Moderator Pierre Mertz Infinera Master-class 4: Updates to ... · Tony started at BT's Research labs and then moved to Alcatel Australia, becoming involved in testing and commissioning

Name:

Company:

Master-class 4: Updates to

Transmission Technology

Ronald Freund Masaaki Hirano Tony Frisch

HHI Sumitomo Electric Xtera

Moderator

Pierre Mertz

Infinera

Copyright © SubOptic2019 Slide 2

Name:

Title:

Email:

Ronald started his career in 1996 as a co-founder of VPI Photonics

and then moved to Fraunhofer Heinrich Hertz Institute, where he is

currently leading the department Photonic Network and Systems

with the focus on high-capacity submarine and core networks, high-

speed access networks as well as satellite and quantum

communication networks (www.hhi.fraunhofer.de/pn).

In 2017 he was appointed Professor for Photonic Communication

Systems at the Technical University of Berlin. He holds a MBA from

RWTH Aachen.

Ronald Freund

Head of Department

[email protected]

http://www.hhi.fraunhofer.de/pn

mailto:[email protected]


Name:

Title:

Email:

Masaaki Hirano

Manger of Optical Fibre Engineering

[email protected]

Hirano joined Sumitomo Electric Industries, Ltd. in 1997 and has

been engaged in R&D on various specialty optical fibres including

low-loss fibre, dispersion-shifted fibre, high-nonlinearity fibre, etc.

He is now a group leader at optical fibre engineering and market

development, responsible for submarine fibres.



Name:

Title:

Email:

Tony started at BT's Research labs and then moved to Alcatel

Australia, becoming involved in testing and commissioning

submarine systems.

A move to Bell Labs gave him experience in terminal design and

troubleshooting, after which he went to Alcatel France, eventually

heading subsea Product Marketing.

He joined Xtera in 2004 and is now CTO.

Tony Frisch

CTO

[email protected]



1. Basics

– Shannon and Nyquist Laws

– Power, cost, complexity

– Coherent transmission

2. Building Blocks

– Constellations

– Amplifiers

– Fibre

– FEC

3. Possible future technology

– SDM

Contents

TF

Long

High-capacity

Systems

Unrepeatered

Systems

(UR)

Regional

Systems

Upgrade

(Legacy)

Systems

Long

High-capacity

Systems


A brief history of transmission

1Gb/s

1Tb/s

1Pb/s

Cable

Capacity

1985 1990 1995 2000 2005 2010 2015 2020 2025

SMF DSF

+

NZDSF DMF Low-loss

Large CoreFuture

Fibre

Dig. CoherentWDMOpt.

Amplification

FEC, DWDMRegenerated

Soft FEC

SDM?


• No major change is subsea amplifier bandwidth

Nearly all system are C-band-only: a few C+L

• Fibre attenuation dropped from 0.17 dB/km to 0.15 dB/km and

Effective area increased to 150 micron2 in the last decade

• Significant improvements in DSP modules

DSP = Digital Signal Processing

• Interesting research on non-linear mitigation,

multi-mode propagation …

Recent developments

TF

Name:

Company:

Basics

TF


TTE

TTE = Terminal Transmission Equipment

SLTE = Submarine Line Terminating Equipment

PFE = Power Feed Equipment

Typical subsea system

TTELoss

A

PFE PFE

Amp

TF

AAmp


• Cost is important

Quite a lot of fixed cost e.g. marine, environmental impact etc.

=> Maximise capacity to optimise cost per bit

• Repairing subsea plant is difficult, so reliability is important

– favours simplicity

• Power has to be fed along cable to subsea amplifiers

– lowest cable resistance currently 0.6 ohm/km

– power is the ultimate practical limit to capacity

Practical constraints

TF


I = In-phase

• On Off Keying (OOK) Q = Quadrature

Phase not important

• Phase Shift Keying (PSK)

Quadrature Amplitude Modulation (QAM)

• Phase and amplitude are modulated

• Points generally represent >1 bit

Coherent transmission

0

1

0180

90

270

Q

I1

1

0001

11 01

0180

90

270

TF


• Add receive signal to Local oscillator

• Convert to electrical signal

• Analogue to Digital Conversion

• Estimate "carrier frequency"

• Separate polarisations

• Compensate for dispersion

• Compensate for non-linearity (optional)

• Decode constellation

• Decode FEC

Coherent detection

~

DSP

Local oscillator

λ/4ADC

ADC

ADC

ADC

Simplified – each detector is

usually differential

Signal

TF


• Constellations with 4-16 points = 2-4 bits per symbol

2 polarisations

x 30 Giga symbols per second

x 2 bits per symbol

= 120 Giga bits per second

20% FEC overhead

100G per wavelength

37.5 GHz grid

2.7 bits/s/Hz

Common constellations

16QAM

QPSK

8QAM

4 bits

2 bits3 bits

TF


More capacity by increasing:

