Modulation Formats Which Approach the Shannon …...Sakamoto, Sano, George Mdlti f tModulation...
Transcript of Modulation Formats Which Approach the Shannon …...Sakamoto, Sano, George Mdlti f tModulation...
Modulation Formats Which Approach the Shannon LimitShannon Limit
Andrew D. Ellis
Photonic Systems Group, Tyndall National Institute / Dept. of PhysicsUniversity College Cork – Irelandy g
www.tyndall.ie
OMM4.pdf © 2009 OSA/OFC/NFOEC 2009
978-1-55752-865-0/09/$25.00 ©2009 IEEE
2 Acknowledgements
• For help with putting the talk together:p p g g– Fatima C. Garcia Gunning, Selwan K. Ibrahim, Jian
Zhao, Emanuel Popovici, Christian Spagnol.• For material and support:
– Sebastien Bigo, Masataka Nakazawa, Itsuro Morita, Akihide Sano Selwan Ibrahim Peter Winzer JianjunAkihide Sano, Selwan Ibrahim, Peter Winzer, Jianjun Yu, Joseph Kahn, Etsushi Yamazaki, Akimasa Kaneko.
• For financial backing
www.tyndall.ie
3 Symbols and Acronyms
EDFA: Erbium doped fibre amplifierWDM: Wavelength division multiplexingFEC: Forward error correctionC+L: Particular amplification wavelengths
C: Channel capacityB: Channel bandwidthN0: Noise spectral density
PDH: Plesiochronous Digital HierarchySDH: Synchronous Digital HierarchyIP: Internet ProtocolQAM: Quadrature Amplitude ModulationSSB: Single side band
Pb: Bit energysnr: Signal to noise ratioQ(x): Q function for a Gaussian random
variableSSB: Single side bandASK: Amplitude shift KeyingPSK: Phase shift keyingm: number of constellation pointsBER: Bit error ratio
Es: Energy value to quantify size of a constellation
Eb: Mean energy of a symbolei: Complex field value
LDPC: Low Density Parity CheckOSNR: Optical signal to noise ratioASE: Amplified spontaneous emissionDSP: Digital signal processingDFB: Distributed feedback laser
�: Wavelengthc: Speed of lightNamps: Number of amplifiersG: Amplifier gainDFB: Distributed feedback laser
MZM: Mach Zehnder modulatorRZ: Return to ZeroDPSK: Differential PSKCS: Carrier suppressedP l P l i ti
p gnsp: Spontaneous emission factorh: Plank’s constant�: FrequencyD: Dispersion
www.tyndall.ie
Pol: PolaraisationOFDM: Orthogonal frequency domain multiplexingCoWDM: Coherent WDMASIC: Application specific integrated circuit
D: Dispersion�: Nonlinear coefficientLeff: Effective length
4 A small selection of related sessions
Short Courses Workshops Tutorials Sessions/Invited Speakers
Communication Digital Coherent FranceschiniCommunication Theory
Digital Coherent Receivers.
Franceschini
FEC Kschischang
Non linear effects Modeling & Design Electronic Signal Capacity Limits of Nakazawa BigoNon-linear effects Modeling & Design Electronic Signal Processing and the Design of Optical Transport Systems
Capacity Limits of Fibers.Impact and Mitigation of Non-Linear Effects.
Nakazawa, Bigo, Fuerst, Shieh, Sun, Sakamoto, Sano, George
M d l ti f t WDM i L H l Si l C i VModulation formats WDM in Long-Haul Modulation Format
Single-Carrier Versus Multiple-Carrier Modulation Formats for WDM Systems
New Technologies Photonic ICs 100Gbit/s for $100 Photonic Crystal Fib
Prevost, Brès, Krauss, K k Kl kiElectronic Circuits Fibers. Kaneko, Klamkin, Morse, Hauske, Yoo, Clarke,
Reinforce Consolidate Build on Highlight
www.tyndall.ie
5 Introduction
• Trends in optical communicationTrends in optical communication• Shannon limit
Generic communication theory– Generic communication theory– Forward Error Correction CodesO ti l S t• Optical Systems– Capacity limit for non-linear systems– Multi-level modulation formats– Breaking the current limit
www.tyndall.ie
6 Maximum capacity (research)
1,000,000
m)
10,000
100,000
ct (T
bit/s
.km
C+LRaman
100
1,000
ce P
rodu
c
WDM
Dispersion ManagementFEC
10
100
Rat
e D
ista
n
OTDM-DDWDM-DDSoliton-DD
EDFA
Coherent Detection
WDM
0
1
83 85 87 89 91 93 95 97 99 01 03 05 07 09 11 13 15 17 19 21 23
Bit
R TDM-DD1550nm
www.tyndall.ie
83 85 87 89 91 93 95 97 99 01 03 05 07 09 11 13 15 17 19 21 23Date
User Demand
100G
1G
10G
1M
10M
100M
Rat
e (b
it/s)
10k
100k
1M
Acc
ess
R
100
1k
www.tyndall.ie
1975 1980 1985 1990 1995 2000 2005 2010 2015
Year
8 Network Architecture
100G 105
PDH SDH/SONET IP
1G
10G
104
s R
ate
1M
10M
100M
Rat
e (b
it/s)
103
re to
Acc
ess
10k
100k
1M
Acc
ess
R
102
Rat
io o
f Cor
100
1k
101
R
www.tyndall.ie
1975 1980 1985 1990 1995 2000 2005 2010 2015
Year
9 Switching Capacities
100G 105
10Tbit/s
1G
10G
1041Tbit/s
40Gbit/s
100Gbit/s
s R
ate
ity
1M
10M
100M
Rat
e (b
it/s)
103
re to
Acc
ess
ed C
apac
10k
100k
1M
Acc
ess
R
102622Mbit/s
Rat
io o
f Cor
155Mbit/s Sw
itch
100
1k
101
2.5Gbit/s10Gbit/s
R
www.tyndall.ie
1975 1980 1985 1990 1995 2000 2005 2010 2015
Year
10 Introduction
• Trends in optical communicationTrends in optical communication• Shannon limit
Generic communication theory– Generic communication theory– Forward Error Correction CodesO ti l S t• Optical Systems– Capacity limit for non-linear systems– Multi-level modulation formats– Breaking the current limit
www.tyndall.ie
11 Shannon Limit
• Shannon’s Second Theorem – The Noisy Channel yCoding Theorem
– C.E.Shannon, “A Mathematical Theory of Communication”, Bell Syst. Tech. J., 27, pp379-423 & pp623-656
– “Reliable communication over a discrete memory-less channel is possible if the communication rate R satisfies R < C, where C is the channel capacity. At rates higher than the capacity, reliable communication is impossible.”communication is impossible.
