Abstract: Tunable semiconductor lasers continue to be in just about everyone’s list of important components for future fiber optic networks. Various designs will be overviewed with particular emphasis on the widely tunable (>32nm) types.
Tunable Semiconductor Lasers
a tutorial
Larry A. Coldren University of California, Santa Barbara, CA [email protected] Agility Communications, Santa Barbara, CA [email protected] 2
Contents ■ Why Tunable Lasers?
■ Basic Tuning Mechanisms
■ Examples of Tunable Lasers
■ Control of the Wavelength
■ Reliability Issues 3
Optical Network Architecture
Core
Edge More bandwidth and services/$Æ Low-cost components and agile architectures 4
Introduction
■ Tunable lasers have been of great interest for some time − Dynamic networks with wavelength reconfigurability − Networking flexibility − Reduced cost − One time provisioning (OTP) and sparing seen as side benefits ■ Current market conditions…. − More cautious approach from carriers and system vendors − OTP and sparing are now the leading applications ■ Tunable lasers are compared with DFB or EML − Important to do “apples to apples” comparison − Functionality − Performance − Total Cost of Ownership 5
Why Tunable Lasers? ■ One time provisioning—inventory and sparing
■ Field re-provisioning—new services without hardware change or truck roll
■ Reconfigurable Optical Add/Drop Multiplexers (ROADM)— Drop and add any channel without demux/mux
■ Wavelength conversion—Eliminates wavelength blocking without OEO line cards
■ Photonic Switching—Eliminates many OEO line cards
■ Wavelength Routing—Use passive optical core 6 Applications – One time provisioning—the universal source
■ Laser is provisioned once only
■ Simplifies manufacturing
■ Drastically reduces inventory
■ Minimizes sparing to a manageable level
■ Simplifies forecasting 7
Applications – Re-provisioning
■ Laser is provisioned many times remotely to set up new services − Seconds timeframe − Point and click or ultimately controlled automatically by software
■ Can only be addressed using a widely tunable laser − Without severe constraints
■ Drastically reduces inventory
■ Simplifies forecasting 8
Applications – Re-configurable OADM
Tunable filter elements
in thru out
drop add
Tunable Rx Tx
■ Drop and Add without Demux and Mux of all channels ■ Must be “hitless” filter tuning ■ Eliminates mux/demux and OEO ■ Tunable lasers are a key enabler 9
Applications – Photonic Switching 1
123 LR Tx LR SR SR Rx 123 Rx Tx OR Rx Tx 4 4 5 5 6
Grey 6 Optics Tunable 8 7 8 7
■ Photonic switches require O-E-O on I/O to prevent blocking ■ Tunability reduces O-E-O requirements in half ■ Requires moderately fast switching (ms) 10
Applications – Wavelength Conversion
■ Intersection of metro rings ■ Wavelengths transition between rings − in optical domain ■ Tunable lasers used to resolve wavelength blocking − Alternative is a bank of fixed wavelength lasers
Node Node 3B Node 3R Node 4B 4R
Node Node 2B Node Node 2R 1B 1R Rx Tx 11 Applications – Wavelength Routing (Optical Packet Switching)
Line cards with Line cards Tunable lasers
T x
N x N Lambda Router (AWG)
■ High capacity, high density router function—need wide tuning ■ Wavelength used to route traffic through passive device ■ For Packets requires very fast switching 12
Contents ■ Why Tunable Lasers?
