U.S. patent application number 12/961442 was filed with the patent office on 2011-06-09 for wavelocker for improving laser wavelength accuracy in wdm networks.
This patent application is currently assigned to VELLO SYSTEMS, INC.. Invention is credited to Chris Wilhelm Barnard, Piotr Myslinski.
Application Number | 20110135301 12/961442 |
Document ID | / |
Family ID | 44082121 |
Filed Date | 2011-06-09 |
United States Patent
Application |
20110135301 |
Kind Code |
A1 |
Myslinski; Piotr ; et
al. |
June 9, 2011 |
Wavelocker for Improving Laser Wavelength Accuracy in WDM
Networks
Abstract
The present invention includes novel techniques, apparatus, and
systems for optical WDM communications. Various wavelocker
apparatus and methods are disclosed that measure the frequency
offsets between signal lasers and reference lasers. The measured
offsets are used to adjust the signal laser frequencies to meet
their target frequencies. The absolute accuracy of the reference
laser frequency is improved by measuring the absorption of the
reference laser by a gas cell with known fixed absorption lines
versus the reference laser frequency. Apparatus and methods are
disclosed to cover scenarios in which the reference laser
polarization is aligned with the signal lasers, as well as those in
which the reference laser polarization is not aligned with the
signal lasers. The wavelocker apparatus may or may not be located
at the same network site as the signal lasers.
Inventors: |
Myslinski; Piotr; (Fremont,
CA) ; Barnard; Chris Wilhelm; (Sunnyvale,
CA) |
Assignee: |
VELLO SYSTEMS, INC.
Menlo Park
CA
|
Family ID: |
44082121 |
Appl. No.: |
12/961442 |
Filed: |
December 6, 2010 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
61267786 |
Dec 8, 2009 |
|
|
|
Current U.S.
Class: |
398/34 |
Current CPC
Class: |
H04J 14/0201 20130101;
H04J 14/0293 20130101; H04B 10/0793 20130101; H04J 14/0287
20130101; H04J 14/02 20130101; H04Q 2011/0081 20130101; H04L 45/28
20130101; H04B 10/572 20130101; H04L 45/50 20130101; H04Q 11/0066
20130101 |
Class at
Publication: |
398/34 |
International
Class: |
H04B 10/08 20060101
H04B010/08 |
Claims
1. A method for improving the accuracy of WDM laser frequencies,
the method comprising the following steps: (a) defining a desired
target frequency to which a signal laser is to be tuned, within a
predefined acceptable range of system tolerance; (b) tuning a
signal laser to an initial frequency which may or may not fall
within the predefined acceptable range of the target frequency; (c)
beating the signal laser with a reference laser by combining and
transmitting onto an optical detector, the light emitted by the
signal laser at the initial frequency with the light emitted by the
reference laser at a reference frequency, thereby creating a beat
signal having a beat frequency equal to the difference between the
signal laser frequency and the reference laser frequency; (d)
scanning the reference laser across a range of frequencies offset
from the target frequency by predetermined amounts, and measuring
the amplitude of the beat signal at a plurality of points within
the range; (e) processing the measured amplitudes of the beat
signal to determine the actual frequency of the signal laser, and
calculating the difference between the actual frequency and the
target frequency; (f) adjusting the frequency of the signal laser
to account for the calculated difference; and (g) repeating steps
(c), (d), (e) and (f) to keep the signal laser within the
predefined acceptable range of the target frequency.
2. The method of claim 1 wherein the frequency of the reference
laser is calibrated by measuring the absorption of the reference
laser light through a gas cell with known absorption peaks at
precise absolute frequencies.
3. The method of claim 1 wherein the light emitted by the reference
laser and the light emitted by the signal laser are linearly
polarized along the same axis with polarization-maintaining
optics.
4. The method of claim 3 wherein a polarization controller is used
to align the polarization of the light emitted by the reference
laser with the polarization of the light emitted by the signal
laser.
5. The method of claim 1 wherein the polarization of the signal
laser and the polarization of the reference laser are not
constrained to be aligned, and the polarization of either the
signal laser or reference laser is scrambled by a polarization
scrambler.
6. The method of claim 5 wherein the beat signal is split by a
polarizing beam splitter into two orthogonally polarized signals
that are detected by two separate optical detectors, and wherein
processing circuitry extracts a corresponding beat signal.
7. The method of claim 1 wherein the desired target frequency is a
subchannel frequency offset from an ITU channel frequency.
8. The method of claim 1 wherein the reference laser is used to
monitor and control the frequencies of a plurality of signal
lasers.
9. The method of claim 8 wherein the reference laser is at a
different location from at least one of the plurality of signal
lasers, and wherein the reference laser communicates with any such
signal laser via a network overhead channel.
10. A system for improving the accuracy of WDM laser frequencies,
the system comprising: (a) a signal laser to be tuned to a desired
target frequency, within a predefined acceptable range of system
tolerance, wherein the signal laser is tuned to an initial
frequency which may or may not fall within the predefined
acceptable range of the target frequency; (b) a coupler that can
combine, and transmit onto an optical detector, the light emitted
by the signal laser at the initial frequency with the light emitted
by a reference laser at a reference frequency, thereby creating a
beat signal having a beat frequency equal to the difference between
the signal laser frequency and the reference laser frequency; and
(c) control circuitry that can: (i) scan the reference laser across
a range of frequencies offset from the target frequency by
predetermined amounts; (ii) measure the amplitude of the beat
signal at a plurality of points within the range; (iii) process the
measured amplitudes of the beat signal to determine the actual
frequency of the signal laser; (iv) calculate the difference
between the actual frequency and the target frequency; (v) adjust
the frequency of the signal laser to account for the calculated
difference; and (vi) repeat steps (i)-(v) to keep the signal laser
within the predefined acceptable range of the target frequency.
11. The system of claim 10 wherein a coupler directs a fraction of
the reference laser light to pass through a gas cell, the gas cell
having known absorption peaks at precise absolute frequencies, and
wherein an optical detector measures the absorption of the
reference laser light by the gas in the gas cell.
12. The system of claim 10 wherein a polarization-maintaining
coupler and polarization-maintaining fibers are used to keep the
polarization of the light from the reference laser aligned with the
polarization of the light from the signal laser.
13. The system of claim 10 wherein a polarization controller with
feedback circuitry is used to align the polarization of the light
emitted by the reference laser with the polarization of the light
emitted by the signal laser.
14. The system of claim 10 wherein the polarization of either the
reference laser or the signal laser is scrambled by a polarization
scrambler.
15. The system of claim 14 wherein the polarization of the light
from the reference laser is linear, the polarization of the light
from the signal laser changes randomly, and the beat signal is
split by a polarizing beam splitter into two orthogonally polarized
signals that are detected by two separate optical detectors, and
wherein processing circuitry extracts a corresponding beat
signal.
16. The system of claim 10 wherein the desired target frequency is
a subchannel frequency offset from an ITU channel frequency.
17. The system of claim 10 wherein the reference laser is used to
monitor and control the frequencies of a plurality of signal
lasers.
18. The system of claim 17 wherein the reference laser is at a
different location from at least one of the plurality of signal
lasers, and wherein the reference laser communicates with any such
signal laser via a network overhead channel.
Description
CROSS REFERENCE TO RELATED APPLICATIONS
[0001] This application claims the benefit, pursuant to 35 U.S.C.
.sctn.119(e), of U.S. Provisional Patent Application No.
61/267,786, filed Dec. 8, 2009, entitled "Subchannel Photonic
Routing, Switching and Protection with Simplified Upgrades of WDM
Optical Networks," which is hereby incorporated by reference in its
entirety.
I. BACKGROUND
[0002] A. Field of Art
[0003] This application relates generally to optical communications
based on optical wavelength-division multiplexing (WDM), and in
particular to subchannel routing, switching, and protection, along
with related techniques that facilitate network upgrades and reuse
of legacy equipment.
[0004] B. Description of Related Art
[0005] 1. Overview
[0006] Optical WDM communication systems transmit multiple optical
channels at different WDM carrier wavelengths through a single
fiber. The infrastructures of many deployed optical fiber networks
today are based on 10 Gb/s per channel. As the demand for higher
transmission speeds increases, there is a need for optical networks
at 40 Gb/s, 100 Gb/s or higher speeds per channel.
[0007] Moreover, there is a need to leverage this higher bandwidth
to realize greater flexibility in routing client signals among
network nodes. For example, increasing the bandwidth of a fiber
channel from 10 Gb/s to 40 Gb/s might enable 4.times.10 Gb/s client
circuits to occupy a channel between two network nodes previously
dedicated to a single 10 Gb/s client circuit. Yet, unless an entire
channel is free to enable all four client circuits to continue
propagating together to a subsequent node on the network, the
desired routing of these four client circuits may not be achievable
without some mechanism for dynamically rerouting individual client
circuits, independent of one another, across different fiber
channels.
[0008] As will become apparent below, there is a need not only for
increased bandwidth, but for sufficient flexibility to divide
and/or combine individual client circuits to achieve desired
routing, switching, concatenation and protection capabilities. Such
flexibility is needed to fully realize the benefit of increasing
the number of available optical circuits in a single fiber.
