Transcript ken_pedrotti

Survey of Short-Reach
Optical Interconnect
Ken Pedrotti
Robert Dahlgren
Presented 10 November 2005
Outline
•
•
•
•
•
•
•
•
Introduction
Pure silica fiber trends
Splicing trends
Connector trends
Active component trends
Transceiver module trends
What to expect in the year 2010?
Questions and discussion
Short Reach Links < 300m
• Optical Internetworking Forum (OIF)
– VSR-1 through VSR-4 classifications
• Gigabit Ethernet
– Multimode 850 nm
• 10 Gigabit Ethernet
– Singlemode or parallel multimode
• Fibre Channel
• Serial HIPPI
• Proprietary links
Main Commercial Trends
• Fiber to the home/curb/pedestal
deployment is moving forward
• Common form factors
• Smaller form factors
• Standards-based specifications
• “Short Reach” and VSR standards
• Convergence of specs at 1 and 10 Gbps
• WDM begat DWDM and CWDM
• Pure silica core fiber less of a niche
• Ribbon fiber and mass splicing/termination
Lesson from the Aircraft Industry
Convergence
• Telecom and datacom PHYs have less
distinction
• Traditional datacom companies like Cisco
are making more carrier-class equipment
• Common medium (often SMF)
• Data rates moving towards convergence
– OC-192, 10G Ethernet, 10G Fibre Channel
• Possible to unify component markets to
take advantage of economies of scale
What Convergence Means
for VSR Links
• 1 and 10 Gbps may be the “sweet spot” for
short links up to 300 m for some time
– 850 nm VCSEL and MMF
– Single and multi-fiber arrangements
• Broad support from industry in volume
• Common electrical interface
• 40 Gbps and 100 Gbps serial technology
too expensive for VSR
• WDM approaches not suitable for VSR
Ribbon Optical Fiber
• Four, twelve, or sixteen fibers
• Usually spaced at 250 mm pitch
• Connectors, splicing more complex
• Cost per splice about 2 ~ 2.5x
• Lower cost per fiber
Pure-Silica Core Fiber Trends
• Several manufacturers now make optical
fiber for high radiation environments.
• Singlemode, step-index multimode,
graded index multimode
• Cost premium ~TBD compared to
standard (e.g. Ge-doped core) fiber of the
same type
• Improved polymer coatings
Pure-Silica Core Fiber Vendors
•
•
•
•
•
•
•
•
•
Fujikura Ltd.
Oxford Electronics
Mitsubishi International
Verrillion Inc.
OFS (formerly Lucent)
CorActive – On custom basis
Sumitomo Electric
3M dropped production
Corning – MMF in development
Evolution of Arc Fusion Splicing
• Active alignment with optical source and
detector at the fiber endfaces
• Local injection into core and detection
• Imaging-based alignment
• Ribbon fiber splicers (mass splicing)
• V-groove (passive) splicers
• Enabled by tighter fiber concentricity specs
• Smaller, lighter, better ergonomics
• Compensation schemes, loss estimation
Improvement of Fiber Concentricity
Enables V-groove Alignment
Active Core Alignment
V-groove Alignment
1985 VINTAGE FIBER
0.64
0.72
0.8
0.72
0.8
0.56
5
0.56
0.48
0.4
0.32
0.24
0
0.16
0.8
0.72
0.64
0.56
0.48
0.4
0.32
0.24
0.16
0
10
0.08
10
15
0
Frequency
20
0.08
0.48
1995 VINTAGE FIBER
30
0
Frequency
0.4
Splice Loss (dB)
1995 VINTAGE FIBER
Splice Loss (dB)
0.64
Splice Loss (dB)
0.32
0
0.8
0.72
0.64
0.56
0.48
0.4
0.32
0.24
0.16
0.08
0
0.24
5
0.16
10
8
6
4
2
0
0.08
Frequency
15
0
Frequency
1985 VINTAGE FIBER
Splice Loss (dB)
Splicing trends
• Continued acceptance of v-groove based
splicers in non R&D applications
• Continuing acceptance of mass fusion
splicing for ribbon fiber
• Continued development of custom splice
programming, e.g. thermal diffusion core
expansion for specialty fiber
• Laser-based fiber stripping and cleaving
needs cost reduction
Mass Fusion Splice of
12-fiber Ribbonized SMF
Images courtesy AFL Telecommunications
Splice Economics
• Splicing cost roughly 20~40 € not including
costs associated with access and
packaging. Can be much higher.
