all optical signal processing using dfb and dbr laser diodes...
TRANSCRIPT
Andualem Yimam
diodesAll optical signal processing using DFB and DBR laser
Academiejaar 2008-2009Faculteit IngenieurswetenschappenVoorzitter: prof. dr. ir. Daniël De ZutterVakgroep Informatietechnologie
Erasmus Mundus Master of Science in PhotonicsMasterproef ingediend tot het behalen van de academische graad van
Begeleider: Promotor: prof. dr. Geert Morthier
i
Permission for usage The author gives the permission to make this work available for consultation and to copy part of the work for personal use. Any other use is bound to the restriction of copyright legislation, in particular regarding the obligation to specify the source when using results of this work. Andualem Ali Yimam June 1st, 2009
ii
Acknowledgements I would like to extend my sincere gratitude to my promoter Prof. dr. ir. Geert Morthier and my supervisor ir. Koen Huybrechts for guiding me through the right direction throughout my work, and helping me whenever I needed it. I am thankful to consortium of Erasmus Mundus Master Program in Photonics, which gave me the opportunity to study in some of the best European institutes. My sincere gratitude goes to ir. Nebiyu Adello for giving me the moral and encouragement while studying in Belgium. Finally, I want to thank my parents for their continuous support and encouragement throughout my life.
iii
Abstract
In this thesis, an all-optical flip-flop operation based on a single DFB laser diode with a CW
injected light and a single DBR laser diode is numerically and experimentally investigated. In
addition, an all-optical 2R regenerator and wavelength converter based on a single DFB laser
diode is achieved numerically.
The work begins by a numerical investigation of an all-optical flip-flop operation by a single
DFB laser diode with a CW light injection. The effect of different laser parameters and operating
conditions on the bistability and switching times of the flip-flop operation is thoroughly
discussed. An experimental demonstration of the flip-flop operation and the effect of some
operating parameters on the switching times is also given. The thesis further extends on the
numerical demonstration of the use of a single DFB laser diode to achieve 2R regeneration and
wavelength conversion at 10Gbit/s. Finally, an all-optical flip-flop based on a single DBR laser
diode is carried out numerically and experimentally.
iv
Numerical and experimental study of the switching times and energies of DFB-laser based All-optical flip-flops
K. Huybrechts1, A. Ali1, T. Tanemura2, Y. Nakano2, G. Morthier1 1: Ghent University – IMEC, INTEC, Photonics Research Group, Sint-Pietersnieuwstraat 41, B-9000 Ghent, Belgium 2: Research Center for Advanced Science and Technology (RCAST), The University of Tokyo, 4-6-1 Komaba, Meguro-ku, Tokyo 153-8904, Japan (Submitted to Photonics in Switching 2009) Abstract: We report on the switching times and energies of a DFB AOFF and their dependence on device and operating parameters. Numerical as well as experimental results are discussed.
Keywords: all-optical flip-flop, switching times, DFB
1. Introduction All-optical flip-flops (AOFFs) could have an important role in future all-optical packet switches [1]. Previously [2,3], we have shown that a robust AOFF can be obtained from a AR-coated, λ/4-shifted DFB laser in which a CW beam is injected. The bistability for a certain power range of the CW beam is then due to the spatial hole burning, induced by the CW beam and affecting the mirror losses. It has been shown before that both the CW beam and the switching pulses can be of relatively arbitrary wavelength (but not too close to the lasing wavelength) and that switching in about 75 ps is possible with switching energies of 200 fJ [2]. In this paper we report on the thorough analysis of the switching times and energies that can be achieved with such an AOFF. Using numerical simulations, we have investigated how the switching times depend on the laser structure and on the operating conditions. The dependence on the operation parameters has been investigated experimentally on a single DFB lasers. The all-optical flip-flop configuration that has been investigated numerically and experimentally is shown in Figure 1. The reset pulses (to switch off the laser) are injected at the same side as the CW beam to induce a carrier density non-uniformity, whereas the set pulses (to switch on) are injected at the opposite side and serve to restore the carrier density uniformity.
Figure 1 : Configuration of the DFB-based all-optical flip-flop
2. Simulations and results
2.1 Brief description of the simulations
The simulations were done using the commercial software package VPI componentmaker (©, [4]). Some typical device parameters used in the simulations are given in the table below. The following device parameters have been varied: differential gain, normalized coupling coefficient, device length, the non-linear gain coefficient and the linear recombination coefficient. Furthermore, bias current, CW injected power and switching pulse power and duration have been investigated. We first simulated the bistability in the laser output power vs. the injected CW power and then studied the switching. A typical switching behaviour is as shown in Figure 2. The switching show the typical ringing behaviour found when switching on the bias current of laser diodes. Due to the fact that the average gain in the off-state is higher than the threshold gain, the ringing is however significantly more pronounced. Using a higher non-linear gain to dampen this ringing didn’t change much either. We define the switch-on time as the time over which the laser power increases from 10 to 90% of the steady state ON value. Reset pulse lengths were 200 ps and set pulse lengths 125 ps.
Figure 2 : Typical switching behaviour of DFB AOFF
2.2 Main results
Rise times and fall times (switch-on and switch-off times resp.) are given vs. the pulse power in Figure 3. The laser has κL=1.2, L=400μm and a bias current of 200mA. Switch-on times are decreasing significantly with increasing differential gain, fall times are independent of it. We investigated also the influence of kL, bias current, length, recombination parameters. Switch-off times decrease mainly with decreasing κL, whereas switch-on times decrease with increasing
v
bias current and differential gain. Furthermore, the required switching energy decreases sharply with decreasing κL.
Figure 3 : Simulated switch-on and switch-off times vs. the pulse power (dark lines are for dg/dN=3 e-20 m2 and grey lines for dg/dN=5 e-20 m2).
3. Experimental results
The measurements were all done on a DFB laser diode from Alcatel-Thales III-V Labs with a length of 400 μm and a κL of 1.6. The laser has a threshold of approximately 30 mA and emits light at 1553 nm. CW light with wavelength of 1543 nm was injected. For a bias current of 150 mA, the CW light injection resulted in a bistable region for input powers between about 6.5 and 7.3 dBm. Figure 4 shows how the switch-on and swich-off times decrease with the energy of the injected pulse. The injected pulses were 150 ps long. With sufficiently energetic switch pulses, switching times of 40 ps can be reached. Part of these switching times may be caused by rise and fall times of the injected pulses though.
Figure 4 : Experimental switch-on and switch-off times
(rise and fall times) vs. the switching energy.
Figure 5 shows how the switching time changes as the CW injected power is varied. As can be seen, there was no significant influence of the CW power in the experiments, as opposed to the numerical results.
. Figure 5 : Experimental switch-on and switch-off times
(rise and fall times) vs. the CW injected power.
4. Conclusion
We have discussed the switching times of a DFB-based AOFF and its dependence on various structural and operational parameters. Both experimental and numerical results have been shown. One of the main conclusions is that theoretically switch-on times of 10 ps and switch-off times of less than 20 ps are possible. Switch-off times and required switch energies decrease with decreasing laser length and decreasing κL. Switch-on times mainly decrease with increasing bias current and increasing differential gain. Parameters like non-linear gain coefficient and recombination rates have very little influence. Required switching energies decrease with decreasing κL.
5. Aknowledgment The authors gratefully acknowledge Alcatel-Thales III/V-labs for providing the DFB laser diodes. This work is supported by the Fund for Scientific Research (FWO), the IAP-project “Photonics@be” and the Erasmus Mundus program. The work of K. Huybrechts is supported by the Institute for the Promotion of Innovation through Science and Technology (IWT).
6. References
[1] H. J. S. Dorren, M. T. Hill, Y. Liu, N. Calabretta, A. Srivatsa, F. M. Huijskens, H. de Waardt, and G. D. Khoe, "Optical packet switching and buffering by using all-optical signal processing methods" Journal of Lightwave Technology 21, pp. 2-12, 2003
[2] K. Huybrechts, G. Morthier, R. Baets, “Fast all-optical flip-flop based on a single distributed feedback laser diode”, Optics Express, 16(15), p.11405-11410, 2008.
[3] K. Huybrechts, T. Tanemura, Y. Nakano, R. Baets, G. Morthier, “Fast 40 Gb/s Optical Packet Switching using an All-Optical Flip-Flop based on a single Distributed Feedback Laser”, OFC, San Diego, USA, 2009.
[4] VPItransmissionmaker, http://www.vpisystems.com
vi
All optical 2R Regeneration and wavelength conversion at 10Gb/s using a single bistable DFB laser diode
K. Huybrechts, A. Ali, G. Morthier Ghent University – IMEC, INTEC, Photonics Research Group, Sint-Pietersnieuwstraat 41, B-9000 Ghent, Belgium (Submitted to Photonics in Switching 2009)Abstract: We report on optical 2R regeneration and wavelength conversion at 10Gb/s using a single bistable DFB laser diode.
Keywords: DFB, 2R regeneration, wavelength conversion
1. Introduction
All optical signal regeneration and wavelength conversion play a crucial role in increasing transmission distances and reducing blocking probability in WDM network nodes in next generation networks [1]. In this paper we show numerically that wavelength conversion and optical regeneration comprising of enhancement of extinction ratio (ER) and signal reamplification (2R) can be achieved at 10Gbps using an AR-coated λ/4-shifted DFB laser diode which is biased above threshold. Previously [2,3], we have shown that a robust AOFF can be obtained from an AR-coated, λ/4-shifted DFB laser in which a CW beam is injected. It has been demonstrated that a bistability for a certain range of the CW beam originates from spatial hole burning and the wavelength of the CW beam as well as the pulses used for switching can be relatively arbitrary as long as it is not too close to the lasing wavelength of the laser. By setting the ‘0’-level of the signal at the left side of the bistability region, we have invesitigated numerically that also optical regeneration and wavelength conversion can be achieved at 10Gb/s. The simulations were done using the commercial software package VPI componentmaker (©, [4]).
2.1 Operation principle The simulation setup used to demonstrate 2R regeneration and wavelength conversion is shown in figure 1. 10Gb/s non-return-to-zero (NRZ) signals is generated using a pseudo random bit sequence (PRBS) generator, such that the ‘0’ level is set at the left side of the bistability region of the DFB laser. The signal to noise ratio of this signal is then controlled by a succession of a variable optical attenuator and an EDFA.The signal then passes through the device and an optical band pass filter is used to obtain the signal at the laser wavelength and the input signal wavelength. With an optical band-pass filter, we obtain the signal at the two different wavelengths and the signal power that arrives at the receiver is adjusted by a variable optical attenuator.
This is useful in plotting the BER vs received power plots. Figure 1: Simulation setup used for the 2R-regeneration and wavelength conversion simulations. VOA: variable optical attenuator, EDFA: erbium doped fiber amplifier, TOF: tunable optical filter, ORx: optical reciever
2.2 2R Regeneration
For 2R regeneration, we use a DFB laser with normalized coupling coefficient кL of 1.8, L=400μm, differential gain of 5 10-20 m2. The drive current of the laser is 130mA. Figure 2 shows eye diagrams for the 10Gb/s back to back NRZ signal and for the output signal after the DFB laser. The wavelength of the signal is 1560nm. The extinction ratio has improved from 7dB to 10dB. More importantly, there is a very clear and significant noise reduction on both the ‘0’s and the ‘1’s. Figure 3 shows the BER for the input and output signals. We have seen almost the same sensitivity for a BER value of 10-9.
Figure 2: a) Eye diagram of a 10Gbit/s NRZ back to back signal with ER of 7.5dB. b)Eye diagram of the regenerated output signal with ER of 10 dB.
a)
b)
EDFA TOF ORx
10Gb/s NRZ PRBS
VOA DFB- LD
VOA
vii
Figure 3 : BER as a function of the signal power incident on the receiver (dashed line is for the back to back signal and solid line is for regenerated output signal.
2.3 Wavelength Conversion
For wavelength conversion simulations, the parameters used for the DFB laser are the same as before, but now we vary the differential gain between 5 10-20 m2 and 6 10-20 m2. The drive current used for the laser is 240mA. The wavelength converted laser output signal has a wavelength of 1569nm while the input signal wavelength is 1560nm. This device is well suited for wavelength conversion as the input signal wavelength can be relatively arbitrary as long as it is not very close to the lasing wavelength of the laser. Figure 4 shows eye diagrams for the 10Gb/s back to back NRZ signal and for the wavelength converted laser output signal. The input extinction ratio is 6.5 dB and the output signal ER is 18dB. This results in an extinction ratio improvement of 11.5dB. This is, however, accompanied by around 3.5dB power penalty in the receiver sensitivity as
Figure 5 : BER as a function of the signal power incident on the receiver (dashed line is for the back to back signal and the others are for the wavelength converted laser output signal (triangles are for dg/dN=6 10-20 m2 and squares for dg/dN=5 10-20 m2). shown in the BER plots in Figure 5 for the back to back and wavelength converted signals. Better receiver sensitivities have been obtained for relatively larger differential gain values. A clear noise reduction can again be observed from the eye diagrams
3. Conclusion
We have numerically demonstrated that both optical signal regeneration and wavelength conversion at 10Gb/s can be performed with a single DFB laser diode. For optical 2R regeneration, a clear noise reduction together with an extinction ratio improvement have been obtained. Wavelength conversion has also been shown to exhibit an extinction ratio improvement of 11.5dB albeit with around 3.5dB power penalty in the receiver sensitivity. Eye diagrams show a clear noise reduction in the wavelength conversion as well. Higher differential gain values show better sensitivity results.
4. Aknowledgment
This work is supported by the Fund for Scientific Research (FWO), the IAP-project “Photonics@be” and the Erasmus Mundus program. The work of K. Huybrechts is supported by the Institute for the Promotion of Innovation through Science and Technology (IWT).
5. References
[1] Yoo, S.J.B, "Wavelength conversion technologies for WDM network applications" IEEE J. Lightwave Technol. 14(6), 955–966,1996.
[2] K. Huybrechts, G. Morthier, R. Baets, “Fast all-optical flip-flop based on a single distributed feedback laser diode”, Optics Express, 16(15), p.11405-11410, 2008.
[3] K. Huybrechts, T. Tanemura, Y. Nakano, R. Baets, G. Morthier, “Fast 40 Gb/s Optical Packet Switching using an All-Optical Flip-Flop based on a single Distributed Feedback Laser”, OFC, San Diego, USA, 2009.
[4] VPItransmissionmaker, http://www.vpisystems.com
1E-111E-101E-091E-081E-071E-061E-051E-041E-031E-021E-01
-27 -25 -23 -21 -19 -17 -15
BER
Received power (dBm)
1E-171E-151E-131E-111E-091E-071E-051E-031E-01
-28 -23 -18 -13
BER
Received power(dBm)
Figure 4: a) Eye diagram of a 10Gbit/s NRZ back to back signal with ER of 6.5dB. b) Eye diagram of the wavelength converted signal with ER of 18 dB.
a)
b)
viii
ALL-OPTICAL FLIP-FLOP BASED ON A SINGLE DBR LASER DIODE Andualem A. Yimam
Promoter: Prof. dr. ir. G. Morthier Supervisor: ir. K. Huybrechts
Ghent University – IMEC, INTEC, Photonics Research Group, Sint-Pietersnieuwstraat 41, B-9000 Ghent, Belgium Abstract: This paper includes the part of the thesis which is not included in the previous two papers submitted to a conference. In this paper, a numerical and experimental investigation of an AOFF based on a single DBR laser diode is presented. Keywords: DBR, AOFF, switching times I. Introduction Driven by the ever increasing need for bandwidth, modern telecommunication networks are undergoing a vast expansion. To that end, the transformation of the current networks where optical-electrical-optical conversions are imposing an electrical bottleneck in the system to all-optical networks is drawing a considerable attention. One of the key functional blocks of these all-optical networks are all-optical flip-flops. Different optical flip flops use different mechanisms of operations and hence each has its own advantages and disadvantages. Flip flops based on coupled Mach-Zehnder interferometer [9], coupled laser diodes [10] and optical feedback between a SOA and DFB laser diode[11] need two relatively large sources for operation. AOFFs which don’t require two active components and without external CW beam can be fabricated making use of saturable absorbers [12], but they have the disadvantage of having a limited maximum repetition rate due to the slow carrier lifetime in the absorber section and they need relatively high energy pulses for switching. Dispersive bistabilities in DFB amplifiers have also been used to achieve flip-flop operation [13], but they require a very tight wavelength control, large switching times and relatively large pulse energies. Another recently proposed alternative is based on ring or disk lasers, in which switching is obtained between the clockwise and counter clockwise direction. Flip-flops based on such lasers can be switched very fast and don’t require an external CW beam [14]. However, fast switching and single mode operation is only obtained if the ring/disk radius is sufficiently small. This makes the fabrication difficult and results in strong, uncontrollable variations of the lasing wavelength. Since the switching is achieved through injection locking, ring/disk lasers also require very tight wavelength control of the switching pulses. Other alternatives proposed in this thesis include AOFF’s based on a single DFB laser with a CW light
injection and a single DBR laser diode. Even though switching with DFB laser diodes with CW light injection doesn’t need very tight wavelength control, needs smaller switching pulse energies, and exhibits smaller switching times [15], AOFF’s based on a DBR laser also have an advantage of working without a CW light injection. Flexibility in tuning of DBR lasers can be exploited to change the lasing wavelength of the lasers to the wavelength of the set and reset pulses.
II. Operation principle All-optical flip-flop operation using DBR laser diodes relies on a bistability in the wavelength versus tuning current characteristics. At the heart of this bistability is an asymmetric gain suppression due to the four wave mixing between the main mode and the side mode [16]. Due to the hysteresis phenomena in the tuning characteristic, the Bragg tuning current corresponding to a mode hop depends on whether Bragg current is increased or decreased. For the hysteresis to be exploited for switching, the laser should be long enough and the reflection spectrum from the Bragg gratings should be broad enough. By biasing the Bragg section with a current in the bistability region, a DBR laser can be forced to be in one of the branches of the bistability. Switching to the second branch can be achieved by using the technique of injection locking with pulses of the same wavelength as the second branch. Switching back to the first branch can be performed by either injection locking with pulses of the same frequency as the first branch or by injecting very short and high power pulses with the same frequency as the second branch through carrier depletion effects.
III. Simulation results A bistability in the wavelength versus tuning current characteristic of a two section DBR laser diode, consisting of a 350 μm active section and a 150 μm Bragg section with a кL of 0.9 has been simulated using a commercial software package VPI componentmaker (©, [17]) and is shown in Figure 1. It shows the relative emission frequency with respect to 193.1 THz obtained for increasing and decreasing Bragg current. By using a Bragg current of 2.6mA and applying 8mW and 400ps long pulses at a relative frequency with respect of 193.1THz of -218GHz, the laser could be switched to the lower branch as shown in Figure 2a through injection locking.
ix
Switching to the upper branch again has been performed by injecting 40mW and 100ps pulses which deplete the carrier density in the gain section making the laser relax back to the upper branch which is characterized by the lowest carrier density. Figure 2b shows filter output of the upper branch at a relative frequency of -155GHz. Fall times are 30ps and rise times are 50ps in this case.