[0009] 2. Single-Wavelength Optical Networks
[0010] Optical fiber has been used as a communication means since
about 1977. Over time, deployed baud rates on a single laser have
increased from 45 MB/s to over 40 Gb/s. Various protocols have been
transmitted across optical fiber, including SONET [GR-253] and
Gigabit Ethernet [IEEE Standard 802.3ae].
[0011] FIG. 1A shows a deployed network 100 that uses OC-48 SONET
add-drop multiplexers 120 at each node, interconnected by a first
fiber optic cable 125 for signals traveling in a clockwise
direction, and a second fiber optic cable 135 for signals
travelling in a counterclockwise direction. At each node (or
network add/drop site) lower-rate client traffic 110 can be added
or dropped, or passed through that node. SONET mappers are used to
map the traffic to the STS-1 virtual containers [described in
Telcordia Standard GR-253], and SONET multiplexers are used to
direct the traffic to the add, drop, or passthrough ports. A pair
of multiplexers can be used on two separate line cards as shown to
provide support for a Unidirectional Path Switched Ring (UPSR), or
a 2-fiber or 4-fiber Bidirectional Line-Switched Ring (BLSR).
[GR-1230 Telcordia Standard describes the SONET BLSR]. The traffic
from a SONET ADM can also be combined with other traffic using
wave-division multiplexing (WDM) to increase the network
capacity.
[0012] FIG. 1B shows a deployed network 150 that uses Gigabit
Ethernet switches 170 at each node, interconnected by a first fiber
optic cable 175 for signals traveling in a clockwise direction, and
a second fiber optic cable 185 for signals travelling in a
counterclockwise direction. At each node incoming Gigabit Ethernet
traffic 160 is mapped to VLANs that are transmitted on the 10 GE
line side. At each node traffic in each VLAN is selected to be
added or dropped, or passed through that node. When GE networks are
deployed in a ring, the standard protocols of STP (Spanning-Tree
Protocol) and RPR (Resilient Packet Ring) can be used to provide
protection. The traffic from a 10 GE switch can also be combined
with other traffic using WDM.
[0013] 3. WDM Networks with Muxponders and Transponders
[0014] Later generations of optical fiber communication systems use
optical amplifiers to increase span and repeater distances and
wavelength-division multiplexing to increase the link capacity or
aggregate bandwidth. WDM networks transmit client traffic from
multiple sources over an optical fiber network. The traffic is
multiplexed on the fiber by transmitting each signal with a laser
set at a different channel on the International Telecommunication
Union (ITU) channel plan defined in Standard G.692. Optical filters
designed to function according to the ITU channel plan are used to
demultiplex the signals and thereby direct each signal to its
designated receiver. These standard ITU channels are hereinafter
referred to simply as "channels."
[0015] Optical signals are transmitted using transponders or
muxponders, and are demultiplexed with fixed optical add-drop
multiplexers (FOADMs), reconfigurable optical add-drop multiplexers
(ROADMs), and/or wavelength selective switches (WSS).
[0016] FIG. 2 shows a currently deployed WDM transponder 200.
Client traffic 210 is connected via a short-reach fiber interface
to client transceivers 215. These are typically pluggable devices
such as an XFP [MSA standard
http://www.xfpmsa.org/cgi-bin/msa.cgi]. After the optical signal is
converted to an equivalent electrical signal (utilizing clock
recovery circuitry 218), it can be processed digitally to
optionally (1) extract performance monitoring information 220, (2)
add channel overhead for remote network management 225, and (3)
encode the data for forward error correction 227. The signal is
then used to modulate light from a fixed or tunable laser on the
WDM grid. The output 230 from the transmitter 229 is then launched
onto the transmission fiber. The transmitted light signal can be
combined with light signals from other WDM transponders on a single
fiber with an optical multiplexer.
[0017] At the receive side of the link, an optical demultiplexer is
used to separate the WDM signals 235 (on the incoming fiber), which
are then converted back into equivalent electrical signals by the
receive circuitry 237 in the transponder. Note that this
transponder requires external means to select the particular
wavelength that is being dropped, though this filter function can
be integrated onto the transponder line card [see, eg, U.S. Pat.
No. 6,525,857]. The electrical signal from the line receiver
(utilizing clock recovery circuitry 239) can be processed digitally
to optionally (1) extract performance monitoring information 241,
(2) drop the channel overhead for remote network management 225,
and (3) correct errors according to the Forward Error Correction
(FEC) algorithm 243. The signal 240 is then returned to the client
equipment via the client-side transceivers 215. As alluded to
above, transponders may utilize clock recovery circuitry 239 to
support different data rates and protocols.
[0018] Typically, the line side optics are designed to operate at
2.7 Gb/s, 10.7-11 Gb/s, or 43 Gb/s with the cost of the components
increasing with bit rate. The line receiver 237 is either a PIN
photodiode or avalanche photodiode. In either case the receiver is
not wavelength specific, so that an optical demultiplexer, or ITU
channel filter, must be placed in front of the receiver to filter
out the designated channel.
[0019] It should also be noted that control plane circuitry and
software 250 is employed to facilitate various transmit and receive
functions of DWDM transponder 200, such as remote network
management 225 (e.g., via the addition or removal of channel
overhead) and the extraction of performance monitoring information
245. In addition, control plane 250 is employed for configuration
of transmission protocols 255 (in concert with clock recovery
circuitry 218) and laser wavelengths 265 (to tune channels via
transmitter 229). Finally, it can detect and handle faults
involving the reception of both client-side (267a) and line-side
(267b) signals.
[0020] FIG. 3 shows a currently deployed WDM muxponder 300. This
module maps lower-rate traffic 310 using a SONET multiplexer
[GR-253], OTN (Optical Transport Network) multiplexer [based on ITU
standard G.709], Ethernet switch, or proprietary digital mapping
and multiplexing 320. The multiplexing may be done with a
commercially available or custom-designed ASIC, or a
custom-designed FPGA. The muxponder 300 has line-side WDM optics
similar to the transponder 200 with a laser (in transmitter 329)
set to a designated channel on the ITU grid and a receiver 337 that
can detect any signal within the ITU channel plan.
[0021] Although the transponder 200 and muxponder 300 can be
designed to transmit signals from different sources and with
different bit rates, the hardware limitations and costs typically
limit the implementation to a specific set of protocols. For
example, a 10 Gb/s transponder may transmit OC-192 or STM-64
signals at 9.95 Gb/s, 10 GbE signals at 10.3125 Gb/s, FC-10 signals
at 10.5 Gb/s, and OTU signals at 10.7 Gb/s. But it may not transmit
data at significantly different data rates such as 2.5 Gb/s or 1.25
Gb/s. This may be a limit of the clock-recovery circuits used,
SERDES (serializer-deserializer) circuits, or the ASIC or FPGA used
to perform the performance monitoring and FEC functions. Similarly,
a muxponder typically supports a subset of data rates and protocols
that are determined by the capabilities of the digital and analog
electronic circuits. The maximum data rate supported by the
transponder and muxponder is typically limited by the analog
circuits on the line side, such as the optical modulator (or
bandwidth of the laser if direct modulation is being used), the
bandwidth of the optical receiver, and the bandwidth of the
transimpedance amplifier used at the receiver.
[0022] WDM network installations have been a compromise between
price and functionality. The cost of the high-speed optics
increases with the line bit rate so that vendors typically
partition their products into different data rates such as 2.5
Gb/s, 10 Gb/s, 40 Gb/s, and 100 Gb/s. The price of WDM ports
dictates that networks be deployed with as much bandwidth per port
as possible. However, this has been offset by transponder prices
increasing with bandwidth (e.g. 40G) so that most WDM lambdas have
bandwidth assignments that are "right sized."
[0023] 4. WDM Channel Plans
[0024] WDM network channel plans usually use a subset of the
wavelengths on the International Telecommunication Union,
Telecommunication Sector (ITU-T) grid. Reference Standard G.692,
which specifies a frequency grid anchored at 192.1 THz with
interchannel spacings at integer multiples of 50 GHz and 100 GHz,
is specified as the basis for selecting channel central
frequencies. For purposes of illustration, the ITU channels will be
referenced as I-210 for 192.1 THz, I-215 for 192.15 THz, etc.
[0025] The number of channels on the ITU grid is limited in most
applications to the gain range available from erbium-doped fiber
amplifiers (EDFAs). Gain-flattened EDFAs are now commercially
available for either the C band (.about.191.8 THz to 196.3 GHz) and
L band (.about.186.9 to 191.4 GHz). Currently a fully-loaded WDM
network can transmit approximately 160 channels-80 channels in the
C band spaced at 50 GHz and 80 channels in the L band spaced at 50
GHz.
[0026] 5. Point-to-Point WDM Links
[0027] FIG. 4 shows a simplified diagram of a point-to-point WDM
network 400 interconnecting two nodes--East Node 410 and West Node
420. Typically two fibers are used--one to transmit from east to
west 402 and one to transmit from west to east 404--but a single
fiber can also be used. Signals from different WDM lasers are
combined via WDM Combiner 415 that can be wavelength-dependent
(using ITU channel filters) or wavelength-independent (using a 1:N
optical splitter).
[0028] A 1:N optical splitter can be based on fused fiber couplers.
It has N input ports and one output port so that optical signals
connected to the input ports are combined in the output fiber with
a nominal power loss of 10*log 10(N) dB for each signal. At the
receive side the signals are demultiplexed via WDM Demultiplexer
417 using optical filters such as thin-film filters.
