Optical Communications and Networking - Zuqing Zhu · Optical Communications and Networking ......
Transcript of Optical Communications and Networking - Zuqing Zhu · Optical Communications and Networking ......
Lecture 5: Active Components 朱祖勍1
Optical Communications and Networking
朱祖勍
Oct. 14, 2019
Lecture 5: Active Components 朱祖勍2
Working Principle of EDFA
Pumping wavelengths: two-level case (1480 nm), three-level case (980 nm).
Lecture 5: Active Components 朱祖勍3
Frequency Response of EDFA
Lecture 5: Active Components 朱祖勍4
Optical Pumping in EDFA
980 nm pumpingHigher degree of population inversion, and low noise insertion.
More efficient, but high-power pump laser is not available.
Pre-amplifier: amplifies the low-power input signal.
1480 nm pumpingNot very efficient and high noise insertion.
High-power pump laser is available, and can yield high signaloutput power.
Post-amplifier: amplifies the signal for retransmission.
Lecture 5: Active Components 朱祖勍5
Structures of EDFA
Wavelength Selective Coupler
Single-Stage ConfigurationErbium-doped
Fiber1550 nm Signal
980 nmPump
DCFErbium-doped
Fiber
1480 nmPumpTwo-Stage Configuration
Lecture 5: Active Components 朱祖勍6
Gain Flattening for EDFAMake gain unique to all WDM channels
Use fluoride glass fiber instead of silica fiberWorse noise performance
Use a filter insider the amplifierFiber Bragg grating,…
Lecture 5: Active Components 朱祖勍7
Noise in EDFA: Amplified Spontaneous Emission
Spontaneous emission in EDFA causes atoms in the higherenergy-level to decay and generate photons with randomphase, frequency, polarization and propagating direction.
Amplified spontaneous emission (ASE): the light wavegenerated by the spontaneous emission gets amplified.
The light wave of ASE becomes the noise in EDFA anddegrades the signal-to-noise ratio (SNR) of the opticalsignal.
Noise figure: the ratio of input SNR to output SNR, it isusually 6 ~ 8 dB in EDFAs.
Lecture 5: Active Components 朱祖勍8
Noise in EDFA: Amplified Spontaneous Emission
Without InputInput at 1550 nm
Under the rule of the conservation of energy, the signal andASE noise compete for power in EDFAs, i.e., EDFA shouldamplify an optical signal before its power is too weak.
Lecture 5: Active Components 朱祖勍9
What does an EDFA look like?
Lecture 5: Active Components 朱祖勍10
Raman Fiber AmplifierReview of stimulated Raman scattering
Due to the interaction between the light wave and the molecules infiber, energy gets transferred from shorter wavelengths to longer ones.
The magic frequency spacing: 13 THz, by pumping an SMF using ahigh-power laser, we can provide gain to the signal whose frequencyis 13 THz below the pump.
Raman fiber amplifier:Nothing but a spool of SMF andhigh-power pump laser(s).
Pumping at 1460-1480 nm isfor amplification at 1550-1600nm.
Lecture 5: Active Components 朱祖勍11
Raman AmplifierBased on stimulated Raman scattering
Energy transfer from shorter wavelengths to longerwavelengthsMagic frequency spacing: 10 THzPumping at 1460-1480 nm for amplification at 1550-1600nm
Pump lasers can be distributedBetter gain flattening when signal and pumpare in the opposite directionsNoise from Rayleigh scattering, lower noisefigure.
Lecture 5: Active Components 朱祖勍12
Structure of Raman Fiber Amplifier
Output Power at 30 dBm or more!
Lecture 5: Active Components 朱祖勍13
Raman Amplifier versus EDFA
EDFA provides gain in 1528 – 1605 nm range, while aRaman amplifier can provide gain at any wavelength.Raman amplifiers can open up more WDM bands.
EDFA works in a discrete manner, while Raman amplifiercan work as a distributed amplifier with the pump attached toone end of the fiber span.