IV. Experimental Results
For the measurements a three section DBR laser with active section length of 600μm, a phase section of 60μm and a Bragg section of 250μm is used. By using a gain current of 100mA, the tuning current has been varied while looking at the emission wavelength with a digital scope at the same time. By doing so, the DBR laser has shown bistability in between Bragg current values of 20mA and 23.5mA. The bistability occurs for two wavelength branches of 1562nm and 1561.4nm. The experimental setup used for the demonstration of the AOFF operation is shown in Figure 3.
The pico-second pulse source produces very strong and 7ps length pulses that make the set pulse train. The
tunable laser and a modulator driven by a NRZ bit pattern generator produce 100ps pulses which are used as reset pulses. The modulators are driven by a NRZ bit pattern generator. The modulator for the pico-second pulse source is used to reduce its 10GHz repetition rate to lower repetition rates. The same signal generator is used to drive the bit pattern generator of the modulators and the pulse source. A succession of an erbium doped fiber amplifier (EDFA) and an attenuator are used to independently control the set and reset pulse powers. A circulator, then, injects this signal to the gain section side of the DBR laser diode and at the same time helps to extract one of the two branches of the bistability regime by connecting a tunable band pass filter at its other terminal. The signal out of this filter is fed to an optical digital sampling scope which displays the flip-flop operation. By using 100ps and 1.2pJ (reset) and 7ps, 6.5pJ (set) pulses for switching at a repetition rate of 350MHz, flip-flop behavior has been demonstrated as shown in Figure 4 albeit with less stable results. The reason for the stability problem has been attributed to the limited extinction ratio of the modulator used to decrease the repetition rate of the pico-second pulse source which cannot suppress the pulses completely. Another possible reason could be the wider spectral linewidth of the very short pulses. Fall times less than 100ps and rise times less than 200ps could be observed.
V. Conclusion
It has been demonstrated both numerically and experimentally that a single standard DBR laser can be used to make a fast and flexible all-optical flip-flop without a need for a CW light injection by using set and reset pulses of different length and energy but with the same wavelength. The problem with the stability of experimental results is attributed to limited extinction ratio of the modulator and wider spectral linewidth of the very short set pulses. Simulation results indicated that switching times decrease with an increase in pulse energy and larger Bragg section lengths lead to larger switching times and require larger pulse energies. Bragg sections with higher кL have shown a pronounced hysteresis and injection locking in such cases could be accomplished with smaller pulse energies. Tunability can be exploited to change the lasing wavelength of the lasers to the wavelength of the set and reset pulses.
DBR-LD
Pico-second pulse source
Pulse generator (10.6GHz)
EDFA
EDFA
ATT
ATT Coupler
EDFA
OBPF
Pattern Generator
Tunable Laser Figure 3: MOD: modulator EDFA: erbium doped fiber amplifier ATT: attenuator OBPF: optical band pass filter
50μW/div
500ps/div Figure 4
Figure 1
Figure 2b Figure 2a
MOD
MOD
scope
OBPF
x
Contents 1. Introduction…..…………………………………………………………………………..…….1
1.1 The Need for All-Optical Signal Processing………………...…………….………….1
1.2 Routing in Optical Networks………………………………………………………….2
1.3 All-Optical Flip-Flops………………………………………………………………...4
1.4 Overview of the Thesis………………………………………………………………..6
1.5 References……………………………………………………………………………..7
2. All-optical Flip-flop Based on a DFB Laser Diode………….……….…………………..........9
2.1 A Brief Introduction to DFB Lasers…………………………………………………..9
2.2 Bistability in a DFB Laser Diode…………………………………………………….13
2.3 Flip-flop Operation in a DFB Laser………………………………………………….16
2.4 Simulation Results……………………………………………………………………18
2.4.1 Simulation setup……………………………………………………………18
2.4.2 Dynamic Flip Flop Operation and the Effect of Different parameters
on the hysteresis curve and switching times……………………………..….19
2.5 Summary…………………………………………………………………………..….33
2.6 References…………………………………………………………………………….34
3. Experimental Investigation of Flip-flop Operation Based on a Single DFB Laser Diode
Injected with CW light injection………………….……………………………………………...35
3.2 Flip-flop operation…………………………………………………………………....36
3.2.1 Experimental setup…………………………………………………………36
3.2.2 Experimental results………………………………………………………..38
3.3 Experimental Investigation of the switching times on operating parameters..............39
3.4 Summary......................................................................................................................40
3.5 References....................................................................................................................41
4. All-optical 2R Regeneration and Wavelength Conversion Using a Single Bistable DFB laser
Diode……………………………………….……………………………………………………. 42
4.1 Simulation Setup…………………………………………………………………….. 42
4.2 2R Regeneration……………………………………………………………..............44
xi
4.2.1 2R Regeneration at 2.5Gb/s………………………………………………..45
4.2.2 2R Regeneration at 10Gbit/s………………………………………………46
4.3 Wavelength Conversion………………………………………………………………47
4.3.1 Wavelength conversion at 2.5Gbit/s………………………………………..48
4.3.2 Wavelength conversion at 10Gbit/s………………………………………...49
4.4 Summary……………………………………………………………………………...51
4.5 References…………………………………………………………………………….51
5. All-optical Flip-flop Based on a Single DBR Laser Diode………………..…………………..52
5.1 Distributed Bragg Reflectors…………………………………………………………52
5. 2 Tunability in DBR Laser Diodes………………………………………………….....55
5.3 Bistability in DBR Lasers…………………………………………………...……..…59
5.4 Flip-flop Operation in a DBR Laser………………………..………………………...61
5.5 Simulation Results……………………………………………………..……………..63
5.5.1 Simulation setup……………………………………………………………63
5.5.2 Dynamic Flip Flop Operation and the Effect of Different Parameters on
the Hysteresis Curve and Switching times…………………………………64
5.6 Experimental Investigation of Flip-flop Operation in a Single DBR Laser Diode…..72
5. 6.1 Bistability…………………………………………………………………..72
5.6.2 Flip-flop Operation…………………………………………………………73
5.6.2.1 Using pulses of the same length but different energies…………..73
5.6.2.2 Using pulses of different energies and different lengths…………75
5.7 Summary……………………………………………………………………………...77
5.8 References…………………………………………………………………………….79
6. Conclusion and Recommendation……………………………………………………………..80
6.1 Conclusion……………………………………………………………………………80
6.2 Recommendations…………………………………………………………………….82
6.3 References…………………………………………………………………………….83
List of Figures…………………………………………………………………………………….84
List of Tables……………………………………………………………………………………..87
xii
List of Abbreviations
WDM Wavelength Division Multiplexing
EDFA Erbium Doped Fiber Amplifier
WWW World Wide Web
IP Internet Protocol
OCS Optical Circuit Switching
OBS Optical Burst Switching
OPS Optical Packet Switching
AWG Arrayed Waveguide Grating
AOFF All-Optical Flip-Flop
SOA Semiconductor Optical Amplifier
DFB Distributed Feedback
CW Continuous Wave
DBR Distributed Bragg Reflector
ER Extinction Ratio
AR Anti-Reflection
TMM Transfer Matrix Method
ASE Amplified Spontaneous Emission
NRZ Non-Return-to-Zero
NZ Return-to-Zero
PRBS Pseudo Random Bit Sequence
BER Bit Error Rate
1
1 1. INTRODUCTION
1.1 The Need for All-Optical Signal Processing
Recently, global telecommunication networks are undergoing a vast expansion following the
introduction of wavelength division multiplexing (WDM), erbium doped fiber amplifier (EDFA),
fiber Raman amplifier, etc. This expansion is driven by the ever increasing need for bandwidth as
people move forward from text-based Email system to the World Wide Web (WWW) to video
broadcasting websites, from voice communication to voice over IP to online video conferencing,
from online chatting to online gaming. To alleviate the ever increasing hunger for bandwidth,
researchers have been devoted in increasing transmission limits [1]. The theoretical bandwidth of
25THz of a typical single mode fiber compared to the 750MHz bandwidth of a coax cable
allowed the evolution of telecommunication to its current state. The fiber is now the backbone of
modern telecom infrastructures. Data throughput in optical fiber has been increased by
wavelength division multiplexing in which multiple signals of different wavelengths coming
from different sources are sent through the fiber at the same time [2].
Apart from high transmission capacity, switching is another crucial functionality in
communication networks [3]. When data is transported from a source to a destination, it travels
through a number of channels via intermediate network nodes. As data passes through these
intermediate nodes, it needs to be routed from the origin to the destination along a prescribed
pathway. Hence, switching is necessary at each intermediate node. Currently, communication
networks consist of network nodes built from electronic boxes (switches and routers) which are
interconnected by optical pipes. Processing in electronic routers at each node is, hence, possible
after the different wavelength channels in the optical fiber are demultiplexed and converted
separately into the electrical domain. Once the channels are processed separately in the electrical
domain, conversion to optical domain and multiplexing are necessary before sending the data
further. Switching in the electrical domain helps to avoid collision and service degradation by
2
using complicated algorithms based on the global network information. Even though switching in
the electrical domain is efficient and uses a mature technology, it suffers a number of
disadvantages at high bit rates and it becomes a bottleneck as the transmission capacity increases
and the problem even worsens when different wavelength channels are present in one optical
fiber ( WDM networks). At high data rates, electronic signal processing is very expensive due to
costly transceiver technologies and packaging and the need for a large number of optical
receivers, modulators and lasers in WDM networks. Besides, it consumes large space and high
power and becomes extremely difficult (physical limitations of electronics). Hence, performing
all the signal processing in the optical domain without conversions to the electrical domain would
alleviate these problems and the need for all-optical networks is justified [4]. In all-optical
networks, routing, buffering and error correction are all done in the optical domain.
1.2 Routing in Optical Networks
Switching in all-optical networks is performed by three basic optical switching technologies.
These include optical circuit switching (OCS) [3], optical burst switching (OBS) and optical
packet switching (OPS)[5]. Figure 1.1 shows a typical optical network that works by optical burst
or packet switching.
Figure 1.1Schematics of an all-optical network (courtesy of [6])
3
In all-optical networks that work by optical burst or packet switching, sending an IP data packet
starts by conversion of electronic data to an optical data. A header will be added to the data
packet before it is added to a local fiber ring by using an add-drop multiplexer. The header
information is used to route the packet through the network. Data coming from different users is
identified based on different wavelengths and multiplexed on the local area fiber ring and then
sent to a core network through an edge router. Routing takes place in the core routers based on
the header information. The packet is then transferred to another edge router which sends the data
to a fiber link with a specific wavelength. At the destination of the data packet, an optical add-
drop multiplexer drops the packet and conversion to an electrical packet takes place. In all-optical
networks that involve packet switching, advanced all-optical signal processing functions
including all-optical header recognition, buffer, switching, wavelength conversion, logic gates
and flip-flop memory are essential. This can be observed in a 1x2 optical packet switch shown in
Figure 1.2 [7].
Figure 1.2: Schematics of a 1×2 all-optical packet switch.
In an optical packet switch there are three functional blocks: an all-optical header processing
block, an all-optical flip-flop memory block, and a wavelength conversion block. When an
optical packet arrives at the optical switch, it is split into two parts. One of the parts is used for
header processing and routing while the other one is delayed to compensate for the time taken to
carry out the header processing functions. On the part going through the header processor, a
comparison between the header of the packet and a predefined list of addresses is used to
determine whether a pulse should be generated or not. If a pulse is generated, it will change the
state of the all-optical flip-flop. The wavelength converter receives the signal coming from the
Output 1
Output 2
State 2 State 1
Wavelength conversion
AWG
Header
Processor Optical Flip- flop
Delay line
4
all-optical flip-flop and the buffered data packet on the other path. Depending upon the output of
the all-optical flip flop, the wavelength converter will either perform wavelength conversion on
the data packet or it will send it without converting its wavelength. An AWG (arrayed waveguide
grating), finally, routes the wavelength channel to the appropriate output port.
1.3 All-Optical Flip-Flops
As discussed above, one of the major functional blocks in packet or burst switched networks is
the all-optical flip-flop (AOFF). An optical flip-flop is a device that can operate in one of two
(more) stable optical states for the same operating conditions and its output depends on its
history. The output states of optical flip-flops change by the injection of optical set and reset
pulses.
Different optical flip flops use different mechanisms of operations and hence each has its own
advantages and disadvantages. A brief overview of different optical flip-flops has been given in
[8]. Some of them include flip-flops that work based on bistable laser diodes, coupled laser
diodes, coupled Mach-Zehnder interferometers, polarization bistable lasers, DFB SOA’s and
optical feedback scheme between a SOA and DFB laser diode.
The most important parameters of interest while discussing AOFFs are their switching times
(10%-90% of the transition time between the different states), the contrast ratio (the difference
between the output power levels of the states), the repetition rate (rate at which two equal
transitions between states occur), the need for a continuous CW light power for operation, the
need for tight wavelength control, fabrication complexity, the pulse energy and pulse length and
the total power consumption of the device.
Flip flops based on coupled Mach-Zehnder interferometer [9], coupled laser diodes[10] and
optical feedback between a SOA and DFB laser diode[11] need two relatively large sources for
operation. AOFFs that make use of coupled Mach-Zehnder interferometer and coupled laser
diodes also suffer from relatively lower repetition rates. In addition, coupled Mach-Zehnder
interferometers require relatively large energy pulses for switching.
5
AOFFs which don’t require two active components and without external CW beam can be
fabricated making use of saturable absorbers [12], but they have the disadvantage of having a
limited maximum repetition rate due to the slow carrier lifetime in the absorber section and they
need relatively high energy pulses for switching.
Dispersive bistabilities in DFB amplifiers have also been used to achieve flip-flop operation [13],
but they require a very tight wavelength control, large switching times and relatively large pulse
energies.
Another recently proposed alternative relies on the use of ring or disk lasers, in which switching
is obtained between the clockwise and counter clockwise direction. Flip-flops based on such
lasers can be switched very fast and don’t require an external CW beam [14]. However, fast
switching and single mode operation is only obtained if the ring/disk radius is sufficiently small.
This makes the fabrication difficult and results in strong, uncontrollable variations of the lasing
wavelength. Since the switching is achieved through injection locking, ring/disk lasers also
require very tight wavelength control of the switching pulses.
While most of the approaches mentioned above are relatively complex or require a difficult active
passive integration, in this work, AOFF has been carried out using single DFB laser diodes with
CW light injection and single DBR laser diodes. While DFB laser diodes need CW light beam for
operation, DBR laser diodes depend on injection locking and hence require a tight control of the
wavelength of the switching pulses. However, these two approaches stand in a reasonably good
position from the alternatives mentioned above due to their mature technology and hence easier
fabrication, their relatively smaller switching times, and their ability to operate at relatively larger
repetition rates. Switching with DFB laser diodes with CW light injection, especially, needs
smaller switching pulse energies and exhibits smaller switching times [15]. The need of CW light
power to make DFB lasers in the bistable region is not a major problem as single CW power can
be used for several flip-flops at the same time introducing only minor additional cost. Flexibility
in tuning of DBR lasers can be exploited to change the lasing wavelength of the lasers to the
wavelength of the set and reset pulses.
6
1.4 Overview of the Thesis
This thesis is organized as follows. In the second chapter, an all-optical flip flop based on a single
distributed feedback (DFB) laser with a CW injected light is discussed. A brief introduction of
DFB lasers is first given. Further, it is shown that these lasers become bistable upon injection of a
continuous wave (CW) light with a wavelength different from the lasing wavelength of the laser
due to spatial hole burning. Switching between the two states of the bistability is accomplished by
using optical pulses. All-optical flip flop operation by using optical pulses is investigated
numerically on a single λ/4 shifted DFB laser with antireflection coated facets working above
threshold. The effect of the different parameters of the laser and operating conditions on the
switching times is discussed in detail.
In the third chapter, bistability and flip-flop operation of λ/4-shifted DFB-lasers with a length of
400 μm and a κL-value of 1.6 is investigated experimentally. The effect of operating conditions
on the switching times is also analyzed experimentally.
In the fourth chapter a single, AR-coated, λ/4-shifted, bistable DFB laser diode will be used to
achieve wavelength conversion and optical regeneration comprising of enhancement of extinction
ratio (ER) and signal reamplification (2R). By setting the ‘0’-level of the signal at the left side of
the bistability region, it is investigated numerically that optical regeneration and wavelength
conversion can be achieved at 10Gbit/s.
In the fifth chapter, an all-optical flip flop based on a single distributed Bragg reflector (DBR)
laser diode is discussed. A brief step by step introduction of DBR lasers is first given. Bistability
in the tuning behaviour of these lasers is exploited to enable switching between two wavelengths
with set and reset pulses of the same wavelength but of different lengths and energies. All-optical
flip-flop operation by using optical pulses is first investigated numerically and the effect of some
operating and device parameters on the switching times is discussed. Then, experimental results
showing the flip-flop operations are also given at the end of the chapter.
Finally, conclusions are drawn and recommendations for future research are also proposed.
7
1.5 References
[1] H. Weber, S. Ferber, M. Kroh, C. Langhorst, R. Ludwig, V. Marembert, C. Boerner, F.
Futami, S. Watanabe, and C. Schubert, “Single channel 1.28 Tbit/s and 2.56 Tbit/s DQPSK transmission," Proceedings of European Conference on Optical Communications, Th 4.1.2, 2005.
[2] L. Leng, S. Stulz, B. Zhu, L. E. Nelson, B. Edvold, L. Gruner-Nielsen, S. Radic, J. Centanni, and A. Gnauck, “1.6-Tb/s (160 x10.7 Gb/s) transmission over 4000 km of nonzero dispersion fiber at 25-GHz channel spacing,” IEEE Photon. Technol. Lett., vol. 15, no. 8, pp. 1153–1155, 2003.
[3] R. Ramaswami and K. Sivarajan, Optical networks: A practical perspective. UK: Morgan Kaufmann Publishers, 1998.
[4] I. Tafur Monroy, E. Van Breusegem, T. Koonen, J. Jennen, C. Peucheret, and E. Zouganeli, “Optical label switched networks: Laboratory trial and network emulator in the ISTSTOLAS project,” IEEE Communications Magazine, vol. 44, pp. 43–51, 2006.
[5] S. J. Ben Yoo, “Optical Packet and Burst Switching Technologies for the Future Photonic Internet,” Journal of Lightwave Technology, vol. 24, no. 12, 2006.
[6] Lectures by Yu Ben, Qian Ying Tang, “Optical Packet Switching”, http://inst.eecs.berkeley.edu/~ee233/sp06/student_presentations/EE233_Optical_paket_switching.pdf, 2006.
[7] PHD Thesis by Nicola Calabretta, “All-optical header processing based on nonlinear gain and index dynamics in semiconductor optical amplifiers,” Technische Universiteit Eindhoven, 2004.