���
����
��
BNPBCo
1log. 2
– At infinite bandwidth
�� o
P
www.tyndall.ie
eNPCo
2log
12 Information Spectral Density
Information Spectral Density of ideal memory-less AWGN channel
n C
/B 20.0
olar
isat
ion
2 0
5.0
10.0 Not Possible
acity
per
po
1.0
0.5
2.0
AllowedHow do we get to this line?
Cap
a
0 10 20 30
0.2
www.tyndall.ie
SNR (dB)
13 Calculation of required SNR
• Linear (coherent) detectionLinear (coherent) detection • Hard decision • Assume memory-less AWGN channel• Assume memory less AWGN channel• Signal-independent noise
– Thermal noise, local oscillator shot noise limited systeme al o se, local osc llato s ot o se l ted syste
• Calculate probability that each bit “escapes”
www.tyndall.ie
14Calculation of required SNR
1: Bit Error RateQuadrature
Decision Threshold
In-phaseab
ility
Signal amplitude
Pro
ba
eP0
12
��
�
�
��
�
��
NeeQ th �
01
2��
�
�
��
�
��
NeeQ th
Signal amplitude
eP0
21 22 �
�
�
�
��
�
� �N
eeQ �
0
21 22 �
�
�
�
��
�
� �N
eeQ
���
��� � 21 ee
Q
www.tyndall.ie
2 22��
2��
�� 02N
Q
15Calculation of required SNR
1: Signal to Noise RatioQuadrature
���
���
2 E
QP s
sE�In-phase
sE sE2
���
���
02N
QPe
NEsnr b
sE2
N00N
PSK or bi-polar ASK
sss
b EEEE �22
AM or uni-polar ASK
ss
b EEE 224
20
�22 22
� snrQPe � 2 � snrQPe
www.tyndall.ie
16 Information Spectral Density
Information Spectral Density of ideal memory-less AWGN channel
10 0
20.0
n C
/B
2.0
5.0
10.0
SSB-B-AM (10-12)
pola
risat
io
1.0
0.5
SSB BPSK (10 12)
1 bit per symbol
acity
per
p
0 10 20 30
0.2SSB-BPSK (10-12)
Cap
a
www.tyndall.ie
SNR (dB)
17 Extending to higher order modulation
sEsE�sE3� sE3
obab
ility
Signal amplitude
Pro
eP 0
21
2 ��
�
�
��
�
� �
Nee
Q� 0
21
2 ��
�
�
��
�
� �
Nee
Q2 � 0
21
2 ��
�
�
��
�
� �
Nee
Q20
21
2 ��
�
�
��
�
� �
Nee
Q� ��
�
�
��
�
� �
0
21
223
Nee
Q��
�
�
��
�
� �
0
21
2Nee
Q
www.tyndall.ie
2 24 4 24 2��
4
18 Extending to higher order modulation
sEsE�sE3� sE3
obab
ility
EEEEEE 59119
Pro
sssssb EEEEEE 54444
���
eP 0
21
2 ��
�
�
��
�
� �
Nee
Q� 0
21
2 ��
�
�
��
�
� �
Nee
Q2 � 0
21
2 ��
�
�
��
�
� �
Nee
Q20
21
2 ��
�
�
��
�
� �
Nee
Q� ��
�
�
��
�
� �
0
21
223
Nee
Q
www.tyndall.ie
2 24 4 24 2��
4
19 Extending to higher order modulation
www.tyndall.ie
20 Calculating modulation scheme performance
1. Calculate minimum distance (e1-e2, 2�Es)( 1 2, s)2. Determine symbol noise level (No)3. Calculate BER using symbol probability and effective
number of boundaries4. Calculate energy per bit Eb as a function of Es
5 Substitute E for E and snr for E /N to give BER as a5. Substitute Eb for Es and snr for Eb/N0 to give BER as a function of snr
6. Calculate ISD from number of levels and symbol yprobability
www.tyndall.ie
21 Signal Constellations
ASK QAM
PSK
www.tyndall.ie
PSK
22Information Spectral Density of common
multi-level modulation formats
SSB-QAMSSB PSKn
C/B
SSB-ASKSSB-Bi-Polar ASKSSB-PSK
pola
risat
io
Increasing number of levels
acity
per
p
SNR (dB)
BER = 10-12
Cap
a
www.tyndall.ie
SNR (dB)
23Impact of Optical Demultiplexing and
Practical Optical Modulation
20% Guard Band for FilteringDouble Side Band Modulation for ASK and PSK
Orthogonal QAMn C
/B
Orthogonal QAM
ASKBi-Polar ASK
QAMPSK
olar
isat
ion
ASK
acity
per
po
BER = 10-12
20% filter margin
Cap
a
www.tyndall.ie
SNR (dB)
24 Fundamental limits1
• Unconstrained modulation (amplitude + phase) with direct detection
C � snrBC
�� 1log2• Phase modulation with coherent detection
� 11log1�� snrC
• Unconstrained modulation (amplitude + phase) with coherent detection
� 1.1log2 2 �� snr
B( p p )
� 0.1log21
2 �� snrBC
www.tyndall.ie
� 2 2B
1: J.M.Kahn, K-P.Ho, JQE 10 2 pp259 (2004)
25 Fundamental Limits2
� ���
����22
1 xErfcxQDefinition
���
����
�
��
snrm
mQmmPe 1
)(log.612 22Bi-polar ASK
�� mm 1
� � �
���
�����
m
ii
mpmpsnrmQmP
2/
1
2
2
1.2)()(log612Uni polar ASK �
� �
���
���� �
m
i
rre
impmpsnr
mQ
mP
1
22
12,)(.