■ Basic Tuning Mechanisms
■ Examples of Tunable Lasers
■ Control of the Wavelength
■ Reliability Issues 13
Generic Single-Frequency Laser
Mode-selection Gain filter Output
mλ/2 = nL Mirror-1 Mirror-2
Gain
Lasing mode Mode-selection filter Possible modes λ 14
Examples of Single-Frequency Lasers gain ■ DFB Grating (mode-selection − All-elements combined and Light Out & distributed mirror) distributed along length Gain region DFB AR HR Gain Rear Mirror ■ DBR Light Out − Elements separated with individual biases DBR
Ext. grating mirror Gain AR ■ External Cavity Light Out − Gain block + external lens & grating
Collimating lens
■ VCSEL Active − Short cavity for mode selection DBR mirrors
Light Out 15
How Tunable Lasers Tune
Mode wavelength: Effective Cavity length mλ/2 = nL Mode number Effective index Wavelength
Relative change in wavelength: Tuned by mode-selection filter (via index or grating angle) ∆λ = ∆ n + ∆ L - ∆m λ n L m Tuned by net cavity index change Tuned by physical length change 16
Generic Tunable Single-Frequency Laser Tunable Gain Mode- output Cavity selection phase filter (∆n) (∆m) (∆L)
Mirror-1 Mirror-2
mλ/2 = nL Gain
Mode-selection Lasing mode filter (∆m) Possible modes(∆n, ∆L) λ 17
Solutions for Tunable Lasers ■ DBR Lasers Gain Phase Rear Mirror Light Out − Conventional DBR (<8 nm)
− Extended Tuning DBR’s (≥ 32 nm) DBR
■ External Cavity Lasers (≥ 32 nm) − Littman-Metcalf/MEMs − Thermally tuned etalon Light Out
Light Out ■ MEMS Tunable VCSEL (< 32 nm) − Optically or electrically pumped
W indow ■ DFB Array (3-4 nm X #DFBs) − On-chip combiner + SOA NEC − Or, off-chip MEMs combiner 8 Microarray MMI DFB-LDs − Thermally tuned S-bent waveguides W indow SOA
Chip size: 0.4 x 2.15 mm 2 18
Contents ■ Why Tunable Lasers?
■ Basic Tuning Mechanisms
■ Examples of Tunable Lasers
■ Control of the Wavelength
■ Reliability Issues 19
Examples of Tunable Lasers ■ Narrowly tunable (not discussed further) − Temperature tuned DFBsÆ ~ 3nm − Narrowly tunable 2 or 3 section DBR lasersÆ ~ 8nm ■ DFB selectable arrays − Select DFB array element for coarse tuning + temperature tune for fine cavity mode tuning − Integrated on-chip combiners + SOAs or off-chip MEMs deflectors ■ External-cavity lasers − External grating reflector for mode-selection filter − Angle-tune mirror for mode selection—coarse tuning − Change length and/or phase section for fine tuning ■ MEMS Tunable VCSELs − Move suspended top mirror by electrostatic or thermal tuning − Single knob tuning for both coarse and fine ■ Widely tunable DBR lasers − Coarse tuning by index tuning of compound mirrors/couplers − Fine tuning by index tuning of phase section − Dual SGDBR or vertical-coupler + SGDBR mode selection filters 20 Wavelength-selectable light sources (WSLs) for wide-band DWDM applications
Window Feature DFB-LD-array-based structure 8 Microarray Wide-band tunability MMI DFB-LDs S-bent Compact & stable waveguides Window SOA Multi-λ locker module Chip size: 0.4 x 2.15 mm2 Performance Schematic of wide-band WSL
WSLs for S-, C-, L- bands (OFC’02) 8 array, ∆λ ~ 16 nm (∆T = 25K) x 6 devices Multi λ-locker integrated Wide-band WSL module (OFC’02) ∆λ ~ 40 nm (∆T = 45K) Multi λ−locker integrated Wide-band WSL module 21 WSLs for S-, C-, L- bands applications - Lasing spectra -
