System engineering of alien wavelengths over the SURFnet
network
Roeland Nuijts, SURFnet, roeland.nuijts@surfnet.nl
Customer Empowered Fiber Networks Workshop, Prague, Czech Republic,
September 13th-14th, 2010
Outline
 Introduction
 Alien wavelength concept, advantages and disadvantages
 Alien wavelengths in the SURFnet NGE (Next-Generation Ethernet) project
•
Alien wavelength for metro-connections using small form factor 10Gb/s
DWDM interfaces on the existing SURFnet network
 Alien wavelength for 40Gb/s long-distance SURFnet CBF connections
•
mix 40Gb/s PM-QPSK and 10Gb/s NRZ-OOK on standard SMF (G.652)
with dispersion compensation
 Conclusions
2
Alien wavelength concept
Rx
Tx
Tx
Rx
(a) conventional closed DWDM system
Rx
Tx
Rx
Tx
Tx
Rx
Tx
Rx
(b) multi-domain DWDM systems
Rx
Tx
Tx
Rx
(c) multi-domain DWDM systems with alien wavelength
3
Alien wavelength advantages
• direct connection of customer equipment  cost savings
• avoid OEO regeneration  power savings
• faster time to service  time savings
• support of different modulation formats  extend network lifetime
4
Alien wavelength challenges
• complex end-to-end optical path engineering in terms of linear (i.e.
OSNR, dispersion) and non-linear (FWM, SPM, XPM, Raman)
transmission effects for different modulation formats
• complicated system integration/functional testing
• end-to-end monitoring, fault isolation and resolution
• end-to-end service activation
5
Application of alien wavelengths in the
SURFnet NGE (Next-Generation Ethernet)
project
•
Huge growth in data-oriented services over the past years
•
Push for low-cost, flexible, hence, ethernet connections
•
Until now Ethernet was transported over SDH/SONET
infrastructure as a means to re-use existing infrastructure
•
With demand for more capacity and longer and high-capacity
connections (WAN instead of LAN) there is now need for Carrier
Ethernet
6
•
SURFnet NGE (Next-Generation Ethernet) project
•
SURFnet NGE project re-uses the existing DWDM layer
SURFnet DWDM network
- after Photonic Evolution project 1Q11
• All implemented ROADMs are of type 1x5 WSSes
• Convert remaining fixed OADM nodes to ROADMs
expensive maybe no time next year
• Zwolle, Enschede, Nijmegen, Wageningen, Delft, Utrecht
ready as of September 10th, 2010
• To be done: Asd1&2, remove fixed OADM in Ehv
• Enables:
• All-optical connection between TUD, TUE and TU
(three Universities of Technology)
• All-optical connection between Aachen (antenna
field in Julich and Astron/JIVE in Dwingeloo)
• All-optical pass-through between Amsterdam
locations to close optical DWDM rings
Alien wavelengths in the metro area
DWDM Architecture SURFnet6/7
CIENA OME6500 and CPL
OME6500
10Gb/s WDM transmitter and
receiver
CPL
DSCM (dispersion
Compensation)
CMD
GMD
9
Form factor improvement – 300pin to XFP
Tunable 50GHz channel spacing 10Gb/s DWDM transponder
300pin MSA transponder
Typical power consumption 10W
Footprint: ±100cm2
10
XFP transponder
Typical power consumption 3W
Footprint: ±14cm2
Example: optical specifications of JDSU XFP
Initial XFP exhibits negative chirp!
Transient chirp
g
(ps)
λ (nm)
D
(ps/nm km)
0
0
λ (nm)
 A negative frequency excursion on the rising edge corresponds to a positive wavelength excursion
which means group delay increases hence velocity decreases
 The opposite occurs on the falling edge
 Both result in pulse compression which counteracts pulse broadening by dispersion, hence more reach
(or dispersion tolerance)
 Two methods to get negative chirp, unbalanced drivers or z-cut Mach-Zehnder modulators
Transmission performance versus dispersion
with negative chirp
Optimum dispersion
around +800ps/nm
In each DWDM ring there are paths from each OADM node to Amsterdam1 and Amsterdam2,
tailored to +800ps/nm as close as possible. Consequently, paths between rings can have up
to +1600ps/nm dispersion and paths between OADM nodes less dispersion. Sufficient
system performance for these dispersion values needs to be verified before deciding to use
low-cost 10Gb/s interfaces in the photonic layer for NGE
Optimum dispersion can not always be achieved in systems due to 2 reasons:
• in systems due to “quantization error” of DCF spools (i.e. DCF10, 20, ….)