12
�
Uni-polar ASK
� � � m
msnrQsnrQ
snrQ
Pe 42
,
2
22112
2
�� �����
��� ��
�PSK (low BER)
www.tyndall.ie1: J.G.Proakis, Digital Communications (4th Edition) McGraw-Hill (2000)
� � otherwisem
SinmsnrQm
log.2log2
22
���
����
����
��� �
26 Bibliography – 1:Communication Theory
1. H.Nyquist, AIEE Trans. 47pp617 (1928)2. C.E.Shannon, Bell Syst. Tech. J., 27 pp379- (1948)3. J.J.Bussgang et.al., Trans. Comm. Systems 12 pp18- (1964)4. J.G.Proakis, Trans. Comm. Tech., 16 pp71- (1968)5 H A Haus JQE QE 23 pp212 (1987)5. H.A.Haus, JQE, QE-23 pp212 (1987)6. S.Haykin, “Digital Communications”, Wiley (1988)7. J.M.Geist, Proc MILCOM 1990, pp768- (1990)8. J.P.Adis et.al., Trans.Inform.Theory, 39 pp184- (1993), y, pp ( )9. J.G.Proakis, Digital Communications (4th Edition) McGraw-Hill (2000) 10. J.Rebola et.al., Proc SPIE 4087 pp49- (2000) 11. Mecozzi et.al., PTL, 13 pp1029- (2001)12. J.M.Kahn, K-P.Ho, JQE 10 2 pp259- (2004)13. K.P.Ho, PTL, 17 pp858- (2005)14. E.Ip et.al., JLT, 23 12 pp4110- (2005)15 E Ip et al Optics Express 16 2 pp753 (2008)
www.tyndall.ie
15. E.Ip et.al., Optics Express, 16 2 pp753- (2008)
27 Introduction
• Trends in optical communicationTrends in optical communication• Shannon limit
Generic communication theory– Generic communication theory– Forward Error Correction CodesO ti l S t• Optical Systems– Capacity limit for non-linear systems– Multi-level modulation formats– Breaking the current limit
www.tyndall.ie
28Error Correction Codes
Current Status
Binary FECencoder
MappingUp-conversion(modulation)
10110 1011010 a+jb
encoder (modulation)
Down-conversion(receiver)
DemappingBinary FEC
decoder
a’+jb’101001010110
www.tyndall.ie
LO
29 Forward Error Correction
Orthogonal QAM
C/B
g
1st Generation: Reed Solomon
Overhead
3rd Generation: Block Turbo Codes2nd Generation: Concatenated Codes
BER = 10-12
20% filter marginCoding Gain
www.tyndall.ie
SNR (dB)
30Error Correction CodesSoft decision decoding
Binary FECencoder
MappingUp-conversion(modulation)
10110 101101 a+jb
encoder (modulation)
• Eg Low Density Parity Check (LDPC) Codes– Within 0.005 of Shannon limit (100% overhead)
Down-conversion(receiver)
DemappingSoft DecisionBinary FEC
decoder
a’+jb’70531710110
www.tyndall.ie
LO
31 Performance of RS & LDPC codes1
100
10-1
10-6
10-2
10-3
rror
Rat
e
Reed Solomon
C
10-4
10 5
Er
LDPC
3 4 5 6 7 8 9 10
10-5
10-7
www.tyndall.ie1: B.Zhou, et.al., Information theory and Applications Workshop, 2008
Eb/N0
32Error Correction Codes
Current Status
Binary FECencoder
MappingUp-conversion(modulation)
10110 101101 a+jb
encoder (modulation)
• Multi-level (c.f. binary) coded LDPC• Additional refinements for soft decision based systems
Down-conversion(receiver)
DemappingLDPC overGalois Field
a’+jb’10110
www.tyndall.ie
LO
33 Short LDPC codes over GF(q) 1
• LDPC over GF(q) easily outperform same length binary codes (length n=720)
• Moreover they perform as well as much longer gbinary codes (n=5760)
www.tyndall.ie1: C.Spagnol Ph.D. Thesis: Aspects of LDPC codes for hardware implementation
34 Bibliography –FEC
1. G.D.Forney et.al., J.Select.Areas Commun., 7pp941- (1989)2. S-Y Chung, IEEE Commun Lett., 5 pp58- (2001)3. P.V.Kumar et.al., “Optical Fiber Telecommunications IV B” Academic (2002)4. T.Mizuochi, OFC2003, paper PD21 (2003)5 B Vasic et al JLT 21 pp438 (2003)5. B.Vasic et.al., JLT, 21 pp438- (2003)6. C.Berrou, IEEE Commun Mag., 41 pp110- (2003)7. C.Spagnol Ph.D. Thesis: Aspects of LDPC codes for hardware implementation8. B.Zhou, et.al., Information theory and Applications Workshop, 2008, , y pp p,
www.tyndall.ie
35 Introduction
• Trends in optical communicationTrends in optical communication• Shannon limit
Generic communication theory– Generic communication theory– Forward Error Correction CodesO ti l S t• Optical Systems– Capacity limit for non-linear systems– Multi-level modulation formats– Breaking the current limit
www.tyndall.ie
36 Signal to Noise Ratio in Optical Systems1
• Capacity limits expressed in terms of SNR = Eb/N0p y p b 0
• Convert to optical signal to noise ratio OSNR =Ps/Pn
RE 2�c
RNEOSNR b
..2.
