Introduction To Optical Networking: From Wavelength Division Multiplexing To Passive Optical Networking
Introduction To Optical Networking: From Wavelength Division Multiplexing To Passive Optical Networking
Introduction To Optical Networking: From Wavelength Division Multiplexing To Passive Optical Networking
Dr. Manyalibo J. Matthews Optical Data Networking Research Bell Laboratories, Lucent Technologies Murray Hill, NJ 07974 USA
A.Harris 2000
M.Matthews
Lucent A la Carte
Akiyama
Quantum Wire Lasers
Tunable Lasers
Matthews
Telecom Lasers
Outline
Introduction Overview of Optical Networking Coarse Wavelength Division Multiplexing Ethernet Passive Optical Networks Conclusions & Future
Types of Networks Fiber, Lasers, Receivers
CO-1
Access Node
CO-n
Metro DMX
EPON node
s Pa e siv DM W
Regional/Metro
PON
Access/Enterprise
DSL, FTTH
Optical Amplifier
WDM Routers
PIN
DWDM: CWDM: TDM: SCM: SMF: MMF: LWPF: DCF: EML: DFB: FP: APD: PIN:
Dense Wavelength Division Multiplexing (<1nm spacing) Coarse Wavelength Division Multiplexing (20nm spacing) Time Division Multiplexing (e.g. car traffic) Sub-Carrier Multiplexing (e.g. Radio/TV channels) Single-Mode Fiber (core~9m) Multi-Mode Fiber (core~50m) Low-Water-Peak Fiber Dispersion Compensating Fiber Externally modulated (DFB) laser Distributed Feedback Laser Fabry-Perot Laser Avalanche Photodiode p-i-n Photodiode
Dispersion:
launch
receive
Types of Dispersion
Chromatic Dispersion is caused mainly by the wavelength dependence of the index of refraction (dominant in SM fibers) Modal Dispersion arises from the differences in group velocity between the modes travelling down the fiber (dominant in MM fibers)
t t t t
Cross-Phase Modulation:
Four-wave Mixing:
First Window
Second Window
ATTENUATION (dB/km)
2.0
1.5
Third Window
1.0
0.5
800
900
1000
1100
1200
1300
1400
1500
1600
1700
WAVELENGTH (nm)
850nm
1310nm
1550nm
First window, second window, third window correspond (roughly) to first, second and third generation optic network technology
First Window @ 850nm High loss; First-gen. semiconductor diodes (GaAs) Second Window @ 1310nm Lower Loss; good dispersion; second gen. InGaAsP Third Window @ 1550nm Lowest Loss; Erbium Amplification possible
Dispersion Characteristics*
3.0
Second Window
Third Window
-30
First Window
-60
-90
-120
800
900
1000
1100
1200
1300
1400
1500
1600
1700
WAVELENGTH (nm)
850nm
1310nm 1550nm Standard SMF has zero dispersion at 1310nm Low Dispersion => Pulses dont spread in time Dispersion compensation needed at 1550nm Limits data transmission rate due to ISI (inter-symbol interference) Dispersion not so important at 850nm Loss usually dominates
Eye opening
850nm
-m lti mu
Cat 3 Cat 5 limit limit
1310nm
g sin e fib de mo le-
1550nm
e od
er fib
Cat 7 limit
1000
10,000
For short reaches (1-2 km), all optics are Gigabit capable For longer reaches (~10 km), only 1310/1550 nm optics are Gigabit capable
Technology Trends
850nm & 1310nm Preferred by high-volume, moderate performance data comm manufacturers
Reason? You need lots of them, they dont need to go far, and youre not using enough fiber ($) to justify wavelength division multiplexing (WDM), I.e. low-quality lasers are OK.
Reason? You dont need lots, but they have to be good enough to transmit over long distances cost of fiber (and TDM) justifies WDM 1550nm is better for WDM
DFB +
gain
cleave
AR coating
FP:
Multi-longitudinal Mode operation Large spectral width high output power Cheap
DFB:
Single-longitudinal Mode operation Narrow spectral width lower output power expensive
Typical FBG-ECL:
gain HR AR Lensed tip FBG T=25C T=85C <1nm grating
-20
-40
-60
-80 1309.0 1309.5 1310.0 1310.5 1311.0 1311.5 1312.0 Wavelength (nm)
?