• Mass splices cost roughly 2 ~ 2.5x for a 12fiber ribbon
• Spliceless ATLAS design tradeoff
• Need lower cost fiber recoating systems
and proof-testers for High-Rel applications
• Splicer manufacturers: Fitel, Fujikura, 3sae
Connector Market Landscape
• Simplex-SC is de-facto standard for telecom.
• Duplex-SC is de-facto standard for datacom.
• Newer “Small Form Factor” connectors vying
for market dominance.
• Several incompatible connectors for ribbon
fiber applications.
Legacy Connectors
ST
Biconic
FC
SMA
ESCON
Images used with permission of Alcoa-Fujikura, Ltd.
“old” FDDI
Ribbon Fiber Connectors:
MT Ferrule Technology
Images used with permission of US Conec, Ltd.
Small Form Factor Connectors
• Driven by smaller front-panel opening, like the
ubiquitous R-45 telephone/ethernet jack.
• Driven by low-cost 100 Mbps and 1 Gbps
ethernet system and cable companies
– High front panel density = low cost/port
• Telcos are looking to replace SC connector
– High front panel density = CO and closet space
• Incorporate cost-saving features
• Incorporate ergonomic features
• Some allow field termination
Some SFF Connectors
MT-RJ
LC
Duplex-LC
MU
Images used with permission of Alcoa-Fujikura, Ltd.
Laser Diode Structures
Most require multiple growth steps
Thermal cycling is problematic for electronic devices
Detector Technologies
Layer Structure
MSM
Simple, Planar,
Low Capacitance
Low Quantum Efficiency
Semiinsulating GaAs
(Metal Semiconductor Metal)
InGaAsP p 5x1018
InGaAs n- 5x1014
InP
n 1x1019
Contact
Absorption
Contact
PIN
Contact
Multiplication
Transition
Absorption
Contact
Substrate
APD
Waveguide
InP
InP
InGaAsP
InGaAs
InP
InP
Absorption Layer
Absorption Layer
Contact layers
Trade-off Between
Quantum efficiency
and Speed
Gain-Bandwidth:
p 1x1018
120GHz
n 5x1016
16
Low Noise
n 1x10
14
Difficult to make
n 5x10
Complex
n 1x1018
Semi insulating
Guide Layers
Key:
Features
High efficiency
High speed
Difficult to couple into
VCSEL status and trends
• VCSELs dominate where DFB laser or
high power is not needed
• Many suppliers at the 1 Gbps level
• Reliability established at 850 nm
• 1300 nm devices have been slow to reach
commercialization
• Low cost visible VCSELs becoming
available at 635 and 650 nm
• TBD 3 Gbps and 10Gbps
Semiconductor Trends
•CMOS and SiGe-BiCMOS has taken over the chip
market up to 10Gb/s rates
•40 Gb/s OC-768 is waiting, with components ready but
deployments few
•As long haul market has softened component
manufacturers have targeted current new developments
at gigabit and 10G Ethernet applications
•Addition of Forward error correction (FEC) drives
maximum bit rate up eg. SONET 9.952Gb/s to 12.5 Gb/s
with a 5-6.5 coding gain
•Chips appearing designed for RZ rather than NRZ
applications
Semiconductor Trends
•Transceivers are including more monitoring and
feedback control elements in chips to reduce part counts
and size
•With CMOS implementations practical at 10Gb/s more
multirate-multiprotocol solutions appearing thus
increasing volumes and lowering product costs
•More network protocol processing in highly integrated
chips
•Transmitters with low speed supervisory tone
modulation inputs
•Equalization and compensation of analog links
•Smaller packaging
•TBD rad-hard electronics
Transceivers conform to standards
•
•
•
•
Under the auspices of IEEE, ANSI, ITU…
Highly technical, dry, and can be political.
