advanced modulation for high data rate optical transmission: 100g and beyond

Advanced Modulation For High Data Rate Optical Transmission: 100G and Beyond

Leigh Wade, Infinera

The State of the Market Today

800G •10Gb/s •NRZ •C-band

80 ch. @ 10G = 800G

More channels

Higher Data Rates

More Spectrum

I will propose that photonic integration is an excellent solution to all three

capacity challenges

Why do we need more than 800G?

Lower Cost per Bit

More Capacity

Higher Speed Services

Fiber Exhaust & Network Economics C

ost

per

Usa

ble

Bit

Time

10G λ

40G λ

100G λ

You want to move to 40G here…

…but what if you hit fiber exhaust here?

Excess cost

Fiber exhaust can force uneconomic network decisions

Double Density Optics Mean Investment Protection and “Option Value”

Conventional 80-96 λ WDM

1 λ per 50 GHz

At 40% bandwidth growth, double-density optics mean two more years to select the lowest cost transmission.

Infinera “Double Density” WDM

1 λ per 25 GHz

800G in the C-band 1.6T in the C-band

Why doesn’t everybody offer Double Density?

Two basic reasons:

WSS ROADMs designed

around 50GHz spacing

Operational challenge of 160 discrete transponders on a single fiber!!

What are you going to see?

Adding a single PIC-based line card, with 10x10Gb/s waves

100Gb/s of capacity for the same effort as one 10Gb/s transponder

Optical Spectrum Analyzer

Stopwatch

Existing 10G waves on the fiber

“Gaps” for additional waves

PICs reduce operational burden by 10x

But the rest of the optical industry does not have access to PICs so…

They are under pressure to move to 40G and 100G as soon as possible

Not necessarily when it’s economical!

Fiber Capacity

Advanced Modulation

Coherent Detection

High Gain FEC

Table Stakes

Core Switching & Grooming 3

Large Scale PICs 2

Photonic Integration 1

Differentiators

100G Technology Features

Complex modulation requires complex optical circuits

Why do I need Complex Modulation?

Optical transmission is about: • Sending high data rates • Over very long distances • For very little money

Our biggest problem is optical fiber: • Loss • Dispersion

• Modal dispersion • Chromatic dispersion • Polarization mode dispersion

• Non-linear effects • Self phase modulation • Cross phase modulation • Four wave mixing

If you stress any one of these variables, the others will respond

For a given modulation type, the gross magnitude of these impairments scales roughly with the square of the symbol rate

Think of a light wave...

Oscillating wave

Wavelength • 1550nm

Frequency • 193.1 THz

“State of the shelf” electronics can process at ~10GHz

Electronics is about 20,000 times “too slow” for direct detection of

“wave properties”

So how do we encode and detect signals on an optical carrier?

Historically, used amplitude modulation

Measures the strength of a large number of waves

On/Off Keying (OOK) may interpret the presence of a signal as a “1”, and the absence of a symbol as a “0”

1 bit per symbol: NRZ Modulation

Laser Modulator

Detector

Tx

Rx

NRZ

Simple modulation technique Easy to implement Low power use But very sensitive to fiber impairments

as bitrate increases • This is what we’re talking about with the “square”

relationship

Increasing power will trigger non-linear effects

Phase Shift Keying

Phase is fundamental property of waves • Two waves in-phase when the peaks & troughs line up

• We say that such waves are coherent

• If non-coherent waves combine we see:

• Reinforcement, cancellation, interference

Interference can be used to extract a lower frequency modulation from a high frequency carrier

In-phase Out of phase Interference patterns

Using Phase to Apply a Signal

LD

Laser generates a constant carrier

The carrier is split into 2

The carriers travel over different paths

S

Can apply a data signal, S, to vary the delay on

one of the arms

When the carriers recombine they will “contain” the data signal encoded as a series of phase changes Tx

Rx Q: How do we recover the data signal at the receiver? Hold that thought!