[0029] 6. WDM Ring Networks
[0030] WDM signals can be transmitted over other topologies, such
as a ring 500 shown in FIG. 5. In this example, and several of the
examples to follow, a single wavelength at each node has been used
to simplify the example. This does not preclude the generalized
case of an arbitrary number of wavelengths at each node. The ring
topology offers the advantage of having two diverse routes between
any nodes on a ring so that any failure on one side of the ring can
be protected with redundant traffic on the other side of the ring.
WDM equipment networks also support channel passthrough at a
node--if an optical filter is not used to drop a particular signal
at that node, then the signal continues around the ring to the next
node.
[0031] Optical filters may be configured to selectively drop
channels at a node. In this case the dropped wavelengths can be
reused for other signals on the next ring segment. This filter
configuration is shown in FIG. 5 where all signals on ring 500 are
directed to a filter (or plurality of filters 515, 525, 535 and
545) to select the dropped channels. Currently deployed WDM
networks route signals using fixed or reconfigurable optical
add-drop multiplexers. In this example, Node 1 510 is configured to
drop I-200, Node 2 520 is configured to drop I-210, Node 3 530 is
configured to drop I-220, and Node 4 540 is configured to drop
I-230.
[0032] WDM signals may also be transmitted on a ring in a broadcast
and select configuration [see, eg, U.S. Pat. No. 7,120,359]. In
this configuration shown in FIG. 6, a portion of the powers of all
signals is tapped off at a node and directed to a filter (or
plurality of filters 615, 625, 635 and 645), to select the dropped
channels. This implementation uses a wideband tap coupler (616,
626, 636 and 646) that directs a fixed fraction of all signals to a
drop port. In this case all signals continue around the ring 600 so
that the dropped wavelength cannot be reused since it would have
interference from the passthrough portion of the signal entering
the node. Furthermore, the return path of the signal on the
protected side of the ring requires a separate wavelength to avoid
interference.
[0033] Although the broadcast and select configuration does not
support channel re-use, it has the advantage that it supports drop
and continue traffic, i.e. traffic that is dropped at more than one
node. It also has the advantage that once the tap coupler is in
place, adding a filter to drop more channels does not interrupt the
passthrough channels. To date, broadcast and select architectures
have been limited by the number of channels supported by the
optical demultiplexers.
[0034] Note that in either configuration of FIG. 5 or FIG. 6, the
drop filter may not provide enough isolation on the passthrough
path. In that case, the drop filters can be cascaded to improve the
isolation. Further note that the diagrams only show one channel
dropped at each node. Typical installations cascade filters at each
node, or use a parallel filter, to drop more than one channel at
each node.
[0035] Another advantage of the broadcast and select architecture
is that it does not reduce the bandwidth available on the line
fiber. All optical filters have a useable passband less than ideal
because of the finite slope of the filter passband. The useable
bandwidth of cascaded filters decreases as more filters are
inserted in the signal path. The graph 700 in FIG. 7 shows the
bandwidth narrowing effect for the case where commercially
available WSS filters are cascaded in a network. Even though this
WSS is designed for 100 GHz ITU channels, it has a useable passband
of 68 GHz. Architectures that minimize the filter insertion in the
optical path therefore have a higher available cumulative
bandwidth.
[0036] WDM equipment is typically deployed in an equipment shelf
that separates the add/drop functionality from the transponders or
muxponders. This offers the service provider the benefits of paying
as they grow, especially since the major expense can be the
transponders and muxponders. This partitioning of WDM equipment 800
is shown in FIG. 8. A practical implementation would use optical
fiber patchcords (802a and 802b, and 804a and 804b) to connect the
discrete modules (add/drop modules 810 and 820, and
transponders/muxponders 830 and 840), but optical backplanes have
also been deployed.
[0037] Note that in FIG. 8 filters 815 and 825 are installed with
larger channel counts than are being used. Over time channels can
be added to the unused filter ports without interrupting the live
services. Because of human error in manually adding the fiber
patchcords (802a and 802b, and 804a and 804b) between the modules
(add/drop modules 810 and 820, and transponders/muxponders 830 and
840), this partitioning can lead to misconnections of the fiber
patchcords. Instead of properly connecting module 810 to
transponder 830 (via fiber patchcords 802a and 802b), as shown in
FIG. 8, these fiber patchcords could mistakenly be misconnected as
shown in WDM equipment 900 in FIG. 9--e.g., connecting
corresponding module 910 to transponder 940 (instead of transponder
930) via corresponding fiber patchcords 902a and 902b. Similarly,
module 920 is mistakenly connected to transponder 930 (instead of
transponder 940) via fiber patchcords 904a and 904b. These mistaken
connections may be difficult to detect, especially if there are two
redundant paths between the terminals.
[0038] Tracing optical connections can be difficult in this
scenario because the multiplexers, amplifiers, and other equipment
may not have means to independently detect each incident channel.
U.S. Pat. No. 5,513,029, however, discloses a method whereby an
optical signal is modulated with a low frequency dither signal to
provide a modulated optical signal having a known modulation depth.
A portion of the optical signal is tapped, and both a total power
and a dither amplitude of the tapped portion of the optical signal
can be measured within the network to provide power levels of the
signals. But this method requires dedicated hardware at all
monitoring points and it cannot detect third-party signals (i.e.,
"alien" signals that are generated by other equipment vendors, but
which may be inserted into a WDM network as long as they are on the
same ITU channel plan and do not interfere with other signals).
[0039] Another method that can be used to monitor signals in the
network is to deploy optical spectrum analyzers at various points
in a network. These can be accessed through the network management
software. However, getting a full view of the network may require
many of these and thus add considerable cost [see, eg, U.S. Pat.
No. 7,002,697]. So there remains a need to improve the end-to-end
visibility of signals in a multichannel optical network.
[0040] 7. Link Budget Rules
[0041] To maintain signal integrity and guarantee a high quality of
service, optical networks transmitting digital signals must
maintain a minimum bit error rate (BER). Well-known factors
affecting the WDM signal BER are received power levels, optical
signal-noise ratio (OSNR), chromatic dispersion (CD),
polarization-mode dispersion (PMD), and nonlinear fiber penalties
such as cross-phase modulation and four-wave mixing. Network design
rules determine the link budget (maximum distance and loss per
fiber span) based on these penalties.
[0042] Guaranteeing the performance and determining the link budget
for an installed network can be costly because determining the
factors listed above may require expensive test equipment.
Furthermore, the tests may have to be run while the network is out
of service so that changes over time after a network is installed
cannot be detected. There is therefore a need to measure the
optical parameters for an installed network, with minimal service
disruption, and minimal extra cost.
[0043] 8. Optical Protection
[0044] Optical networks often require protection against equipment
failures or fiber cuts. One good way of protecting traffic is to
provide two or more redundant paths between the end points with a
protection algorithm that selects traffic from one of the redundant
paths. Standard protection algorithms are the SONET Unidirectional
Path Switched Ring (UPSR) and Bidirectional Line-Switched Ring
(BLSR). The UPSR provides dedicated protection (each working
circuit has a protection circuit of equal bandwidth) and the BLSR
provides shared protection (the protection bandwidth equals the
total working bandwidth in a single fiber).
[0045] Dedicated and shared protection both require means to (1)
bridge traffic at the transmit end onto the redundant paths, and
(2) select traffic at the receiver from one of the redundant
paths.
[0046] Shared protection also requires a means to manage the
passthrough traffic at the intermediate nodes. Examples of shared
protection schemes can be found in U.S. Pat. Nos. 7,046,619 and
7,499,647, and U.S. Pat. App. No. 2007/0025729.
[0047] Various forms of optical protection have been proposed and
implemented, with the most common being a dedicated 1+1 protection
with a 1.times.2 optical switch in front of the receiver. Although
shared optical protection offers significant bandwidth savings, its
practicality is limited by the requirement of maintaining link
budget rules for all possible protection paths.
[0048] 9. Larger WDM Networks and WSS
[0049] FIG. 10 shows a typical network layout 1000 of a service
provider. The ring topology is commonly used in WDM networks
because it provides the lowest-cost means of offering protected
services. A ring network requires that all nodes have at least two
connections to separate neighboring nodes. Ring networks may have
spurs hanging off them to serve a small number of isolated nodes
that have only a single fiber span connected to another node. Ring
networks may be interconnected as shown in FIG. 10 with traffic
passing between the rings at one or more intersecting nodes (nodes
A and B). Many deployed networks with similar layouts need to pass
traffic from a spur to a node on the same ring (node C to D), from
a spur to a node on a different ring (node C to E), and between
nodes on different rings (node D to E). A 1.times.N
Wavelength-Selective Switch (WSS) can be used to direct traffic
between N nodes with direct optical connections [see, eg, U.S. Pat.
No. 7,492,986].
[0050] WSS-based filters are currently much more expensive than
fixed filters. Ring interconnections can also be done with fixed
optical filters, but those connections cannot be reconfigured
remotely, so that network upgrades require technicians to go to the
interconnecting sites and manually reconfigure the connections on
the fixed filters.