[8] PHD Thesis by Wouter D’Oosterlinck, “Non-Linear Behaviour in a Semiconductor Optical Amplifier and Laser Diode Feedback Scheme,” Ghent University, Engineering Faculty, Information Technology (INTEC) Group, 2007.
[9] Y. Liu, R. McDougall, J. Seoane, E. Kehayas, M. T. Hill, G. Maxwell, S. Zhang, R. Harmon, F. Huijskens, L. Rivers, P. V.Holm-Nielsen, J. M. Martinez, J. Herrera, F. Ramos, J. Marti, H. Avramopoulos, P. Jeppesen, A. M. J. Koonen, A. J. Poustie, and H. J. S. Dorren, “Characterization of hybrid integrated all-optical flip-flop memory,” in LEOS 2006, pp. 943–944, 2006.
[10] M. T. Hill, H. de Waardt, G. D. Khoe, and H. J. S. Dorren, “All-optical flip-flop based on coupled laser diodes,” IEEE J. Quantum Electron., vol. 37, no. 3, pp. 405–413, 2001.
[11] W. D’Oosterlinck, F. Ohman, J. Buron, S. Sales, A. P. Pardo, A. Ortigosa-Blanch, G. Puerto, G. Morthier, and R. Baets, “Alloptical flip-flop operation using a SOA and DFB laser diode optical feedback combination,” Opt Express, vol. 15, no. 10, pp. 6190–6199, 2007.
[12] K. Takeda, M. Takenaka, M. Raburn, X. Song, J. Barton, and Y. Nakano, “Single mode and dynamic all-optical flip-flop operation of multimode interference bistable laser diodes with distributed Bragg reflectors,” in Proc. ECOC 2006, p. Th 1.4.4, 2006.
[13] D. N. Maywar, G. P. Agrawal, and Y. Nakano, "All-optical hysteresis control by means of cross-phase modulation in semiconductor optical amplifiers," J. Opt. Soc. Am. B - Opt. Phys. 18, 1003-1013 (2001).
[14] B. Li, M. Memon, G. Mezosi, G. Yuan, Z. Wang, M. Sorel, S. Yu, ‚“All-Optical Response of Semiconductor Ring Laser to Dual-Optical Injections”, IEEE Phot. Techn. Lett., Vol. 20, pp. 770-772, May 2008.
8
[15] Koen Huybrechts, Geert Morthier and Roel Baets, “Fast all-optical flip-flop based on a single distributed feedback laser diode,” OPTICS EXPRESS 11405, Vol. 16, No. 15,2008.
[16] A. P. Bogatov, P. G. Eliseev, and B. N. Sverdlov, “Anomalous Interaction of Spectral Modes in a Semiconductor Laser ,” IEEE J. Quantum Electron.,Vol. QE-11, No. 7, 1975.
[17] VPItransmissionmaker, http://www.vpisystems.com
9
2 ALL-OPTICAL FLIP-FLOP BASED ON A DFB LASER DIODE
In this chapter, an all-optical flip-flop based on a single distributed feedback (DFB) laser with
CW light injection is discussed. A brief introduction of DFB lasers is first given. Further, it is
shown that these lasers become bistable upon injection of a continuous wave (CW) light with a
wavelength different from the lasing wavelength of the laser due to spatial hole burning.
Switching between the two states of the bistability is accomplished by using optical pulses.
All optical flip flop operation by using optical pulses is investigated numerically on a single λ/4
shifted DFB laser with antireflection coated facets working above threshold. The effect of the
different parameters of the laser and operating conditions on the switching times is discussed in
detail.
2.1 A Brief Introduction to DFB Lasers
Distributed feedback (DFB) semiconductor lasers were developed during the 1980s and are now
one of the main building blocks in modern optical communication networks and are used
routinely for WDM lightwave systems [3]. Such lasers are the workhorse of demanding optical
communication.
The feedback in DFB lasers, as the name implies, is not localized at the facets but is distributed
throughout the cavity length. This is achieved through an internal built-in grating that leads to a
periodic variation of the mode index. Feedback occurs by means of Bragg diffraction, a
phenomenon that couples the waves propagating in the forward and backward directions [1].
Mode selectivity of the DFB mechanism results from the Bragg condition: the coupling occurs
only for wavelengths near λ B satisfying:
Λ = m(λB/2neff) (2.1)
10
where Λ is the grating period, neff is the average mode index, and the integer m represents the
order of Bragg diffraction. The coupling between the forward and backward waves is strongest
for the first-order Bragg diffraction (m = 1).
The Bragg propagation constant then becomes:
β0=m / Λ (2.2)
Analysis of DFB lasers can be done by writing the field as a sum of the forward and backward
propagating waves [2].
E z R z exp jβz S z exp jβz (2.3)
Where R(z) and S(z) are right and left propagating waves with a wave vector β close to the wave
vector at Bragg wavelength, β0.
Considering a periodic variation of the refractive index in the propagation direction according to:
n(z)=neff+Δncos(2 β0z), (2.4)
assuming a slow variation of R(z) and S(z) with z, neglecting second order derivatives of R and
S, and noting that Δβ=β-β0 (the real part of the Bragg deviation) is very small compared to β0, one
can arrive at the coupled-mode equations: R j∆β g R jкS (2.5)
S j∆β g S jкR (2.6)
Where к π∆λ
, is the coupling coefficient and g0 is the gain for the field in a DFB laser with
non-reflecting facets. If R and S are known at z=0, the coupled-mode equations can be solved for
R and S[2] and their values at z=L where L is the length of the laser will be:
R L cosh γL ∆βγ
sinh γL R 0 к γL Sγ
(2.7)
S L к γL Rγ
cosh γL ∆βγ
sinh γL S 0 (2.8)
Where γ к ∆β jg
A transfer matrix can, hence, be developed as:
R LS L F кL, g, ∆β R 0
S 0 (2.9)
W
T
a
T
t
W
s
n
F
A
w
t
F
(
Where matri
The oscillati
and becomes
The net gain
the intensity
Where Ґ is th
solution to
normalized i
Figure 2.1: L
As can be se
wavelengths
the main mo
Figure 2.2:
(c)Typical o
ix elements f
ion condition
s:
γL
n that must b
y gain 2g0. Th
2g
he confinem
equation (2
intensity end
Loci of DFB
een in Figure
s placed sym
ode.
a) Distribu
output spectru
follow equat
n holds when
L coth γL
be supplied t
hat is
g0=Ґgth-αi
ment factor, g
.10) is disp
d loss and th
B modes in th
e 2.1, modes
mmetrically a
(a)
uted feedbac
um from a D
Norma
tions 2.7 and
n setting ele
j ∆βL
to compensa
gth is the thre
played in Fi
he normalized
he (ΔβL, 2g0
s are located
around the B
ck (DFB) la
DFB laser. Ia
alized detuning
d 2.8.
ement 2,2 of
jg L
ate for the po
eshold gain a
gure 2.1. It
d detuning.
0L) plane for
d symmetrica
Bragg wavele
aser structur
a: injection c
g, ΔβL
Ideal Lasi
f the transfer
(2.10)
ower emitted
(2.11)
and αi is the
t shows the
r various val
ally around
ength. This
(b)
re. (b) Idea
urrent.
ng Emission
matrix to be
d from the en
internal loss
relationship
lues of кL
ΔβL and las
is shown in
(c)
al lasing em
Optical Power
e equal to ze
nds is equal
s. A numeric
p between t
sing occurs f
Figure 2.2 f
mission outpu
11
ero
to
cal
the
for
for
ut.
12
This degeneracy can be lifted by creating a phase shift in the middle of the grating. Due to
symmetry, the round-trip phase at Bragg wavelength is . Hence, by introducing a ΔL=λ/(4neff)
phase shift in the middle of the grating, the round- trip over this part corresponds to a phase of
and makes the entire phase over a round trip 2 at the Bragg wavelength. These are called λ/4
shifted DFB’s and are capable of providing much larger gain margin than conventional DFB
lasers [1].
For a λ/4 shifted DFB laser with AR-coated facets, the lasing condition can be found from a
transfer matrix [2]:
F(кL)=F(кL/2)Fmid( /2)F(кL/2) (2.12)
Where Fmid( /2) is the transfer matrix for the phase shift introduced in the middle of the grating.
The oscillation condition can again be evaluated to be:
γ coth L j ∆β jg к (2.13)
This is shown in Figure 2.3 below and we can see that there is a mode exactly at the Bragg
wavelength with the smallest end loss and hence this mode will lase.
Figure 2.3: Loci of λ/4 shifted DFB modes in the (ΔβL, 2g0L) plane for various values of кL.
Normalized detuning, ΔβL
13
As can be seen in Figure 2.1 and Figure 2.3, the end losses are higher for low values of кL and
for wavelengths far from the Bragg wavelength as the grating doesn’t provide efficient feedback
for these situations. 2.2 Bistability in a DFB Laser Diode
Optical bistability refers to the situation in which two stable optical output states are associated
with a single input state depending on the history of the device [4].
Optical bistability in a DFB laser diode can be achieved by injection of a continuous wave light
into the laser. The origin of the bistability is from the strong influence of the carrier distribution
on the threshold characteristics of the laser.
When an external CW light is injected into the cavity of the laser diode, the dynamics of the
carrier density N in the active layer of the gain medium can be described by the following rate
equation:
N I
VN Ґ S
VҐ S
V (2.14)
Where I stands for the injection current, q the elementary charge constant, V the volume of the
active region (V=wdL, with w, d and L the width, thickness and length of the active region
respectively), the carrier life time in the active region, Slas, and Sinj the z dependent photon
numbers corresponding to the laser power and signal power respectively, glas and ginj the gain at
the lasing wavelength and the signal wavelength respectively and Ґ is the confinement factor.
The laser power and signal power can be represented as:
P z S
L (2.15)
P z S
L (2.16)
14
Assuming a linear approximation for the gain, it can be put as:
g v g a N N A N B (2.17)
g v g a N N A N B (2.18)
Where v is the group velocity, g and g is the material gain at the laser and signal
wavelengths, a is the differential gain coefficient and N0 the transparency carrier density.
For a CW operation at a constant current I, the carrier density rate equation reduces to:
IV
N Ґ A N B P Ґ A N B P (2.19)
As can be seen from equation (2.19), a z dependent stimulated emission causes a variation of the
carrier density along the length of the laser. Hence, if the injected CW light undergoes a strong
variation along z, it results in a strong spatial variation of the carrier density. A variation in the
carrier density in turn causes a variation in the effective index along z accordingly [4]. This
variation in the effective refractive index can be analyzed by the transfer matrix method (TMM)
where the laser cavity is divided into a finite number of discrete cells within which the laser
parameters are constant [4].
The interaction between carrier and photon populations which has a strong influence on the
spatial distribution of the effective refractive index is named the spatial hole burning. Bistability
in DFB laser diodes can be achieved by a difference in the spatial hole burning in the laser in the
two stable states when an external CW light is injected into it.
As has been discussed in section 2.1, lasing for a λ/4 shifted DFB laser occurs at the Bragg
wavelength and this lasing mode is characterized by the lowest Bragg deviation Δβ. A spatial
variation in the effective index will lead to a Bragg deviation which will not be constant along the
length of the laser anymore.
15
∆β n NΛ
(2.20)
From Figure 2.3, we can see that an increase in the Bragg deviation leads to higher end losses
and hence higher threshold gain. A strong variation of the Bragg deviation along the laser
degrades the distributed feedback and raises the threshold gain.
The DFB flip flop can be used for broadband operations as the only requirement in the
wavelength of the signal is that it should not be too close to the lasing wavelength to prevent
reflection and to keep the non-uniformity in the injected light power along the length of the laser.
The λ/4 shifted DFB lasers used are AR coated to prevent reflection of the injected light and to
maintain the dependence of the lasing threshold on the Bragg deviation.
When a CW light is injected into the laser, a bistability occurs in which two stable states can be
identified.
State 1-In this state, the DFB laser is lasing and the gain is clamped and relatively small. As long
as the wavelength of the CW light is not too close to the Bragg wavelength, it will undergo a
small amplification as it passes by and the laser is essentially an amplifier with a small clamped
gain. Hence, its impact on the lasing is negligible and the DFB laser goes on lasing. In this state,
there is a relatively uniform laser and signal power along the length of the laser and this will lead
to a relatively low and uniform Bragg deviation. A uniform and small Bragg deviation then
causes a low threshold gain and the laser keeps on lasing with a low and clamped gain.
State 2-In this state, the DFB laser is switched off and the external light undergoes a very high
amplification as the gain is relatively large. A large gain causes a rising power of the signal as it
goes across the cavity of the laser resulting in a strong spatial non-uniformity of the carriers due
to spatial hole burning. As discussed previously in this section, this will alter the effective index
and hence the Bragg deviation across the cavity. This will ultimately increase the losses and the
threshold gain will be large. The DFB laser then keeps on being switched off.
16
As these two stable states exist for a given range of CW light power, a bistability exists for the
laser power. Figure 2.4a, shows simulations done by a commercial software package (VPI),
showing the carrier density distributions when the laser is lasing and when it is switched off and
Figure 2.4b shows bistability of the laser power occurring for a certain range of CW light input
power.
a) b)
Figure 2.4 a) Longitudinal distribution of the carriers in the DFB-laser. Dashed: lasing state; Solid: non-lasing state. b) Simulation results of the bistable behavior of a λ/4-shifted and AR-coated multiquantum well DFB-laser with length 400 μm, кL 1.2 and active layer thickness 40 nm. The laser is electrically pumped with I/Ith = 4 and Ith = 42.5 mA. 2.3 Flip-flop Operation in a DFB Laser
Once, the DFB laser is biased by a CW light injection to operate the device in the region of
bistability, switching from one state to another can be achieved by applying short and strong
positive pulses on opposite sides of the laser.
Switching from lasing state to the non-lasing state is achieved by injecting a short and strong
pulse at the side of the laser where the CW light is injected. This will move the operation point of
the laser out of the hysteresis curve by creating a non-uniform carrier distribution across the
cavity and thus increasing the lasing threshold. The laser will remain switched off due to
increased threshold gain.
17
To switch back to the lasing state, a short and strong pulse should be injected on the other side of
the device. This will restore the uniformity of the carrier density across the cavity and will reduce
the threshold gain. Ultimately, the laser begins to lase again.
A flip flop operation can, therefore, be achieved by injection of short and strong pulses as shown
in Figure 2.5. Figure 2.6 shows a typical dynamic flip flop in action when set and reset pulse
trains are injected.
Figure 2.5 An illustration showing how the CW light and pulses are injected into a DFB laser for all optical flip flop operation (courtesy of [5]).
Figure 2.6 Simulation of the all-optical flip-flop behaviour in a single DFB-laser for кL=1.2 Idc=240mA, L=400μm, Linear gain coefficient = 3 10-20m2.
18
2.4 Simulation Results
2.4.1 Simulation Setup
A commercial software package called VPI componentmaker, based on the transmission line
model (TLLM) is used to investigate the static and dynamic behavior of the device. VPI, using
coupled rate equations to describe the DFB laser diode, not only provides a relatively complete
model and hence a more accurate representation of a real DFB laser but also includes all relevant
parameters and gives the opportunity to see the effects of all parameters of interest in the
simulations.
The following scheme is used to investigate the bistability. CW light is injected into a quarter
wave shifted DFB laser diode and a tunable optical filter filters out the laser output and feeds it to
a 2D display. The 2D display shows how the laser output changes as function of the input CW
signal. This is done for both increasing and decreasing CW signal levels by the variable optical
attenuator so that the bistability can be seen.
Figure 2.7: Scheme for investigating bistability using the VPI. VOA: variable optical attenuator, OBPF: optical band pass filter.
The setup used in the simulations to realize the flip flop operation is shown in Figure 2.8. CW
input power and the reset pulse train are fed into the combiner whose output is injected to one
side of the DFB laser and another set pulse train is injected on the opposite side of the laser. A
tunable optical filter separates the laser output and a scope shows switching behavior of the laser
signal based on the pattern on the pulse trains.
2D
DFB-LDTunable Laser
(CW light) VOA OBPF
19
Figure 2.8: Simulation setup for flip flop dynamic operation. TOF: tunable optical filter. 2.4.2 Dynamic Flip Flop Operation and the Effect of Different parameters on the hysteresis curve and switching times Some typical and common device parameters used in the simulations are shown in the following table.
units Active region type MQW Active region width 2.5e-6 m Active region thickness MQW 0.04e-6 m Active region thickness SCH (half thickness (one side thickness) of the separate confinement heterostructure)
0.21e-6 m
Current injection efficiency 1 Nominal wavelength 1.5525246e-6 m Group effective index 3.7 Internal loss 3000 1/m Confinement factor (MQW) 0.07 Confinement factor ( SCH) 0.56 Optical coupling efficiency 1 Facet reflectivity 1e-12 Facet reflectivity phase shift at left and right 0 degrees Grating model Coupling Grating stopband frequency 442.6e9 Hz Grating phase shift 90 degrees Reference carrier density 2e24 m-3 Linear recombination 0 1/s Bimolecular recombination 1e-16 m3/s Auger recombination 1.3e-41 m6/s Carrier capture time constant 70e-12 s Carrier escape time constant 140e-12 s Initial carrier density 1e24 m-3 Gain model Linear Transparency carrier density 1.5e24 m-3 Gain shape model Flat Non-linear gain coefficient 1e-23 m3 Non-linear gain time constant 500e-15 s Linewidth factor (MQW) 3 Noise inversion parameter 2
Table 2.1: The different common parameters of the DFB laser used in the simulations. Other parameters of interest are specified while discussing the effect of parameters on the bistability and switching times.
RESET
Tunable Laser (CW light)
Combiner DFB LD
SET
TOF
Scope
20
In the simulations for flip flop operations, switching off of the DFB laser is almost instantaneous
and fall times as small as 15ps have been recorded and details are given in the next section.
However, the switching on operation shows overshoot and ringing behaviour as shown in Figure
2.9.The overshoot and ringing are due to the fact that the average gain in the device just before
switching on is higher than the threshold gain.The ringing behaviour after the switch on of the
laser diode represents relaxation oscillations. This is similar to the typical ringing behaviour
shown by laser diodes when switching on the bias current. However, the ringing is more
pronounced in this case due to a relatively large average gain in the off state of the device than
the threshold gain.
In recording the rise times, the overshoot and ringing phenomena impose difficulty as to exactly
determine the time gap taken by the signal to reach to 90% of the final value. Rise times as small
as 8ps have been found.
Figure 2.9: Simulations showing overshoot and ringing while switching on of the laser diode. Parameters for the DFB laser are кL=1.2 Idc=240mA, L=400μm, differential gain = 310-20m2.
Simulations using the VPI showing the effect of different parameters on the hysteresis curve and
the switching time are discussed in detail below.
21
a) The effect of linear gain coefficient (Differential gain)
In order to see the effect of the differential gain, we have used a constant кL value of 1.2
(к=3000/m and L=400µm) at a laser drive current of 240mA and the effect of two different linear
gain coefficient values of 3 10-20 and 5 10-20m2 on the bistability curve and switching times are
investigated.
