Workshop on 40 Gig Networks: The Actual World PMD Challenge · 3/25/2007 · PMD as Bottleneck for 40Gbit/s Long-Haul Systems Introduction § Detailed survey of all usable fiber

OFC Workshop 1

OFC 2007, Anaheim, CA

PMD as Bottleneck Problem for the Introduction of 40Gbit/s

and Future 100Gbit/s Ethernet into German WDM Backbone

W. Weiershausen, D. BreuerW. Weiershausen, D. BreuerW. Weiershausen, D. BreuerW. Weiershausen, D. Breuer

TTTT----Systems, Deutsche TelekomSystems, Deutsche TelekomSystems, Deutsche TelekomSystems, Deutsche Telekom

Workshop on

40 Gig Networks: The Actual World

PMD Challenge

OFCNFOEC-2007 Workshop "40Gb/s Networks and the PMD Challenge" - Complete Workshop Program at www.fiberwork.net/ofc.asp

www.fiberwork.net/ofc.asp

OFC Workshop 2

Outline

■ DT’s network dimensions

■ Vintage fibers in DT’s network

■ Bottleneck for 40 Gbit/s (and 100 GEthernet) backbone

introduction

■ Results from measured PMD dynamics in DT’s network.

Requirements definition for active PMDC



OFC Workshop 3

German Fiber Network: The DistancesAre 40G/100GE as 40GBaud/100GBaud possible ?

Statistical analysis of paths in DT Network

Parameter Value

Max path length 951 km

Min path length 36 km

Mean path length 413.5 km

Standard deviation path length 201.6 km

Max length of 90% of paths 675 km

Max path hops 7

Mean path hops 2.85

Standard deviation path hops 1.45

Max hops of 90% of paths 4.5



OFC Workshop 4

Future Option for Fully Transparent Leased Line OTH Channel FeedThrough Between Regional Metro Rings via Backbone Link

Hamburg

BerlinHannoverBremen

Norden

Essen

Köln

Düsseldorf

Frankfurt

NürnbergMannheim

Karlsruhe

Stuttgart

UlmMünchen

LeipzigDortmund

Hamburg

BerlinHannoverBremen

Norden

Essen

Köln

Düsseldorf

Frankfurt

NürnbergMannheim

Karlsruhe

Stuttgart

UlmMünchen

LeipzigDortmund

Transit bypass

OADM OADM OADM OADM orororor

Band Band Band Band BypassBypassBypassBypass

1000 km max.

1500 km max.



OFC Workshop 5

Mean and variance of PMD coefficient as a function of

cable installation year

0,00

0,05

0,10

0,15

0,20

0,25

0,30

0,35

0,40

0,45

19

86

19

87

19

88

19

89

19

90

19

91

19

92

19

93

19

94

19

95

19

96

19

97

19

98

19

99

20

00

20

01

Cable Vintage

PM

D [

ps/√

km

]

Mean Valueper year

Variance

Region 2

PMD specification introduced

as standard process

Mean Value 0,13

Region 1

Mean Value 0,32

Region 3

Mean Value 0,052

PMD in DT Network: Cable Vintage Dependence



OFC Workshop 6

Percentage of fibers in different PMD classes and

cable installation years

0,0%

0,1%

1,0%

10,0%

100,0%

<=

0,0

2

>0,0

4 to <

=0,0

6

>0,0

8 to <

=0,1

>0,1

2 to <

=0,1

4

>0,1

6 to <

=0,1

8

>0,2

to <

=0,2

2

>0,2

4 to <

=0,2

6

>0,2

8 to <

=0,3

>0,3

2 to <

=0,3

4

>0,3

6 to <

=0,3

8

>0,4

to <

=0,4

2

>0,4

4to

<=

0,4

6

>0,4

8 to <

=0,5

>0,5

2 to <

=0,5

4

>0,5

6 to <

=0,5

8

>0,6

to <

=0,6

2

>0,6

4 to <

=0,6

6

>0,6

8 to <

=0,7

PMD-coefficient [ps/√ km]

Perc

enta

ge

<= 1991 (Region 1)

1992 to 1998 (Region 2)

>= 1999 (Region 3)

MeanValue0,052

MeanValue0,13

MeanValue0,32

> 0

,7

PMD in DT Network: Cable Vintage Dependence



OFC Workshop 7

■ Less than 50% of fibers

usable for 40 Gb/s

■ Adaptations needed

■ Enhanced receiver

robustness

■ PMD compensation

■ Equalizer

■ New modulation

formats, with multilevel

symbols and reduced

symbol rate.