• wavelength dependence of dispersion in transmission and compensating fibers
Calculate 10Gb/s wavelengths for NGE*
Network diagram
24
1
7
LP+IP traffic
23
Al
21
22
6
5
3
19
2
18
17
8
13
16
9
10
+
Alr01
Amf01
Asd01
Asd02
Bd01
Ddt01
Dgl01
Dt01
Ehv01
Es01
Gn01
Gv01
Ht01
Hvs01
Ledn01
Mt01
Nm01
Rt01
Tb01
Ut01
Wg01
Zl01
r0
1
Am
f0
1
As
As
d0
d0
1
2
14
21
29
Dg
Bd Dd
t0
l0
01
1
1
Eh
Dt
v0
01
1
G
Es
G
v0
01 n0 1
1
Hv Le d
Nm
Ht
n0 Mt0
s0
01
01
1
1
1
Tb
01
Rt
01
1
7
81
45
19
7
80
7
7
8
2
46
9
22
9
9
44
23
38
16
14
28
18
26
14
14
7
12
7
21
7
16
18
21
20
25
1
18
14
4
4
1
22
23
14
7
14
21
1
Zl
15
27
7
3
01
7
14
7
14
1
14
37
35
1
1
1
9
39
2
14
7
21
12
11
Solved for required 10Gb/s wavelength connections and with
minimum number of interfaces
* Joint effort with Anteneh Beshir at the TUD (Delft Technical University)
14
7
14
78
21
14
14
14
W
Ut
01 g 01
7
1
7
23
1
28
2
14
Required 10Gb/s wavelengths for
SURFnet NGE - only LightPath traffic
Wavelengths
15
Nodes
1
2
3
4
5
6
Total
Amsterdam1
4
4
5
5
5
5
28
Amsterdam2
5
5
5
5
5
5
30
Leiden
0
2
0
2
2
2
8
Den Haag
0
2
2
2
0
2
8
Delft
4
4
4
4
4
4
24
Utrecht
0
2
0
2
2
2
8
Hilversum
0
0
0
2
0
2
4
Rotterdam
2
2
0
2
0
2
8
Dordrecht
2
0
0
0
0
2
4
Breda
0
2
2
2
0
2
8
Tilburg
0
0
2
0
0
2
4
Eindhoven
3
3
3
3
3
0
15
Den Bosch
2
2
0
0
0
2
6
Maastricht
0
2
2
0
0
-
4
Nijmegen
3
3
3
3
3
0
15
Wageningen
2
2
2
0
0
2
8
Amersfoort
0
0
0
2
2
2
6
Enschede
0
2
0
0
2
2
6
Zwolle
3
3
4
4
4
4
22
Almere
0
2
0
0
0
2
4
Groningen
-
-
2
2
0
2
6
Dwingeloo
-
-
0
0
2
2
4
Total
30
42
36
40
34
48
230
1. 230 10Gb/s interfaces required
2. 82 different wavelength paths
required  simulated optical
transmission performance of all
82 wavelengths in order to
verify whether these work and
to check whether the first
assessment of the FOM was
correct
Simulation results of transmission performance
- dispersion, received power and OSNR
25
Frequency
20
15
10
Dmin (ps/nm)
Dmax (ps/nm
5
0
Total Dispersion (ps/nm)
25
Frequency
20
15
OSNR
Calculated OSNR was well above ROSNR
(Required OSNR) for each of the 82 paths
10
5
-18.00
-17.500
-17.00
-16.500
-16.00
-15.500
-15.00
-14.500
-14.00
-13.500
-13.00
-12.500
-12.00
-11.500
-11.00
-10.500
-10.00
-9.500
-9.00
0
Measured OSNR of each 10Gb/s wavelength in
the SURFnet network was well above the
ROSNR
Received optical power level (dBm)
16
Required performance can be delivered by the new low-cost 10Gb/s interfaces!