ASE noise is not an issue for Raman amplifiers. In Ramanamplifiers, fluctuations in pump power will cause the gain tovary and generate crosstalk.
Lecture 5: Active Components 朱祖勍14
Considerations for Raman Amplifiers
For Raman amplifiers, it is important to keep the pump at aconstant power to reduce crosstalk.
Having the pump propagate in the opposite direction canhelp reduce noise insertion and crosstalk dramatically, sincethe negative effects are averaged over the propagation timeover the fiber.
Lecture 5: Active Components 朱祖勍15
Semiconductor Optical Amplifier
Semiconductor optical amplifiers (SOA) are another type ofoptical amplifiers other than the fiber amplifiers.
SOAs are not as good as EDFAs for use as opticalamplifiers.
SOAs have other important applications in switches andwavelength conversion devices.
The understanding of SOAs is key to the understanding ofsemiconductor lasers.
The SOA is essentially a pn-junction.
Lecture 5: Active Components 朱祖勍16
Review of pn-Junction
A semiconductor is a material that haselectrical conductivity to a degreebetween that of a metal and that of aninsulator.
A semiconductor that has excess ofelectrons is an n-type semiconductor,and p-type semiconductor has adeficiency of electrons.
A pn-junction can be formed by putting n-type and p-type semiconductors together.
Lecture 5: Active Components 朱祖勍17
Review of pn-Junction
During the forming of a pn-junction, the holes in the p-typesemiconductor moves to the n-type one, while theelectrons in the n-type semiconductor moves in theopposite direction.
Depletion region: a region whose charge carriers aredepleted, is non-conductive.
Lecture 5: Active Components 朱祖勍18
Review of pn-Junction- Forward Biased
In forward biased case, the p-type is connected with thepositive terminal and the n-type is with the negative one.
The carriers (holes in p-type and electrons in n-type) arepushed toward the junction, which reduces the width of thedepletion region.
Electric current is invoked.
Lecture 5: Active Components 朱祖勍19
Review of pn-Junction- Reverse Biased
In reverse biased case, the p-type is connected with thenegative terminal and the n-type is with the positive one.
The carriers (holes in p-type and electrons in n-type) arepushed away from the junction, which increases the widthof the depletion region.
No electric current.
Lecture 5: Active Components 朱祖勍20
What is a Band-Gap?
A semiconductor consists of two bands of electron energy-levels: a band of low energy-levels called the valence bandand a band of high energy-levels called the conductionband. Electrons in the conduction band can conduct current.
These two bands are separated by an energy differencecalled the band-gap. No energy-levels exist in the band-gap.
To get to the conduction band, the electron has to gainenough energy to jump the band-gap.
Consider a p-type, there is only a very small concentrationof electrons in the conduction band at thermal equilibrium.
Lecture 5: Active Components 朱祖勍21
Band-Gap and Conductivity
In metals, the conduction band and valence band overlap, and thereforethere are always electrons to carry current.
In semiconductors, the conduction and valence bands are separated bya small band-gap. At room temperature, there is enough thermal energyto allow occasional electron jump, which provides the semiconductorlimited conductivity.
In insulators, the band-gap is too large to let the electrons jump, andtherefore there is no conductivity.
Lecture 5: Active Components 朱祖勍22
Population Inversion in pn-Junction
When no voltage is applied to the pn-junction, there is onlyminority carrier concentrations (electrons in p-type andholes in n-type).
When the pn-junction is forward biased and the bias voltageis sufficiently high, the increased minority carrierconcentrations result in population inversion, and the pn-junction becomes an optical amplifier, i.e., SOA.
In SOAs, the lowest optical frequency that they can beamplified is determined by the band-gap Eg, as hfc > Eg.
SOAs provide bandwidths on the order of 100 nm.
Lecture 5: Active Components 朱祖勍23
Light Emission in Semiconductors
Spontaneous Emission + Stimulated Emission
Lecture 5: Active Components 朱祖勍24
SOA versus EDFA
Bandwidth: SOAs achieve much larger bandwidth thanEDFAs. Signals in 1300 nm and 1550 nm ranges can beamplified simultaneously using SOAs.