PMD in ps/sqr(km)

Mean Value 0,079 ps/sqr(km)

0

0,01

0,02

0,03

0,04

0,05

0,06

0,07

0,08

0,09

0,0

2

0,0

4

0,0

6

0,0

8

0,1

0,1

2

0,1

4

0,1

6

0,1

8

0,2

0,2

2

0,2

4

0,2

6

0,2

8

0,3

0,3

2

0,3

4

0,3

6

0,3

8

0,4

0,4

2

0,4

4

0,4

6

0,4

8

0,5

< 5 ps < 7.5 ps > 7.5 pspspsps

< 2.5 ps

Class 1 Class 2 Class 3 Class 4

nnnn

PMD as Bottleneck for 40Gbit/s Long-Haul Systems Introduction



OFC Workshop 8


§§ Detailed survey of all usable fiber combinations Detailed survey of all usable fiber combinations

to install 40 to install 40 Gbit/sGbit/s in DTin DT’’s backbone network: s backbone network:

Best case approachBest case approach

§§ Result is that less than 50% of the routes can be Result is that less than 50% of the routes can be

used for 40GBit/sused for 40GBit/s

§§ Solutions: Detect and remove dominating highSolutions: Detect and remove dominating high--

PMD cable segments or using robust WDM (e.g. PMD cable segments or using robust WDM (e.g.

low low baudratebaudrate techniques)techniques)

Real usability of fibers is much worse than Real usability of fibers is much worse than

expected from overall statistics:expected from overall statistics:

§§ typically: 40 km link statisticstypically: 40 km link statistics

§§ More realistic: Link Design Value (LDV) for real More realistic: Link Design Value (LDV) for real

linkslinks

§§ Realistic: Real connections: Dominating bad Realistic: Real connections: Dominating bad

fiber segmentsfiber segments



OFC Workshop 9

time [h]time [h]

Long Term DGD Field Measurements: Hinge Behavior in DT’s Network

48 24 0

Zeit [h]48 2

4 0

Zeit [h]

2.28

054

0.40

0.62

0.95

0.28

0.340.16

0.180.13

0.15

synchronoushinge

PMD [ps]



OFC Workshop 10

Sa Su Mo Tu We Th Fr

52

53

1

2

LongtermMeasurements of PMD Fluctuations DT Fiber Infrastructure



OFC Workshop 11

Measured Fast SOP/DGD Changes in DT‘s Network

304 306 308 310 312 314 316-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

Time in ms

Sto

ke

s p

ara

me

ter

0 100 200 300 400 500 600-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

Tim e in m s

Sto

ke

s p

ara

me

ter



OFC Workshop 12

Dynamic PMD Behavior: Fast SOP/PSP/DGD Fluctuations (ms)

0 50 100 150 200 250 300 3500

200

400

600

800

1000

1200

1400

1600

1800Histogramm der maximalen Momentangeschwindigkeit aller Ereignis ; Skalierung 5rad/s

rad/s

Anz

200 rad/s =7,8 ms per 90°

100 rad/s = 15,6 ms per 90°

0 50 100 150 200 250 300 3500

0,0001

0,001

0,01

0,1

Geschw. [rad/s]

PD

F [s/rad]

Ln-Ansicht der PDF der max|S'(t)| pro Ereignis ; Skal. 5 rad/s

§§LongtermLongterm measurements on DT link over 18 monthsmeasurements on DT link over 18 months

§§Further studies followFurther studies follow

§§Critical region: some msCritical region: some ms

§§PMDC should be able to track strong SOP and PSP excursions, as wPMDC should be able to track strong SOP and PSP excursions, as well ell

as its full DGD dynamics within some milliseconds.as its full DGD dynamics within some milliseconds.