20 10 S-band C-band L-band 0 -10 -20 -30 -40 Intensity (dBm) -50 S 1 S2 C1 C2 L1 L2 -60 Wafer 1 Wafer 2 -70 1470 1490 1510 1530 1550 1570 1590 1610 W avelength (nm )
∆λ 16 nm (∆T 25K) @15 - 40 ~ ℃ 6 devices → 135 channels @100-GHz ITU-T grid SMSR > 42 dB
Pf > ~ 10 mW @ IDFB= 100 mA, ISOA= 200 mA 22
Fujitsu DFB Array Integrated Tunable Laser
0.5 × 1.8 mm
Fujitsu Laboratories Ltd. 23
Fujitsu Wavelength Tuning Characteristics
Temperature tuning Spectra at 32 wavelengths
Fujitsu Laboratories Ltd. 24
Santur Switched DFB Array 1 mm
12 element DFB array, each temperature tuned 3nm for 36nm total tuning range - only one laser on at a time MEMS mirror couples the selected laser to fiber
Advantages: DFB characteristics (optical quality, reliability, wavelength stability) No SOA, tuning sections, phase-sensitive mechanics High yield, low cost passive alignment (MEMS does the rest) Built-in shutter/VOA 25
Santur 20 mW Module Performance
z Full band tunability (36nm C-band, 42nm L-band) z Built-in wavelength locker (25GHz channel spacing) z >50dB SMSR, 2MHz linewidth z Typical tuning time ~ 2sec z Resistant to shock and vibe with no servo (10G causes < 0.2dB fluctuation in power) 26 Intel External-Cavity Approach (acquired from New Focus)
• Double sided external cavity laser design, well known in test and measurement applications • Temperature tuned etalon replaces mechanical tuning device • No moving parts, but challenging packaging requirements 27
Littman-Metcalf Cavity (after New Focus)
HR Laser Diode Coating Chip Collimating Lens
AR Retroreflector Coating
Pivot Point
Wavelength Laser Tuning Output
Diffraction Grating 28 Iolon External-Cavity Laser with MEMs Mirror Movement 29
Tunable VCSELs (optically pumped)
■ Cortek-Nortel-Bookham?
■ Component technologies − MEMS − Thin Film − InP Laser − Packaging
■ Advantages: − High Power − Wide Tuning Range − Continuously Tunable 30 31
Agere “Narrowly” Tunable DBR/SOA/EAM
EA-DBR Operation Tuning Current High Low
L Gain Tuning EA Modulator Bragg mirror select FP mode Tuning current moves Bragg mirror λ A Five Stage Bell Labs Design
Gain DBR Optical Detector Modulator Section Mirror Amplifier “Power Monitor”
Many more 7 – 10 nm designs 32 Extended tuning range: SSGDBR--NEL
Phase modulated gratings 33 Extended tuning range: GCSR--ADC-Altitun
SGDBR + GACC
30
25
Gain Coupler Phase Reflector S-DBR 20 400µm 600µm 150µm 900µm 15 ] A p-InP 10 8 Reflector current [m current Reflector 5 n-InP 7 06 1515 1525 1535 1545 1555 1565 QWs structures λg = 1.3 µm λg = 1.38 µm 5 λg = 1.55 µm 4 ] A 3 2
Coupler current [m current Coupler 1 0 1515 1525 1535 1545 1555 1565 Wavelength [nm] Agility’s Extended Tuning Range Technology: Widely Tunable SGDBR Lasers 35
Sampled Grating Tunable Lasers 10 Front Rear 2 4 1 3 Amplifier Mirror Gain Phase Mirror 0
-10
-20
Light Out -30 Laser Emission, dBm Emission, Laser Relative Power (dB) -40
-501 Back ■ Front 5-10X Tuning Range of DBR 0.8 ■ Reliable, Manufacturable InP Technology 0.6 ■ Can Cover C band, L band or C + L 0.4 Mirror Reflectivity Mirror 0.2
0 1530 1540 1550 1560 1570 Wavelength (nm) 36
Advantages of Monolithic Integration Widely Tunable SG-DBR Laser with integrated SOA and EAM
Front Rear Modulator Amplifier Mirror Gain Phase Mirror
Light Out
EA Modulator SOA SG-DBR Laser
Advantages: smaller space (fewer packages) lower cost (fewer package components) lower power consumption (lower coupling losses) high reliability (fewer parts) 37 Fast Wavelength Switching of SGDBR Lasers Packet Switching Applications 100