[0051] Furthermore, many deployed optical networks have difficulty
managing WDM traffic through on these paths so that the traffic may
be segmented by electro-optical conversions at the intersecting
nodes (A, B, F). These electro-optical conversions add cost and
complexity to the network while reducing reliability. However,
all-optical routing between rings and from spurs to rings requires
that the network be engineered so that the link budget rules are
met by the intra-ring signals, and that individual powers be
managed at the intersecting nodes.
[0052] 10. Subchannel Multiplexing
[0053] Various forms of subchannel modulation have been proposed as
a means to reduce the dispersion penalties associated with high bit
rate transmission in optical fibers (see, eg, WO 2009/105281) and
increase spectral efficiency (see, eg, U.S. Pat. No. 6,525,857).
These "subchannels" (eg, subchannels of ITU channels) are typically
generated by microwave modulators or comb generators with a single
laser. Examples of optical comb generators are described in U.S.
patent application Ser. No. 12/175,439, entitled "Optical
Wavelength-Division-Multiplexed (WDM) Comb Generator Using a Single
Laser" and filed on Jul. 17, 2008, which is incorporated by
reference herein. These subchannels are closely spaced relative to
the source laser and are not independently tunable across a wide
wavelength range, i.e. they are tuned in parallel as the source
laser is tuned. Although an embodiment of one of the previously
referenced patent applications (WO 2009/105281) proposes the use of
more than one laser to generate the subchannels, such lasers are
constrained to operate in parallel within a single ITU G.692
window.
[0054] Lower-rate subcarriers support a simplified upgrade of an
installed DWDM network. For example, a legacy 2.5 Gb/s network may
have transmitters with a reach of 600 km. When that network is
upgraded to 10 Gb/s, dispersion compensators may have to be
installed, since the reach of the 10 Gb/s transmitter may be only
80 km. Installing dispersion compensation and amplifiers to
compensate for their loss can be very disruptive since operators
may have to break the traffic multiple times and at multiple sites.
If four subcarriers are used instead, with each subcarrier
transmitting at 2.5 Gb/s to get 10 Gb/s composite bandwidth, they
can have comparable dispersion-limited reach to the installed 2.5
Gb/s channels. The use of subcarriers therefore provides system
operators with a means of upgrading an installed WDM network to
increase the network capacity without having to change the
dispersion map.
[0055] There is thus a need for an improved implementation of
subchannels (eg, using independently tunable lasers to generate
independent subcarrier frequencies) that will not only increase
bandwidth and spectral efficiency by enabling multiple client
circuits to be assigned to respective subchannels of a single ITU
channel, but will also allow those client circuits to be divided
and/or combined with one another and assigned independently to
subchannels within and across ITU channels. Such flexibility is
needed, as noted above, to achieve desired routing, switching,
concatenation and protection capabilities, and thus fully realize
the benefit of increasing the number of available optical circuits
in a single fiber.
[0056] 11. Network Upgrades
[0057] Even with the ability to upgrade the capacity without
installing additional dispersion compensators, adding or removing
channels from a DWDM network can be disruptive to the live traffic
because the channels can propagate through shared components such
as amplifiers and attenuators that act upon the total power. For
example, if an attenuator output is being controlled to a certain
output power, doubling the channel count will cause the power per
channel to be cut in half. This drop in power could cause bit
errors. System operators have a need therefore for control (eg, via
software) over channel changes in a WDM network in a manner that is
minimally disruptive to the live channels.
[0058] 12. Management Cards
[0059] WDM network equipment (e.g., equipment 1100 shown in FIG.
11) is typically installed in a shelf 1110 with one or two
management cards 1120 (MGT) and various line cards 1125. The
equipment 1100 is typically managed with a client-server element
management system (EMS) consisting of one or more clients, such as
client 1130, and EMS Server 1140. The EMS connects through a
private or public IP network (via Router 1150) to the management
cards 1120.
[0060] FIG. 12 illustrates how two management cards 1220a and 1220b
in equipment shelf 1200 can be deployed in an active/standby
configuration to improve network robustness. The standby MGT 1220b
takes over the management function if there is any hardware or
software failure on the active MGT 1220a. This configuration
typically uses two ethernet planes (1235 and 1245) on the backplane
so that any line card can communicate with either management card.
A handshaking protocol between the management cards is used to
determine which is the active MGT at any given time. On each line
card there is a switch to select which ethernet bus is used for
communications.
[0061] This configuration requires control of the software versions
running on the MGT microprocessors. They run the same version to
ensure compatibility in the event of a switchover from active to
standby. The configuration and status databases on the operative
MGT are constantly backed up on the backup MGT so that when a
failure occurs the backup MGT can take over the management as
quickly as possible, and without any service interruptions.
[0062] 13. OSC Options and Routing Protocols
[0063] WDM equipment typically requires that the EMS have a
management connection to all remote nodes for functions such as
provisioning equipment, reporting faults, downloading software
upgrades, and retrieving and reporting performance metrics. The MGT
also employs a management connection to remote nodes for end-to-end
provisioning, controlling protection switching, and reporting
remote performance and faults. For these functions, current WDM
equipment deploys an optical service channel (OSC) that is outside
of the ITU-T G.692 spectral window, i.e. at 1510 nm or 1620 nm.
[0064] Control messages and status can be transmitted from the MGT
card to the OSC card over the backplane, and then transmitted
optically by the OSC to the remote node where it is routed to the
remote MGT card over the remote backplane.
[0065] Adding the filters to add and drop the OSC channel add loss
and cost to the network. The OSC can be eliminated if channel
overhead is inserted into the signals, but the typical channel
overhead bandwidth (500 kb/s) is much lower than the typical OSC
channel bandwidth (100 Mb/s). There is therefore a need for
improved in-band communications channels that provide the necessary
bandwidth without adding cost.
[0066] 14. Optical Switches Interconnected with WDM Links
[0067] Switching matrices are used in a telecommunications network
to direct traffic from multiple inputs to multiple outputs. An
electrical crossbar switch has a matrix of switches between the
inputs and the outputs. If the switch has M inputs and N outputs,
then a crossbar has a matrix with M.times.N cross-points or places
where the "bars" cross. A given crossbar is a single layer,
non-blocking switch. Collections of crossbars can be used to
implement multiple layer switches. A Clos network is a kind of
multistage switching network, first formalized by Charles Clos in
1953 [see, eg, Charles Clos (March 1953), "A study of non-blocking
switching networks," `Bell System Technical Journal` 32 (5):
406-424]. The Clos network provides a practical multi-stage
switching system that is not limited by the size of the largest
feasible single crossbar switch. The key advantage of Clos networks
is that the number of crosspoints (which make up each crossbar
switch) required can be much fewer than if the entire switching
system were implemented with one large crossbar switch. Although
VLSI technology has enabled very large switching matrices in
electronics [see, eg, U.S. Pat. No. 6,714,537], the switch size is
still limited at very high bandwidths.
[0068] WDM links can be used to interconnect large electro-optic
switches, as illustrated in FIG. 13. Optical crossconnect switches
based on MEMS [see, eg, U.S. Pat. No. 6,574,386] have also provided
a means of switching at the optical layer, but these switches may
need wavelength demultiplexers to switch individual wavelengths.
Large crossconnect switches 1310 provide the connectivity required
to support large traffic demands and WDM links 1320 provide the
bandwidth between the switches.
[0069] Note that this architecture shown in ring 1300 can be costly
because O-E-O conversions may be required at each switch and
bandwidth is being used to send traffic to and from the centralized
switches. Also, the cost of such switches increases with the number
of ports and bandwidth per port so that a network based on switches
that support traffic bandwidth >1 Tb/s combined with high
bandwidth WDM links can have a very high cost. Furthermore, an
all-optical switch can have high loss, so that it requires
expensive optical amplifiers to compensate for the loss. There is
therefore a need for an optical network architecture that supports
many high-bandwidth inputs and outputs (>500) with non-blocking
switching and minimal O-E-O conversion for the switching.
[0070] 15. Network Management and Management Sublayers
[0071] Network functionality can be described by the 7-layer OSI
model. Optical networking equipment resides mainly at the lowest
layer, the Physical Layer. For the purposes of describing WDM
networks in general and the current invention in particular, the
Physical Layer can be divided into sublayers 1400 as shown in FIG.
14.
[0072] Except for the wavelength assignment and detection 1431, all
of the sublayers shown are optional. For example, transponders do
not necessarily provide electrical mapping, multiplexing, or
protection switching.
[0073] The electrical sublayers 1420 include: [0074] The mapping
sublayer, where client data is received and mapped to available
bandwidth according to the mapping protocol used. [0075] The
multiplexing sublayer, where electrical data is selectively added,
dropped, or passed through. [0076] The protection switching
sublayer, which can provide protocol-based protection, e.g. UPSR or
BLSR protection for SONET-mapped signals, or STP or RPR protection
for ethernet-mapped signals. [0077] The next 2 sublayers are
typically implemented according to the ITU G.709 standard that
defines OTN frame formats. Path trace and CRC checks can be
inserted into the OTN frame for receive side monitoring of the
signal source and signal quality respectively. [0078] The lowest
electrical sublayer provides forward-error correction encoding and
correction.