1. Number of constellation points

2. Symbol rate

3. Number of wavelengths

Publications

up to 256QAM = 8 bits/symbol

up to 80 Gbaud

up to 300

Main trend / objective

TF


Demonstrations

20T

40T

2012

Year

2010 2014 2016 2018 2020

60T

80T

Capacity

per fibre

Installed systems (Infinera, MAREA)

(NEC & Google, FASTER)

Laboratory

TF


• For a bandwidth of B (Hz)

the maximum symbol rate is:

• B symbols per second

• Assumes a sin(t) pulse shape

t

• Latest modules shape pulses using DACs to approximate this

[ DAC = Digital to Analogue Converter ]

• Higher symbol rate = larger bandwidth = more noise

Nyquist’s law

-0.30

0.20

0.70

1.20

-3 -2 -1 0 1 2 3

Am

plit

ude

Time

TF


• Maximum information capacity is: B·log2( 1 + SNR )

• Approximately (optical): 2B·log2( )

• Ideal; assumes a linear system

• Different constellations

need different SNR

Shannon’s law

Popt

Noise

QPSK

16QAM

BPSK

TF


BPSK

QPSK

8QAM

16QAM

32QAM

64QAM

0

3

6

9

12

15

18

1 2 3 4 5 6 7

Required O

SN

R (

dB

) re

fative

to B

PS

K

Bits per symbol / Relative symbol rate

Shannon - SNR

Constellations

Shannon - BW

Shannon’s law

TF


QPSK, 8QAM & 16QAM

0

5

10

100

Capacity per wavelength

200150

Relative

OSNR

(dB)

Limits capacity unless SNR is just right

TF


Intermediate modulation schemes

Ideal case

0

5

10

100

Capacity per wavelength

200150

Relative

OSNR

(dB)

130 170

Maximises capacity for a given SNR

TF


• 5 x 7 inch modules

• Offer up to 800G per wavelength

– capacity set by line design

• Steps of 25G

• Flexibility also on symbol rate,

WL grid and FEC parameters

• Most suppliers integrate

2 wavelengths in the same module

Maybe more in the future

Available modules

Source: Acacia

Communications

TF


0.18

0.19

0.20

0.21

0.22

1450 1500 1550 1600 1650

Att

en

ua

tio

n (

dB

/km

)

Wavelength (nm)

Attenuation of "standard" single-mode fibre

Fibre and optical bands

S LC

Currently

not

used

MH


• Low attenuation 0.150-0.156 dB/km available

• See OP-18 for details of 0.14 dB/km

• Negligible cabling loss– Index profile

– Multi-layer coatings

• Low non-linearity– Large effective area 130-150 micron2

– High chromatic dispersion ~20 ps/km/nm

Fibre

MH

Hard secondary resin

Soft primary resin

Glass

Name:

Company:

Break for questions

TF

Break


• Initially Reed-Solomon, then Product codes, now Soft Decision

• Q limit (typically at BER = 1E-13)

11 dB (RS hard decision)

8 dB (Product hard decision)

5 dB (Soft decision)

• 5 dB means correction of around 1 error per 25 bits (average)

Forward Error Correction (FEC)

TF


Hard-decision FEC

• Decide what symbol was detected

• Apply FEC decoding

• Fails if error probability is too high

Soft-decision FEC

• Uses the symbol value before

decoding to determine if an error

is likely

Principle of soft-decision FEC

12

0010

11 01

TF

Q ≈ 6


0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

70.0%

80.0%

90.0%

100.0%

0 1 2 3 4 5 6 7 8 9

Code O

verh

ead

Q Limit (dB)