Electronic Optical signal at final ITU +/- ~10 GHz Trigger Switching time = 10 ns Voltage Voltage 80
Channel 50 on 60
Channel 50 off
Count 40
Channel 10 on 20 Channel 10 off Light Power PowerLight Light Light Light Power Power
0 102030405060708090100 0 Time (ns) 0 5 10 15 20 25 30 35 40 45 SwitchingTime (ns) Current source rise time can be designed for application. Inherent laser limit is in ~ 2-10 ns range. Thermal transients can complicate rapid switching. 38
SG-DBR Laser with Integrated SOA High Power Widely Tunable Laser: 20 40 dB 10 13dBm
0
-10
-20
-30
Fiber Coupled Power (dBm) -40 L-band 1570 1580 1590 1600 1610 Wavelength (nm) >100 50 GHz ITU Channels Fiber coupled power = 13dBm = 20mW SMSR > 40 dB SOA: Power leveling, blanking, and VOA w/o degradation of SMSR Channel switching time (software commandÆverified channel) < 10 ms 39
RIN & Linewidth Dependence on Power
RIN vs. SOA Current White FM Noise Density vs. λ -130 5 13
SOA Current 12 40 mA, 5.7 dBm 4 -135 60 mA, 7.5 dBm 80 mA, 8.6 dBm Output Power (dBm) 100 mA, 9.2 dBm 11 120 mA, 9.8 dBm 3 150 mA, 10.5 dBm -140 10
2 RIN (dB/Hz)
Linewidth (MHz) 9 -145 1 8
-150 0 7 0 2000 4000 6000 8000 10000 1525 1530 1535 1540 1545 1550 1555 1560 1565 Frequency (MHz) Wavelength (nm) ■ RIN is only weakly dependent on output power (SOA current). ■ Linewidth is less than 2.5 MHz across all wavelengths − Scales with Laser Power as expected. 40 SGDBR-SOA-EAM Transmission Characteristics 5 1528 nm 1560 nm OC-48 Std. SMF 4 350 km
error rate, dB 3 -10 1540 nm PRBS 231-1 at 2.5 Gb/s 2 275 km 4th order Bessel- Thomson filter SONET mask with 25% 1 margin 200 km 1550 nm Unfiltered 2.7 Gb/s 0 1.53 1.54 1.55 1.56 1.57
Dispersion Penalty @ 10 Wavelength, µm Dispersion penalty at 10-10 errors/s error rate for 200, 275, and 350 km of standard SMF for 38 ITU channels sampled across C-band. 41 SGDBR-SOA-EAM
RF-ER, Pave, & VOA Operation 6 15 3 dBm ~1550 nm
14 RF Extinction (dB)Ratio 5.5
13 2 dB 5 12
4.5 11 0 dB Time AveragedTime (dBm) Power
4 10 192 193 194 195 196 Optical Frequency (THz) -3 dB Ave. power >5 dBm and RF ER > 10 dB across C-band Output power dynamic range of ~10 dB w/ small change in SMSR and Wavelength (open loop operation) 42
OC-192 Operation of EAM 31 PRBS 2 -1, Vp-p = 3V
Integration technology compatible with higher bit rates > 10 dB RF ER across C-band Not optimized, improvements to come 43
MZ-SGDBR (UCSB)
Curved waveguides 200µm
MMI Length:96µm
Light Out
Width: 9µm Taper:20µm 44
Extinction & Chirp: MZ-SGDBR (UCSB)
• > 20 dB extinction Chirp parameter as function of with 2V drive DC extinction curve for 550µm 2 -5
• Negative chirp when -10 0 increasing reverse bias Alpha 1525nm Insertion Loss(dB) Alpha 1545nm ‘turns on’ modulator -15 -2 α -20
∆neff (real) 2∆φ -4 = = chirp ∆n (imag) ∆αL -25
eff Chirp Parameter
-6 -30 Measured by the (Power (dBm) 1525nm Power (dBm) 1545nm Devaux method -8 -35 -4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 Arm #1 DC Bias (V) 45
MZ-SGDBR RF Performance: Lumped (UCSB)
• BCB for low capacitance 10Gbit/s 15 • Lumped drive– can improve Eye 10 -1 with traveling wave electrodes PRBS
0
-2
BCB -4
Normalized S21 Normalized -4V Bias -6
-8 0 5 10 15 20 Frequency (GHz) 46
Contents ■ Why Tunable Lasers?