Figure 2.10 shows the hysteresis curve for the two cases. We can see that higher linear gain
coefficient values lead to a broader bistability curve. A more stable laser operation at the higher
linear gain coefficient has resulted in demanding higher CW light power to switch off the laser.
Figure 2.10 Simulations showing hysteresis curves for different linear gain coefficients.
Using a CW light power of 1.78mW for both cases, the laser is in the bistability region and its
switching performances were investigated. Fall and rise times have been recorded for both cases
and are shown in Figure 2.11. The reset pulse length is 200ps and that of the set pulses is 125ps.
The switching times are recorded by varying the pulse power so that one can see how the fall and
rise times change for stronger pulses.
0.0
10.0
20.0
30.0
40.0
50.0
60.0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Out
put l
aser
Pow
er (m
W)
CW light power (mW)
Linear Gain Coeff.=3*e-20m^2.Linear Gain Coeff.=5*e-20m^2.
Differential gain (dg/dN)=3 10-20m2
Differential gain (dg/dN) =5 10-20m2
22
Figure 2.11 Fall and rise times for different linear gain coefficient values.
From the simulations, switching could be observed starting from a certain pulse power at which
the pulses are strong enough to disturb the carrier distribution in the cavity and as shown in
Figure 2.11, the rise and fall times decrease when we increase the pulse power. This effect,
however, continues for a certain range of pulse power and the switching times become almost
constant afterwards.
The effect of the differential gain is more pronounced in the rise times than the fall times. A more
stable laser operation at the higher linear gain coefficient has reduced the rise times.
b) The effect of normalized coupling coefficient, кL
To see the effect of the normalized coupling coefficient, two approaches are analyzed. In the first
case, the effect of different кL values are considered for the same injection current while in the
second case different injection currents are used for the same laser output power.
Case1: For the same drive current of 240mA, simulations are done for кL=1.2 (к=30/cm,
L=400μm) and кL=1.8(к=45/cm, L=400μm).
Figure 2.12 shows the hysteresis curves for the two кL values. The hysteresis curve for lower кL
value is narrow and shifts to the left. This is basically due to the increase in the threshold gain for
smaller кL values. An increase in threshold gain results in an increased amplification of the
0
10
20
30
40
50
60
70
0 10 20 30
Tim
e in
ps
Pulse Power in mW
Fall times for Linear gain coeff=3*e-20m2
Fall times for Linear gain coeff=5*e-20m2
0
2
4
6
8
10
12
14
16
0 10 20 30
Tim
e in
ps
Pulse Power in mW
Rise times, Linear Gain Coeff.=3e-20m2
Rise times, Linear gain coeff.=5e-20m2
Fall times, differential gain =310-20m2
Fall times, differential gain.=5 10-20m2
Rise times, differential gain.=3 10-20m2
Rise times, differential gain.=5 10-20m2
23
injected CW light when the laser is lasing at a clamped gain. Hence, smaller CW light power gets
amplified and a switching off occurs sooner for lower кL. The bistability curve as a result will be
narrower and shifts to the left.
Figure 2.12 Simulations showing hysteresis curves for different normalized coupling coefficients
Using CW light power of 1.78mW for the case of кL=1.2 and 5.65mW for the case of кL=1.8,
switching performances were measured. As is demonstrated in the previous case, reset pulse
lengths of 200ps and set pulse lengths of 125ps were used in the first case. In the case of кL=1.8,
switching off was not possible till the energy of the pulses was increased very much and
switching off could be achieved for 400ps pulses. The set pulses were 125ps long.
Figure 2.13 Fall and rise times for different normalized coupling coefficients (кL)
0
10
20
30
40
50
60
0 2 4 6 8 10 12 14
Out
put L
aser
Pow
er (m
W)
CW light Power (mW)
kl=1.8
.
kl=1.2
.
кL=1.2
кL=1.8
0
10
20
30
40
50
60
70
0 10 20 30 40
Tim
e (p
s)
Pulse power (mW)
Fall times for kL=1.2
Fall times for kL=1.80
2
4
6
8
10
12
14
0 10 20 30 40
Tim
e (p
s)
Pulse Power (mW)
Rise times at kL=1.8
Rise times at kL=1.2
Fall times: кL=1.2
Fall times: кL=1.8
Rise times: кL=1.8
Rise times: кL=1.2
24
Due to the much broadened bistability curve at кL=1.8 which originates from the lower threshold
gain, we can observe in Figure 2.13 that stronger and longer pulses are required for switching off.
This is accompanied by increased fall times. We can also observe that larger кL leads to larger
rise times but the changes are not pronounced.
Case 2: For the same laser power of 30mW at no CW light injection, simulations are done for
кL=1.2 (к=30/cm, L=400μm) with drive current of 130mA and кL=1.8(к=45/cm, L=400μm) with
a drive current of 140mA. The hysteresis curves are shown in Figure 2.14 and similarly as for
case 1, a rise in the threshold gain justifies the narrowing and left-shifting of the hysteresis curve
for кL=1.2.
Figure 2.14 Simulations showing hysteresis curves for different normalized coupling coefficients for the same laser output power at no injection of CW light.
Using CW light power of 1.48mW for кL=1.2 and 5.9mW for кL=1.8, switching characteristics
have been analyzed. In addition to the large CW light requirement, higher кL lasers need strong
reset pulses. Similar trends in switching times compared to case1 have been recorded.
A reduction in the reset pulse energy for higher кL can be achieved by using the right edge of the
bistability curve, but this highly degrades the contrast ratio between the two stable states.
DFB lasers with smaller кL values where the impact of the spatial hole burning is higher are
generally easy to switch off.
0
5
10
15
20
25
30
0 2 4 6 8 10
Out
put L
aser
Pow
er(m
W)
CW light Power (mW)
kl=1.8,Idc=140mA, P=30mW
.
kl=1.2,Idc=130mA,P=30mW
.
кL=1.8, Idc=140mA, P=30mW
кL=1.2, Idc=130mA, P=30mW
25
c) Lower кL values for switching
In order to see the effect of lower кL values on the bistability and switching times a drive current
of 240mA, and a differential gain of 5 10-20m2 are used for two different кL values of 0.8
(к=20/cm and L=400μm) and 1.2 (к=30/cm and L=400µm).
Figure 2.15 shows the hysteresis curves for the two normalized coupling coefficients and as
expected the lower кL is characterized by a narrower and left shifted hysteresis due to increased
threshold gain.
Figure 2.15: Simulations showing hysteresis curves for different normalized coupling coefficients to see effect of a low coupling coefficient Using a CW light power of 0.6mW for кL=0.8 and 1.78mW for кL=1.2, rise and fall times have been recorded and are shown in Figure 2.16.
0
10
20
30
40
50
60
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Lase
r Out
put p
ower
(mW
)
CW light power (mW)
kl=0.8
..
kl=1.2
..
кL=0.8
кL=1.2
26
Figure 2.16: Fall and rise times for different normalized coupling coefficients (кL).
We can see from Figure 2.16 that switching is possible using narrow bistability curves and this
reduces the power requirement of the CW light and the pulses used. But, they require a careful
and tight control over the CW light power as well as the pulse energies. It also reduces the
switching times. The change in the switching times is more pronounced for the switching off.
It has been possible to use just 75fJ of pulses for switching ON with switching times around 9ps
for the initial overshoot. Fall times of around 15ps are also obtained. This shows that DFB lasers
with lower кL values are better suited for flip flop operation.
If smaller linear gain coefficients are used together with small кL, the hysteresis curve will be
very narrow and may not be recognized and making the exact value of CW light power to choose
difficult. Hence, relatively larger linear gain coefficients are better to work with for smaller кL
values.
d) The effect of the laser drive current
To see the effect of the drive current, we have investigated the hysteresis curves and the
switching times by using drive currents of 130mA and 240mA. In these cases, we have used
кL=1.2 with linear gain coefficient of 5 10-20m2.The resulting hysteresis curves are shown in
Figure 2.17.
0
10
20
30
40
50
60
70
0 5 10 15 20 25 30
Tim
e (p
s)
Pulse power (mW)
Fall times for kL=1.2
Fall times for kL=0.8
8.8
9
9.2
9.4
9.6
9.8
10
10.2
0 5 10 15 20 25
Tim
e (p
s)
Pulse power(mW)
Rise times for kL=1.2
Rise times for kL=0.8Fall times: кL=1.2
Fall times: кL=0.8
Rise times: кL=1.2
Rise times: кL=0.8
A
C
e
m
o
U
h
Figure 2
As can be se
CW light po
effect by the
more stabiliz
of the hyster
Using a CW
have been re
125ps long.
Figure 2
Lase
rOut
putP
ower
(mW
)
0
10
20
30
40
50
60
70
80
0
Tim
e (p
s)
.17: Simulat
een in Figure
ower needed
e injected li
zed laser op
resis curve.
W light powe
ecorded and
.18: Fall and
0
10
20
30
40
50
60
0
Lase
r Out
put P
ower
(mW
)
10
Pulse
tions showin
e 2.17, when
d to switch o
ight that sw
eration at hi
er of 1.4mW
are shown in
d rise times f
1 2CW light pow
20
power (mW)
Fall times At
Fall times At
Fall times
Fall times
ng hysteresis
n the drive c
off the laser
witches the la
igher drive c
W for Idc=130
n Figure 2.1
for different
3wer in mW
Idc=13..Idc=24..
30
t Idc=130mA
t I dc=240mA
Idc=
Idc=2
s: Idc=130mA
s: Idc=240mA
s curves for d
current of the
diode also
aser sooner
current has c
0mA and 1.8
8.The reset p
drive curren
4
30mA
40mA
400
2
4
6
8
10
12
14
16
18
20
0
Tim
e (p
s)
130mA
240mA
different driv
e DFB laser
decreases du
when the d
caused the w
8mW for Idc
pulses are 20
nts.
5
10
Puls
ve currents
diode decre
ue to the ca
drive current
widening and
=240mA, sw
00ps and the
20
e power (mW)
Rise times
Rise times
Rise time
Rise time
eases, the inp
rrier depleti
t is smaller.
d right shifti
witching tim
e set pulses a
0 30
at Idc=240mA
at Idc=130mA
es: Idc=240mA
es: Idc=130mA
27
put
on
A
ng
mes
are
0
28
Figure 2.18 shows that the rise times are more sensitive to the drive currents and a more
stabilized laser operation at higher drive currents has resulted in smaller rise times.
e) The effect of the injected light power position in the hysteresis curve
In order to see the effect of the position on the hysteresis curve of the injected CW light power on
the switching, parameters of кL=1.1 (к= 27.5/cm and L=400µm), differential gain=5 10-20m2and
drive current of 110mA for a laser power of 30mW have been used and two points on the
hysteresis curve have been chosen for the CW light power as shown in Figure 2.19. The points
are around 0.5mW (on the left edge of the hysteresis) and 0.75mW (on the right edge of the
hysteresis).
Figure 2.19 Bistability curve to see effect of position on the hysteresis curve on the switching
The fall and rise times recorded are shown below in Figure 2.20. For the CW light at the right
edge of the hysteresis the reset pulses are 125ps long and the set pulses are 225ps long and for the
left edge, the reset pulses are 550ps long the set pulses are 125ps long.
0
5
10
15
20
25
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Out
put L
aser
Pow
er(m
W)
CW light power(mW)
29
Figure 2.20: Fall times to see differences while using right edge and left edge of bistability for switching.
At кL values around 1.1 where the bistability curve is some how flat, one can try different points
in the hysteresis curve without losing too much contrast ratio between the two states on the right
edge.
We can see that when the injected CW input power used is on the left edge of the bistable region,
the reset pulse energy rises. A decrement in the set pulse energy is also recorded. This is due to
the fact that the set and reset pulses have to make the laser diode move out of the hysteresis curve
temporarily to switch between the two states. Hence, when the CW input power is situated near
the right edge of the hysteresis curve, smaller reset pulse energies are enough to push the laser out
of the hysteresis temporarily and shift to the off state than when using CW input power on the left
edge of the bistability and vice versa. This is also accompanied by a change of the switching
times based on the position of the CW light in the hysteresis curve. Operating at the left edge of
the bistability reduces the rise times and operating at the right edge of the bistability leads to
smaller fall times.
f) The effect of device length
To see the effect of the device length on the hysteresis curve as well as the switching times, three
different lengths of 700, 400 and 300μm have been used for the same кL value of 1.2, a drive
0
20
40
60
80
100
120
140
160
0 2 4 6 8 10
Tim
e (p
s)
Pulse Power (mW)
Fall times forCW at right edge of bistability
Fall times for CW at left edge of bistability
0
2
4
6
8
10
12
14
16
18
20
0 5 10 15 20 25
Tim
e (p
s)
Pulse power (mW)
Rise times for CW at right edge of bistabilityRise times for CW at left edge of bistability
30
current of 240mA and a linear gain coefficient of 5 10-20m2. Therefore the coupling coefficients
for the three cases become 17.4/cm, 30/cm and 40/cm respectively.
The hysteresis curves are shown in Figure 2.21.
Figure 2.21: Bistability curves to see effect of device length on switching.
As we see in Figure 2.21, a longer device length has a narrower and left shifted hysteresis curve.
The reason behind this is that longer lengths accompanied by smaller к (hence same кL), provide
longer space for the injected CW light to be amplified as it goes by the laser diode in the
longitudinal direction. In addition to the increased amplification across the laser diode, it also
makes the spatial variation of the carrier density more pronounced so that the spatial hole burning
effect will be stronger in the longer devices. Hence a relatively smaller CW light injection is
required which makes the hysteresis curve narrower and shifted to the left.
Using CW light power of 1.4, 1.76 and 2.1mW for the lengths of 700, 400 and 300μm
respectively switching has been simulated for reset pulse lengths of 200ps and set pulse lengths
of 125ps and is shown in Figure 2.22.
0
10
20
30
40
50
60
70
0 1 2 3 4 5
Out
put L
aser
Pow
er (m
W)
CW Laser Power (mW)
l=700um, K=17.14/cm...l=400m, K=30/cm...l=300um, K=40/cm...
L=700μm, к=17.14/cm
L=400μm, к=30/cm
L=300μm, к=40/cm
31
a) b)
Figure 2.22: Fall times (a) and rise times (b) for different device lengths and same кL.
Figure 2.22 shows that smaller device lengths (for the same кL) are characterized by smaller fall
and rise times. This may originate from the time spent by the pulses as they propagate along the
length of the laser diode i.e disturbing the carrier density uniformity (switching off) and later
restoring the uniformity (switching on) are quicker in smaller device lengths.
g) The effect of linear recombination rates
Using кL=1.2 (к=30/cm and L=400μm), linear gain coefficient of 3 10-20m2 and a drive current of
240mA, two different linear recombination rates of zero and 5108 /s have been investigated.
When we increase the linear recombination rate, the threshold current increases and to work on
the same laser output power at no CW light injection the drive current should be increased as
shown in Figure 2.23.
0
10
20
30
40
50
60
70
0 10 20 30 40 50
Tim
e in
ps
Pulse power in mW
L=700um, k=17.14/cm
L=400um, k=30/cm
L=300um, k=40/cm
0
2
4
6
8
10
12
14
16
0 5 10 15 20 25 30
Tim
e in
ps
Pulse Power in mW
L=700um, k=17.14/cm
L=400um, k=30/cm
L=300um, k=40/cm
Fall times
Rise times
L=700μm, к=17.14/cm
L=400μm;к=30/cm
L=300μm, к=40/cm
L=700μm,к=17.14/cm
L=400μm, к=30/cm
L=300μm, к=40/cm
32
Figure 2.23 Hysteresis curves for different linear recombination rates
Using a CW light power of 1.75mW and reset pulses of 250ps and set pulses of 125ps long,
simulations on the switching times have been recorded and are shown in Figure 2.24.
Figure 2.24: Fall and rise times with and without linear recombination.
When we introduce a linear recombination at a certain injection current the laser output power
will decrease as the threshold current increases but this has barely changed the shape and position
of the bistability curve. When we increase the current for same laser power at no CW injection,
the bistability curve almost fits to the first case and shows almost the same switching
performance.
0
10
20
30
40
50
60
0 1 2 3 4 5 6
Out
put l
aser
pow
er (
mW
)
CW light power in mW
At Idc=240mA with linear recombination.
At Idc=240mA without Linear recombination.
At Idc=258.5mA current with Linear recombination.
0
10
20
30
40
50
60
70
0 5 10 15 20
Tim
e (p
s)
Pulse power (mW)
Fall times with a linear recombination of 5*e(8)/sFall times without linear recombination
0
2
4
6
8
10
12
14
16
0 10 20 30
Tim
e (p
s)
Pulse power (mW)
Rise time with out linear recombination
Rise time with linear recombination
Fall times with linear recombination Fall times without linear recombination
33
h) The effect of non-linear gain coefficient
To see the effect of the non-linear gain coefficient, a кL=1.2 (к=30/cm and L=400μm), linear
gain coefficient of 3 10-20m2 are used for non-linear gain coefficients of 10-23m3 and 2 10-23m3.
For the same laser output power at no injection drive currents of 240mA and 242mA are used for
the two cases respectively. The hysteresis curves are shown in Figure 2.25.
Figure 2.25: Hysteresis curves for different non-linear gain coefficients.
An increment in the non-linear gain coefficient from 10-23m3 to 2 10-23m3 has resulted in a slight
drop of power at the same current and when we increase the current a bit (by 2mA), we end up
almost the same bistability curve. Using a CW light power of 1.75mW and reset and set pulse
lengths of 200 and 125ps respectively switching has been investigated and both cases have shown
almost similar switching times.
2.5 Summary
It has been demonstrated numerically in this chapter that a bistability occurs in a single, λ/4
shifted DFB laser with antireflection coated facets working above threshold when a CW light is
injected into it. The bistability can be observed keeping the wavelength of the CW light not too
close to the Bragg wavelength. Once the DFB laser is in the bistability regime, switching from
lasing state to the non-lasing state could be achieved by injecting a short and strong pulse at the
side of the laser where the CW light is injected thereby creating a non-uniform carrier distribution
along the cavity and switching back to the lasing state could be carried out by injecting a short
0
10
20
30
40
50
60
0 1 2 3 4 5
Lase
r out
put p
ower
(mW
)
CW light power (mW)
NL gain coeff=1e(-23)m^3
.
NL gain coeff=2e(-23)m^3
.
NL gain coeff.=10-23m3
NL gain coeff.=2 10-23m3
34
and strong pulse on the other side of the device which restores the uniformity of the carrier
distribution.
It has been shown numerically that theoretically switch-on times of 10 ps and switch-off times of
less than 20 ps are possible.
A numerical investigation of the effects of device and operating parameters on the hysteresis
curve and switching times has resulted in the following conclusive points.
Switching times decrease with increasing pulse energies.