Strict rules of operation, balloting to approve.
Communication protocol and Physical Layer
– SONET, Fibre Channel, ATM, Gigabit Ethernet…
• Optical connector intermatability standard
– Duplex-SC, duplex-LC, MT-RJ, SG…
• Environmental
– Telecordia, Product Safety, Military, EMC…
Example OC-192 10 Gbps
VSR Standards (OIF)
Designation
Max
Reach
Fiber
Type
Number
of fibers
(HalfDuplex)
Data
Rate
Laser
Type
Wavelen
gth
VSR-1
300 m
MMF
12
1.25
gbps
VCSEL
array
850 nm
VSR-2
600 m
SMF
1
10 Gbps FP
1310 nm
VSR-3
300 m
MMF
4
2.5
Gbps
VCSEL
array
850 nm
VSR-4
300 m
MMF
1
10 Gbps VCSEL
850 nm
Transceivers conform to MSAs
• Standards bodies only define the minimum
necessary requirements for interoperability.
• Multi-source Agreements (MSAs) between
manufacturers describe common features
outside of the standard, e.g. module pinout
• Generates consensus and critical mass
without violating anti-trust guidelines.
• Electrical connector/formfactor standards
– 1x9, GBIC, GLM, 2x5, 2x10, SFP…
– 200pin, 300pin, XFP, Xenpak…
Example MSA Form Factors
10 Gigabit Small Form Factor Pluggable MSA
XFP Applications:
•OC192/STM-64 9.95 Gb/s
•10 Gigabit FC 10.5 Gb/s
•G.709 10.7 Gb/s
•10 Gigabit Ethernet 10.3 Gb/s
•Smaller space and lower cost alternative to parallel-optics VSR.
XFP Value Propositions
•Protocol Agnostic - "any application, any rate".
•Allows 16 XCVRs on a typical 19" rack with 23mm pitch density.
•Single footprint for all links.
•Less than 1/3 the power and size of an MSA with parallel interface.
•Hot plugable.
XFI (10 Gigabit Serial Electrical Interface) Electrical Signaling
•Supports 12" of FR4 with one connector
•Low EMI and power due to nominal 500 mV differential drive.
•Slew control for improved Signal Integirty and lower EMI.
•TX and RX signals each are a 100 Ohm differential pair, AC coupled for simplicity.
Xenpak
Xpak
http://www.xenpak.org/
http://www.xfpmsa.org/
http://www.x2msa.org/
X2
SFF Transceiver Showing
Duplex-LC Receptacle
Image courtesy Picolight, Inc.
XAUI Electrical Interface
•
•
•
•
•
•
•
•
Defined in IEEE 802.3ae, section 42
Extends the Media Independent Interface
Fewer pins than full parallel I/O
Quartet of differential pair per direction
3.125 Gbps per lane
XAUI chips resets jitter accumulation
XAUI chips establish lane order
XAUI chips eliminates lane-to-lane skew
Transceiver Trends
• More intelligence and RAM in modules
• Price erosion of 10 Gbps modules
• Garden variety 1 Gbps modules at nearcommodity pricing
• Equalization of optical dispersion
• Vcc of 3.3V (and lower) rather than 5V
• Better EMI and ESD margins
• Gbps modules for polymer optical fiber
• Special BiDi modules for FTTx
What to expect in 2010
for VSR Data Links
•
•
•
•
•
•
•
•
Fiber used in shorter and shorter links
1300 nm VCSELs
Silicon Photonics
Resonant microcavity devices
Few new connectors
Electronic dispersion compensation
More intelligence in transceivers
Inexpensive mini fusion splicers
Questions
• Is one gigabit/second technology
adequate for the lifetime of the detector?
• Radiation hardness of VCSELs and
commercially-available transceivers
• Can we use 1310/850 BiDi module to aid
with photobleaching?
• Is it economically feasible for spliceless
design made completely of pure silica fiber
• Suitability of photonic crystal fiber