Output power: The gains and output powers achievable withEDFAs are higher than those with SOAs.
Noise figure: SOAs insert more noise than EDFAs.
Crosstalk: SOAs introduce severe crosstalk when they areused in WDM systems.
Lecture 5: Active Components 朱祖勍25
Structure of an SOA
Higher forward bias voltageresults in larger amplificationbandwidth of SOA.
Anti-reflection coatings on itsfacets are used to avoid lasingin SOAs.
Lecture 5: Active Components 朱祖勍26
Crosstalk in SOA
Amplification is based on carrier population inversion, andWDM channels compete for the carrier concentration (gain).
Consider an SOA whose input is the sum of two wavelengthchannels, the gain seen by one channel varies with theintensity of the other channel.
This crosstalk phenomenon depends on the spontaneousemission lifetime from the high energy-level to the low-energy level.
Lecture 5: Active Components 朱祖勍27
Carrier Lifetime
Carrier lifetime: average time that an electron takes to jumpfrom the high energy-level to the low energy-level andinvokes a spontaneous emission.
If the carrier lifetime is large enough compared to the rate offluctuations of power in the input signals, the electronscannot make the transition from the high energy-level to thelow energy-level in response to the fluctuations – nocrosstalk.
SOA’s carrier lifetime: ~ ns => Gb/s
EDFA’s carrier lifetime: 10 ms => 0.1 kb/s
Lecture 5: Active Components 朱祖勍28
Carrier Lifetime
CarrierDensity
Time
Creation of Population Inversion
Carrier Depletion
CarrierLifetime
The carrier density determines the gain seen by the input signal.
Lecture 5: Active Components 朱祖勍29
Crosstalk in SOA
One input signal gets amplified => carrier depletion =>population inversion (carrier density) gets changed => theother signal’s gain gets changed
Light intensity on one WDM channel can modulate the gainof the other channel: cross-gain modulation
Carrier density can also affect the refractive index of themedia.
Light intensity on one WDM channel can modulate thephase of the other channel: cross-phase modulation
Lecture 5: Active Components 朱祖勍30
Utilize the Crosstalk in SOA
Wavelength Conversion: imprint the data modulation onone WDM channel to the other one.
Cross-gain modulation: input an optical signal withintensity modulation and a CW light without anymodulation on another wavelength, SOA can imprint theintensity modulation on the signal light to the CW lightwith inversed logic.
Lecture 5: Active Components 朱祖勍31
Cross-Gain Modulation
2.5 Gb/s Input
2.5 Gb/s Output
Lecture 5: Active Components 朱祖勍32
Utilize the Crosstalk in SOA
Cross-phase modulation: SOA can imprint the intensitymodulation on the signal light to the CW light with eitherinversed or non-inversed logic.
We just need to use a interferometer to convert the phase-modulation to intensity-modulation transition.
Lecture 5: Active Components 朱祖勍33
Experiment with Cross-Phase Modulation
DFB-LDPC
Mod
ModDriver
BERT
EDFA BPF
Attenuator2
SOA-MZI
DFB-LD50 m PM
Fiber
PC
Isolator
Agilent Lightwave Converter
Attenuator1
1550.32 nm8.0 dBm
25 dB attenuation
1555.08 nm5.8 dBm
(a) (b)
Attenuator3
BPFEDFAJDSUBPF
DCBlock
LPF
Receiver
(c)
Lecture 5: Active Components 朱祖勍34
XPM Output
Eye-diagram Measurements
10 Gb/s Input 10 Gb/s Noisy Input
2R (Reamplification and Reshaping) was achieved
Lecture 5: Active Components 朱祖勍35
Carrier Effect in EDFA
Packetized Input ~0.25 ms
EDFA’s Output
EDFA’s carrier effect can distortions on burst-mode opticalinputs.We need to clamp the gain of the EDFA to avoid distortion.