[0079] Of the optical sublayers 1430, the highest sublayer 1431,
maps the signal from the client onto a specific wavelength that is
routed over the network by fixed or tunable optical filters. The
optical protection layer provides redundant optical paths from the
source to destination and a means for bridging the traffic onto the
redundant paths and selecting the received signal from one of the
redundant paths according to alarms and signaling in the network.
The lowest optical sublayer provides multiple point-to-point
connection between two points according to the provisions in the
higher layers.
[0080] Managing a WDM network requires that the network management
system (NMS) have a management link 1440 from the NMS server to all
of the optical network elements. The network connections can be
provided by an external IP network, or with dedicated overhead
channels that are provisioned on the optical network. The overhead
channel may be mapped directly to one of the deployed wavelengths,
or it may be transmitted over the OTN overhead channel, e.g. GCC0
in G.709, or in an unused section of the higher-layer protocol's
overhead channel.
[0081] Software on the WDM equipment is required to configure,
monitor, maintain, and report on all of the functions shown in FIG.
14. Adding new optical functionality requires adding new management
software at the appropriate sublayer.
[0082] 16. Ensuring Wavelength Accuracy in WDM Networks
[0083] In WDM networks, the laser wavelength (or frequency) must be
maintained within a certain accuracy so that there is no
interference between neighboring channels, and there are no
penalties from laser-filter misalignment.
[0084] As is the case with all electronic and optical components,
the performance characteristics of the lasers employed in DWDM
systems change with temperature and with time. In particular, the
frequency of emitted laser light changes due to ambient temperature
variations (typically from -5 degC. to 65 degC.) and due to
aging.
[0085] WDM laser frequencies are maintained to a first order by
controlling the temperature of the laser by mounting the laser on a
thermoelectric cooler (TEC). Etalons may also be integrated into
the laser cavity to provide a second-order correction. Currently
deployed WDM lasers have an accuracy that is adequate for 50 GHz
spacing. There is currently a need for more accurate means of
controlling laser frequencies to space the WDM channels as close
together as possible.
II. SUMMARY
[0086] Various embodiments of the current invention are disclosed
herein, including techniques, apparatus, and systems for optical
WDM communications that employ tunable lasers to generate
respective subcarrier frequencies which represent subchannels of an
ITU channel to which client signals can be mapped. Client circuits
can be divided and combined with one another before being mapped,
independent of one another, to individual subchannels within and
across ITU channels.
[0087] Novel techniques are employed (at the subchannel
level/layer) to facilitate the desired optical routing, switching,
concatenation and protection of the client circuits mapped to these
subchannels across the nodes of a WDM network, resulting in a
significant increase in the number of optical circuits in a fiber,
and thus in the overall bandwidth and spectral efficiency of the
WDM network.
[0088] Network architectures and subchannel transponders,
muxponders and crossponders are disclosed that map client signals
to a set of subchannel frequencies. In one embodiment, these
architectures employ two levels of frequency mapping and two
cascaded optical filters (one for filtering WDM channels and one
for filtering subchannels). Additional methods of multiplexing
channels and subchannels by means of polarization multiplexing and
related feedback control electronic systems are also disclosed.
Selectively mapping client signals to a subset of the subchannels
facilitates network functions such as broadcast and select
transmission, arbitrary concatenation, optical source routing,
shared optical protection, and simplified network reconfiguration
at a significantly lower cost than is required for currently
deployed WDM networks.
[0089] Subchannel muxponders are disclosed that measure network
characteristics such as optical signal to noise ratio, chromatic
and polarization mode dispersion, power levels, and bit error
rates. Highly accurate wavelocker circuits are also disclosed that
enable the equipment to provide very dense subchannels with
accurate control.
[0090] Embodiments of the current invention extend existing WDM
network designs by adding a new sublayer to the WDM network
architecture between the FEC encode layer and the wavelength
assignment layer. Novel means of mapping, multiplexing, switching,
and managing sublayer services are described in a common format
that scales from small 1 GE and 2.5G access networks to large
regional networks and long-haul networks with capacity scalable to
17 Tb/s. Novel means of connecting spur traffic to a ring, and
interconnecting optical rings without O-E-O conversion, are also
disclosed. Moreover, these techniques are designed so as to enable
standard ITU-T G.692 based (and other legacy) hardware to be
reused.
[0091] Designs for subchannel transponders, muxponders and
crossponders are disclosed, where client services are mapped to
several subchannels within an ITU channel (as well as across ITU
channels, and combined with other client services employing
different signal protocols). These subchannel devices support
subchannel routing, restoration and protection, and direct
measurement of the most significant optical parameters, such as
power, OSNR, chromatic dispersion, and polarization-mode
dispersion.
[0092] Embodiments of line cards that support these novel
architectures are also described.
[0093] Multilayer routing protocols are disclosed that enable
network operators to easily map services to available bandwidth,
while maintaining full visibility of the deployed channels and
available bandwidth. Means for adiabatically adjusting the network
capacity are described to ensure minimal planned or indeliberate
service interruption. A novel OSC routing protocol is described to
manage such a network with minimal cost overhead. Other embodiments
are disclosed that enable networks to be upgraded from being
ITU-channel based to subchannel based.
[0094] Because subchannel lasers require a higher degree of
accuracy than ITU channel lasers, methods are disclosed for more
accurately controlling lasers, based on a heterodyne measurement
against a reference tunable laser that can be continuously
calibrated with a reference gas cell.
III. BRIEF DESCRIPTION OF THE DRAWINGS
[0095] FIGS. 1A and 1B illustrate optical ring networks that
utilize OC-48 SONET add/drop multiplexers and Gigabit Ethernet
switches, respectively.
[0096] FIG. 2 is a block diagram of a WDM transponder.
[0097] FIG. 3 is a block diagram of a WDM muxponder.
[0098] FIG. 4 is a block diagram of a point-to-point WDM network
employing two fiber-optic cables.
[0099] FIG. 5 is a block diagram of a WDM Ring network with drop
filters and add couplers at each node.
[0100] FIG. 6 is a block diagram of a Broadcast and Select WDM Ring
network with drop filters and add couplers at each node.
[0101] FIG. 7 is a graph illustrating the effect of cascaded ROADMs
or WSS filters on usable C-band bandwidth in an optical
network.
[0102] FIG. 8 is a block diagram illustrating the partitioning of
WDM equipment functionality between transponder modules and
add/drop filters.
[0103] FIG. 9 is a block diagram illustrating common misconnections
between WDM transponder modules and add/drop filters.
[0104] FIG. 10 is a block diagram illustrating interconnected WDM
Ring networks with spur nodes.
[0105] FIG. 11 is a block diagram illustrating how typical WDM
equipment is installed and managed in a shelf with management cards
and line cards.
[0106] FIG. 12 is a block diagram illustrating a shelf
configuration of WDM equipment designed for redundant
management.
[0107] FIG. 13 is a block diagram illustrating a WDM Ring network
employing large crossconnect switches at each node, interconnected
via WDM links.
[0108] FIG. 14 is a block diagram illustrating electrical and
optical sublayers of WDM equipment residing at the physical
(lowest) layer of the 7-layer OSI model.
[0109] FIG. 15 is a block diagram of one embodiment of a subchannel
muxponder of the present invention.
[0110] FIG. 16 is a block diagram of one embodiment of multichannel
clock recovery circuits in a subchannel muxponder of the present
invention.
[0111] FIG. 17 is a block diagram of one embodiment of a subchannel
muxponder of the present invention with an electrical crossconnect
switch.
[0112] FIG. 18 illustrates one embodiment of an overlay of the
subchannels of the present invention on 100-GHz ITU channels and
filters.
[0113] FIG. 19 illustrates the characteristics of one embodiment of
cyclical filters of the present invention.
[0114] FIG. 20 illustrates the filtering of subchannels in one
embodiment of the present invention, where an ITU filter is
followed by cyclical filters.
[0115] FIG. 21 illustrates the filtering of subchannels in one
embodiment of the present invention, where cyclical filters are
followed by an ITU filter.
[0116] FIG. 22 illustrates one embodiment of an overlay of the
subchannels of the present invention on 50-GHz ITU channels and
filters.
[0117] FIG. 23 illustrates one embodiment of a pre-emphasis on the
subchannels of the present invention to counteract penalties from
the edge of ITU channel filters.
[0118] FIG. 24 illustrates one embodiment of cyclical filters of
the present invention with cascaded interleavers.
[0119] FIG. 25 illustrates one embodiment of a cyclical filter of
the present invention made with an array waveguide grating
(AWG).
[0120] FIG. 26 is a graph illustrating the frequency offset from
the optimum AWG design frequency (for an AWG cyclical filter of the
present invention).
[0121] FIG. 27 is a graph illustrating the shift in the AWG
temperature setpoint (for an AWG cyclical filter of the present
invention) with the ITU channel number.
[0122] FIG. 28 is a graph illustrating the shift in the AWG
temperature setpoint (for an AWG cyclical filter of the present
invention) with the ambient temperature.
[0123] FIGS. 29A and 29B are graphs illustrating changes in
superimposed AWG transmission spectra (for an AWG cyclical filter
of the present invention) of 4 subchannels for ITU channels 50 and
60 with changes in ambient temperature (65 degC.).
[0124] FIGS. 30A and 30B are graphs illustrating changes in
superimposed AWG transmission spectra (for an AWG cyclical filter
of the present invention) of 4 subchannels for ITU channels 30 and
40 with changes in ambient temperature (-5 degC.).