0 ���
nmnmsGbitdBdBdB OSNRSNR 1.0,1550,/6.104�
• Noise accumulation well known
� �� �� 1 hnGNP l i iN � �� �....1. hnGNP spampsonpolarisatiN
Add input insertion loss to gain
www.tyndall.ie1: J.M.Kahn, K-P Ho, JQE 10 2, pp259 (2004)
Add input insertion loss to gain
37 Capacity limit for a 2,000km optical system
8Hz/
pol)
~ Damage threshold O ti t ll tiHz/
pol)
6
8
y Li
mit
(b/s
/H ~ Damage threshold
Limit ~ 6 b/s/Hz
Optimum constellationPSK signalDirect Detection
8
10
12
14
y Li
mit
(b/s
/H
2
4
ctra
l Effi
cien
cy
2
4
6
ctra
l Effi
cien
cy2000km, 80km spacing, 4.5dB noise figure, 20% guard bands,
50GHz channel spacing
1.00.5 2.00.2 5.00.1 10.0
Transmitted Power Density (W/THz)Spe
c
1´ 104100 200 500 1000 2000 5000
System Reach (km)S
pec
80km spacing, 4.5dB noise figure, 1dB input loss, 500mW over C band, 50GH h l i50GHz channel spacing 50GHz channel spacing
But optical communication links are characterised by a distributed nonlinear response
www.tyndall.ie
38Origin of fundamental capacity limit for
optical fibres1,2,3
T 1
Rx-2
R 3
DSP
Tx-1 Rx-3
Rx-1Tx-2Tx-3
Rx-1x N DSP
• Per channel DSP– Dispersion (Chromatic + Polarisation)– Coherent Detection– Self Phase Modulation Correction3
• But not non-linear mixing between signal & ASE (early dispersion maps)But not non linear mixing between signal & ASE (early dispersion maps)• Multiple transmitters
– Don’t actually communicate– Are not always co-located
Inter channel nonlinearity gives the fundamental limit
www.tyndall.ie1: P.P.Mitra, J.B.Stark, Nature, 411, pp1027, (2001)2: L.G.L.Wegener et.al., Physica D, 189, pp81, (2004)3: R-J, Essiambre et.al., ECOC 2008, We1E1(2008)
• Inter channel nonlinearity gives the fundamental limit
39Example of simple receiver based
nonlinearity compensation.
CoherentCoherentReceiver
www.tyndall.ieK.Kikuchi, Optics Express, 16 2 pp889- (2008)
40 Cross phase modulation
• Signal processing to mitigate SPM• Dispersion management to reduce signal –ASE interaction
All other channels act as noise so rces to a elength of interest• All other channels act as noise sources to wavelength of interest• Appropriate Dispersion Map
– High local dispersion to degrade phase matching– Dispersion management to control PAPDispersion management to control PAP– Sufficient residual dispersion per span– Dispersion map is key
Quadrature
In-phase
www.tyndall.ie
41 Impact of cross phase modulation
• Random walk process– Sub-linear accumulation with number of amplifiersp
• Depends on dispersion map• Proportional to fibre non-linearity and effective length
effa
XPM
LNDBI
���
����
2ln.2
..2�
�
• Adds noise to the signal
IP
PePPPP XPM
S
���
�������
�
1 sNXPMNN PePPPP ���
������� 1
• Leaches power from the signalP
www.tyndall.ie1: P.P.Mitra, J.B.Stark, Nature, 411, pp1027, (2001)
XPM
SIP
sS ePP�
�
42 Four Wave Mixing1
• Similar treatment to XPM• Is dominant for PSK with certain dispersion maps (constant intensity)
DB �� ���
1cnqp
effa
XPM
LNDBI
���
����
2ln.2
..2�
����
2
0,
1qp
qpA
FWM
NI � 2222
22
.2 cpqfD
Dpq
�� ��!