■ Basic Tuning Mechanisms
■ Examples of Tunable Lasers
■ Control of the Wavelength
■ Reliability Issues 47
Control Issues
■ Finding the desired channel − Look-up tables vs. channel counting? − Is global wavelength monitor required? − Must look-up tables be updated over life?
■ Staying on the desired channel − Is locker required to meet spec? − Is single knob control from locker sufficient over life? 48
Generic Tunable Single-Frequency Laser Tunable Gain Mode- output Cavity selection phase filter (∆n) (∆m) (∆L)
Mirror-1 Mirror-2
mλ/2 = nL Gain
Mode-selection Lasing mode filter (∆m) Possible modes(∆n, ∆L) λ 49
Control comparison across types
Laser λcoarse λfine Amplitude VOA
DFB Array/SOA Varray(j) T Igain(j) ∆ISOA
DFBs/MEMs VM1, VM2(j) T Igain(j) VM1, VM2(j)
SGDBR/SOA Im1, Im2 Iφ ISOA ∆ISOA
Ext. Cavity VMθ VML, Iφ Igain VMshutter
VCSEL/MEMs VM1 V*M1 Igain ------50
Iolon Control Scheme for Ext. Cavity Laser 51
Agility Control of SG-DBR Lasers
Control Circuitry Wavelength Locking Mirror Control ■ DWDM DBR Power Control − Power Control − Temperature Control(fixed) − Wavelength Locking − Mirror Control (Locking?)
Light Out SG-DBR
■ DWDM DFB comparison SOA Front Mirror Gain Phase Rear Mirror − Power Control − Temperature Control − Wavelength Locking 52
Contents ■ Why Tunable Lasers?
■ Basic Tuning Mechanisms
■ Examples of Tunable Lasers
■ Control of the Wavelength
■ Reliability Issues 53
Wavelength Reliability
■ It’s not enough to just put out the right power in a single mode for a long time (old criterion) ■ Prior to end-of-life of a multi-channel DWDM source, power & wavelength must be in spec. ■ Intimately linked to wavelength control (or lack of it) − Finding the desired channel − Look-up tables vs. channel counting? − Is global wavelength monitor required? − Must look-up tables be updated over life? − Staying on the desired channel − Is locker required to meet spec? − Is single knob control from locker sufficient over life? ■ If look-up tables must be updated, how can this be done reliably? 54
What causes the wavelength to change
Tuned by mode-selection filter (via index or grating angle) ∆λ = ∆ n + ∆ L - ∆m λ n L m Tuned by net cavity index change Tuned by physical length change
Physical Causes, assuming a fixed look-up table:
∆n – Changes in internal temperature, Tint, or carrier lifetime, τc ∆L – Physical movements—solder relaxation, MEMs charging ∆m – ∆n of DBR, ∆θ of ext. grating, or MEMs charging 55 Critical issues for wavelength stability
Laser Variables in Table *Critical ∆λ issues
DFB Array/SOA j, Ig(j), T, ISOA ∆n(Tint) Å ∆Ig
DFBs/MEMs j, Ig(j), T, VM1(j), VM2(j) ∆n(Tint) Å ∆Ig
SGDBR/SOA Im1, Im2, Iφ, ISOA ∆nDBR(τc) Æ ∆m
Ext. Cavity VMθ, VML, Iφ, Igain, VMshut ∆L(VM), ∆m(VM), ∆n(Tint) Å ∆Ig
VCSEL/MEMs VM1, Ig ∆L(VM)
*Requiring table update or global channel locator 56
Estimated Open-loop Wavelength Shifts