§§ In case of lowerIn case of lower--speed tracking, active PMDC can introduce oscillating speed tracking, active PMDC can introduce oscillating

intrinsic element DGD which can lead to complete system outage uintrinsic element DGD which can lead to complete system outage until ntil

retrackingretracking

§§Passive PMD suppression from robust system design, like DQPSK loPassive PMD suppression from robust system design, like DQPSK low w

baudratebaudrate modulation format is a reasonable alternative to actively modulation format is a reasonable alternative to actively

controlled PMDC controlled PMDC

SOP dynamics



OFC Workshop 13

Summary

■ DT’s network equipped with three classes of vintage fibers

■ Problems to install 40 Gbit/s in the backbone network

■ Temporal statistics in DT network has been measured (& is

on further study): high speed requirements for active PMDC

■ Robust low baud rate transmission systems are a key for

logical 40G and 100GEthernet employment in the backbone

(DT=600 km, ULH=1000 km)



OFC Workshop 14

BACKUP



OFC Workshop 15

Percentage of fibers suitable for 2.5 Gb/s, 10 Gb/s and 40 Gb/s over various link length (Additional PMD contribution due to OLAs, DCF’s neglected)

400 km600 km

800 km1000 km

40 Gbit/s

R2

R3

99

,7

99

,4

98

,9

98

,5

91

,6

89

,6

87

,7

86

,8

71

,6

63

,1

58

,7

53

,7

0

20

40

60

80

100

Bit

Rate

40 Gb/s

10 Gb/s

2.5 Gb/s

Pe

rce

nta

ge

[%]

Link Length




Probability Distribution with Strongly Inhomogeneous Birefringence Distribution

homogeneous

mode-coupling :

Maxwellian PDF

of DGD vs. λλλλ

SMF with one highly birefringent

segment: shift and deformation

of PDF of DGD vs. λλλλ

Decrease of σ / τσ / τσ / τσ / τ

σ / τ

σ / τ

σ / τ

σ / τ

PMD of single HiBi-segment

DGD / ps

DGD / ps

PD

F

PD

F



OFC Workshop 17

Fu

nctio

na

lity

Assessment of solutions against PMD (for bitrates of 40Gbit/s and beyond)

Low costHigh cost

Poor

function-

ality

Good

function-

ality

Solutions against PMD

Solution Approaches

n Limit transmission speed per

channel to 10 Gbit/s

n Limit optically transparent

distance by using O/E/O

conversion (transponders)

n Optical or electrical

compensation schemes

n Develop modulation schemes

with less bandwidth

requirements

n Develop codes for correction of

PMD caused errors

n Combine solutions 3, 4, 5

n Install new fibres

7

7

1

1

2

2

5

5

6

6

3

3

4

4

Economics

Performance limitation

Network transformation

Type of approach Robust technology

enhancements

Polarisation-Mode

Dispersion (PMD)

Cost (CapEx)



OFC Workshop 18

n Pure receiver tolerance, NRZ (5 min/year outage) 10%…12% PMD/Tbit*

n 1+1 protection scheme +4%

n RZ instead of NRZ +4% (med.size duty cycle)

n FEC inband (SDH/SONET) +1% (more if noisy)

n FEC outband Reed Solomon(7%OH),ITU-T G.709 7+10%

n Super-FEC / double stage FEC (10dB OSNR gain) 7+15% ca.

(1st-order approximation, only, regarding vertical eye closure due to horizontal eye closure)

n Nonlinear Kerr trapping (PM-XPM) +10%

n 1st-order PMD-compensator +20%

n 1st+higher-order PMDC (diff.No.of degr.of freedom) +25%...+30%...>+30%

n New fibres (LDV>0.1ps/sqrt(km)) replacing ITU-T standard 0.5 ps/sqrt(km)

7++40%** (++ of already obtained)

n DQPSK (and other multi-level formats) ++10% (++ of already obtained)

n Further: FFE+DFE, MLSE, λ-protection/restoration, FEC+Scramblers

Typically Expected Gain fromDifferent PMD Mitigation Approaches

**) the values are obtained by a comparison of relaxed PMD impact at the receiver due

to the use of better fibres with the effective impact mitigation obtained by using,e.g., PMDC.

*) PMD is the average <DGD>, Tbit is the bit period.

For the Rx tolerance limit the system outage (>1 dB BER penalty) of 5 min/year (DT internal standard) was applied. See also ITU-T G.691.

+10% +13% +% depends +??%



Documents

Workshop on 40 Gig Networks: The Actual World PMD Challenge · 3/25/2007 · PMD as Bottleneck for 40Gbit/s Long-Haul Systems Introduction § Detailed survey of all usable fiber