�
qpqp
Dpq
,2,1
• XPM typically dominates and is considered hereafter• XPM typically dominates and is considered hereafter.
www.tyndall.ie1: J.M.Kahn, K-P Ho, JQE 10 2, pp259 (2004)
43 Capacity limit with non-linear transmission
8
z/po
l)
G l
Linear
6
mit
(b/s
/Hz
Coherent Detection
Goal
Non-Linear
4
icie
ncy
Lim
Linear
2
pect
ral E
ffi
Direct Detection
Non-Linear
1.00.5 2.00.2 5.00.1 10.0
Transmitted Power Density (W/THz)
Sp Direct Detection
www.tyndall.ie
Transmitted Power Density (W/THz)
2000km, 80km spacing, 4.5dB noise figure, 50 GHz channels at 50 baud
44 Numerical simulations
l) 8y
Lim
it (b
/s/H
z/po
4
6
8Linear
Spe
ctra
l Effi
cien
c
2
4Non-Linear
www.tyndall.ie
Transmitted Power Density (W/THz)0.01 0.1 1 10 100
R-J Essiambre et.al., OFC2008 paper OTuE1 (2008)
45 Bibliography –Nonlinearity and its mitigation
1. D.M.Pepper et.al., Opt.Lett., 5 pp59- (1980) 2 J P G d t l O t L tt 15 1351 (1990)2. J.P.Gordon et.al., Opt.Lett., 15, pp1351 (1990)3. S.Watanabe et.al., JLT, 14 3 pp243- (1996)4. A.Mecozzi et.al., PTL, 12, pp392, (2000)5. X.Liu et.al., Opt. Lett., 27, pp1616 (2002)6. K-P.Ho et.al., JLT, 22 pp779 (2004)7. D-S Ly-Gagnon, Proc OECC/COIN2004, paper14C3-3 (2004)8. K-P.Ho, “Phase modulated optical communication systems”, Springer, (2005)9 G Charlet et al ECOC 2006 paper Th4 3 4 (2006)9. G.Charlet et.al., ECOC 2006, paper Th4.3.4 (2006)10. G.Zhu et.al., PTL, 18 pp1007 (2006)11. D.Boivin et.al., J.Opt.Soc.Am.A B 23 pp2019 (2006)12. K.Kikuchi et.al., OFC 2007, paper OTuA2 (2007)13 A J L PTL 19 1556 (2007)13. A.J.Lowery, PTL, 19 pp1556 (2007)14. A.J.Lowery, Optics Express, 15 pp12965 (2007)15. E.Ip et.al., Optics Express, 16 2 pp753- (2008)16. K.Kikuchi, Optics Express, 16 2 pp889- (2008)
www.tyndall.ie
p p pp ( )
46 Bibliography –Non-linear Limit
1. P.P.Mitra &J.B.Stark, Nature, 411 pp1027 (2001)2. J.B.Stark et.al., Opt.Fiber Technol., 7 pp275- (2001)3. J.Tang, JLT, 19 pp1104- (2001)4. E.E.Narimanov et.al., JLT 20 pp530- (2002)5 K S Turitsyn Phys Rev Lett 91 20 pp203 (2003)5. K.S.Turitsyn, Phys.Rev.Lett., 91 20 pp203- (2003)6. L.G.L.Wegener et.al., Physics D, 189 pp81- (2004)7. J.M.Kahn et.al., J.Sel.Top.Quantum Electron, 10 2 pp259- (2004)8. J.M.Kahn et.al., ECOC 2005 paper Th2.2.1 (2005), p p ( )9. E.B.Desurvire, JLT, 24 12 pp4697- (2006)10. R-J Essiambre et.al., OFC2008 paper OTuE1 (2008)11. R-J, Essiambre et.al., ECOC 2008, We1E1(2008)
www.tyndall.ie
47 Introduction
• Trends in optical communicationTrends in optical communication• Shannon limit
Generic communication theory– Generic communication theory– Forward Error Correction CodesO ti l S t• Optical Systems– Capacity limit for non-linear systems– Multi-level modulation formats– Breaking the current limit
www.tyndall.ie
48 Progress towards the capacity limit
Coherent Detection Direct Detection
ol)
ol)
Linear
ficie
ncy
(b/s
/Hz/
po
ficie
ncy
(b/s
/Hz/
po
span
Non-linear
Linear
Non-linear
Spe
ctra
l Eff
Spe
ctra
l Eff span
Transmission distance (Mm) Transmission distance (Mm)
www.tyndall.ie
Solid lines calculated for; 80km spacing, 4.5dB noise figure, 50 GHz channels at 50 Gbaud, Optimum powerDots plotted for notable experimental results, per polarisation
49Direct Detection 1: Filtered CS-RZ DPSK2
8,000km 0.8 b/s/Hz
• Long Haul Transmission (mitigate non-linearity)
Im
– RZ format – dispersion management
• For spectral efficiencyDPSK for sensitivity benefit
Re
– DPSK for sensitivity benefit– Pre-filtering for SSB– FEC
DFB MZM 50/100 GHz ol)
PSKSSB-PSK
DFB MZM 50/100 GHzInterleaver
ncy
(b/s
/Hz/
p
ASK
ectra
l Effi
cien
www.tyndall.ie1: I.Morita, N.Edagawa, ECOC 2003, PDP page 60 (2003)
Spe
SNR (dB)
50Direct Detection 2: Coherent WDM1
1,200km 1 b/s/Hz
• High Capacity Transmission – Low symbol rate (impairments)
Im
• For spectral efficiency– Orthogonal Carriers– FEC
�mod
Re
Laser MZM MZM
~
������
modmodmodmodmod
mod
ol)
CoWDM
�mod
ncy
(b/s
/Hz/
p
ASK
ectra
l Effi
cien
www.tyndall.ie1: T. Healy, et.al., ECOC’07, Mo1.3.5 (2007).