Alien 40Gb/s wavelength transmission on
SURFnet CBF connections
JOINT SURFnet/NORDUnet
40Gb/s PM-QPSK alien
wavelength DEMONSTRATION
40G
10G
W
S
S
Copenhagen
End-to-end FoM = 1400
(a couple of dB margin over BOL OSNR limit - set against nonlinearities and
potentially adverse effect from filter concatenation [4])
416km TWRS
Alcatel-Lucent
(with dispersion compensation)
640km TWRS
Nortel
W
S
S
Hamburg
(without dispersion compensation)
40G
40G
17
Hamburg
W
900GHz
S
S
W
S
S
10G
Amsterdam
40G
10G
40G alien wave
10G
5x10Gb/s @ 50GHz
350GHz
5x10Gb/s @ 100GHz
Alien 40gb/s wavelength on 10Gb/s-5x100km
DWDM system using standard G.652 fiber and DCFs
34% precompensation
95% mid-span
compensation
10G
60% postcompensation
10G
10x10Gb/s NRZ-OOK
40G
10G
W
S
S
W
S
S
40G
10G
1x40Gb/s PM-QPSK
10x10Gb/s NRZ-OOK
5x100km SMF
1
10
40Gb/s
PM-QPSK
50GHz
1
50GHz
 0 channels guard band
18
10
•Fairly unfavorable dispersion map due to
zero-dispersion crossing at every span
and hence high XPM efficiency
40Gb/s
PM-QPSK
1
P
10
50GHz
50GHz
10
ROSNR 40Gb/s (arb. units)
Simulation results 40Gb/s alien wavelength
5.00
4.00
3.00
4DP
2.00
3DP
1.00
2DP
DP
.00
-1.00
.00
 0 channels guard band
.500
1.00
1.500
2.00
2.500
Power per channel 40Gb/s (arb. units)
• The ROSNR of the 40Gb/s alien wavelength increases With increasing power level of
the 10Gb/s NRZ-OOK channels, starting at about 4P, probably due to XPM
• Increasing the power per channels of the 40Gb/s alien wavelength in the range where
we conducted these simulations does not seem to affect (improve) the ROSNR so SPM
(Self-Phase Modulation) does not affect the 40Gb/s alien wavelength
• Best performance when 40Gb/s channel is stronger than 10Gb/s channels
19
Conclusions
-
We have investigated using low-cost 10Gb/s DWDM interfaces for the
SURFnet NGE project by using a heuristic model to determine the
required wavelength topology and a transmission propagation model
to determine the required performance
-
Simulation results show that new low-cost 10Gb/s XFPs deliver
sufficient performance to be used for the NGE project and these
results suggest they can be connected to the existing SURFnet
DWDM layer
-
Preliminary simulation results of 40Gb/s PM-QPSK transmission on
a DWDM system with standard (i.e. G.652 SMF) and DCFs and
equipped with 10Gb/s DWDM signals show that power of the 10Gb/s
channels should be well below the power of the 40Gb/s channel in
order to avoid XPM
20
Acknowledgements
Some of the research leading to these results has received funding from the European
Community’s Seventh Framework Programme (FP7/2007-2013) under grant agreement nº
238875 (GÉANT)
21
Thanks for your attention!
Questions?
roeland.nuijts@surfnet.nl
+31-30-2305 305
22
What limits system performance?
ASE (Amplified Spontaneous Emission)
r() = 2 h n n sp (G() – 1)
-
Amplifiers are used to overcome fiber losses.
Optical Noise is added by each amplifier.
Engineering rules usually defined for equal spans (e.g. 20 x 20dB) which is not
the case in the real fiber networks
23
Slide courtesy of Kim Roberts, Nortel
OSNR (Optical Signal-to-Noise Ratio) - Simple formula
repeater
SMF
OA
OA
NF1
Pin,1
Pin,2
Description
h
Planck's constant (J•s)
c
speed of light (m/s)
R
OSA resolution BW (Hz)
Pin
Input power (W)
Nsp
Noise Figure (linear)
OA
NFN-1
Pin,3
Parameter
DCF
OA
NF3
N
2 h n R N sp , j
1

OSNR j 1
Pin, j
24
SMF
OA
NF2
Pin,N-1
Calculated OSNR (dB)
Tx
SMF
DCF
49
47
45
43
41
39
37
35
33
31
29
27
25
NFN
Pin,N
Rx
ƒ
OSNR
Resolution bandwidth = 0.1nm
0
40
80
120
160
200
240
280
320
360
400
Distance (km)
Simple formula, accurate to within a few tenths of a dB
but sensitive information needs to be provided to fiber
suppliers, which equipment vendors don’t like:
• NF of amplifiers
• launch power per channel
• minimum required OSNR => need simplification
New method to quantify fiber
link quality, FoM (Figure of Merit)
In order to quantify optical link grade, we propose a new method of representing
system quality: the FOM (Figure of Merit) for concatenated fiber spans

FOM
N
 Lj

 10
10



Lj, span losses in dB
N, number of spans
j 1
A
120km
120km
120km
80km
80km
80km
Total 600km
B
C
80km
80km
100km
100km
100km
A
B
C
25
120km
120km
80km
100km
FOM
5504
5504
1897
120km
100km
100km