-18-


5.4 Packaging issues In order to have a transportable set-up for doing field experiment, we plan to gather the different discrete devices in a limited space. The schematic space occupation of the emitter is illustrated in Fig.23, with the arrangement of the different emitter devices.

Figure 23: Planned packaging of the discrete devices forming the emitter. The real size is about a A4 paper sheet.

The corresponding surface is approximately covering an A4 paper size. The same surface will be required by the receiver setup. External elements are still required, but they only consist in the electrical data input or output (e.g. PRBS generator for the emitter, and data analyzer for the receiver), and the different power supplies (for the on-board laser current and temperature control, for the DPSK rest point control, and for the active RF photodetector and for the RF driver). We expect to be able to perform various field experiments with such transportable emitter / receiver modules. 6. Conclusions In this deliverable the developed hybrid and monolithic integrated transmitter/receiver modules were presented in terms of their technical features, the chaos generation properties and the synchronization performance. Regarding their technical features, it is highlighted that in this devices there is the possibility of adjusting crucial parameters – such as the feedback strength and the phase of the internal field – that determine the dynamics emitted and the synchronization performance. In terms of their chaos generation properties, the plethora of diverse dynamics generated by these devices was demonstrated. In terms of the synchronization performance in a closed-loop configuration, the monolithic integrated devices are much more stable compared to the hybrid integrated counterparts due to the accuracy in the cavity length at both transmitter and receiver modules. A stability analysis was carried out highlighting the effect of temperature fluctuations on the synchronization error. Finally, the results coming from the study of the first packaged devices were presented illustrating the increased stability and the synchronization dependence on the exact matching of feedback phase at both ends.

-19-


REFERENCES

1. M. Peil, I. Fischer and W. Elsäßer, “A short external cavity semiconductor laser cryptosystem”, Comptes

Rendus Physique, v. 5, n. 6, pp. 633-642, July-August 2004. 2. T. Heil, I. Fischer, and W. Elsäßer and A. Gavrielides, “Dynamics of Semiconductor Lasers Subject to

Delayed Optical Feedback: The Short Cavity Regime”, Phys. Rev. Lett. 87, 243901 (2001) 3. R.J. Jones, P.S. Spencer, J. Lawrence, and D.M. Kane, “Influence of external cavity length on the

coherence collapse regime in laser diodes subject to optical feedback”, IEE Proceedings in Optoelectronics, v. 148, n. 1, p. 7-12, February 2001.

4. R. Lavrov, M. Peil, M. Jacquot, V.S. Udaltsov, L. Larger, and John Dudley, ''Electro-optic delay oscillator with non-local non linearity: optical phase dynamics, chaos, and synchronization'', submitted to Phys. Rev. E, April 2009

5. R. Lavrov, M. Peil, M. Jacquot and L. Larger, ''Electro-optic phase chaos for 10Gb/s chaos communications'' CLEO Munich 2009

-20-

photonic integrated components applied to secure chaos

Documents