MZI

Component Complexity

Tx Rx

Part 1 The Transmitter

ODB Modulation (Optical Duo-Binary)

Laser MZ Modulator

Detector

Tx

Rx

ODB

First generation 40G modulation scheme

Phase & Amplitude based modulation • Requires MZ modulator

• Can use simple, direct detection

Much more tolerant of dispersion

Limited reach

Widely used by 1st Gen 40G • Stratalight, Mintera

1 bit per symbol: DPSK

Most basic phase modulation technique

Differential technique allows phase slips to be ignored

Used by OpNext & Mintera, and their OEMs

AKA: BPSK, where local oscillator coherent detection is used

Re{Ex} 1 0

2 bits per symbol: Quadrature PSK

Advanced modulation, 4 phase states = 2 bits

More bits per symbol

2 bits per symbol: Quadrature PSK

Advanced modulation, 4 phase states = 2 bits

More bits per symbol

0,0

0,1

1,1

1,0

3 bits per symbol: 8-PSK ...And higher orders of modulation

8 phase states = 3 bits

Twice as complex, but only 50% more bits

1,0,0

0,0,1

1,1,0

1,0,1 0,0,0

0,1,1

1,1,1

0,1,0

For discrete implementations, 8-PSK seems to be too complex

The Law of Diminishing Returns Phase States vs Component Complexity

Let’s set a circuit complexity factor of 1, to be the equivalent of a simple DPSK transponder

DPSK (D)QPSK 8-PSK 16-QAM 32-QAM 64-QAM

16x

Bit/Hz

9

8

7

6

5

4

3

2

1

9

8

7

6

5

4

3

2

1

Co

mp

lexi

ty F

acto

r

Is there a better way to get to more bit/Hz?

32x

PM-QPSK, 4 bits per “symbol”

Im{Ex}

Re{Ex}

Im{Ex}

Re{Ex}

Im{Ey}

Re{Ey}

Two Polarizations

X-Polarization

Y-Polarization

Implementing Phase Modulation Using Discrete Optical Components...

S

S

Implementing Phase Modulation Using Discrete Optical Components...

This is QPSK...

S1

S2

Im{Ex}

Re{Ex}

This structure called a “Super Mach Zehnder”

This is a PM-QPSK Transmitter

PBS LD

X Polarizations

Y Polarizations

Component Complexity

Tx Rx

Part 2 The Detector

Let’s cut to the chase…

The only practical, long haul 100G implementations will be required to use Coherent Detection

What is it, and why is it useful?

What is “coherent detection”?

Physics definition • A detection technique that is based on the phase properties

of the carrier

• If you are using a phase-based detector, you could claim to be implementing coherent detection

…however…

Practical definition • The market has now come to expect a “coherent detector” to

make use of sophisticated, digital signal processing (DSP) algorithms

Conventional WDM Detection

PD

Mixture of waves on fiber… …wideband detector

How do we select the channel we want to detect?


Wavelength demux

PD

Mixture of waves on fiber… …wideband detector

Direct conversion of photons into electrons that “look like” bits

…11010110…


Wavelength demux

PD

Summary of “Conventional WDM Detection”

Wideband Photodetector (PD) is used

To prevent inter-channel interference, a wavelength demux is used to spatially separate channels

Modulation technique allows minimal Rx circuit complexity – essentially “direct detection”

No additional signal processing normally required

ADC DSP

Coherent WDM Detection

PD LO

We could take a mixed signal that uses a phase-based modulation technique

Use a local oscillator to choose the “color” we want to “detect”

ADC DSP

Coherent WDM Detection

PD LO …11010110…

Convert the photons to electrons

Convert the “analog electrons”

into “digital electrons”

Clean it all up!

If you need to detect 5 from 1…n, then choose a local oscillator tuned to 5

Local oscillator does not carry a signal – simply a continuous beam of light

But it is non-coherent with the received signal (ie. it is out of phase)

Use an array of interferometers to “measure” the interference patterns

Convert the interference patterns into an electronic signal, and “process it”

Why phase-based modulation?