• Higher coding gain needs more overhead = more processing

• Overhead starts to reduce data capacity as Q limit decreases

FEC limits

SD

HD

TF


• EDFA

• NF ~4.5 dB

+ Compact

+ Low consumption

– Bandwidth limited to C-band, ~40 nm

• Most common in subsea (and terrestrial) systems

Fibre-based amplifiers

PEDF

RF


• C + L using couplers

• NF ~4.5 dB

+ Essentially the same technology

+ Bandwidth ~70 nm

– Loss of splitter and coupler

– L amplifier generally more complex than shown

• Only one supplier offering this


PEDF

PEDF

C

L

RF

Splitter Combiner


• Distributed Raman

• Amplification occurs in the cable

• NF 0-3 dB

+ Low noise

+ Bandwidth up to 100 nm

– Needs several high power Raman pumps

• Not used in subsea amplifiers, but good for unrepeatered systems


C

P

Raman

P P

RF

different

wavelengths


• EDFA + Raman

• NF ~2.5 dB

+ Low noise

+ Bandwidth ~70 nm

– Needs Raman pump as well EDFA Pump – similar powers

• Only one supplier offering this


PEDF

C

P

Raman

RF


• SOA

• Anti-reflective coating to prevent lasing

• NF 6-10 dB

+ Compact

+ Low consumption

+ Large bandwidth >100 nm

– Noise performance not so good 6-10 dB

• Used in non-telecoms applications e.g. fibre-gyroscopes

Semiconductor Optical Amplifier

Semiconductor

Diode

RF


• Optical Monitoring

• Gain Flattening Filter (GFF)

• Isolator for stability

• Loop-back / OTDR path

Passive mechanism for fault location

Optical amplifier extras

PEDF

M

GFF

Loop-back

OTDR

Break

TF


• ALL amplifiers generate noise

• Noise spectral density = G·hv·NF G = Gain

(for 2 polarisations) h = Planck’s constant

EDFA NF ~3 (4.5 dB) v = Frequency

EDFA-DRA ~2 (3 dB) NF = Noise Figure

• After N amplifiers

• Total noise ≈ N·G·hv·NF (assumes all sections are the same)

Amplifier noise

TF


• Chromatic dispersion: different wavelengths travel at different speeds

• Reduces non-linear interaction between different wavelengths

• Most significant effect is Self-Phase Modulation (SPM)

Dispersion and non-linearity

TF

A few km

20 ps/nm/km

Cross-Phase Modulation

(XPM)


0.0

1.0

2.0

3.0

4.0

0 2 4 6 8

Q c

ha

nge

(dB

)

Amplifier output change (dB)

• Benefits of increasing power are limited by non-linear effects

Non-linear effects

Real example

Linear

Measured

TF


• Several solutions

1. Large core fibre

2. Modification of transmitted constellation

3. Non-linear compensation in receiver

Reducing the effects of non-linearity

TF


• Increase mode field diameter: reduce power density

• Needs careful design to avoid increasing micro-bending sensitivity

1. Large-core fibre

0

0.1

0.2

0.3

0.4

0.5

0.6

-3 -2 -1 0 1 2 3

Position across fibre

Power

More power in

outer part of fibre

MHM

icro

-be

nd

ing

loss

at 1

550

nm

[dB

/km

]

1

0.1

0.01100 110 120 130 140 150 160 170

Effective Area, Aeff [mm2]


• Increases the level of power that can be used

Improvement due to large-core fibre

MH

-60

2

4

6

8

-4 -2 0

110mm2

+4

Launched Signal Power; Pch [dBm/ch]

Q c

ha

ng

e [d

B]

+2

150mm2

+0.9 dB


• Geometric (GCS): Modify the position of the points

• Probabilistic (PCS): Modify the relative probabilities of different points

• Need to consider both non-linear performance and carrier recovery

2. Improved constellations

Constellation shaping

Square Geometric Carrier recovery

TF

Probabilistic


• Back propagation

• s → f(s + n1) s is signal, n is noise

• s → f(f(s + n1) + n2) f is a non-linear transfer function

→ f(f(f(s + n1) + n2) + n3)

→ f(f(f(f(s + n1) + n2) + n3) + n4) ...

• Compensation requires applying f-1 multiple times to derive s

• This process known as "Backward propagation"

• Very complex, but some good alternatives being developed

3. Non-linear compensation

s

n1

f

n2 n3 n4

RF

f f


• Using neural network

• Nice demonstration by NEC and Google on the FASTER cable system

• Treats the transmission system as a "black box“

• Initial training allows the network to learn what output should be

based on knowing what was transmitted

Then apply this to unknown traffic

• Not as complex as digital back propagation

3. Non-linear compensation

s

n1

f

n2 n3 n4

RF


1. Large core fibre commonly used solution

2. Constellation shaping slightly more complex DSP

3. Non-linear compensation much more complex DSP

0.5-0.8 dB improvement today

• Can be used in combination

Comparison

RF

Name:

Company:

Possible Future Technology

RF

Break


• Cable capacity proportional to:

number of Fibre-pairs x cores x modes x Bandwidth x log2(1 + SNR)

• Improved amplifiers Bandwidth

• Enhanced non-linear compensation SNR

• Conjugate transmission / reflection SNR

• SDM (Spatial Division Multiplexing) Fibre pairs

– Multi-core fibres Fibre cores

– Multi-mode fibres Fibre modes

Possible technologies

RF


P P P

• Possible with Raman – greater than 100 nm demonstrated

• Good noise figure

• BUT

– more pumps

– more power

– more complexity

• A good solution where power is not a limitation

Example: Unrepeatered systems

Even greater bandwidth amplifiers

P

Raman

P P

RF


Not easy to find practical realisation

Optical Phase Conjugation (OPC)