Lecture 5: Active Components 朱祖勍36
Gain-Clamping (GC) in EDFA
Data*
ModDFB-LD
ParBERT
70%
30%ATT
OATTBPF
BPF
(a) (b)EDFA
70%
30%
1555.32nm
EDFA’s OutputWith GC
Lecture 5: Active Components 朱祖勍37
Fundamentals of Semiconductor Laser
Lecture 5: Active Components 朱祖勍38
Longitudinal Modes
Facets: the two end mirrors ofthe cavity.
The wavelength is within thebandwidth of the gain medium.
The length of the cavity is anintegral multiple of half thewavelength in the cavity(longitudinal modes).
Laser: put the gain medium in acavity and cause resonation.
Lecture 5: Active Components 朱祖勍39
Multiple-Longitudinal Laser (MLM)
Multiple-longitudinal laser (MLM): the laser oscillatessimultaneously in several longitudinal modes.
MLMs have large spectral widths~10 nm in reality
For high-speed optical communication systems, the spectralwidth of the source must be as narrow as possible tominimize the effects of chromatic dispersion.
Thus, it is desirable to design a laser that only oscillates in asingle-longitudinal mode.
Lecture 5: Active Components 朱祖勍40
Single-Longitudinal Laser (SLM)
Single-longitudinal mode oscillation can be achieved byusing a filtering mechanism in the laser that selects thedesired wavelength.
Side-Mode Suppression Ratio: determines the level towhich the other longitudinal modes are suppressed,compared to the main mode.
~30 dB in reality
SLM Lasers: Distributed-Feedback (DFB) lasers, Distributed-BraggReflector (DBR) lasers, External Cavity Lasers, Vertical Cavity Surface-Emitting (VSCEL) lasers and etc.
Lecture 5: Active Components 朱祖勍41
SLM Laser Spectrum
Lecture 5: Active Components 朱祖勍42
Distributed-Feedback (DFB) Lasers
Distributed-Feedback (DFB) lasers achieve single-longitudinal lasing with a Bragg grating (corrugationwaveguide) within the gain region of the cavity.
The corrugation waveguide acts as a filter to select thelongitudinal mode for lasing.
Lecture 5: Active Components 朱祖勍43
Distributed-Feedback (DFB) Lasers
DFB lasers are required in almost all high-speed opticaltransmission systems.
A DFB laser is usually packed with a thermoelectric (TE)cooler and a photo-detector attached to its rear facet.
The TE cooler is necessary to maintain the laser at a constantoperating temperature to stabilize its output wavelength.
The photo-detector monitors the optical power leaking out of the rearfacet, which is proportional to the laser’s output power.
The packaging of a DFB laser contributes a significant partof the device’s overall cost.
Lecture 5: Active Components 朱祖勍44
Distributed Bragg Reflector (DBR) Lasers
Distributed Bragg Reflector (DBR) lasers achieve single-longitudinal lasing with a Bragg grating (corrugationwaveguide) outside the gain region.
The main advantage of DBR lasers is that the gain region isdecoupled from the wavelength selection region, and thus itis possible to control both regions independently.
Lecture 5: Active Components 朱祖勍45
External Cavity Lasers
Suppression of oscillation at more than one longitudinalmode can also be achieved by using another cavity, calledexternal cavity, following the primary gain cavity.
The net result is that laser is only capable of lasing at thosewavelengths that are resonant wavelengths of both theprimary and external cavities.
Filters for external cavity:Fabry-Perot filter
Diffraction grating
Fiber Bragg grating
…
Lecture 5: Active Components 朱祖勍46
Vertical Cavity Surface-Emitting Laser (VCSEL)
In an MLM laser, the frequency spacing between twoadjacent longitudinal modes is c/2nl, where l is the cavitylength and n is the refractive index.
If we decrease the cavity length, an SML can be achievedwhen the frequency spacing increases such that only onelongitudinal mode occurs within the gain bandwidth.
With such a small vertical cavity, mirrors can be formed onthe top and bottom surfaces.
The laser emits from one of the surfaces.