Laser Critical ∆λ issues ∆λ @ EOLgain (No table update)
DFB Array/SOA ∆n(Tint) Å ∆Ig 40GHz (10GHz/SOA feedback)
DFBs/MEMs ∆n(Tint) Å ∆Ig 40GHz
SGDBR/SOA ∆nDBR(τc) Æ ∆m<10GHz
Ext. Cavity ∆L(VM), ∆m(VM), 100GHz (MEMs charging)
∆n(Tint) Å ∆Ig
VCSEL/MEMs ∆L(VM) 1000GHz
• Only SGDBR lands on correct mode near EOL Open-loop • Others require global channel monitor or the like 57 Effects of SGDBR Mirror Aging: Measurement
7.5
0.09
6.5 0.08
5.5 0.07
0.06 4.5 240 Hours 576 Hours 0.05
rtBM 1080 Hours 3.5 1416 Hours 0.04 2088 Hours 2304 Hours 2.5 0.03 2640 Hours
0.02 1.5
0.01 0.5 0.51.52.53.54.55.50.00 0123456 rtFM
t=96 Hours t=2304 Hours ■ Corresponds to > 100 yrs of operation ■ Aging gives fixed amount of root current increase to provide a shift in the “mode map” to higher current . 58 Very High SGDBR Wavelength Stability and Reliability
3 2 1 0 1 2 3 104 ~ 106 Device Hours MTTF ~ ~350yrs 350 yrs measured. σ ∼~ 0.56 0.56 Ea=0.55 eV, n=2 λFITS @ 25 yrs < 1 Very low Bragg Wavelength 3 10 Aging Rates < 0.5 pm/ year at worse case.
102 Gain and SOA sections have Mirror Life Time (yrs) similar MTTF and failure distribution. Mirrors - Experimental 101 Mirrors - Least Squares Fit OK for open-loop operationÆ 0.115 205080959999.9 no mode hops or incorrect Cumulative Failures, % channels 59
SGDBR Laser/SOA FITs vs. Time
250 • Open-loop failure rate vs. time 200
150 • Gain section determines
EOL 100 Failure Rate (FITs) Rate Failure
• Closed-loop mirror control 50 has also been implemented to monitor any drift 0 0 5 10 15 20 25 Operating Time (yrs)
2*Mirror Failure Rate Amplifier Only Failure Rate Gain Only Failure Rate Total Failure Rate 60
SGDBR vs DFB Chip Reliability
10000 Use Current Density ■ Historically, DBR Reliability (~4kA/ cm2) WAS Poor… 1000
100 ■ Defects in the grating area, DFB EOL found to be primary cause of *Mawatari, 1999 DBR failure. 10
1 ■ Improvement to re-growth Agility, 2002 (InP/InP) and minimal grating area of SG-DBR, allow 0.1
equivalent or better WavelengthLasing Shift (pm/Year) 0.01 performance vs. DFB’s. 110100 DBR Current Density (kA/cm2)
*Mawatari et al, “Lasing Wavelength Changes Due to Degradation in Buried Heterostructure DBR Laser”, Journal of Lightwave Technology, v.17, no.5 1999 61
Summary
■ Tunable lasers can reduce operational costs ■ Narrowly tunable versions have some short term inventory/sparing cost advantages but newer full-band types offer many further opportunities ■ Several configurations have emerged for current applications ■ Monolithic integration offers significant potential for reducing size, weight, power, & cost ■ Wavelength control issues still exist for many configurations. Look-up table updating and/or global channel monitors are necessary in some cases. ■ Reliability has been proven for the SGDBR version without any updating of the look-up tables or need for channel searching