Spe
SNR (dB)
51Direct Detection 3: PM-QPSK1
240km 1.6 b/s/Hz/pol
• For spectral efficiency– RZ DPSK for sensitivity benefit
Pre filtering to approach SSB
Im
– Pre-filtering to approach SSB– DQPSK for enhanced efficiency– Optical equaliser– FEC
Re
C
DFB MZM 50/100 GHzInterleaver
ol)
PSKSSB-PSK
MZM �/2
ncy
(b/s
/Hz/
p
ASK
ectra
l Effi
cien
www.tyndall.ie1: A.H.Gnauck, OFC 2007, PDP19 (2007)
Spe
SNR (dB)
52 Direct Detection Modulation formats
• When amplifier spacing taken into accountAll th lt 2 f th li it
Direct Detection– All three results ~ x2 from the limit– Independent of
• Bi-polar / Uni-polar modulation• Single or dual quadrature ol
) 80km100km
Amplifier Spacing
• Polarisation multiplexing• Key features
– FEC– Techniques approaching SSB fic
ienc
y (b
/s/H
z/po 100km
40kmLinear
N liTechniques approaching SSB • 80% there using RZ pre-filtering• 100% there using Coherent WDM
• Promising approachFEC
Spe
ctra
l Eff Non-linear
– FEC– Polarisation multiplexed (x2)– Dual Quadrature (x2)– Coherent WDM (x2)
Transmission distance (Mm)
www.tyndall.ie
( )• Little scope beyond ~2 b/s/Hz/pol
53 Coherent Detection
8
Hz)
Linear
6
Lim
it (b
/s/H
Coherent DetectionNon-Linear
4
Effi
cien
cy L
Significant room for performance enhancement
Linear
2
Spe
ctra
l E
Direct DetectionNon-Linear
1.00.5 2.00.2 5.00.1 10.0
Transmitted Power Density (W/THz)
S
www.tyndall.ie
Transmitted Power Density (W/THz)
2000km, 80km spacing, 4.5dB noise figure, 50 GHz channels at 50 Gbaud
54Coherent Detection 1: PSK
1,200km 3 b/s/Hz
Im• Long Distance High Data Rate– Constant Intensity Format
Re– Increased constellation size– Narrow Line-width Laser– FECFEC
ECL MZMMZM �/2
PMAxes to add and legend (all slides)
pol)
ency
(b/s
/Hz/
ppe
ctra
l Effi
cie
www.tyndall.ieM.Seimetz et.al.,, ECOC 2007 Tu1E1
Sp
SNR (dB)
55Coherent Detection 2: OFDM
4,160km 5.6 b/s/Hz
Orthogonal Frequency Domain Multiplexing• Data spread amongst many orthogonal carriers• Orthogonal carriers give SSB performance (less
overhead)• Modulate each subcarrier with QAM constellation• Increase constellation size electronically• FEC
Double side band
pol)
ncy
(b/s
/Hz/
pec
tral E
ffici
en
www.tyndall.ie
Spe
SNR (dB)
H. Takhashi et.al., , ECOC 2008 PDP Th3E4
56Coherent Detection 3: QAM
150km 10b/s/Hz
• Ultra high Spectral Density– Maximum QAM Constellation
• With Niquist filtersUl li id h l– Ultra narrow line-width laser
– FEC
ol)
cy (b
/s/H
z/po
ectra
l Effi
cien
www.tyndall.ie
Spe
SNR (dB)M. Yoshida, Optics Express, 16 2 pp829 (2008)
57 Coherent Detection Modulation Formats
• Performance2 f– > x2 performance gap
– Relative performance degrades with reach
• Key Features
/Hz/
pol) Linear
– All polarisation multiplexed• Often comes for “free”
– Multi-level modulation essential– PSK performs well ra
l Effi
cien
cy (
b/s Non-linear
S pe o s well• Is performance limited by
nonlinear effects?– Orthogonal modulation and/or
spectrum control usedS
pect
rp
• Optimum Solution– Still to be established– Perhaps
256 QAM over OFDM over CoWDM
Transmission distance (Mm)
www.tyndall.ie
• 256 QAM over OFDM over CoWDM
58 Latest Capacity Distance Results
• Coherent detection already outstripping direct detection• Record result is coherently detected OFDM
1,000,000
)1,000,000
)
• Post Deadline Papers?– > 0.1 Pbit/s.Mm
10,000
100,000uc
t (Tb
it/s.
km)
10,000
100,000uc
t (Tb
it/s.