Lecture 5: Active Components 朱祖勍47
Vertical Cavity Surface-Emitting Laser (VCSEL)
Since the gain region only has a very short length, veryhigh reflectivity is required for the laser to oscillate.
Such high reflectivity is difficult to obtain with metallicsurfaces, and thus a stack of alternating low- and high-index dielectrics is used to serve as a high-reflective mirror.
Advantages of VCSEL: simple and more efficient fibercoupling, easy packaging and testing, low-cost and etc.
We can fabricate and package multiple lasers in one time.
Lecture 5: Active Components 朱祖勍48
Structure of VCSEL
Lecture 5: Active Components 朱祖勍49
Mode-Locked Lasers
Mode-locked lasers can generate narrow optical pulses.
In an MLM laser, if each of its longitudinal mode oscillates atrandom phases, the output intensity is near-constant in time.
If each mode operates with the same phase, the modes willperiodically all constructively interfere with each other andproduce optical pulses in time, i.e., mode-locked.
The time interval between two pulses of is 2nl/c.
The most common means of achieving mode lock is bymodulating the gain of the laser cavity with a period of 2nl/c.
Lecture 5: Active Components 朱祖勍50
Mode-Locked Lasers
Applications:Nonlinear OpticsNuclear fusion…
Lecture 5: Active Components 朱祖勍51
Wavelength-Tunable Lasers
For WDM systems, lasers operating at different wavelengthsare highly desirable.
The inventory and sparing issues associated with stockingmultiple fixed-wavelength lasers cause high cost and affecteverybody from laser manufacturers to network operators.
Tunable lasers are one of the key enablers of reconfigurableoptical networks. They provide the flexibility to choose thetransmit wavelength at the source.
An ideal tunable laser should be able to tune rapidly over awide tuning range and its output should be stable.
Lecture 5: Active Components 朱祖勍52
Wavelength-Tuning Mechanism
Injecting current into a semiconductor laser causes achange in the refractive index of the material and in turnchanges the lasing wavelength.
Very short tuning time, a few ps (10-12 second).
A tuning range of 10-15 nm in a DFB laser operating in 1550 nm range.
Change the temperature of a semiconductor laser canchange the lasing wavelength too.
Very slow tuning time, a few second to stabilize.
We need 10 degrees to tune 1 nm wavelength.
Mechanical tuning, such as change the gain cavity lengthwith force or select output from a laser array.
Lecture 5: Active Components 朱祖勍53
Wavelength-Tunable Laser with DFB Laser Array
Use an array of wavelength-differentiated DFB lasers andturn on one of them at any time.
We can fabricate and package DFB lasers in a singlepackage and use it as a wavelength-tunable laser.
Only inject current to one DFB laser in the array at any time.
Lecture 5: Active Components 朱祖勍54
Wavelength-Tunable Laser with DBR Laser
A conventional DBR laser only has one electrode to injectcurrent to or generate forward biasing in the gain region.
If we add another electrode to inject a separate current intothe Bragg region, we can change the filtering response ofthe Bragg region and make the laser wavelength-tunable.
The tuning is then discrete to jump among longitudinalmodes. We have a two-section DBR laser.
Lecture 5: Active Components 朱祖勍55
Wavelength-Tunable Laser with DBR Laser
To obtain continuous tuning, we need to add the third phasesection in the DBR laser, i.e., three-section DBR laser.
Injecting current into the phase region allows use to obtaincontrol of cavity mode spacing and then the outputwavelength can be tuned continuously.
The tuning range is 10-15 nm, and we need more tricks toincrease the tuning range.
Lecture 5: Active Components 朱祖勍56
Sampled Grating DBR Laser
A sampled grating DBR (SG-DBR) laser has two gratings,one in the front and one in the back.
In order for lasing to occur, we need to have an overlapbetween the two reflection peaks of the Bragg gratings.
Even though the tuning range of each reflection peak islimited to 10-15 nm, combining two sets of reflection peaksresults in a much larger tuning range.
Lecture 5: Active Components 朱祖勍57
Sampled Grating DBR Laser