[0125] FIG. 31 is a graph illustrating superimposed AWG
transmission spectra (for an AWG cyclical filter of the present
invention) of 4 subchannels for 40 ITU channels (channels 20-60) at
ambient temperature (-5 degC.).
[0126] FIG. 32 is a graph illustrating superimposed AWG
transmission spectra (for an AWG cyclical filter of the present
invention) of 160 subchannels for 40 ITU channels (channels 20-60)
at ambient temperature (-5 degC.).
[0127] FIG. 33 is a graph illustrating superimposed AWG
transmission spectra (for an AWG cyclical filter of the present
invention) of 160 subchannels for 40 ITU channels (channels 20-60)
at ambient temperature (65 degC.).
[0128] FIG. 34 is a graph illustrating how a shift in channel
spacing (for an AWG cyclical filter of the present invention) can
be achieved by a change in ambient temperature.
[0129] FIG. 35 is a top view of a top enclosure for an AWG wafer
(chip) embodying an AWG cyclical filter of the present
invention.
[0130] FIG. 36 is a bottom view of a top enclosure for an AWG wafer
(chip) embodying an AWG cyclical filter of the present
invention.
[0131] FIG. 37 is a component view of the thermo-mechanical design
of an AWG wafer (chip) embodying an AWG cyclical filter of the
present invention.
[0132] FIG. 38 is a schematic diagram of a high-precision
electronic circuit to control the temperature of an AWG wafer
(chip) embodying an AWG cyclical filter of the present
invention.
[0133] FIG. 39A illustrates one embodiment of a channel plan for
subchannels of the present invention with polarization
multiplexing.
[0134] FIG. 39B is a block diagram of one embodiment of a receive
circuit to demultiplex polarization-multiplexed subchannels of the
present invention.
[0135] FIG. 40A is a block diagram of an existing implementation of
polarization multiplexing in a DWDM network.
[0136] FIG. 40B is a block diagram of a novel embodiment of the
feedback control electronics in the dithering scheme employed in
the implementation of polarization multiplexing presented in FIG.
40A.
[0137] FIG. 40C is a block diagram of a novel embodiment of the
feedback control electronics in the dithering scheme employed in
the implementation of polarization multiplexing presented in FIG.
40A for f.sub.dith-2 substantially lower than f.sub.dith-1.
[0138] FIG. 40D is a block diagram of one embodiment of a
polarization tracking scheme of the present invention with three
dithering frequencies.
[0139] FIG. 40E is a block diagram of one embodiment of the
feedback control electronics for the polarization multiplexed
system presented in FIG. 40D.
[0140] FIG. 40F is a block diagram of one embodiment of the
feedback control electronics for the polarization multiplexed
system presented in FIG. 40D for f.sub.dith-2 substantially lower
than f.sub.dith-2.
[0141] FIG. 40G is a block diagram of one embodiment of
polarization multiplexing feedback control electronics applied to a
subchannel-based DWDM system of the present invention.
[0142] FIG. 40H is a block diagram of an alternative embodiment of
polarization multiplexing feedback control electronics applied to a
subchannel-based DWDM system of the present invention.
[0143] FIG. 40I is a block diagram of one embodiment of a
polarization tracking scheme of the present invention without
dithering lasers on the transmit end.
[0144] FIG. 40J is a block diagram of one embodiment of a
polarization tracking scheme of the present invention that enables
polarization matching of added signals to passthrough signals.
[0145] FIG. 41 is a block diagram of one embodiment of the mapping
of client services to subchannels of the present invention.
[0146] FIG. 42 is a block diagram of one embodiment of mapping
lower-rate client services to subchannels of the present
invention.
[0147] FIG. 43 is a block diagram of one embodiment of mapping 40G
client services to subchannels of the present invention.
[0148] FIG. 44 is a block diagram of a 4-node WDM network with one
embodiment of subchannel muxponders of the present invention at
Node 1 and lower-rate transponders or muxponders at Nodes 2, 3 and
4.
[0149] FIG. 45 is a block diagram of one embodiment of
software-controlled 1.times.2 switches of the present invention to
selectively direct traffic to the East or West side of a WDM ring
network.
[0150] FIG. 46 is a block diagram of one embodiment of
software-controlled 1.times.3 switches of the present invention to
selectively direct traffic to the East or West side (or broadcast
to both sides) of a WDM ring network.
[0151] FIG. 47 is a block diagram illustrating the upgrading of a
10 Gb/s legacy ITU-channel network to employ subchannels of the
present invention on ITU channels 192.1 THz and 192.2 THz, while
maintaining the legacy 10 Gb/s service on ITU channel 192.3
GHz.
[0152] FIG. 48 is a block diagram illustrating the high-capacity
transmission resulting from one embodiment of 10G subchannel
muxponders of the present invention.
[0153] FIG. 49 is a block diagram of one embodiment of a monitor
channel filter of the present invention in an optical network.
[0154] FIG. 50 is a block diagram of one embodiment of a circuit of
the present invention to measure net dispersion of a fiber link due
to subchannel delay times.
[0155] FIG. 51 is a graph illustrating the effect of subchannel
spacing in the present invention on phase detector voltage.
[0156] FIG. 52 is a graph illustrating the measurement of
polarization-mode dispersion based upon the delays between
orthogonal subchannels of the present invention.
[0157] FIG. 53 is a graph illustrating the result of a
software-controlled circuit of the present invention used to
monitor the bit error rate (BER) in an optical network as a channel
frequency is tuned.
[0158] FIG. 54 is a data structure of one embodiment of a
diagnostic spreadsheet employed by the present invention that lists
device registers as well as expected and actual values.
[0159] FIG. 55 is a block diagram of one embodiment of an element
management system (EMS) of the present invention with distinct
shelves for legacy and new products.
[0160] FIG. 56 is a block diagram of one embodiment of an element
management system (EMS) of the present invention managing a shelf
running two software versions in parallel.
[0161] FIG. 57 is a block diagram illustrating an embodiment of the
present invention in which management data is optionally routed
throughout an optical network.
[0162] FIG. 58 is a block diagram illustrating the addition of a
subchannel management layer of the present invention to existing
WDM management layers.
[0163] FIG. 59 is a block diagram illustrating the fiber
interconnections in one embodiment of a 3-node optical ring network
of the present invention with degree-2 nodes (i.e., which connect
to 2 other nodes).
[0164] FIG. 60 illustrates one embodiment of a simple routing table
of the present invention for intra-node connections for a degree-2
optical node.
[0165] FIG. 61 illustrates one embodiment of an interconnect
routing table of the present invention for the 3-node optical
network illustrated in FIG. 59.
[0166] FIG. 62 illustrates a linear representation of the fiber
connections of the 3-node optical network illustrated in FIG.
59.
[0167] FIG. 63 is a block diagram of a subchannel ring network of
the present invention with subchannel routing.
[0168] FIG. 64 is a block diagram illustrating the fiber
interconnections in one embodiment of a 4-node optical ring network
of the present invention with degree-2 nodes (i.e., which connect
to 2 other nodes).
[0169] FIG. 65 illustrates one embodiment of a subchannel
interconnect map of the present invention for the 4-node optical
network illustrated in FIG. 64 with subchannel routing.
[0170] FIG. 66 illustrates one embodiment of a subchannel bandwidth
map of the present invention for the 4-node optical network
illustrated in FIG. 64 with subchannel routing.
[0171] FIG. 67 illustrates the highlighted protected connection in
the subchannel bandwidth map illustrated in FIG. 66.
[0172] FIG. 68 illustrates 9 available subchannels (between Node 1,
Port 2 and Node 2, Port 5) in the subchannel bandwidth map
illustrated in FIG. 66.
[0173] FIG. 69 illustrates various embodiments of subchannel
payloads of the present invention resulting from the mapping of
client services to subchannels.
[0174] FIG. 70 is a block diagram illustrating the mapping of
client services (10.times.1G Ethernet switch cards) to a subchannel
of the present invention.
[0175] FIG. 71 illustrates one embodiment of a bandwidth map for
the subchannel services illustrated in FIG. 70.
[0176] FIG. 72 illustrates one embodiment of a portion of a service
status table of the present invention (listing performance metrics
for the traffic at Node 1) for the subchannel services illustrated
in FIG. 70.
[0177] FIG. 73 illustrates one embodiment of the configuration of a
subchannel muxponder of the present invention with red/blue filters
that direct the traffic to one of two add ports and one of two drop
ports.
[0178] FIG. 74 illustrates how the subchannel muxponder of FIG. 73
can be deployed as a switchable subchannel crossponder of the
present invention such that traffic can be redirected away from a
span for node insertion.
[0179] FIG. 75 illustrates how the subchannel crossponder of FIG.
74 can be used to bridge traffic on two diverse spans to implement
protection switching in one embodiment of the present
invention.
[0180] FIG. 76 illustrates an alternative implementation of
protection switching in the present invention.
[0181] FIG. 77 illustrates an alternative implementation of the
bridge and switch functions that implement protection switching in
the present invention.
[0182] FIG. 78 is a block diagram of one embodiment of a subchannel
crossponder of the present invention.
[0183] FIG. 79 is a block diagram illustrating the deployment of a
subchannel muxponder of the present invention in a dedicated
protection mode.