km)
10
100
1,000
Dis
tanc
e Pr
od
OTDM-DDWDM DD10
100
1,000
Dis
tanc
e Pr
od
OTDM-DDWDM-DDS lit DD
0
1
10
Bit
Rat
e D WDM-DD
Soliton-DDTDM-DD
0
1
10
Bit
Rat
e D Soliton-DD
TDM-DDOFDM/CoWDMCoherent
www.tyndall.ie
83 85 87 89 91 93 95 97 99 01 03 05 07 09 11 13 15 17 19 21 23Date
83 85 87 89 91 93 95 97 99 01 03 05 07 09 11 13 15 17 19 21 23Date
59 Bibliography – Modulation Formats
1. R.C.Steele, Electron.Lett., 19 pp69- (1983)2. S.Norimatsu et.al., PTL, 4 pp765- (1992)3. S.Walklin et.al., JLT 17 11, pp.2235– (1999)4. B.Wedding et.al., . ECOC 1998, pp523– (1998)5 M I H t l PTL 13 8 881 (2001)
24. G.Charlet et.al., OFC 2008, paper PDP3 (2008)25. A.H.Gnauck et.al., JLT 26 1 (2008)26. N.Kikuchi et.al., JLT 26 1 (2008)27. Y.Mori et.al., ECOC 2008, Paper Tu.1.E.4 5. M.I.Hayee et.al., PTL, 13 8, pp881– (2001)
6. T.Nakamura et.al., OECC 2002, pp554– (2002)7. I.Morita, N.Edagawa, ECOC 2003, PDP page
60 (2003)8. X.Liu et.al., ECOC 2003, pp.1010– (2003)9 T Nakamura et al 22 pp733 (2004)
, , p(2008)
28. M.Nakamura et.al., ECOC 2008, Paper Tu.1.E5 (2008)
29. M.Nakazawa, ECOC 2008, Paper Tu.1.E.1 (2008) 9. T.Nakamura et.al., 22 pp733– (2004)
10. S.Tsukamoto et.al., OFC2005 Paper PDP29 (2005)
11. K.Sekine et.al., Electron. Lett., 41, pp430 (2005)
12. S.K.Ibrahim et.al., IEEE J. OF Select. Topics in
( )30. T.Sakamoto et.al., ECOC 2008, Paper
Tu.1.E.3 (2008)31. T.Tokle, OFC 2008, paper OMI1 (2008)32. P.J.Winzer et.al., ECOC2008, paper Th.3.E.5
(2008), pQuantum Electron., 12 4 (2006)
13. T.Pfau et.al., PTL, 18 pp1907- (2006)14. R.A.Griffin et.al., OFC2002, paper WX6 (2006)15. K.Kikuchi, OFC2006, paper OTuI4 (2006)16. M.Nakazawa et.al., Electron.Lett., 42 pp710
(2008)33. M.Yoshida et.al., ECOC 2008, Paper Mo.4.D5
(2008)34. M.Yoshida et.al., Optics Express, 16 2 pp829-
(2008)35 J Yu et al ECOC 2008 paper Th 3 E3 (2008)pp
(2006)17. Y.Han et.al., Electron.Lett., 42 2 (2006)18. J.Hongo et.al., PTL, 19 pp638- (2007)19. Ip et.al., JLT, 25, pp2675- (2007)20. J.Hongo et.al., OFC2007, Paper OMP3 (2007)
35. J.Yu et.al., ECOC 2008, paper Th.3.E3 (2008)36. X.Zhou et.al., OFC 2008, paper PDP1 (2008)
www.tyndall.ie
21. A.H.Gnauck, OFC 2007, PDP19 (2007)22. M.Serbay et.al., OFC 2007 paper OThL2
(2007)23. M.Seimetz et.al.,, ECOC 2007 Tu1E1
60 Bibliography – OFDM & CoWDM
1. V.Jungnickel et.al., Proc Vehicular Technology, p861-, (2005)2. A.D.Ellis et.al., PTL, 17, pp 504- (2005)3. W.Shieh et.al., OFC 2007, paper OMP2 (2007)4. N.Cvijetic et.al., OFC 2007, paper OTuA5 (2007)5 A Lowery et al OFC 2007 paper OTuA4 (2007)5. A.Lowery et.al., OFC 2007, paper OTuA4 (2007)6. T.Healy et.al., ECOC’07, paper Mo1.3.5 (2007)7. F.C.Garcia Gunning et.al., CLEO Europe 2007, paper CI8-5 (2007)8. W.Shieh et.al., Optics Express, 15 pp9936 (2007), p p , pp ( )9. E.Ip et.al., Optics Express, 16 2 pp753- (2008)10. S.L.Jansen et.al., OFC 2008, paper PDP2 (2008)11. H.Takhashi et.al., ECOC 2008 PDP Th3E4 (2008)12. Q.Yang et.al., OFC 2008, paper PDP7 (2008)13. E.Yamada et.al., OFC 2008, paper PDP8 (2008)
www.tyndall.ie
61 Introduction
• Trends in optical communicationTrends in optical communication• Shannon limit
Generic communication theory– Generic communication theory– Forward Error Correction CodesO ti l S t• Optical Systems– Capacity limit for non-linear systems– Multi-level modulation formats– Breaking the current limit
www.tyndall.ie
62How to increase the capacity further:
Optimisation of current technology
ain
(dB
)
/s/H
z/po
l)
~3dB / bit Com
p
16QAM
PSK, QAM
net c
odin
g g
Effic
ienc
y (b
/
128 QAM
8PSK
lexity of Implem
e
PSK
QPSK8PSK
16QAM
bit / b l/ l
Req
uire
d
Spe
ctra
l E
SNR (dB)
entation
PSK
PSK, QAM
bits/symbol/polSNR (dB)
• Multi-level modulation formats with coherent detection• Required FEC strength increases significantly with signal bits
– Maximise number of orthogonal channels first • Polarisation• Quadrature
S b i
www.tyndall.ie
• Subcarrier
– Expect significant (�100%) overhead for 128 QAM and above
63The Ultimate Quest
Maximum capacity for fixed installed plantz/
pol)
ding
Throughput of multilevel transmission at fixed SNRAssuming required overhead = 70 x achievable raw BER
@ SNR suitable for ASK transmission at 10-12
PSK QAM
cien
cy (b
/s/H
z
ymbo
l afte
r cod
PSK, QAM64QAM
16PSK
Spec
tral E
ffic
ASK, QAM
Net
bits
per
sy
• Fixed Plant – Terminal Changes Only
S
SNR (dB) Bits per symbol transmitted
– Fixed SNR– Modulation format changes– Coding (FEC) changes
www.tyndall.ie
64How to increase the capacity further:
Amplifier optimisation
• Amplifier bandwidth– Capacity scales ~linearly with bandwidth ��
�
����
��
BNPBC 1log. 2
• Amplifier Noise figure– Modest capacity increase ~ Log2 noise reduction
) pol)
��
�� BNo
cy (b
/s/H
z/po
l
ency
(b/s
/Hz/
p
ectr
al E
ffici
enc
Amplifier Bandwidth:96 THz (2 cycles per bit)4.8 THz (40 cycles per bit / C-band)1 THz Sp
ectr
al E
ffici
e
Amplifier Noise Figure:4.5dB Typical3dB Ideal0dB Phase sensitive*
Spe
Power spectral density (W/THz)
S
Power spectral density (W/THz)
www.tyndall.ie
2000km, 80km spacing, (4.5dB noise figure), 50 GHz channels at 50 Gbaud, (5THz amplifier bandwidth), 0.2dB/km loss
* Phase sensitive amplifier: 0dB noise figure, but quantum noise for attenuation ~ ½ of 3dB noise figure case1
1: L.Thylen et.al., Channel capacity of optical fibres, Private communication (2002)
65How to increase the capacity further:
Multi-tone transmission
• Non-linear index effects may be compensated at the terminals.