Transmitter ReceiverFiber Fiber

EDFA EDFA

Fiber Fiber

EDFA EDFA

Signal

field, A

λ

Conjugated

signal field, A*

λ

Transmitter ReceiverFiber Fiber Fiber Fiber

EDFA EDFA EDFA EDFA

Low launch

powers

Linear

distortion

High launch

powers

Nonlinear

distortion

Transmitter ReceiverFiber Fiber Fiber Fiber

EDFA EDFA EDFA EDFA

Low launch

powers

Linear

distortion

High launch

powers

Nonlinear

distortion

RF


Phase conjugation using non-

degenerate intermodal FWM

Idler

Invers

e g

rou

p v

el.

Pump2 Pump1

Signal

Mode 1Mode 2

Phase matching

Mode 2: Average

Mode 1: Average Energy conservation:

𝜔𝑝1 +𝜔𝑝2 = 𝜔𝑠+ 𝜔𝑃𝐶

𝜔

𝜔𝑝2 𝜔𝑝1

𝜔𝑃𝐶 𝜔𝑠

Broadband operation can be achieved using dispersion-engineered waveguide

EDFA

Pump 2

32-Gbaud

QPSK Transmitter

Coherent

Receiver

× N

EDFA SSMF

80 km

× N

EDFA SSMF

80 km

PC

Pump 1

ECLSiP

UB-PSP

ECL

Filter

LP01

LP11a

De

mu

x

Mu

x

LP11a

LP01

Dual-pump Optical Phase Conjugator

RF


PEDF

L

+ Saves fibre

– Adds couplers

– More complex L amplifier

+ Saves power

+ Simpler

– More fibre (or cores)

C + L or C + C?

PEDF

C

PEDF

PEDF

C

C

TF


• Power reduction is significantly greater than Q change

-60

2

4

6

8

-4 -2 0 +4Launched Signal Power; Pch [dBm/ch]

Q c

hange [d

B]

+2

110mm2Effective area is not critical

150mm2

10

Also reduce power

TF

Low attenuation is valuable


• Affects consumption by changing the power needed from the amplifier

• Example: 100 km spacing

0.01 dB/km reduction

= 1.00 dB per section

= 20-25% more amplifiers for the same total consumption

Attenuation

TF


• Highly desirable to maintain cable size

Cost, Handling, Length that can be loaded …

Typical cable

Source: Nexans ~3 mm

Loose packing

17-18 mm

TF


Coating

Fibre

Possible SDM routes

0.25 mm

Individual Fibre

Multi-Mode Fibre

3

Multi-core Fibre

2

TF

36 Fibres

0.5 mm

1


• OP-10 "Low-loss multi-core fibers for submarine transmission"

Possible SDM fibres

Single-core fibre

0.15 dB/km

SiO2

2 core fibre

0.16 dB/km

SiO2

4 core fibre

0.16 dB/km

SiO2

Needs MIMO

Multi-mode fibre

0.23 dB/km

GeO2+SiO2

Needs MIMO

MH


• More compact, so will fit in existing tubes

• Core interaction is an issue

• Will probably need individual amplifiers

Cladding pumped amplifier doesn’t solve:

Gain flattening

Loop-back / OTDR coupling

Multi-core fibres

NF 1 dB

worse

Pumppower

TF


• By changing the core, a fibre

can propagate several modes

• Each could carry different data

• It is relatively easy to create

multi-mode EDF amplifiers,

which avoids problems

coupling fibre to amplifier

• BUT

• What will the performance of

multi-mode fibre be like?

Micro-bending?

Mode coupling?

• How are the modes coupled

into the fibre and separated at

the end?

Multi-mode fibres

TF

Name:

Company:

Sum-up

TF


• New types of amplifier Not soon

• Larger core fibre No; SDM instead

• Better non-linear mitigation No; SDM instead – upgrades?

• Improved FEC Getting harder to improve?

• More flexible constellations Better granularity

• Lower consumption DSP Yes

• Spatial Division Multiplexing (SDM) Yes

• Multi-core fibre Some challenges

• Multi-mode fibre Mode mixing?

Summary – to provoke questions!

TF

Don’t take too seriously

Please remember to complete the session

survey in the conference mobile app

Thanks for listeningWe hope it was useful, interesting

Documents

Moderator Pierre Mertz Infinera Master-class 4: Updates to ... · Tony started at BT's Research labs and then moved to Alcatel Australia, becoming involved in testing and commissioning