[0184] FIG. 80 is a block diagram illustrating the deployment of a
subchannel muxponder of the present invention in a shared
protection mode.
[0185] FIG. 81 is a block diagram illustrating the deployment of a
subchannel muxponder of the present invention in a shared
protection mode where the protection subchannels are regenerated at
an "intermediate node" (not directly connected to a cut fiber).
[0186] FIG. 82 is a block diagram of one embodiment of a
distributed subchannel switching network of the present invention
with up to N (number of subchannels) interconnects.
[0187] FIG. 83 is a block diagram of a logical mesh, provided by
the distributed subchannel switching network illustrated in FIG.
82, in which routing is controlled by tuning subchannel
frequencies.
[0188] FIG. 84 is a graph illustrating a DWDM signal laser of the
present invention beating at a fixed frequency with a narrow
optical carrier while an oscillator laser frequency is scanned.
[0189] FIG. 85 is a graph illustrating a DWDM signal laser of the
present invention beating at a fixed frequency over a broad
spectrum while an oscillator laser frequency is scanned.
[0190] FIG. 86 is a block diagram of one embodiment of an optical
frequency stabilization scheme of the present invention for one
DWDM signal laser with polarizations of both signal and oscillator
lasers aligned.
[0191] FIG. 87 is a graph illustrating typical absorption lines of
a hydrogen cyanide (H.sup.13C.sup.14N) gas cell.
[0192] FIG. 88 is a graph illustrating the dependence of the P16
absorption line spectral position on gas pressure for hydrogen
cyanide (H.sup.13C.sup.14N).
[0193] FIG. 89 is a graph illustrating the dependence of the P16
absorption line FWHM ("Full Width at Half Maximum") linewidth on
gas pressure for hydrogen cyanide (H.sup.13C.sup.14N).
[0194] FIG. 90 is a graph illustrating the absolute frequency
accuracy of hydrogen cyanide (H.sup.13C.sup.14N) absorption line
positions.
[0195] FIG. 91 is a graph illustrating the FWHM ("Full Width at
Half Maximum") linewidths of hydrogen cyanide (H.sup.13C.sup.14N)
absorption lines.
[0196] FIG. 92 is a graph illustrating the spectral shape of the
P16 absorption line of hydrogen cyanide (H.sup.13C.sup.14N) at a
pressure of 13 kPa (measured with a 1 pm scanning step at
approximately 74 MHz).
[0197] FIG. 93 is a graph illustrating calibration of the
oscillator laser frequency setpoints by reference frequencies of
the absorption cell.
[0198] FIG. 94 is a block diagram of one embodiment of an optical
frequency stabilization scheme of the present invention for one
DWDM signal laser (with polarizations of both signal and oscillator
lasers aligned) and a reference absorption cell.
[0199] FIG. 95 is a block diagram of one embodiment of an optical
frequency stabilization scheme of the present invention where
variations of the oscillator laser optical power are measured and
used as a reference signal.
[0200] FIG. 96 is a block diagram of one embodiment of an optical
frequency stabilization scheme of the present invention where the
absolute accuracy of DWDM signal laser frequency is increased by
measuring the frequency prior to data modulation.
[0201] FIG. 97 is a block diagram of one embodiment of an optical
frequency stabilization scheme of the present invention where there
are no restrictions on the state of polarization of the signal
laser (and a single mode fiber can be used for all optical
connections except from the Tx Laser to the Data Modulator).
[0202] FIG. 98 is a block diagram of one embodiment of an absolute
wavelength stabilization scheme of the present invention for
multiple DWDM signal lasers propagating in dedicated fibers.
[0203] FIG. 99 is a block diagram of one embodiment of an absolute
wavelength stabilization scheme of the present invention for
multiple DWDM signal lasers propagating in a single fiber.
[0204] FIG. 100 is a block diagram of one embodiment of a frequency
monitoring scheme of the present invention where the DWDM spectrum
in a network node (i.e., all incoming and outgoing optical spectra
in all fibers of the node) are monitored with ultrahigh absolute
accuracy.
[0205] FIG. 101 is a block diagram of one embodiment of a circuit
of the present invention to measure the beat signal between a
reference tunable laser and an array of signal lasers.
[0206] FIG. 102 is a block diagram of one embodiment of a circuit
of the present invention to measure the beat signal between a
reference tunable laser and an array of signal lasers with a
polarization controller.
[0207] FIG. 103 is a block diagram of one embodiment of a circuit
of the present invention to measure the beat signal between a
reference tunable laser and an array of signal lasers with a
polarization scrambler.
[0208] FIG. 104 is a block diagram of one embodiment of a
subchannel muxponder of the present invention with integrated DWDM
transceivers.
IV. DETAILED DESCRIPTION OF THE CURRENT INVENTION
A. Subchannel Architecture
[0209] As noted above, the current invention employs subchannels to
increase the number of optical circuits in a single fiber, and
leverages those subchannels to fully realize the benefits of the
increased bandwidth by facilitating the desired optical routing,
switching, concatenation and protection of the client circuits
mapped to those subchannels. To illustrate how this subchannel
architecture can be implemented in a WDM network, one embodiment of
a subchannel muxponder is described, followed by descriptions of
the corresponding channel plans, filters and associated hardware
and software used to map client circuits to (and extract them from)
subchannels across various nodes of a WDM network.
[0210] 1. Subchannel Muxponder
[0211] One embodiment of a subchannel muxponder 1500 is shown in
FIG. 15. In this embodiment the subchannel muxponder maps data from
client traffic onto four subchannels. Data from the client traffic
is terminated with client optics, 1501, typically pluggable devices
such as an XFP, Xenpak, or SFP.
[0212] After the optical client signal is converted to an
equivalent electrical signal on the subchannel muxponder, each
subchannel's electrical signals can be processed digitally to
optionally (1) extract performance monitoring information, (2) add
channel overhead for remote network management, and (3) encode the
data for forward error correction. This can be done with the
SERDES-FEC-SERDES block, 1502 (SERDES=Serializer-Deserializer). The
10 Gb/s deserializer converts the data to parallel streams of
lower-rate data that are then processed by an FEC device.
[0213] The data is thereby mapped to a subchannel within an FEC
frame. Management overhead can optionally be inserted into one or
more of the FEC frames. Moreover, in one embodiment, block 1502 can
also monitor a client signal's overhead bytes to extract various
data, such as a "destination tag" (e.g., an Ethernet address, IP
address, VLAN ID, MPLS tag, etc.). The header information can be
relayed to the management software that uses the destination tag to
determine the destination port. The management software can then
provision the subchannel frequency to the frequency assigned to
that destination port.
[0214] Controlling wavelength switching in an optical network based
on destination tags can, in one embodiment, proceed as follows:
[0215] 1. Network operator provisions switching mechanism, e.g.
Virtual Local Area Network identifiers [0216] 2. Client receiver
detects source address (SA) and destination address (DA) [0217] 3.
NE broadcasts SA to other nodes over OSC or in-band overhead [0218]
4. NE broadcasts DA to other nodes paired with receiver ITU channel
and subchannel. ITU channel can be a fixed value (determined by
fixed drop filter) or range of values if ROADM is being used.
[0219] 5. The nodes distribute the SA/DA information to build up a
network-wide distributed switch table [0220] 6. Given an associated
DA with ITU channel and subchannel, a signal appearing at a client
port causes the source node to tune its subchannel frequency
wavelength to the correct subchannel frequency [0221] 7. If
required, ITU channel filter ROADM add/drop/passthrough channels
are tuned accordingly [0222] 8. Client Rx periodically checks the
SA/DA and triggers a wavelength change when the DA changes
[0223] Returning to FIG. 15, an optical modulator 1504 modulates a
CW laser beam to produce a modulated laser beam that carries the
respective lower speed electronic signals 1505. Each tunable laser
1503 is set to an ITU G.692 frequency, with an offset dependent on
the subchannel. The different electronic-to-optical conversion
units 1506 are configured to have different lasers 1503 at
different subchannel frequencies that may or may not be within the
same ITU G.692 channel. The subchannel lasers 1503 can therefore be
(a) assigned to different subchannels within different ITU G.692
windows, and (b) be transmitted to different receive nodes that
have different ITU channel filters.
[0224] This embodiment can be distinguished from subcarrier
multiplexing [such as was described in U.S. Pat. No. 6,525,857]
where a single laser is deployed for a group of subcarriers. Here,
each subchannel has its own independently tuned and modulated
laser, and each subcarrier can carry independent protocols.
Moreover, there are no restrictions at the transmit side on the
frequency spacing between subchannels, and each subchannel can be
transmitted in a different ITU channel.
[0225] The modulation of each subchannel can be selectively chosen
to be one of many different types of modulation such as Non-return
to Zero, Duobinary, or Differential Quadrature Phase Shift Keying.
Modulation formats with a narrow spectral width, such as duobinary
and DQPSK, are favored because their spectra must pass through the
narrow-band filter 1525 at the receive side. If duobinary
modulation is used, a precoder and low-pass filter 1507 are
inserted in the data path. The precoder is used such that the
recovered signal is identical to the transmitted signal. For a
duobinary signal of a bandwidth of B, the low-pass filter passband
is set to approximately from 0.2 B to 0.3 B, in which the
electrical baseband modulation signal swings from -V.sub..pi. to
+V.sub..pi. (with the modulator biased at a minimum point). The
modulation signal is then fed into the optical modulator 1504 to
control the optical modulation which produces the optical WDM
signal 1514.