B k d ti– Backwards propagation– For OFDM1
– Multiple coherent channels2
• Increase bandwidth of channel• Increase bandwidth of channel
pol)
10TH
ency
(b/s
/Hz/
p5THz
10THz
Spec
tral
Effi
cie
1GHz
100GHz1THz
3THz
www.tyndall.ie1: A.J.Lowery, Optics Express 15 20 pp12965 (2007)2: E.Yamazaki, WS-1, ECOC 2008
S
Power spectral density (W/THz)2000km, 80km spacing, 4.5dB noise figure, 10THz amplifier
66How to increase the capacity further:Multi-wavelength optical regeneration
• Multi-wavelength all optical regeneration � Breaks up noise accumulation– Dispersion managed Mamyshev regenerator1p g y g– Quasi continuously filtered regenerator2
– Bi-directional / dual polarisation3
– Parametric effects4
s/H
z/po
l)
Number of
Effic
ienc
y (b
/s
Regenerators3216842
Spec
tral
T i i R h (k )
21
www.tyndall.ie
2000km, 80km spacing, 4.5dB noise figure, 5THz amplifier
Transmission Reach (km)1: M. Vasilyev et.a., OFC 2005, OME62.2: B.Cuenot et.al., Optics Express 15 18 pp11492 (2007)3: L.Provost et.al., ECOC 2007, Tu4.5.14: B.Cuenot, OAA 2006, OWA4
67How to increase the capacity further:
Fibre Design
SMF-28TM LongLineTM
Fiber1Hollow Core PCF3
(prediction)
Loss (dB/km) 0.2 0.185 .13
Non-linear coefficient (W-1km-1)
~ 1 ~ .75 ~0.01(99% air propagation)
Waveband (nm) 1550 1550 1900Waveband (nm) 1550 1550 1900
m) 10Gb/s
12
10/Hz/
pol)
Dis
tanc
e (M
m 40Gb/s
80 to 160 Gb/s
10
8
6
4
ISDxD contours
Effic
ienc
y (b
/s
Information Spectral Density (b/s/Hz)0 1 2 3 4
2
Spec
tral
E
Power spectral density (W/THz)
www.tyndall.ie1: M. Bigot-Astruc, Mo.4.B1, ECOC’082 :G.Charlet, Th3Ee, ECOC’083: P.J.Roberts et.al., Optics Express 13 1 pp236 (2005)
2000km, 80km spacing, 4.5dB noise figure, 50 GHz channels at 50 Gbaud
68 The Ultimate Capacity?
• Transmission link fabricated from hollow core PCF• Optical amplifiers replaced by ideal WDM regenerators• Ultimate limit ~ 17 b/s/Hz/pol
– ~170 Tbit/s in a 40nm bandwidthwith a higher bandwidth 1 Pbit/s Mm– with a higher bandwidth ~ 1 Pbit/s.Mm
s/H
z/po
l)
Ideal regen, ideal PCF
Effic
ienc
y (b
/s
Optical amplifiers, SMF
Spec
tral
E
Power spectral density (W/THz)
Optical amplifiers, SMF
www.tyndall.ie
2000km, 80km spacing, 4.5dB noise figure, 50 GHz channels at 50 Gbaud
69Conclusions:
Reaching the limits of communication capacity
• What do we need to do (system design)?– Dispersion management– FEC
O th l i
• What do we need to make (components)?– Integrated arrays– High speed DSP, ASICs
N li idth l– Orthogonal carriers– Coherent detection– QAM– Non-linear compensation
– Narrow linewidth lasers– Linear modulators and drivers
• And to extend the limit…– Amplifier for waveband extension 10,000
100,000
1,000,000
(Tbi
t/s.k
m)
64 QAM CO-OFDM ?
Amplifier for waveband extension– Improved transmission medium– Undiscovered modulation techniques– Multi-wavelength regeneration 10
100
1,000
,
ate
Dis
tanc
e Pr
oduc
t (
OTDM-DDWDM-DDSoliton-DDTDM-DD
www.tyndall.ie
0
1
83 85 87 89 91 93 95 97 99 01 03 05 07 09 11 13 15 17 19 21 23Date
Bit
Ra TDM DD
OFDM/CoWDMCoherent
70
• Thank You
Ph t i S t G• Photonic Systems Group, Tyndall National Institute & Department of Physics, University College Cork, Ireland.
physics.ucc.ie/photonics/photonicsJobs.htm
www.tyndall.ie