The Detector Requires a Complex Optical Circuit Example: For PM-QPSK Modulation…

PBS

LO PBS

PD

PD

PD

PD

The signals that come out of the PD array are “analog and dirty”

PM-QPSK Signal

Two very different functions in the detector

Phase state extraction

Signal processing

• Separate the polarization components

• Create interference against a reference laser (local oscillator)

• Separate the phase components

• PD & A/D conversion

• Compensate for local oscillator instability

• Compensate for static CD

• Compensate for dynamic PMD

How do we implement these functions?

• Separate the polarization components

• Create interference against a reference laser (local oscillator)

• Separate the phase components

• PD & A/D conversion

• Compensate for local oscillator instability

• Compensate for static CD

• Compensate for dynamic PMD

Sophisticated optical circuit

(PIC)

Sophisticated digital signal processing

(DSP)

A Coherent Detector Schematic (For one wavelength only)

Incoming carrier (2 polarizations, each with 4 phase states)

ADC A/D Converter AMZ Adjustable Mach Zehnder DSP Digital Signal Processor LO Local Oscillator PD Photo Detector PS Polarization Splitter

LO

PD

PD

PD

PD

ADC

ADC

ADC

ADC

DSP

AMZ

AMZ

AMZ

AMZ

Optical Circuit Electronic Circuit

PBS

PBS


LO

PD

PD

PD

PD

ADC

ADC

ADC

ADC

DSP

AMZ

AMZ

AMZ

AMZ


PBS

PBS


1

Step 1: Take the two optical sources – signal and local oscillator



LO

PD

PD

PD

PD

ADC

ADC

ADC

ADC

DSP

AMZ

AMZ

AMZ

AMZ


PBS

PBS


2

Step 2: Separate the X and Y polarizations



LO

PD

PD

PD

PD

ADC

ADC

ADC

ADC

DSP

AMZ

AMZ

AMZ

AMZ


PBS

PBS


3

Step 3: Generate a set of interference patterns in the SMZ array



LO

PD

PD

PD

PD

ADC

ADC

ADC

ADC

DSP

AMZ

AMZ

AMZ

AMZ


PBS

PBS


4

Step 4: Convert optical signals to analog electronic signals



LO

PD

PD

PD

PD

ADC

ADC

ADC

ADC

DSP

AMZ

AMZ

AMZ

AMZ


PBS

PBS


5

Step 5: Convert analog to digital and process


Coherent Detection – Pros and Cons

Pros: • Operates over the existing fiber plant and amp chains

• Outstanding reach performance

• Closest thing to achieving 40G and 100G with same reach as 10G NRZ

• Significant pilot test results indicate it really does work!

Cons: • Potential non-linear interaction with 10G NRZ in same fiber

• The “cure” is managing launch power

• Probably represents the practical complexity limit for discretes

• State of the shelf DSP technology draws too much power to allow for large scale implementations (ie. multiple waves in one modules)

• Solution is to use emerging 40µm DSP technology

• DSP operation probably eliminates the chance of future line side interop

Complex modulation requires complex optical circuits

So remember…

Where have we seen this problem before?

In the 1950s computers were made from individual transistors, resistors and capacitors...

…today?

http://images.google.com/imgres?imgurl=www.acte.dk/pages_dk/produkter/billeder/capacitor.gif&imgrefurl=http://www.acte.dk/pages_dk/produkter/yageo.asp&h=140&w=180&prev=/images?q=capacitor&svnum=10&hl=en&lr=&ie=UTF-8&oe=UTF-8

http://images.google.com/imgres?imgurl=www.suntan.com.hk/Metallized Polycarbonate Film Capacitor.jpg&imgrefurl=http://www.suntan.com.hk/product.html&h=286&w=413&prev=/images?q=capacitor&start=40&svnum=10&hl=en&lr=&ie=UTF-8&oe=UTF-8&sa=N

http://images.google.com/imgres?imgurl=www.toroid.se/inductor.jpg&imgrefurl=http://www.toroid.se/prod.htm&h=183&w=276&prev=/images?q=inductor&svnum=10&hl=en&lr=&ie=UTF-8&oe=UTF-8

The electronics industry controlled component complexity with large scale integration

We know the same thing works for optical components – we did it 5 years ago!