1-2nm grating
FBG-ECL output
Typical FP output
-11
Power (dB)
Narrow FBG bandwith limits output ~1nm for extended reach or WDM applications. Simple design (AR-coated FP, XBR, butt-coupled FBG) Mode-hop free operation over 070C
11 11 11 11 1111 11 11
-11
-11
-11
-11 11 11
wavelength (nm)
FP drift ~ 0.3nm/oC
ave dependence 1111 . nm/C
Wavelength(nm)
1 1
1 1
1 1
Temperature ( C)
DWDM: High channel count, narrow channel spacing Temp-stablized DFBs required Temp-stablized AWGs required (typically) 1480nm CWDM: Low channel count, large channel spacing Uncooled DFBs can be used Filters can be made athermal 1260nm xWDM?: Moderate channel count, moderate channel spacing FBG-ECL or Temp-stablized DFBs required Filters can be made athermal suitable for athermal WDM PON! 1480nm
18 channels (O,E,S,C,L) 3.2nm (400GHz) >100 channels (C+L+S) 20nm
1610nm
1610nm
1610nm
CWDM Lasers
16 uncooled, directly modulated CWDM lasers (DMLs) rated for 2.5 Gb/s direct modulation (cheap! - $350 a piece) NRZ-modulation at 10 Gb/s (careful laser mounting; no device selection)
2.5-Gb/s DML
50 line
47 chip resistor
Outside Plant
Homes/Businesses
Internet
PON
IP Video Services
NO Active Elements in Outside Plant Enable triple-play services Simple & cheap
Choices of PONs
Architecture/Layout
OLT ONU
Upstream Multiplexing
ONU ONU ONU
WDM:simple, expensive
ONU ONU ONU
DFB
EPON
10G Ethernet Or up to 6 1GbE
Metro Network
FP
Metro Edge
optical splitter
Broadcast Video VOD Voice/IP Services
Note on Lasers: -Use DFB at headend (shared) -Use FP at Homes (not shared)
ONU Design
PON 1.25G BM BiDi Xcvr SERDES (w/CDR) GigE uplink
watchdog1 watchdog0
CHILD BOARD
FPGA w/ Embedded Processor Packet memory
FPGA
CPU
TX EPON MAC
Mux
Packet Memory
Demux
RX
PARENT BOARD
Control Parser
Serial Port
ONU
PON
OLT Design
watchdog1
GigE uplink
SERDES (w/CDR)
EPON core MPCP core Grant List Gate Generator RX GMII TX TX EPON MAC RTT table Memory manager Queue manager RTT Processor Report processor Report table
Mux
Packet Memory
Demux
RX
Control Parser
FPGA
CPU
Edge Router
OLT
2
1
1 2
O N U
3 2
Downstream: continuous, MAC addressed Uses Ethernet Framing and Line Coding Packets selected by MAC address QOS / Multicast support provided by Edge Router
3 2
O N U O N U O N U
O N U
3
Control Reports
Edge Router
OLT
3 3
1
2
2
3 3
Upstream: Some form of TDMA ONU sends Ethernet Frames in timeslots Must avoid timeslot collisions Must operate in burst-mode BW allocation easily mapped to timeslots
O N U
3 3
BURSTMODE TRANSMITTERS
Data Clock Tx FIFO Encoder Serializer Transmitter Physical Media
Prebias
Optical output
1 0 off Ith
current
Modulation current
BURST-MODE RECEIVERS
Data Clock
Rx FIFO
Decoder
Deserializer
CDR
Limiting Amp
Receiver
Reset
PROBLEM OF FAST CDR LOCKING GAIN LEVELING & DYNAMIC RANGE OF OPTICAL RECEIVER
IMPACT ON EFFICIENCY
Cascaded PON
ONU 1 OLT 1:4 1:8
. . .
ONU 2
guardband
Throughput Efficiency
1 11 . 1 1 11 . Utilisation 0. 0 1 11 . 0. 0 0 11 . 1 1111 1111 1111 A G C + C D R + L A S E R O N /O FF ( n s ) 0. 0
Our current situation Standa rd GE transc eivers Burst-mode transceivers
O P C T L T S I F R H O E T I D F O K S N L P ST T SM
Data
C R C
Ethernet
IP
64 Bytes
TCP
~1460 Bytes
Conclusions
Optical Networking getting closer and closer to end user For Metro, CWDM is lowest cost solution, but must be improved to handle 10Gbps PON systems could deploy in mass over next 1-2 years, with EPON one of the leading standards Lasers dominate cost, therefore useful to study physics of low-cost laser structures! THANK YOU VERY MUCH! (Domo Arigato Gozaimashita!)
Spare Slides
Typical p-i-n receivers w/ ~150nA current noise, 1.25Gbps, R~1 -27dBm (about 1 W) Typical 1310nm FP lasers 0dBm output power (about 1mW)
D (ps/nm.km)
(nm)
Reach (km)
1 1 1 1 1 1 1 1 1 1 1 11 . 11 . 11 . 11 . 11 .
L = ( DB )
k ln k 1 mpn
11 . 11 .
= BDL
Where k is the MPN coeficient, dependent on mode power correlations.
Reach dependent on quality of laser (k factor) (another) Reason why asymmetry in PONs (e.g., 155/622Mbps) are favored GigE? Worst-case isnt quite fair statistical model shows most fiber-laser combinations, D<3ps/nmkm, k<0.5.
REDUCING MPN
Dispersion Compensation at OLT
Additional Loss, some cost One-size wont fit all, SMF 0 ~ 1300-1325nm
Input waveguides
Output waveguides
TM, y
} core layer