Higher differential gain leads to a broader bistability regime and smaller rise times.
The hysteresis curve for lower кL values is narrow and shifts to the left. Fall times and
required switch energies decrease with decreasing кL.
Higher drive currents lead to widening and right shifting of the hysteresis curve. Rise
times are more sensitive to the drive currents and higher drive currents result in smaller
rise times.
A longer device length has a narrower and left shifted hysteresis curve. Smaller device
lengths are characterized by smaller fall and rise times.
Operating at the left edge of the bistability reduces the rise times and operating at the right
edge of the bistability leads to smaller fall times.
Parameters like non-linear gain coefficient and recombination rates have very little
influence on the hysteresis curve and switching times.
2.6 References
[1] G.P Agarwal, “Fiber-Optic Communication Systems,” 3rd Edition, Wiley Series in Microwave and Optical Engineering, 2002.
[2] Markus C. Amann, Jens Buus, “Tunable Laser Diodes”, Artech House, 1998. [3] G. P. Agrawal, “Nonlinear Fiber Optics,” Academic Press, San Diego, CA, 2001. [4] C. Ferreira Fernandes, “Hole-burning corrections in the stationary analysis of DFB laser
diodes,” Materials Science and Engineering, B74, 75–79, 2000 [5] K. Huybrechts, G. Morthier, R. Baets, “Fast all-optical flipflop based on a single
distributed feedback laser diode”, Optics Express, 16(15), p.11405-11410, 2008.
35
3 EXPERIMENTAL INVESTIGATION OF FLIP-FLOP OPERATION IN A SINGLE DFB LASER DIODE WITH CW LIGHT INJECTION For the experimental investigation of the DFB flip flops, λ/4-shifted DFB-lasers with a length of
400 μm and a κL-value of 1.6 provided by Alcatel-Thales III/V-labs have been used. The DFB
lasers used have a reflectivity of 10-4 at both facets and are antireflective coated to prevent
reflection of the injected light so that the dependence of the lasing threshold on the Bragg
deviation is maintained. The threshold current of the DFB lasers is approximately 30mA and their
emission wavelength is 1553nm.
3. 1 Bistability The hysteresis curve for the same DFB lasers has already been measured earlier [1]. They used
the experimental setup as depicted in Figure 3.1.
Figure 3.1 Setup used to investigate bistability in DFB laser diodes during injection of a CW light at one side of the laser. VOA: variable optical attenuator, OBPF: optical band pass filter. A CW-light generated by a tunable laser at a wavelength of 1543nm is injected into the DFB
laser. A variable optical power attenuator controls the power level of the CW-light. An optical
band pass filter is used to obtain the signal at the DFB laser emission wavelength (1553.7 nm).
By using the injected CW-light power as a reference into the power meter, the laser output versus
the CW-light power will give the hysteresis curve for the device. Coupling from fiber to the DFB
chip has been carried out by using a lensed fiber.
The hysteresis curve has been measured [1] for different drive currents of the laser and is shown
in Figure 3.2.
Power meter DFB-LDTunable Laser VOA OBPF
36
Figure 3.2: Hysteresis curves for different injection currents (courtesy of [1])
As can be seen in Figure 3.2, the hysteresis curves for the different injection currents found
experimentally show widening and right shifting upon increment of the injection current. This
matches with the simulation results discussed in the previous chapter.
A decrement in the injection current is accompanied by a smaller CW light for switching off due
to the carrier depletion effect by the injected light which can switch the laser sooner when the
drive current is smaller. The right shifting and widening upon increment in the drive current is
attributed to a more stabilized laser operation at higher drive currents. The shift in the lower
boundary of the hysteresis may be due to heating and non-radiative recombination as pointed out
in [1]. As can be seen in Figure 3.2, extinction ratio up to 35dB can be reached and no bistability
could be observed for smaller drive currents (≤100mA).
3.2 Flip-flop operation 3.2.1 Experimental setup The experimental setup used for the demonstration of all-optical flip flop operation of the device,
is shown in Figure 3.3.
37
Figure 3.3: Experimental setup used for the measurements of the DFB laser diode based all-optical flip-flop. EDFA: erbium doped fiber amplifier, ATT: attenuator, OBPF: optical band pass filter. For the generation of the set and reset pulses, a pico-second pulse source with a repetition rate of
10GHz is used. The pico-second pulse source produces pulses with 7ps length. To decrease the
repetition rate of the pulse source, an electro-optical modulator is used. The modulator is driven
by a NRZ bit pattern generator. The same signal generator is used to drive the bit pattern
generator of the modulator and the pulse source.
The pulse train from the modulator is then split into two identical pulse trains by a 3dB coupler.
These pulse trains will be used as set and reset pulses. To compensate for the losses in the
modulator and coupler and give flexibility to independently control the set and reset pulse
powers, a succession of an erbium doped fiber amplifier (EDFA) and an attenuator are used on
both pulse train branches.
An optical band pass filter is used on the set pulse train path to filter out part of the amplified
spontaneous emission (ASE) noise of the EDFA. An optical fiber delay line on this path helps to
create a time delay between the set and reset pulse trains. A variation in the length of this fiber
EDFA ATT
SCOPE
DFB-LD
OBPF
OBPF
EDFA ATT
Coupler
Tunable Laser ATT
Coupler
Pico-second pulse source
Pulse Generator (10.6GHz)
Pattern Generator
Modulator
38
delay line can be exploited to change the time of arrival of the set pulses. The set pulse train is
finally injected on the right side of the DFB chip.
The lower 3dB coupler combines the reset pulse trains with the CW input light. As the CW light
should be in the bistable regime, its power is controlled by an attenuator before it is combined
with the reset pulses. A circulator injects the signal from the coupler to the DFB and at the same
time helps to extract the laser output power by connecting a tunable band pass filter at its other
terminal. The tunable optical filter is used to separate the CW light power from the laser output
power. The signal out of this filter is fed to an optical digital sampling scope which displays the
flip-flop operation. Polarization wheels are used to control the polarization of the injected light
and a peltier element and thermistor are used to control the temperature of the DFB laser diode. 3.2.2 Experimental results
By using the set-up depicted in Figure 3.3, flip flop operation has been obtained as shown in
Figure 3.4. The drive current of the laser is 150mA. The wavelength of the CW light is 1543nm
and it has been adjusted to be in the bistability region.
By combining the CW light with the reset pulses with energies around 190fJ, switching off was
possible by disturbing the uniformity of the carrier density distribution as shown in Figure 3.4.
The set pulses with energies around 75fJ are injected on the right side of the DFB chip. These
pulses restore the uniformity of the carrier density distribution and hence switch on the laser
again.
Experiments conducted for a repetition rate of 1.25GHz are shown in Figure 3.4 (a). The fall
times are around 40ps while the rise times are around 80ps. The larger rise times unlike the
simulation results may be due to slight depletion of the laser after the set pulses are injected and
carrier density needs to be built up in the cavity [1]. The wavelength of the pulses used and the
CW light can be relatively arbitrary as long as it is not very close to the lasing wavelength of the
laser. We were also able to see the flip flop operation at 3GHz repetition rate and this is shown in
Figure 3.4 (b). Fall times in this case are around 30ps and rise times are around 70ps.
39
(a)
(b) Figure 3.4 (a) Flip flop operation with a repetition rate of 1.25GHz (b) Flip flop operation with a repetition rate of 3GHz
3.3 Experimental Investigation of the switching times on operating parameters
To see the effect of the pulse energies and the position of the CW light power in the bistabilty on
the switching times, measurements have been conducted on the same DFB laser diodes [2]. The
measurement set-up is similar to Figure 3.3 with the pico-second pulse source replaced by a
0.2mW/div
500ps/div
50ps/div
100ps/div
50μW/div
50ps/div
t
C
F
i
i
p
r
F
b
n
3
λ
d
r
i
o
tunable laser
CW light po
Figure 3.5 (
injected puls
increases. Th
ps long. By
reached. Par
Figure 3.5 (b
be seen, ther
numerical re
Figure 3.
a) rise and
3.4 Summar
λ/4-shifted D
different inj
right shifting
injecting CW
opposite fac
r. A bias cu
ower between
(a) shows h
ses. It can b
his trend ha
y using suffi
rt of these sw
b) shows how
re was no sig
esults.
(a)
5 : Experime
d fall times v
ry
DFB-lasers w
ection curre
g upon incre
W light and
cet. By using
urrent of 150
n about 6.5 a
how the swi
be observed
s also been
iciently ener
witching time
w the switch
gnificant infl
)
ental switch
vs. the switch
with a length
ents and the
ement of the
reset pulses
g reset and s
0mA has bee
and 7.3dBm
itch-on and
that the swi
found in the
rgetic set an
es may be ca
hing time ch
fluence of the
-on and swit
hing energy.
h of 400 μm
hysteresis c
e injection c
s on one sid
set pulses w
en used and
[2].
swich-off t
tching times
e simulation
nd reset puls
aused by rise
hanges as the
e CW power
tch-off times
b) rise and f
m and a κL-v
curves foun
current. Flip
de of the DF
with energies
a bistability
times chang
s decrease a
results. The
ses, switchin
e and fall tim
e CW injecte
r in the expe
(b)
s (courtesy o
fall times vs
value of 1.6 h
d experimen
p-flop operat
FB and by i
s of 190fJ an
y has been fo
ge with the
as the energy
e injected pu
ng times of
mes of the inj
ed power is v
eriments, as o
)
of [2]).
s. the CW inj
have shown
ntally show
tion has bee
njecting set
nd 75fJ resp
found for inp
energy of t
y of the puls
ulses were 1
f 40 ps can
njected pulse
varied. As c
opposed to t
jected power
n bistability f
widening a
en obtained
pulses on t
pectively, fli
40
put
the
ses
50
be
es.
can
the
r.
for
and
by
the
ip-
41
flop operation has been demonstrated with 30ps fall time and 70ps rise time at a repetition rate of
3GHz. Switching times decrease as the energy of the pulses increases.
3.5 References
[1] K. Huybrechts, G. Morthier, R. Baets, “Fast all-optical flip-flop based on a single distributed feedback laser diode”, Optics Express, 16(15), p.11405-11410, 2008.
[2] K. Huybrechts, A. Ali, T. Tanemura, Y. Nakano, G. Morthier, “Numerical and experimental study of the switching times and energies of DFB-laser based All-optical flip-flops”, Submitted for photonics in switching 2009.
42
4 ALL OPTICAL 2R REGENERATION AND WAVELENGTH CONVERSION USING A SINGLE BISTABLE DFB LASER DIODE
As signals propagating in optical networks suffer from degradations due to accumulated noise
from erbium doped amplifiers, dispersion and jitter, all optical regeneration techniques are
required at intermediate distances. All optical signal regeneration plays a crucial role in
increasing transmission distances with a considerable cost and transparency by avoiding the
electronic bottleneck [1].
Wavelength conversion is another critical functionality in modern telecommunication networks
following their transformation driven by all-optical technologies for wavelength division
multiplexing (WDM) and wavelength routing applications. In such high capacity and dynamic
WDM networks, which offer a whole range of networking functions such as the ability to add or
drop channels and wavelength routing and switching, blocking due to wavelength contention at
network nodes can be reduced by wavelength conversion. All-optical wavelength conversion
addresses a number of key issues in WDM networks including transparency, interoperability and
network capacity [2, 3].
In this chapter a single, AR-coated, λ/4-shifted, bistable DFB laser diode will be used to achieve
wavelength conversion and optical regeneration comprising of enhancement of extinction ratio
(ER) and signal reamplification (2R). By setting the ‘0’-level of the signal at the left side of the
bistability region, we have investigated numerically that optical regeneration and wavelength
conversion can be achieved at 10Gbit/s.
4.1 Simulation Setup
The simulation setup used to demonstrate 2R regeneration and wavelength conversion is shown
in Figure 4.1. The simulations were done using the commercial software package VPI
componentmaker (©, [4]). Input non-return-to-zero (NRZ) signals are generated using a pseudo random bit sequence
(PRBS) generator, such that the ‘0’ level is set at the left side of the bistability region of the DFB
43
laser. The signal to noise ratio of this signal is then controlled by a succession of a variable
optical attenuator and an EDFA. The signal then passes through the device and an optical band
pass filter is used to obtain the signal at the laser wavelength and the input signal wavelength.
The signal power that arrives at the receiver is adjusted by a variable optical attenuator. This is
useful in plotting the BER vs received power plots.
Figure 4.1: Simulation setup used for the 2R-regeneration and wavelength conversion simulations. VOA: variable optical attenuator, EDFA: erbium doped fiber amplifier, TOF: tunable optical filter, ORx: optical reciever
In the simulations for the 2R regeneration and wavelength conversion, characterization of the
device has been carried out both qualitatively using eye diagrams and quantitatively using bit
error rate (BER) values at the input and output.
Eye diagrams help to see whether there is a clear signal improvement by comparing them at the
input and output of the device. Together with signal improvements in the eye diagram, the device
will also be evaluated in terms of the extinction ratio (ER) change at the input and output.
Another important figure of merit that gives a quantitative description of how well the device is
operating is the BER value at the input and output. BER of the signal as a function of the incident
power to the receiver for the back to back signal (bypassing the device) and the signal after the
device (at a wavelength of the input signal for the 2R regeneration and at the wavelength of the
laser signal for the wavelength conversion) will be compared. The BER curves in the simulations
are calculated by a chi-squared statistics which is more accurate than a Gaussian distribution in
dealing with a received optical signal with optical noise (such as ASE noise from optical
EDFA TOF ORx
10Gb/s NRZ PRBS
DFB LD VOA VOA
44
amplifiers). Sensitivity improvements are considered by evaluating the difference in the required
received power to obtain a BER of 10-9.
4.2 2R Regeneration As stated in the introduction, the 2R regeneration simulations are done by setting the ‘0’-level of
the signal at the left side of the bistability region. By using the left side of the bistability for the
‘0’ level, the nonlinear transfer characteristic of the device presents a sharp threshold to switching
which improves the extinction ratio and the relatively flat transfer function above and below the
threshold power allows reshaping of the input signal as well as noise removal.
To give a clear insight into the method of operation, a typical bistability curve for the signal
injected at one side of a DFB laser with normalized coupling coefficient кL of 1.2, a length of
400 μm and a differential gain of 5 10-20 m2 is shown in Figure 4.2. The drive current of the laser
is 240 mA.
Figure 4.2 Bistability curve for a signal injected at one side of a DFB laser with кL of 1.2, a length of 400 μm and a differential gain of 5 10-20 m2. Crosses show the position for the ‘0’ levels. We can see from Figure 4.2 that relatively flat regions on either side of the bistability region allow significant noise reduction and the sharp transfer function will be exploited in improving the signal ER. Hence, the device is well suited to 2R regeneration.
0
20
40
60
80
100
120
0 1 2 3 4 5
Out
put p
ower
(mW
)
Input power (mW)
45
For the 2R simulations, although better ER can be found for DFB lasers with smaller кL due to an
increased threshold gain at lower data rates of 2.5Gbit/s, they suffer more power penalties in
sensitivity at higher data rates of 10Gbit/s compared to those with relatively larger кL.
4.2.1 2R Regeneration at 2.5Gb/s Using a DFB laser with normalized coupling coefficient кL of 1.2, a length of 400 μm and a
differential gain of 5 10-20 m2, 2R regeneration has been investigated for 2.5Gbit/s input NRZ
data. The drive current of the laser is 240mA. Figure 4.3 shows eye diagrams for the 2.5Gbit/s
back-to-back NRZ signal and for the output signal after the DFB laser. The wavelength of the
signal is 1560 nm.
Figure 4.3: a) Eye diagram of a 2.5Gbit/s NRZ back to back signal with ER of 5.7dB. b) Eye diagram of the regenerated output signal with ER of 14 dB.
From the eye diagrams of the input back to back NRZ signal and the signal after the device, we
can see that the signal undergoes an improvement in the ER from an input ER of 5.7dB to output
extinction ratio of 14dB. This results in an extinction ratio improvement of 8.3dB. Apart from
that, we could also notice a very clear and significant noise reduction on both the ‘0’s and the
‘1’s.
In addition to the eye diagrams, a more quantitative description of the signal improvement can be
seen from the BER of the input and output signals. This is shown in Figure 4.4. A clear
improvement in the receiver sensitivity of around 4.6dB can be observed for a BER value of 10-9.
46
Figure 4.4: BER as a function of the signal power incident on the receiver for the back to back signal and the signal after device. 4.2.2 2R Regeneration at 10Gbit/s Using a DFB laser with normalized coupling coefficient кL of 1.8, a length of 400 μm and a
differential gain of 5 10-20 m2, 2R regeneration has been investigated for 10Gbit/s input NRZ
data. The drive current of the laser is 130mA. Figure 4.5 shows eye diagrams for the 10 Gb/s
back-to-back NRZ signal and for the output signal after the DFB laser. The wavelength of the
signal is 1560 nm.
Figure 4.5: a) Eye diagram of a 10Gbit/s NRZ back to back signal with ER of 7dB. b) Eye diagram of the regenerated output signal with ER of 10 dB.
1E-191E-171E-151E-131E-111E-091E-071E-051E-031E-01
-30 -28 -26 -24 -22 -20 -18 -16 -14 -12 -10
BE
R
Received power (dBm)
Back to back BER
Signal after device BER
47
The eye diagrams in Figure 4.5 show that the 10Gbit/s input NRZ signals undergo an ER
improvement from 7dB for the back to back signal to 10dB for the signal after the device. This
gives a 3dB improvement in the ER. More important than this, we can observe a very clear noise
reduction of both the ‘0’s and the 1’s.
Figure 4.6 shows the BER for the input and output signals. One can see that the back to back
signal and the signal after the device show almost the same sensitivity at BER of 10-9.
Figure 4.6: BER as a function of the signal power incident on the receiver for the back to back 10Gbit/s NRZ signal and the signal after device. 4.3 Wavelength Conversion As the laser signal power is at a different wavelength than the input signal with albeit an inverted
shape, all optical wavelength conversion is possible. Again just like the 2R regeneration, the
simulations for the wavelength conversion are carried out by setting the ‘0’ level of the signal at
the left side of the bistability region. Hence, the sharp threshold imposed by the non-linear
characteristic will be exploited to improve the ER and the flatness of the transfer function at
either side of the bistability will be used to reduce the noise significantly.
1E-111E-101E-091E-081E-071E-061E-051E-041E-031E-021E-01
-27 -25 -23 -21 -19 -17 -15
BE
R
Received power (dBm)
Back to back BER
Signal after device BER
48
This device is well suited for wavelength conversion as the input signal wavelength can be
relatively arbitrary as long as it is not very close to the lasing wavelength of the laser. Hence,
wavelength conversion can be achieved for a wide range of input signal wavelengths within the
gain bandwidth of the device.