Small Scale vs Large Scale Photonic Integration

Small Scale… • Operates on a single wavelength

• Primarily used to address manufacturing cost

If it works for one wave, why not…

CPUs with 2-8 cores GPUs with 200-800 cores!!

Infinera 100G Transmission Differentiators 500G, Large Scale, Monolithic PIC Implementation

500G Tx PIC

500G Rx PIC

Number of channels 5 x 100G

Monolithic InP Chips 2

Optical elements > 600

“Gold Box” Replacements > 100

Fiber Replacements > 400

COST

SIZE

POWER CAPACITY

RELIABILITY

54 © 2011 Infinera Corporation Confidential & Proprietary

How much capacity can actually be used?

Fat Pipes Are Not Enough

100 Gb/s Transmit

100 Gb/s Receive

PICs enable cost-effective OEO

100Gb/s to 1Tb/s “WDM system on a chip”

Affordable access to digital domain

Photonic Integration


Infinera 100G Transmission Differentiators PICs Enable Pervasive Digital Switching

1001

0101

0101

1010

1101

0101 0101 1010 1101 0101

Enables “digital” functionality

Integrated switching at every node

High functionality Digital ROADM

Dramatic network simplification

100101011101010000101011 100101010101101011010101

110101000010101110010101 001010111011010110010101

Inte

grat

ed P

ho

ton

ics

Inte

grat

ed P

ho

ton

ics

Optical (O) Electrical (E) Optical (O)

Trib

Integrated Switching + WDM




1001

0101

0101

1010

1101

0101 0101 1010 1101 0101

100101011101010000101011 100101010101101011010101

110101000010101110010101 001010111011010110010101

Inte

grat

ed P

ho

ton

ics

Inte

grat

ed P

ho

ton

ics

Pervasive Digital Switching

Integrated Switching + WDM


10010101110101010000 10010101010110101011

10010101110101010000 10010101010110101011 end-end service

Software-based “Ease-of-Use”

Digital OTN switching at every node

Unconstrained bandwidth everywhere

Lowest cost per switched Gb/s



Solving The 100G Muxponder Tax

The Problem: • Backbone waves move to 100G, but service demands still 10G or lower

• All-optical ROADMs have no inter-wavelength, or sub-wavelength grooming capability → 100G muxponders!

How big is the “Muxponder Tax”

in a real 100G network?

A B All services must go A ↔B

10GbE

10GbE

10GbE

Mu

xpo

nd

er

10GbE

Mu

xpo

nd

er

ROADM Network

A C ↔

A D ↔

B C ↔

B D ↔

Require Extra, Partially Filled

Muxponder Pairs

Service Demands:

© 2011 Infinera Corporation Confidential & Proprietary 59

100

90

80

70

60

50

40

30

20

10 Dep

loye

d C

apac

ity

(%)

Rev

enu

e G

ener

atin

g (%

) 100

90

80

70

60

50

40

30

20

10

100G Muxponder

50%

40G Muxponder

66%

Infinera Digital ROADM

92%

Infinera National Network Model Summary

• Large N. Am. Network Model: 33,084 route km, 47 core WDM links • About 10 Tb/s of customer service demands (network traffic volume)


Summary of Network Efficiency

A “Perfect Storm” is emerging in terms of network bandwidth efficiency: • Wavelength speeds moving to 100Gbit/s

• Majority of services demands remaining at 10Gbit/s or less for near-term

• All-optical ROADMs have no effective way to offer contentionless wavelength conversion and sub-wavelength grooming in the core

• Muxponders are simply point-point aggregators and do not do grooming

The result is that a Service Provider may need to purchase 2X Network Capacity for 1X Service Revenue

The solution is an Integrated Digital OTN Network with: • End to End, Any to Any service capability

• Integrated OTN switching and grooming in the core

• End to End intelligent optical control plane

Bandwidth Virtualization

8Tb/s

More channels

Higher Data Rates

More Spectrum

Beyond 8Tb/s?