4.3.1 Wavelength conversion at 2.5Gbit/s
Again using a DFB laser with normalized coupling coefficient кL of 1.8, a length of 400 μm and
a differential gain of 5 10-20 m2, wavelength conversion has been investigated for 2.5Gbit/s input
NRZ data. The drive current of the laser is 240mA. The wavelength converted laser output signal
has a wavelength of 1569nm while the input signal wavelength is 1560nm. Figure 4.7 shows eye
diagrams for the 2.5Gb/s back-to-back NRZ signal and for the wavelength converted laser output
signal after the DFB laser.
Figure 4.7: a) Eye diagram of a 2.5Gbit/s NRZ back to back signal with ER of 6dB. b) Eye diagram of the wavelength converted laser output signal with ER of 18 dB. We can see from the eye diagrams in Figure 4.7 that there is an ER improvement from 6dB for
the back to back signal to 18dB for the wavelength converted laser signal. Hence there is an ER
improvement of 12dB. Besides, there is a clear reduction of the intensity noise. On the other
hand, there seems to be significantly increased timing jitter.
49
Figure 4.8 shows the BER for the back to back and wavelength converted signals. An
improvement in the receiver sensitivity of around 1dB can be observed at a BER of 10-9.
Figure 4.8: BER as a function of the signal power incident on the receiver for the back to back 2.5Gbit/s NRZ signal and wavelength converted signal. 4.3.2 Wavelength conversion at 10Gbit/s Using the same DFB laser parameters as the previous case and the same drive current of 240mA,
wavelength conversion for 10Gbit/s input NRZ signals has been carried out. The differential gain
is now varied between 5 10-20 m2 and 6 10-20 m2. Again, the wavelength converted laser output
signal has a wavelength of 1569nm while the input signal wavelength is 1560nm. Figure 4.9
shows eye diagrams for the 10Gb/s back to back NRZ signal and for the wavelength converted
laser output signal.
1E-411E-371E-331E-291E-251E-211E-171E-131E-091E-051E-01
-25 -24 -23 -22 -21 -20 -19 -18
BE
R
Received power (dBm)
Back to back BER
Wavelength converted signal BER
50
Figure 4.9: a) Eye diagram of a 10Gbit/s NRZ back to back signal with ER of 6.5dB. b) Eye diagram of the wavelength converted signal with ER of 18 dB. The input extinction ratio is 6.5dB and the output signal ER is 18dB. This results in an extinction
ratio improvement of 11.5 dB. This is, however, accompanied by around 3.5dB power penalty in
the receiver sensitivity as shown in the BER plots in Figure 4.10 for the back to back and
wavelength converted signals. Better receiver sensitivities have been obtained for relatively larger
differential gain values. A clear noise reduction can again be observed from the eye diagrams.
Figure 4.10: BER as a function of the signal power incident on the receiver for the back to back and wavelength converted signals
1.00E-17
1.00E-15
1.00E-13
1.00E-11
1.00E-09
1.00E-07
1.00E-05
1.00E-03
1.00E-01
-30 -25 -20 -15 -10 -5 0
BE
R
Received power(dBm)
back to back BER
wavelength converted BER (dg/dn=5e-20m2)
wavelength converted BER (dg/dn=6e-20m2)
(dg/dN=5 10-20 )m2
(dg/dN=6 10-20 m2)
51
4.4 Summary It has been numerically demonstrated that both optical signal regeneration and wavelength
conversion at 2.5Gbit/s and 10 Gbit/s can be performed with a single DFB laser diode.
For optical 2R regeneration, a clear noise reduction together with an extinction ratio improvement
have been obtained. For 2.5Gbit/s input NRZ signals, ER improvement of 8.3dB together with an
improvement in the sensitivity of 4.6dB at a BER of 10-9 have been found and for 10Gbit/s input
NRZ signals, ER improvement of 3dB and almost the same sensitivity at a BER of 10-9 have been
obtained.
Wavelength conversion for 2.5Gbit/s input NRZ signals has shown an ER improvement of 12dB
along with a sensitivity improvement of 1dB at a BER of 10-9. Wavelength conversion has also
been shown to exhibit an extinction ratio improvement of 11.5dB albeit with around 3.5dB power
penalty in the receiver sensitivity for 10Gbit/s input NRZ signals. Eye diagrams show a clear
noise reduction in the wavelength conversion as well. Higher differential gain values show better
sensitivity results. 4.5 References [1] J. Nakagawa, M.E. Marhic, L.G. Kazovsky, “All-optical 3R regeneration technique
using injection locking in gain switched DFB-LD,” Electronics Letters, 37,no. 4, pp 231-232, 2001
[2] S.J.B. Yoo,, "Wavelength conversion technologies for WDM network applications," IEEE J. Lightwave Technol. 14(6),955–966,1996.
[3] N. Yan, E. Tangdiongga, H. D. Jung, I. Tafur Monroy, H. deWaardt, A. M. J. Koonen, “Regenerative all-optical wavelength multicast for next generation WDM network and system applications,” Photon Netw Commun, 15:1–6, 2008
[4] VPItransmissionmaker, http://www.vpisystems.com
52
5 ALL-OPTICAL FLIP-FLOP BASED ON A SINGLE DBR LASER
DIODE In this chapter, an all-optical flip-flop based on a single distributed Bragg reflector (DBR) laser
diode is discussed. A brief step by step introduction of DBR lasers is first given. Bistability in the
tuning behavior of these lasers is exploited to enable switching between two wavelengths with set
and reset pulses of the same wavelength but of different length and energies.
All optical flip flop operation by using optical pulses is first investigated numerically and the
effect of some operating and device parameters on the switching times is discussed. Experimental
flip-flop operation is finally given at the end of the chapter.
5.1 Distributed Bragg Reflectors
Unlike DFB lasers, the feedback in DBR lasers doesn’t take place in the active region. The
wavelength dependent cavity gain in DBR lasers is achieved by using a periodic structure as
reflector at one or both ends of the laser structure.
The coupled mode equations involving the right and left propagating fields of the periodic
structure for DBR lasers can be solved just like the analysis done for DFB lasers in chapter 2.
Following similar procedures, the coupled mode equations for the Bragg section of a DBR laser
can be put as:
R j∆β α R jкS (5.1)
S j∆β α S jкR (5.2)
53
Where R(z) and S(z) are right and left propagating waves with a wave vector β close to the wave
vector at Bragg wavelength( β0 ), Δβ=β-β0 is the real part of the Bragg deviation, к π∆λ
is the
coupling coefficient and α0 is the loss coefficient for the field in the grating region.
If R and S are known at z=0, the coupled-mode equations can be solved for R and S and their
values at z=L where L is the length of the grating will be:
R L cosh γL ∆β αγ
sinh γL R 0 к γL Sγ
(5.3)
S L к γL Rγ
cosh γL ∆β αγ
sinh γL S 0 (5.4)
Where γ к ∆β jα
If we take a right propagating field with S(L)=0, the field reflection coefficient can be solved
from (5.4) to be
r SR
кγ γL
γL ∆β αγ γL
(5.5)
The power reflection coefficient thus becomes : R r , and this is plotted as a function of the
normalized detuning, ΔβL, in figure 5.1 for three different normalized coupling coefficients.
Figure 5.1 Power reflectivity as a function of normalized detuning for different normalized
coefficients of a grating of length L=100μm and α0=0.
-8 -6 -4 -2 0 2 4 6 80
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Normalized detuning
Pow
er re
flect
ivity
кL=3
кL=1
кL=0.5
54
We can see from Figure 5.1 that the reflectivity becomes stronger for larger coupling coefficients
and becomes weaker for larger deviation from the Bragg wavelength. As the Bragg deviation
becomes large, the reflections do not add up in phase, hence lower reflectivity. One can also see
that the bandwidth of the periodic structure i.e the range over which large reflectivity can be
obtained is roughly 2кL.
Considering a two section DBR laser as shown in Figure 5.2 with a grating of length LB and
active region of length LA and considering a perfect power transition between the active region
and the grating section, the gain condition for the laser becomes:
R R exp 2g LA 1 (5.6)
Where R and R are the power reflections at the grating and at the other end of the active region
and g is the net modal gain.
Figure 5.2: A two section DBR laser with an action region and a grating section.
The net modal gain in the active section then follows from (5.6):
gLA
lnR
lnR
(5.7)
The second term in (5.7) is the equivalent end loss which represents the power leaving the active
region. The equivalent end loss is plotted in Figure 5.3 and we can see that a strong reflection at
λB has resulted in the smallest modal gain at zero Bragg deviation. The cavity losses are
minimum for the longitudinal mode closest to λB and increase substantially for other longitudinal
modes.
Active layer Corrugated dielectric structure
DBR AR coating
55
Figure 5.3: Equivalent end loss as a function of normalized detuning (Bragg deviation) for кL=1,
L=100μm and α0=0.
Neglecting variations in material gain and loss, the difference in cavity gain between a mode at
the Bragg wavelength and a mode at a wavelength λ is given by [1]:
∆gLAln R B
R (5.8)
Mode spacing for wavelengths near the Bragg wavelength is given by
∆λ BLA A L B
(5.9)
Where n A and n B are effective indices for the active and bragg section respectively. L is the
effective grating length:
L L кLкL
(5.10)
-8 -6 -4 -2 0 2 4 6 80
2
4
6
8
10
12
14
16
Normalized detuning
Equi
vale
nt e
nd lo
ss
56
5. 2 Tunability in DBR Laser Diodes
Tuning the lasing wavelength of a DBR laser can be achieved in two ways: by either changing
the cavity gain characteristic or the position of the comb-mode spectrum. A more flexible tuning
can, thus, be possible by a simultaneous variation of both parameters.
The cavity gain curve of a DBR laser is given by [1]:
g λ ґg λ α α λ (5.11)
Where ґ is the confinement factor and g λ , , and α λ are the active medium gain, internal
loss and mirror loss respectively.
From (5.11), we can notice that a spectral shift in g λ can be achieved by either using the
wavelength dependence of the active medium gain or by using a tunable wavelength selective
mirror. While tuning g λ , the condition of gain clamping imposes a slight amplitude variation
on the gain curve as g λ should be zero at the dominant laser mode. This kind of tuning leads to
hopping from one longitudinal mode to the next as the comb-modes are spectrally fixed and the
laser wavelength remains constant as long as the shift in the peak wavelength is less than half of
the mode spacing. Tuning by shifting the cavity gain is, thus, discontinuous.
Tuning is also possible by keeping the peak gain wavelength fixed but shifting the comb-mode
spectrum. This can be attained by changing the optical length of the laser cavity (neffL).For a
cavity length of L, the parameter to play with will be neff. A change in neff leads to a shift in the
mode wavelengths according to[1]:
∆λ ∆
, (5.12)
Where mode i is considered and ∆n is the index change and n , is the effective group index.
Considering an equal amount of shift for the comb modes in the tuning range, the mode
wavelengths change linearly with the shift in the comb modes within the interval between two
57
modes so that stepwise continuous tuning regimes can be obtained. In this case, a downward
mode jump equal to the mode spacing occurs to the next higher order mode at each border of the
tuning regimes.
In the case of two section DBR lasers made by integrating longitudinally an amplifying section
with a tunable Bragg reflector as shown in Figure 5.2, a current (IB) injected into the Bragg
section can control the spectral shape of the cavity gain by tuning the Bragg wavelength through
carrier-induced refractive index changes [2]. Thus, only discontinuous tuning using variations of
the Bragg wavelength is possible.These lasers lack the comb-mode spectrum tuning and are
suited only for the discontinuous wavelength tuning comprising mode jumps from one
longitudinal mode to the next. The current injected into the active region of two section DBR
lasers controls the optical gain and the optical power. Two different structures can be realized for
such lasers either by making the Bragg section purely passive or with optical gain.
A representative tuning characteristic for a two section DBR laser with a passive Bragg section is
shown in Figure 5.4.
Figure 5.4: Wavelength versus Bragg current of tunable two-section DBR laser with passive
tuning section. Active section length is 298μm and Bragg section length is 250μm [1].
As the Bragg tuning current increases, an increase in the injected carriers causes index reduction
in the region and this will reduce the effective refractive index of the transverse modes shifting
the entire comb-mode spectrum to shorter wavelengths.
U
t
h
p
I
c
w
F
a
B
c
C
e
s
F
Unlike two
tunable phas
has an activ
passive phas
Ip through ca
change the lo
wavelength b
Figure 5.5:L
and Bragg re
By simultan
cavity gain
Changes in
emission wa
shows a qua
Figure 5.6: T
section DBR
se shifter to p
e section tha
se shift secti
arrier- induc
ocation of th
by the curre
Longitudinal
eflector secti
neously cont
peak wave
the gain cu
avelength in
alitative tunin
Tuning curve
R lasers, thr
perform the
at controls t
ion that dete
ced refractiv
he cavity gai
nt IB again th
cross-sectio
ions.
trolling Ip th
elength, a c
urrent (Ia) al
n such lasers
ng curve for
es for a three
ree section D
comb-mode
the optical g
ermines the l
ve index cha
in peak wave
hrough carri
on of a three
hat shifts th
ontinuous tu
lso result in
s is depende
a three secti
e section DB
DBR lasers i
e spectrum sh
gain and opti
location of t
nges and a B
elength whic
ier-induced r
e section DB
he comb-mo
uning is po
wavelength
nt on gain,
ion DBR las
BR laser with
incorporate
hift. As show
ical power b
the comb-mo
Bragg reflec
ch is approxi
refractive ch
BR laser cons
de spectrum
ossible in th
h changes d
phase and B
ser for a cons
h constant ga
an additiona
wn in Figure
by an injecte
ode spectrum
ctor section
imately equa
hanges.
sisting of ga
m and IB tha
hree section
due to heatin
Bragg curren
stant Ia[3].
ain current
Passive wav
al longitudin
e 5.5, this las
ed current Ia
m by a curre
that is used
al to the Bra
ain, phase sh
at controls t
n DBR lase
ng [1]. Henc
nts. Figure 5
(Ia) [3].
veguide
58
nal
ser
a, a
ent
to
agg
hift
the
ers.
ce,
5.6
59
5.3 BISTABILITY IN DBR LASERS
All-optical flip flop operation using DBR laser diodes relies on a bistability in the wavelength
versus tuning current characteristics. At the heart of this bistability is an asymmetric gain
suppression due to the four wave mixing between the main mode and the side mode.
Bogatov et. al.[4] showed that there is anomalous asymmetric interaction between closely spaced
spectral modes of a semiconductor laser due to stimulated scattering of the laser radiation by
dynamic inhomogeneities of the electron density. This effect leads to an asymmetric gain
suppression where the longer wavelength spectral mode of two neighboring modes dominates.
Due to the hysteresis phenomena in the tuning characteristic, the Bragg tuning current
corresponding to a mode hop depends on whether Bragg current is increased or decreased. This
effect is attributed to nonlinear four wave mixing due to adjacent modes beating causing a
modulation in different parameters of the laser. As pointed out in [5,6], the dominant result of this
situation is carrier density pulsation and this in turn generates a sharp and asymmetric gain
variations around the lasing wavelength.
A clearer picture of the hysteresis can be observed by looking at the tuning of a Bragg filter
(δνBνB ν∆ν P
, where νB is the Bragg frequency and ∆ν P is the mode spacing) from mode 0 to
mode 1 and back to mode 0, and the asymmetric modal gain [7]. This is shown in Figure 5.7. One
can see from Figure 5.7(a) that at the beginning at which δνB=0, carrier density pulsation causes
mode -1 to experience a higher gain than mode 1. In Figure 5.7(b), injection of Bragg current has
increased νB and δνB=0.7, but due to its low asymmetric gain, mode 1 doesn’t reach threshold. In
Figure 5.7(c), δνB=0.75 and the asymmetric gain becomes centered on mode 1, allowing it to
reach threshold due to a strong gain. Further increasing the Bragg current makes δνB=1 and
minimum losses at mode 1 [Figure 5.7(d)]. Now, by decreasing the Bragg current δνB=0.6, but
mode 0 doesn’t reach threshold due to low asymmetric gain [Figure 5.7(e)]. Further reduction of
the Bragg current makes δνB=0.55 [Figure 5.7(f)] and mode 0 reaches threshold due to a high
asymmetric gain. The difference in δνB (between 0.75 and 0.55) up on increasing and decreasing
the Bragg current results in the hysteresis.
60
Figure 5.7: Schematic description of FP modes selection by Bragg grating (courtesy of [7]). Tuning of the Bragg filter ( δνB
νB ν∆ν P
) from mode 0 to mode 1 and then back to mode 0 is represented by the cavity loss and the asymmetric modal gain (crosses) spectra versus mode position. Parabolic curve is for cavity loss, horizontal dashed line is for linear gain and curve with crosses is for non-linear gain.
A bistability in the wavelength versus tuning current characteristic of a two section DBR laser
diode, consisting of a 350 μm active section and a 150 μm Bragg section with a кL of 0.9 has
been simulated using a commercial software package VPI componentmaker (©, [8]) and is shown
in Figure 5.8. It shows the relative emission frequency with respect to 193.1 THz obtained for
increasing and decreasing Bragg current.
Parabolic curve: cavity loss
Horizontal dashed line: linear gain
Curve with crosses: non-linear gain
61
Figure 5.8: Bistability in the emission wavelength versus Bragg tuning current for a two section
DBR laser (with 350 μm active section, 150 μm Bragg section with a кL of 0.9 ).
DBR lasers with longer active section cavities undergo enhanced non-linear effects due to closer
longitudinal modes leading to broader bistability curves [5]. Larger grating coupling coefficients
also enforce four wave mixing by strengthening longitudinal side modes and thus their beating
effects. The hysteresis phenomena is generally more pronounced when the power reflectivity of
the Bragg reflector, the linewidth enhancement factor and the power coupling efficiency between
the active and the Bragg sections are larger [9].
5.4 FLIP FLOP OPERATION IN A DBR LASER By biasing the Bragg section with a current in the bistability region, a DBR laser can be forced to
be in the upper branch. Switching to the lower branch can then be achieved by using the
technique of injection locking. Applying pulses of the same frequency as the lower branch forces
the laser to switch to the lower branch through injection locking.
Switching back to the upper branch can be performed in two different approaches:
1) Injection locking with pulses of the same frequency as the upper branch.
2) By injecting very short and high power pulses with the same frequency as the lower
branch itself.
-300
-250
-200
-150
-100
-50
0
0 1 2 3 4 5 6 7
Rel
ativ
e em
issi
on fr
eque
ncy
(GH
z)
Bragg current (mA)
Increasing Bragg current
Decreasing Bragg current
62
While injection locking is the obvious reason for the first approach, the second approach uses the
fact that the upper branch in the bistability corresponds to a lower carrier density in the gain
section compared to the lower branch. Hence, by using a short and strong pulse, the carrier
density in the gain section can be depleted making the laser relax back to the upper branch which
is characterized by the lowest carrier density.
A flip flop operation can, therefore, be achieved by injection of pulses as shown in Figure 5.9.
Figure 5.10(a) shows a typical switching in action when 8mW and 400ps long pulses at a relative
frequency with respect of 193.1THz of -218GHz act to switch the laser with the specification
indicated in Figure 5.8 and a Bragg current of 2.6mA to the lower branch through injection
locking and 40mW and 100ps pulses at the same relative frequency of -218GHz use the second
approach to relax back to the upper branch at a relative frequency of -155GHz. Figure 5.10(b)
shows the filtered output of the laser signal and its flip-flop operation corresponding to the upper
branch at a relative frequency of -155GHz.
Figure 5.9: An illustration showing how the pulses are injected into a DBR laser for all optical flip-flop operation .
Figure 5.10: (a)Switching between the upper and lower branches using 400ps, 8mW(reset) and 100ps, 40mW(set) pulses in a two section DBR laser (with 350 μm active section, 150 μm Bragg section with a кL of 0.9, Bragg current of 2.6mA ). (b) Flip flop operation of the Filtered output power at the upper branch with a relative frequency of -155GHz.
RESET
SET
(a) (b)
63
5.5 SIMULATION RESULTS 5.5.1 SIMULATION SETUP
Just like the simulations in the DFB lasers, the simulations for the DBR flip-flop have been
carried out by using the commercial software package, VPI componentmaker(©, [8]), based on
the transmission line model (TLLM) to investigate the static and dynamic behavior of the device.
The simulations have been done for a two section DBR laser. The following scheme shows the
setup used to investigate the bistability. The emission frequency of the laser is scanned by a
spectrum analyzer as a function of increasing and decreasing Bragg current by using a variable
electrical attenuator.
Figure 5.11: Scheme for investigating bistability using the VPI (IB: Bragg current, Ia: gain current, VEA: variable electrical attenuator, OFMeter: optical frequency meter, 2D: two dimensional display).
The setup used in the simulations to realize the flip flop operation is shown in Figure 5.12. The
second approach for switching explained in the previous section where set and reset pulses of
different energies and length but of the same frequency as the lower branch in the bistability are
injected to the DBR laser. A tunable optical filter is used to obtain signals at the upper and lower
branch wavelengths.
OFMeter
Ia
IB
Bragg section Gain section
VEA 2D
64
Figure 5.12: Simulation setup for flip flop dynamic operation. 5.5.2 Dynamic Flip Flop Operation and the Effect of Different parameters on the hysteresis
curve and switching times The major parameters of the two section DBR lasers used in the simulations are given in the
following table.
Active Section Parameters units Active region type MQW Active region width 2.5e-6 m Active region thickness MQW 0.04e-6 m Active region thickness SCH (half thickness (one side thickness) of the separate confinement heterostructure)
0.21e-6 m
Current injection efficiency 1 Nominal wavelength 1.5525246e-6 m Group effective index 3.7 Internal loss 3000 1/m Confinement factor (MQW) 0.07 Confinement factor ( SCH) 0.56 Optical coupling efficiency 1 Facet reflectivity 0.32 Facet reflectivity phase shift at left and right 0 degrees Reference carrier density 1e24 m-3 Linear recombination 0 1/s Bimolecular recombination 1e-16 m3/s Auger recombination 1.3e-41 m6/s Carrier capture time constant 70e-12 s Carrier escape time constant 140e-12 s Initial carrier density 1e24 m-3 Gain model Linear Linear gain coefficient 30e-21 m2 Transparency carrier density 1.5e24 m-3
Combiner
Tunable Optical Filter
Scope
RESET
SET
Bragg section Gain section
Ia IB
65
Gain shape model Parabolic Non-linear gain coefficient 1e-23 m3 Non-linear gain time constant 500e-15 s Linewidth factor (MQW) 3 Noise inversion parameter 2
(a)
Grating Section Parameters units Active region type Bulk Active region width 2.5e-6 m Active region thickness 200e-9 m Current injection efficiency 1 Nominal wavelength 1.5525246e-6 m Group effective index 3.7 Internal loss 1000 1/m Confinement factor 0.3 Optical coupling efficiency 1 Facet reflectivity 1e-12 Facet reflectivity phase shift at left and right 0 degrees Grating model Coupling Grating stopband frequency 0 Hz Grating phase shift 0 degrees Gain coupling 0 Grating Reference carrier density 2e24 m-3 Linear recombination 2 e8 1/s Bimolecular recombination 1e-16 m3/s Auger recombination 1.3e-41 m6/s Initial carrier density 1.1e24 m-3
(b) Table 5.1 Parameters used in the simulation for the two section DBR laser. (a) Gain section, (b) Grating section. Other parameters of interest are specified while discussing the effect of parameters on the bistability and switching times. In the simulations, the signal at the upper branch of the bistability could easily be filtered out
since the pulses are the same frequency as the lower branch. Hence, rise times and fall times for
the signal to change between 10% and 90% of the maximum value have been investigated. Fall
times as small as 20ps and rise times around 40ps have been found. The switching times decrease
with an increase in pulse energy. Switching times for a two section DBR laser with 350 μm
active section, 150 μm Bragg section with a кL of 0.9, Bragg current of 2.6mA using 400ps
(reset) and 100ps (set) pulses for the upper branch in the hysteresis shown in Figure 5.8 have
been recorded and are shown in the Figure 5.13.
66
Figure 5.13: Fall and rise times of signal at a wavelength of 1554.5nm ( upper branch in the hysteresis in Figure 5.8) for a two section DBR laser with 350 μm active section, 150 μm Bragg section with a кL of 0.9. The effect of some operating conditions and laser parameters on the switching times are
discussed below.
a) The effect of different Bragg current positions in the hysteresis on the switching times
Considering the two section DBR laser discussed above, two different Bragg currents near the
right edge of the hysteresis (near hopping point while increasing the Bragg current) and left edge
of the hysteresis (near hopping point while decreasing Bragg current) were taken and switching
times have been compared. Using a Bragg currents of 2.6mA (near the right hopping position)
and 1.97mA (near the left hopping position), reset pulses of length 400ps and set pulses of 100ps,
switching times for the signal at the upper branch have been recorded and are plotted in Figure
5.14.
0
10
20
30
40
50
60
70
80
0 5 10 15 20 25 30
switc
hing
off
tim
e (p
s)
pulse power(mW)
Fall times
0102030405060708090
0 10 20 30 40 50 60 70 80
switc
h on
tim
e (p
s)
pulse power(mW)
Rise times
67
Figure 5.14: Switching times for Bragg currents near right and left hopping points in the
hysteresis
As can be seen from Figure 5.14, taking Bragg currents near the right hopping point, reduces both
the pulse energies and the switching times corresponding to the smaller wavelength difference
between the two branches on the right side of the hysteresis.
b) The effect of Bragg section length keeping normalized coupling coefficient (кL) constant
To see the effect of the Bragg section length keeping кL constant, two different situations have
been investigated at L=150μm (the case previously discussed with к=60/cm) and L=300μm with
к=30/cm. Figure 5.15 shows the hysteresis curves for the two situations.
020406080
100120140160180200220
3 8 13 18 23
switc
hing
tim
e(ps
)
pulse power(mW)
Fall times (Bragg current near right hopping point)
Fall times ( Bragg current near left hopping point)
0
20
40
60
80
100
120
140
30 40 50 60 70 80
switc
hing
tim
e(ps
)
pulse power(mW)
Rise times (Bragg current near right hopping point)
Rise times (Bragg current near left hopping point)
68
Figure 5.15: Hysteresis curves for different Bragg section lengths
Using a Bragg currents of 2.6mA for the first case (L=150μm and к=60/cm) and 2.2mA for the
other case (L=300μm and к=30/cm), signals at the upper branch with relative frequencies with
respect to -193.1THz of -155GHz and -220GHz respectively have been filtered out after
switching with 600ps pulses (to switch to lower branch) and 75ps pulses(to switch to upper
branch again). Figure 16 shows fall and rise times for both cases.
Figure 5.16: Fall and rise times to see effect of Bragg section length keeping кL constant
-350
-300
-250
-200
-150
-100
-50
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
Rel
ativ
e em
issi
on fr
eque
ncy(
GH
z)
Bragg current (mA)
L=300μm,к=30/cm.L=150μm,к=60/cm.
0
10
20
30
40
50
60
70
80
90
0 5 10 15 20 25 30
Switc
hing
tim
e(ps
)
.
pulse power(mW)Fall times for к=60/cm, L=150μm
Fall times for к=30/cm, L=300μm
0
20
40
60
80
100
120
30 35 40 45 50 55 60 65
switc
hing
tim
e (p
s)
pulse power(mW)Rise times for к=60/cm, L=150μmRise times for к=30/cm, L=300μm
69
Figure 5.16 shows that larger Bragg section lengths at constant кL lead to high pulse power
requirement and larger switching times.
c) The effect of normalized coupling coefficient (кL)
To see the effect of кL, two situations are assessed: к=60/cm, L=150μm (with кL=0.9) and
к=80/cm, L=150μm (with кL=1.2). The hysteresis curves for the two situations are shown in
Figure 5.17.
Figure 5.17: Hysteresis curves for кL=0.9 and кL=1.2
We can see from Figure 5.17 that larger grating coupling coefficients lead to a pronounced
hysteresis. As discussed before in section 5.3, larger coupling coefficients enforce four wave
mixing by strengthening longitudinal side modes and thus their beating effects resulting in a
pronounced hysteresis.
Using a Bragg currents of 2.6mA for кL=0.9 and 3.6mA for кL=1.2, 600ps pulses have been used
to switch to the lower branches and 75ps pulses to switch back to the upper branches. Figure 5.18
shows the switching times for the signal corresponding to the upper branch after filtering.
-300
-250
-200
-150
-100
-50
0
0 1 2 3 4 5 6 7
Rel
ativ
e Em
issi
on fr
eque
ncy(
GH
z)
Bragg Current(mA)
k=80/cm, L=150μm (кL=1.2).k=60/cm, L=150μm (кL=0.9).
70
Figure 5.18: Fall and rise times for кL=0.9 and кL=1.2 We can see from Figure 5.18 that fall times are more or less similar, but injection locking at
higher кL values allows using smaller pulse power for switching. As the switching back doesn’t
use injection locking but rather by using a shorter and stronger pulse at the same frequency as the
pulse used previously, the higher кL has led to stronger pulse requirement to switch back to the
original state.
d) Using injection locking with different frequency pulses for switching
For a two section DBR laser (with 350 μm active section, 150 μm Bragg section with a кL of 0.9,
Bragg current of 2.6mA ), switching between the upper and lower branches using injection
locking with different frequency pulses have been carried out. Corresponding to the upper and
lower branches, 10mW pulses at relative frequencies with respect to 193.1THz of -157.5GHz and
-214GHz respectively have been used for switching. The switching between the two branches is
shown in Figure 5.19. 400ps pulses were used to switch to lower branch and 75ps pulses were
used to switch to upper branch again.
0
10
20
30
40
50
60
70
80
0 3 6 9 12 15 18 21 24 27
Switc
hing
tim
e (p
s)
pulse power(mW)
Fall times (к=80/cm,L=150μm)Fall times (к=60/cm,L=150μm)
0
10
20
30
40
50
60
70
80
90
30 40 50 60 70 80
Switc
hing
tim
e(ps
)
pulse power(mW)
Rise times, к=60/cm, L=150μm
Rise times, к=80/cm,L=150μm
71
Figure 5.19: Switching using injection locking in both directions. a) Switching between upper and lower branches. b) Filtered signal corresponding to upper branch at relative frequency of -155GHz. C) Filtered signal corresponding to lower branch at relative frequency of -214GHz.
As this approach uses injection locking in both directions, filtering the pulses at the required
signal becomes a problem in measuring rise times.
(a)
(c) (b)
72
5.6 Experimental Investigation of Flip-flop operation in a single DBR laser diode
For the measurements a three section DBR laser with active section length of 600μm, a phase
section of 60μm and a Bragg section of 250μm is used.
5. 6.1 Bistability
It has been discussed in section 5.3 of this chapter that all-optical flip flop operation using DBR
laser diodes relies on a bistability in the wavelength versus tuning current characteristics whose
origin is the asysmmetric gain suppression due to four wave mixing between the main mode and
the side mode.
Hence, to investigate the flip-flop operation we need first to see the range of Bragg currents over
which the laser is bistable in the wavelength versus tuning current characteristics. By using a gain
current of around 100mA, the tuning current has been varied while looking at the emission
wavelength with a digital scope at the same time. By doing so, the DBR laser has shown
bistability in between Bragg current values of 20mA and 23.5mA. The bistability occurs for two
wavelength branches of 1562nm and 1561.4nm. Hence, a bistability curve can be sketched
roughly from this information as shown in Figure 5.20. Coupling from fiber to the DBR chip has
been carried out by using a lensed fiber.
Figure 5.20: Hysteresis curve for the DBR laser used in the measurements
λ (nm)
IB (mA) 20 23.5
1562
1561.4
73
5.6.2 Flip-flop operation Two different approaches have been followed to investigate the flip-flop operations of the DBR laser. 5.6.2.1 Using pulses of the same length but different energies Experimental set-up The first experimental setup used for the demonstration of all-optical flip flop operation of the device, is shown in Figure 5.21.
Figure 5.21: Experimental setup used for the measurements of the DBR laser diode based all-optical flip-flop using pulses of the same length but different energies (at same wavelength). EDFA: erbium doped fiber amplifier, ATT: attenuator, OBPF: optical band pass filter A tunable laser and a modulator driven by a NRZ bit pattern generator are used to generate a
pulse train. The NRZ bit pattern generator is used to produce pulses of 100ps length . The pulse
train from the modulator is then split into two identical pulse trains by a 3dB coupler. These pulse
trains will be used as set and reset pulses. To compensate for the losses in the modulator and
coupler and give flexibility to independently control the set and reset pulse powers, a succession
of an erbium doped fiber amplifier (EDFA) and an attenuator are used on both pulse train
branches. An optical delay device on the set pulse train path produces a time delay between the
set and reset pulse trains.
EDFA ATTDBR-LD
OBPF
SCOPE
Optical Delay
Coupler
Pulse Generator (10.6GHz)
Pattern Generator
Modulator Tunable Laser EDFA
OBPF
ATTCoupler
74
The next 3dB coupler combines the reset and set pulse trains. An optical band pass filter helps in
filtering out part of the amplified spontaneous emission (ASE) noise from the EDFA’s in both
branches. A circulator, then, injects this signal to the gain section side of the DBR laser diode and
at the same time helps to extract one of the two branches of the bistability regime by connecting a
tunable band pass filter at its other terminal.
The signal out of this filter is fed to an optical digital sampling scope which displays the flip-flop
operation. Polarization wheels are used to control the polarization of the injected light and a
peltier element and thermistor are used to control the temperature of the DBR laser diode. Experimental results
Using the set-up in Figure 5.21, flip flop operation was obtained as shown in Figure 5.22 albeit
with a pronounced instability. The gain current used is around 100mA. Corresponding to the
bistability in Figure 5.20, Bragg current in between 20mA and 23.5mA is used. Switching from
the lower branch with a wavelength of 1561.4nm to the upper branch with a wavelength of
1562nm was accomplished by pulses of wavelength 1562nm and energies around 1pJ making use
of injection locking. By making the set pulses relatively stronger (around 1.5pJ) but of the same
wavelength of 1562nm, switching on is possible but with less stable results. The set pulses make
use of the lower carrier density of the lower branch in the wavelength vs. tuning current
characteristic compared to the upper branch for switching back to the lower branch by depleting
the carrier density in the gain section of the laser. Hence, for the laser to relax back to the lower
branch, very short and strong pulses are needed. In this case, both the set and reset pulses are the
same length and this may be a reason for the unstable flip flop behavior.
Even though the results are unstable, from the flip-flop behavior seen in Figure 5.22, fall times
less than 100ps (Figure 5.22 (a)) and rise times less than 200ps (Figure 5.22 (b) can be noticed.
Repetition rate used is 1.25GHz.
75
Figure 5.22: Flip flop operation with a single DBR laser diode with pulses of the same length but
different energies (at same wavelength) Scope in average mode.
5.6.2.2 Using pulses of different energies and different lengths
Experimental set-up
In the second approach, the experimental set-up as depicted in Figure 5.23 is used. The major
difference in this set-up from the previous one is the use of the pico-second pulse source for the
generation of very short and strong set pulses. Two modulators are needed for the tunable laser
and the pico-second pulse source. The modulators are driven by a NRZ bit pattern generator. The
modulator for the pico-second pulse source is used to reduce its 10GHz repetition rate to lower
repetition rates. The pico-second pulse source produces very strong and 7ps length pulses that
make the set pulse train. The tunable laser and the modulator driven by a NRZ bit pattern
generator produce 100ps pulses which are used as reset pulses. The same signal generator is used
to drive the bit pattern generator of the modulators and the pulse source. The function of the other
devices used in the set-up has already been discussed in the previous approach.
(a) (b)
(c) (d)
500ps/div 500ps/div
1ns/div 1ns/div
50μW/div
500μW/div
76
The wavelength of both the set and reset pulse train is 1562nm. The same mechanism as the
previous case where injection locking for switching off and carrier density depletion in the gain
section by strong and short pulses for switching on are used in this case as well.
Figure 5.23: Experimental setup used for the measurements of the DBR laser diode based all-optical flip-flop using pulses of the different length and energies. EDFA: erbium doped fiber amplifier, ATT: attenuator, OBPF: optical band pass filter.
Experimental results
By using the set-up depicted in Figure 5.23, flip-flop operation was investigated with less stable
results as shown in Figure 5.24. The gain current used is around 100mA. Corresponding to the
bistability in Figure 5.20, Bragg current in between 20mA and 23.5mA is used again. Switching
from the lower branch with a wavelength of 1561.4nm to the upper branch with a wavelength of
1562nm was accomplished by 100ps, 1.2pJ pulses with a wavelength 1562nm making use of
injection locking. Very short (7ps) and strong pulses (6.5pJ) from the pico-second pulse with a
wavelength of 1562nm deplete the carrier density in the gain section making the laser relax back
to the state of lowest carrier density (lower branch). But, the flip-flop operation investigated was
unstable and this instability may arise from the limited extinction ratio of the modulator which
DBR-LD
EDFA ATT
EDFA ATT
Coupler
Pico-second pulse source
Pulse Generator (10.6GHz)
Pattern Generator
Tunable Laser
Modulator
Modulator
OBPF
SCOPE
EDFA
OBPF
77
cannot suppress the pulses completely. The very short pico-second pulses having a wider spectral
linewidth in their spectrum may also contribute to the instability.
Even though the results are less stable, from the flip-flop behavior seen in Figure 5.24 (c), fall
times less than 100ps and rise times less than 200ps can be noticed. Repetition rate used is
350MHz.
(a) (b)
(c)
Figure 5.24: Flip flop operation with a single DBR laser diode with pulses of different length and
energies (at same wavelength).
5.7 Summary
All-optical flip flop operation using a single DBR laser diode relying on a bistability in the
wavelength versus tuning current characteristics has been demonstrated numerically and
experimentally.