Gridless Super-Channels

Even more complex modulation!

L-Band S-Band E-Band O-Band

Outside the scope of this discussion

What’s changed so far

Since the advent of DWDM…

now

Phase Modulation

Coherent Detection

ITU Frequency Grid

Intensity Modulation

Direct Detection

ITU Frequency Grid


What Comes Next For Terabit Transport?


…so what has to change

Phase Modulation

Coherent Detection

ITU Frequency Grid

Intensity Modulation

Direct Detection

ITU Frequency Grid

Quadrature Amplitude Modulation (QAM)

Coherent Wave Combining and Separation

Grid-less FlexChannels


Advanced Modulation Formats

Pol-Mux QPSK Pol-Mux

8-QAM

Pol-Mux 16-QAM

IM-DD

PM- BPSK

1.6 8 12 16 24

C-Band Capacity (Tb/s)

0

0.2

0.4

0.6

0.8

1

1.2

Cap

acit

y *

Re

ach

Pro

du

ct




Quadrature Amplitude Modulation

Coherent Wave Separation


On-Off Keyed Modulation

Direct Detection

ITU Frequency Grid




Single Carrier vs Multi-Carrier

Goal: Create a 1Tb/s unit of transmission capacity

How?

Option 1: Build a single-carrier 1Tb/s

channel

Option 2: Build a multi-carrier 1Tb/s

“super-channel”


1Tb/s Single Carrier: The A/D Converter Problem

1 2 4 6 8 10 12

1

2

3

4

5

6

7

8

9

10

OSN

R P

enal

ty (

dB

)

Number of bits per symbol

PM-BPSK 640GBaud

PM-QPSK 320GBaud

PM-8QAM 210GBaud

PM-16QAM 160GBaud

PM-32QAM 128GBaud

PM-64QAM 105GBaud

By 2014 commercial ADCs are expected to operate at ~64GBaud

wavelength demux

DWDM Direct Detection

PD

Spacing on the fiber needed between waves: “Guard Bands”

Spatially separate the

channels using a

wavelength demux


wavelength demux

Spatially separate the

channels using a

wavelength demux

DWDM Coherent Detection

Spacing on the fiber needed between waves:

“Guard Bands”

ADC DSP PD LO

Use a local oscillator to choose the “color” we want

to “detect” to match the demux port color


How 1Tb/s Might Look… Conventional WDM vs FlexChannels

Guard bands to allow for individual wavelength demux

Fewer guard-bands

25% increase in useable amplifier spectrum

Conventional Per-Channel WDM Filtering

1Tb/s

Multi-Carrier FlexChannel

1Tb/s





Quadrature Amplitude Modulation

Coherent Wave Separation


On-Off Keyed Modulation

Direct Detection

ITU Frequency Grid


FlexChannels Increase Total Fiber Capacity More complex modulation → more capacity per fiber

PM-QPSK

8-QAM

16-QAM

1Tb/s

12 Tb/s

18 Tb/s

25 Tb/s


Reach, Spectral Efficiency, and Co-Existence

1Tb/s PM-8QAM FlexChannel

1Tb/s PM-16QAM FlexChannel

10x100G PM-QPSK 1Tb/s PM-QPSK

FlexChannel

or

A E

B C

D


Summary: The Key Technologies For 1Tb/s Are Well Understood

But the implementation of those technologies will be critical to allowing service providers to differentiate their products and services

Advanced Modulation

Coherent Processing

Advanced FEC

Foundation Features

Large Scale PICs 1

FlexCoherent™ Modulation 2

Pervasive, Switched DWDM 3

Differentiators


Thank You! [email protected]