By biasing the Bragg section with a current in the bistability region, switching from one branch to
another is possible by injection of pulses of the same wavelength as the other branch through
injection locking and switching back to the previous branch is possible by injection of short and
50μW/div
100μW/div
500ps/div 500ps/div
500ps/div
78
strong pulses through carrier depletion. Fall times as small as 20ps and rise times around 40ps
have been found numerically for a two section DBR laser. The effect of some device parameters
and operating conditions on the switching has been investigated numerically and some conclusive
points can be drawn.
The switching times decrease with an increase in pulse energy.
Working with Bragg currents near the right hopping point of the hysteresis reduces both
the pulse energies and the switching times.
Larger Bragg section lengths at constant кL lead to high pulse power requirement and
larger switching times.
Larger grating coupling coefficients lead to a pronounced hysteresis. Injection locking at
higher кL values allows using smaller pulse energies for switching.
Experimental investigation of bistability and flip-flop operation on a three section DBR laser with
active section length of 600μm, a phase section of 60μm and a Bragg section of 250μm has been
carried out with two different setups.
In the first setup, pulses of the same length (100ps) but different energies are used. Switching
from one branch to the other is possible with 1pJ pulses by injection locking and switching back
to the lower branch is possible by 1.5pJ pulses by carrier depletion at a repetition rate of 1.25GHz
with albeit less stable results. The problem with the stability could be attributed to the use of the
same length pulses.
In the second setup, 100ps, 1.2pJ pulses are used to switch from one branch to the other by
injection locking while very short (7ps) and strong(6.5pJ) pulses are used to switch back to the
the previous branch by carrier depletion at a repetition rate of 350MHz with again less stable
results. The reason behind the stability in this case could be due to the limited extinction ratio of
the modulator used to decrease the repetition rate of the pico-second pulse source which cannot
suppress the pulses completely. Another possible reason could be the wider spectral linewidth of
the very short pulses. Fall times less than 100ps and rise times less than 200ps could be observed.
79
5.8 References [1] Markus C. Amann, Jens Buus, “Tunable Laser Diodes”, Artech House,1998. [2] G.P Agarwal, “Fiber-Optic Communication Systems,” 3rd Edition, Wiley Series in
Microwave and Optical Engineering, 2002. [3] Lectures by Prof. Geert Morthier, “Lasers,” Dept. Inform. Technol.,Gent Univ., Ghent,
Belgium [4] A. P. Bogatov, P. G. Eliseev, and B. N. Sverdlov, “Anomalous Interaction of Spectral
Modes in a Semiconductor Laser ,” IEEE J. Quantum Electron.,Vol. QE-11, No. 7, 1975. [5] Helene Debregeas-Sillard, Catherine Fortin, Alain Accard, Olivier Drisse, Estelle
Derouin, Frederic Pommereau, and Christophe Kazmierski, “Nonlinear Effects Analysis in DBR Lasers:Applications to DBR-SOA and New Double Bragg DBR,” IEEE J.Quantum Electron., Vol. 13, No. 5, 2007.
[6] A. Uskov, J. Mørk, and J. Mark, “Wave mixing in semiconductor laser amplifiers due to carrier heating and spectral hole burning,” IEEE J.Quantum Electron., Vol. 30, No. 8, pp. 1769–1781, 1994.
[7] G. Sarlet, “Tunable laser diodes for WDM communication. Methods for control and characterisation,” Ph.D. dissertation, Dept. Inform. Technol.,Gent Univ., Ghent, Belgium, 2001.
[8] VPItransmissionmaker, http://www.vpisystems.com. [9] D. Syvridis, G. Guekos, S. Pajarola, M. Tsilis, “Large optical bistabiltiy and self
pulsations in a 3 section DBR laser diode,” IEEE Phot. Techn. Lett., Vol. 6, pp. 594-596, 1994.
80
6 CONCLUSION AND RECOMMENDATION
6.1 Conclusion
Driven by the ever increasing need for bandwidth, modern telecommunication networks are
undergoing a vast expansion. To that end, the transformation of the current networks where
optical-electrical-optical conversions are imposing an electrical bottleneck in the system to all-
optical networks is drawing a considerable attention. In the heart of these networks, all-optical
flip-flops, all-optical 2R regenerators and wavelength converters play a crucial role.
In this thesis, an all-optical flip flop operation based on a single DFB laser diode with a CW
injected light and a single DBR laser diode has been numerically and experimentally investigated.
In addition, an all-optical 2R regenerator and wavelength converter has been numerically
investigated based on a single DFB laser diode.
A numerical investigation of the bistability and flip-flop operation of an AOFF based on a single
λ/4 shifted DFB laser with antireflection coated facets and CW injected light has been carried out
in chapter two. It has been demonstrated numerically that theoretically switch-on times of 10ps
and switch-off times of less than 20ps are possible. A thorough analysis of various device
parameters and operating conditions on the bistability and switching times has also been
presented. It has been found that switch-off times and required switch energies decrease with
decreasing laser length and decreasing кL. Switch-on times mainly decrease with increasing bias
current and increasing differential gain. Parameters like non-linear gain coefficient and
recombination rates have very little influence. Required switching energies decrease with
decreasing кL.
An experimental investigation on the flip-flop behavior of λ/4-shifted DFB-lasers with a length of
400 μm and a κL-value of 1.6 has been given in chapter 3. Hysteresis curves show widening and
81
right shifting upon increment of the injection current. By using reset and set pulses with energies
of 190fJ and 75fJ respectively, flip-flop operation has been demonstrated with 30ps fall time and
70ps rise time at a repetition rate of 3GHz. It has been found that switching times decrease as the
energy of the pulses increases.
It has been numerically demonstrated in chapter 4 that both optical signal regeneration and
wavelength conversion at 10 Gb/s can be performed with a single DFB laser diode. For optical
2R regeneration, a clear noise reduction together with a 3dB improvement in the ER has been
obtained. Wavelength conversion has also been shown to exhibit an extinction ratio improvement
of 11.5dB albeit with around 3.5dB power penalty in the receiver sensitivity. Eye
diagrams show a clear noise reduction in the wavelength conversion as well. Higher differential
gain values show better sensitivity results.
An AOFF based on a single DBR laser diode has been investigated numerically and
experimentally in chapter 5. Simulations on a two section DBR laser show that fall times as small
as 20ps and rise times around 40ps can be achieved. It has also been found that switching times
decrease with an increase in pulse energy and larger Bragg section lengths lead to larger
switching times and require larger pulse energies. Bragg sections with higher кL have shown a
pronounced hysteresis and injection locking in such cases could be accomplished with smaller
pulse energies.
Experimental investigation of bistability and flip-flop operation on a three section DBR laser with
active section length of 600μm, a phase section of 60μm and a Bragg section of 250μm has been
carried out with two different approaches: using pulses of the same length but different energies
and using pulses of different lengths and energies. In the first approach, 100ps pulses with
energies of 1pJ(reset) and 1.5pJ (set) are used for switching at a repetition rate of 1.25GHz and
flip flop behavior has been demonstrated but with less stable results. The instability could be
attributed to the use of the same length pulses. In the second approach, 100ps and 1.2pJ(reset)
and 7ps, 6.5pJ (set) pulses have been used for switching at a repetition rate of 350MHz and flip-
flop behavior has been demonstrated again with less stable results. The reason for the stability
problem in this case has been attributed to the limited extinction ratio of the modulator used to
82
decrease the repetition rate of the pico-second pulse source which cannot suppress the pulses
completely. Another possible reason could be the wider spectral linewidth of the very short
pulses. Fall times less than 100ps and rise times less than 200ps could be observed.
6.2 Recommendations
The work described in this thesis can be further extended in the following aspects:
For the DFB laser diode:
While it has been numerically demonstrated that a single DFB laser diode can be used to
achieve optical 2R regeneration and wavelength conversion at 10Gbit/s, it will be
interesting to see how the device performs experimentally.
It is also recommended to see the performance of the device regarding other optical signal
processing functionalities. An example is signal format conversion where the device can
be used to change a RZ bit stream into a NRZ bit stream. This can be accomplished by
splitting the RZ pulse train into two and delaying one branch with a time equal to the
length of a single bit and injecting this pulse train with the CW light as reset pulses and
injecting the other branch of RZ pulse train on the opposite facet as set pulses. The output
of the laser will give a NRZ bit pattern.
Packaging the device or integrating the device on a photonic circuit could help to increase
stability of the fiber to chip coupling, hence, improving the device performance.
For the DBR laser diode:
In the experimental investigation of the flip-flop operation, more stable flip-flop behavior
could be found by using a modulator of higher ER that will suppress the pulses from the
pico-second pulse source well.
Just like the DFB’s, increasing the stability of the fiber to chip coupling can improve the
device performance.
83
While AOFF’s based on a DFB laser diode with a CW light injection have already been
used in optical packet switching [1], it is also interesting to see how an AOFF based on a
single DBR laser diode would perform in such applications.
6.3 References
[1] K. Huybrechts, T. Tanemura, Y. Nakano, R. Baets , G. Morthier, “Fast 40 Gb/s Optical
Packet Switching Using an All-Optical Flip-Flop based on a single Distributed Feedback
Laser,” Optical Fiber Communication Conference and Exposition (OFC), OMU4, United
States, 2009
84
List of Figures 1.1 Schematics of an all-optical network [6]………………………………………………...2 1.2 Schematics of a 1×2 all-optical packet switch…………………………………………...3 2.1 Loci of DFB modes in the (ΔβL, 2g0L) plane for various values of кL……………….…11 2.2 a) Distributed feedback (DFB) laser structure. (b) Ideal lasing emission output. (c)
Typ-ical output spectrum from a DFB laser. Ia: injection current…………………….….11 2.3 Loci of λ/4 shifted DFB modes in the (ΔβL, 2g0L) plane for various values of кL……..12 2.4 a) Longitudinal distribution of the carriers in the DFB-laser. Dashed: lasing state;
Solid: non-lasing state. b) Simulation results of the bistable behavior of a λ/4-sh- ifted and AR-coated multiquantum well DFB-laser with length 400 μm, кL 1.2 and active layer thickness 40 nm. The laser is electrically pumped with I/Ith = 4 and Ith = 42.5 mA………………………………………………………………………....16
2.5 An illustration showing how the CW light and pulses are injected into a DFB laser for all optical flip flop operation (courtesy of [5])……………………………………....17
2.6 Simulation of the all-optical flip-flop behaviour in a single DFB-laser for кL=1.2 Idc=240mA, L=400μm, Linear gain coefficient = 3 10-20m2……………………………...17
2.7 Scheme for investigating bistability using the VPI. VOA: variable optical attenuator, OBPF: optical band pass filter…………………………………………………………....18
2.8 Simulation setup for flip flop dynamic operation. TOF: tunable optical filter…………...19 2.9 Simulations showing overshoot and ringing while switching on of the laser diode.
Parameters for the DFB laser are кL=1.2 Idc=240mA, L=400μm, differential gain = 310-20m2………………………………………………………………………………..20
2.10 Simulations showing hysteresis curves for different linear gain coefficients…………....21 2.11 Fall and rise times for different linear gain coefficient values…………………………..22 2.12 Simulations showing hysteresis curves for different normalized coupling coefficients....23 2.13 Fall and rise times for different normalized coupling coefficients (кL)…………………23 2.14 Simulations showing hysteresis curves for different normalized coupling coefficients
for the same laser output power at no injection of CW light…………………………….24 2.15 Simulations showing hysteresis curves for different normalized coupling coefficien-
ts to see effect of a low coupling coefficient………………………………………….….25 2.16 Fall and rise times for different normalized coupling coefficients (кL)………………....26 2.17 Simulations showing hysteresis curves for different drive currents……………………...27 2.18 Fall and rise times for different drive currents…………………………………………..27 2.19 Bistability curve to see effect of position on the hysteresis curve on the switching……..28 2.20 Fall times to see differences while using right edge and left edge of bistability for
switching………………………………………………………………………………….29 2.21 Bistability curves to see effect of device length on switching……………………….…..30 2.22 Fall times (a) and rise times (b) for different device lengths and same кL……………...31 2.23 Hysteresis curves for different linear recombination rates……………………………….32 2.24 Fall and rise times with and without linear recombination……………………………....32 2.25 Hysteresis curves for different non-linear gain coefficients……………………………..33 3.1 Setup used to investigate bistability in DFB laser diodes during injection of a CW
light at one side of the laser. VOA: variable optical attenuator, OBPF: optical ba- nd pass filter……………………………………………………………………………....35
3.2 Hysteresis curves for different injection currents (courtesy of [1])…………………..….36 3.3 Experimental setup used for the measurements of the DFB laser diode based all-
85
optical flip-flop. EDFA: erbium doped fiber amplifier, ATT: attenuator, OBPF: optical band pass filter……………………………………………………………………37
3.4 (a) Flip flop operation with a repetition rate of 1.25GHz (b) Flip flop operat- ion with a repetition rate of 3GHz………………………………………………………..39
3.5 Experimental switch-on and switch-off times (courtesy of [2]). a) rise and fall ti- mes vs.the switching energy. b) rise and fall times vs. the CW injected power…………40
4.1 Simulation setup used for the 2R-regeneration and wavelength conversion simu- lations. VOA: variable optical attenuator, EDFA: erbium doped fiber amplifier, TOF: tunable optical filter, ORx: optical receiver………………………………………..43
4.2 Bistability curve for a signal injected at one side of a DFB laser with кL of 1.2, a length of 400 μm and a differential gain of 5 10-20 m2. Crosses show the position for the ‘0’ levels………………………………………………………………………….44
4.3 a) Eye diagram of a 2.5Gbit/s NRZ back to back signal with ER of 5.7dB. b) Eye diagram of the regenerated output signal with ER of 14 dB……………………...…45
4.4 BER as a function of the signal power incident on the receiver for the back to ba- ck signal and the signal after device…………………………………………………...…46
4.5 a) Eye diagram of a 10Gbit/s NRZ back to back signal with ER of 7dB. b) Ey- e diagram of the regenerated output signal with ER of 10 dB……………………………46
4.6 BER as a function of the signal power incident on the receiver for the back to back 10Gbit/s NRZ signal and the signal after device…………………………………....47
4.7 a) Eye diagram of a 2.5Gbit/s NRZ back to back signal with ER of 6dB. b) E- ye diagram of the wavelength converted laser output signal with ER of 18 dB………….48
4.8 BER as a function of the signal power incident on the receiver for the back to back 2.5Gbit/s NRZ signal and wavelength converted signal…………………………...49
4.9 a) Eye diagram of a 10Gbit/s NRZ back to back signal with ER of 6.5dB. b) E- ye diagram of the wavelength converted signal with ER of 18 dB………………………50
4.10 BER as a function of the signal power incident on the receiver for the back to back and wavelength converted signals………………………………………………….50
5.1 Power reflectivity as a function of normalized detuning for different normalized coefficients of a grating of length L=100μm and α0=0……………………………….…..53
5.2 A two section DBR laser with an action region and a grating section……………...…...54 5.3 Equivalent end loss as a function of normalized detuning (Bragg deviation) for
кL=1, L=100μm and α0=0……………………………………………………………….55 5.4 Wavelength versus Bragg current of tunable two-section DBR laser with passive t-
uning section. Active section length is 298μm and Bragg section length is 250μm [1].....57 5.5 Longitudinal cross-section of a three section DBR laser consisting of gain, phase sh-
ift and Bragg reflector sections…………………………………………………………...58 5.6 Tuning curves for a three section DBR laser with constant gain current (Ia) [3]…….…58 5.7 Schematic description of FP modes selection by Bragg grating (courtesy of [7]).
Tuning of the Bragg filter ( δνBνB ν∆ν P
) from mode 0 to mode 1 and then back to mode 0 is represented by the cavity loss and the asymmetric modal gain (cro- sses) spectra versus mode position. Parabolic curve is for cavity loss, horizontal dashed line is for linear gain and curve with crosses is for non-linear gain…………...…60
5.8 Bistability in the emission wavelength versus Bragg tuning current for a two se- ction DBR laser (with 350 μm active section, 150 μm Bragg section with a кL of 0.9 )...61
5.9 An illustration showing how the pulses are injected into a DBR laser for all opt-
86
ical flip-flop operation …………………………………………………………………62 5.10 (a)Switching between the upper and lower branches using 400ps, 8mW(reset) and
100ps, 40mW(set) pulses in a two section DBR laser (with 350 μm active section, 150 μm Bragg section with a кL of 0.9, Bragg current of 2.6mA ). (b) Flip flop op- eration of the Filtered output power at the upper branch with a relative frequency of -155GHz………………………………………………………………………………….62
5.11 Scheme for investigating bistability using the VPI (IB: Bragg current, Ia: gain curre- nt, VEA: variable electrical attenuator, OFMeter: optical frequency meter, 2D: two dimensional display)……………………………………………………………………...63
5.12 Simulation setup for flip flop dynamic operation…………………………………......…64 5.13 Fall and rise times of signal at a wavelength of 1554.5nm ( upper branch in the hyst-
eresis in Figure 5.8) for a two section DBR laser with 350 μm active section, 150 μm Bragg section with a кL of 0.9…………………………………………………………....66
5.14 Switching times for Bragg currents near right and left hopping points in the hystere- sis……………………………………………………………………………………...….67
5.15 Hysteresis curves for different Bragg section lengths……………………………...……68 5.16 Fall and rise times to see effect of Bragg section length keeping кL constant………...…68 5.17 Hysteresis curves for кL=0.9 and кL=1.2…………………………………………...…...69 5.18 Fall and rise times for кL=0.9 and кL=1.2…………………………………………..…..70 5.19 Switching using injection locking in both directions. a) Switching between upper
and lower branches. b) Filtered signal corresponding to upper branch at relative fre- quency of -155GHz. C) Filtered signal corresponding to lower branch at relative frequency of 214GHz……..…………………………………………………………...….71
5.20 Hysteresis curve for the DBR laser used in the measurements……………………….....72 5.21 Experimental setup used for the measurements of the DBR laser diode based all-
optical flip-flop using pulses of the same length but different energies (at same wavelength). EDFA: erbium doped fiber amplifier, ATT: attenuator, OBPF: opt- ical band pass filter……………………………………………………………………….73
5.22 Flip flop operation with a single DBR laser diode with pulses of the same length but different energies (at same wavelength) Scope in average mode……………………75
5.23 Experimental setup used for the measurements of the DBR laser diode based all- optical flip-flop using pulses of the different length and energies. EDFA: erbium doped fiber amplifier, ATT: attenuator, OBPF: optical band pass filter…………………76
5.24 Flip flop operation with a single DBR laser diode with pulses of different length and energies (at same wavelength)…………………………………………………...….77
87
List of Tables 2.1 The different common parameters of the DFB laser used in the simulations.
Other parameters of interest are specified while discussing the effect of par- ameters on the bistability and switching times…………………………………………...19
5.1 Parameters used in the simulation for the two section DBR laser. (a) Gain se- ction, (b) Grating section. Other parameters of interest are specified while dis- cussing the effect of parameters on the bistability and switching times…………………65