[0226] The light from each subchannel is then combined optically
with a polarization combiner, 1:4 coupler, or subchannel
multiplexing filter 1520. In the illustrated example, the optical
polarization of each signal is controlled so that two optical WDM
channels next to each other in frequency are orthogonally polarized
to each other. The optical WDM channels in the same polarization
are directed into beam combiners 1511 and 1512 to produce a
combined signal with optical channels in the same polarization. Two
such beam combiners 1511 and 1512 are used, one for each
polarization. The combined signals from the beam combiners 1511 and
1512 are directed into a polarization-maintaining directional
coupler 1513, to produce an output signal that combines all
subchannels 1514 so that any two adjacent subchannels have
orthogonal polarizations.
[0227] Having adjacent channels launched at orthogonal
polarizations minimizes crosstalk penalties at the receiver.
However, when the adjacent lasers on the subchannel transceiver are
allowed to be set to arbitrary ITU G.692 channels and subchannels,
the adjacent subchannel at a receiver may be transmitted from
completely different source locations. In this case it is very
difficult to maintain orthogonal polarizations between the adjacent
signals. To minimize penalties in this case, an optional
polarization scrambler 1518 can be inserted in the path to reduce
the crosstalk penalties. Regardless of the means to control
polarization of the adjacent subchannels, the system impact of
adjacent-channel crosstalk must be quantified and accounted for
during the network engineering. Typically, the crosstalk penalty
leads to a slight increase in the required Optical Signal-to-Noise
Ratio (OSNR) at the receiver.
[0228] A variable optical attenuator (VOA) 1515 combined with a tap
coupler and monitor photodiode 1516 can optionally be used to
control the output power of the combined subchannels. If the output
is too low an optional optical amplifier 1517 could also be
inserted in the transmit path. This output signal is transmitted
through a single fiber connected to the line output port that is
connected to an optical network. Note that it is preferable to use
a VOA to control the output power, rather than adjusting the laser
power. Keeping the laser power fixed in time (after initial
calibration) simplifies the control circuits that maintain the
transmit eye quality. The optical network transmits the subchannels
from the transmit node through optical fiber waveguides to the
receive node.
[0229] On the receive side an optical amplifier and/or variable
attenuator can be used to control the received power. In this
example, the ITU G.692 channel WDM demultiplexer 1521 is used to
receive the light from the network and select the subchannels in a
single ITU channel to be directed to the Line Input port of the
subchannel transceiver.
[0230] On the receive side of the transceiver an optional optical
amplifier 1522 (e.g. an EDFA) can be used to amplify the received
signal. An optical attenuator 1523 with tap coupler and monitoring
photodiode 1524 after the amplifier 1522 can be used to ensure that
the amplified signal does not overload the photodetectors 1530.
Furthermore, control software can be used to control the variable
optical attenuator VOA 1523 so that the optical power incident on
each photodetector 1530 is kept very close to the ideal incident
power of the photodetector 1530, thereby optimizing system
performance. Preferably, EDFA 1522 is kept at high gain (hence low
noise and high optical signal to noise ratio) and uses the VOA 1523
to optimize the received powers.
[0231] Note that the EDFA 1522 and VOA 1523 at the receiver can be
shared among the subchannels, dedicated to a single subchannel, or
not used at all. These choices depend on the requirements for the
reach of the subchannel muxponder and the cost targets.
[0232] The composite signal containing the subchannels is then
directed to an Ultra-dense WDM filter 1525 that separates the
subchannels to output paths 1526. The cyclical filter 1525
described below requires that the subchannel spacing be equal to
the ITU frequency spacing (in GHz) divided by an integer M. In the
case when the bit rate per subchannel is on the order of 11 Gb/s,
the channel spacing is typically set at 10 or 12.5 GHz.
[0233] Multiple optical detectors 1530 are used to respectively
receive and detect the separated optical subchannel signals, with
one subchannel per detector, to produce electronic signals 1535
that are directed back to the FEC blocks 1502. Each electronic
signal path may include an electrical equalizer that is typically
integrated into the photodetector's transimpedance amplifier or
deserializer. The equalizer can mitigate the eye distortion, either
due to static band-limiting effects caused by the electrical or
optical pre-filtering in the optical transmitter module, or due to
fiber chromatic dispersion.
[0234] The SERDES-FEC-SERDES block 1502 then decodes the FEC frame,
corrects the errors according to the FEC algorithm, demaps the
data, and optionally provides performance monitoring information on
the data that is directed to the client transmit optical
transceivers 1501.
[0235] Note that each subchannel has independent clock recovery and
generation circuitry to support different data rates and protocols.
One embodiment of these circuits is shown in detail in FIG. 16.
[0236] Circuit 1600 in FIG. 16 shows one embodiment of independent
clock circuits for the subchannel timing of the SERDES-FEC-SERDES
block. Each client-side clock recovery unit (CRU) 1630 has a
multi-rate reference clock 1620 that can be set independently for
each subchannel's reference clock rate. A transmit digital
phase-locked loop (PLL) 1610 for each subchannel multiplies the
recovered client clock by a factor that provides the FEC rate
clock. Similarly, the line side SERDES has a multi-rate reference
clock 1640 for each subchannel receiver CRU 1660, and a receive PLL
1650 is used to convert the subchannel FEC rate clock to the
subchannel client rate. The reference clocks 1620 and 1640 can also
be used as the source clocks to transmit maintenance signals, such
as an OTN AIS (Alarm Insertion Signal) when the client services are
in an alarm state or out of service. Note that running each client
service on an independent subchannel maintains the end-to-end
synchronization of each client service. Compared to implementations
that use electronic multiplexing to combine 10 Gb/s client services
into a native 40 Gb/s service, this invention provides a distinct
advantage for applications such as SONET and Synchronous Ethernet
that require end-to-end synchronization of the client signals.
[0237] The subchannel muxponder (e.g., subchannel muxponder 1500
from FIG. 15) supports capacity upgrades of ITU channel-based
networks. The transmission symbol rate (e.g., 10 Gbaud) per
subchannel is equivalent to an existing low-data rate (e.g., 10
Gb/sec), which is already running on the incumbent infrastructure.
This limits signal degradations caused by network impairments such
as chromatic dispersion (CD), polarization-mode dispersion (PMD),
and amplifier noise within the incumbent optical fiber
infrastructure. Therefore, if the fiber network has been designed
to work at line rates of 10 Gb/s the network infrastructure
(amplifiers and dispersion compensation modules) need not be
changed when implementing the subchannel optical transceiver since
the bit-rate dependent penalties of the subchannel and ITU channel
are equivalent. In this embodiment, the subchannel muxponder can be
used to increase the available bandwidth in an ITU channel by four
times without changing or modifying the network. Furthermore,
allowing the control software to provision the subchannel laser
wavelengths at different ITU channels enables the subchannel
muxponder to perform optical routing based on the subchannel
frequencies.
[0238] To optimize optical performance it may be required to
balance the powers of the subchannels. This can be accomplished
after the subchannel muxponder is activated by turning on one laser
at a time, recording the power on the tap photodiode of each laser
and applying an offset to each laser to compensate for the power
differences. This balancing can be done at low output power with
the VOA at or near full attenuation.
[0239] FIG. 17 shows another embodiment of a subchannel muxponder
1700 with an electronic crossconnect. The crossconnect switch 1750
is added between the SERDES and external FEC blocks and provides
further switching and routing functionality as described below.
Note that the SERDES, FEC, and crossconnect functions can be
integrated into a single VLSI device 1760.
[0240] 2. Channel Plans
[0241] An example of a channel plan 1800 followed by the disclosed
design is shown in FIG. 18. In this embodiment, the carriers 1810
are spaced 12.5 GHz apart centered around ITU channels 1820 spaced
at 100 GHz. There is a guard band 1830 between the ITU channels
1820 where no carriers are present. This allows for bandwidth
narrowing of the ITU channel as it passes through optical elements
in the transmission network.
[0242] In this embodiment of the present invention, tunable lasers
are used to generate each subchannel, so that any client signal can
be transmitted on any of the subchannels.
[0243] The receiver demultiplexing required in this network
consists of two stages. The first stage is comprised of fixed,
tunable, or reconfigurable ITU-T G.692 channel filters that may be
built with technologies such as (but not limited to) thin-film
filters, Array Waveguide Gratings, MEMS arrays, or diffraction
gratings. The second stage of demultiplexing in this network
consists of a narrow-band cyclical or tunable filter to select one
of the subchannels within the ITU passband. Example of a cyclical
filter are the Array Waveguide Grating and cascaded
interleavers.
[0244] A common characteristic of cyclical filters in this network
design is that the nth subchannel in each ITU window is directed to
the nth output port, as shown in channel plan 1900 in FIG. 19, and
implemented by cyclical filter 1910.
[0245] To fully separate each carrier, the cyclical filter is
cascaded with an ITU channel filter as shown in FIG. 20. In this
example filtering architecture 2000, a standard 100-GHz ITU channel
filter 2010 is used as the first filtering stage. This design
therefore supports an upgrade of an installed WDM system that uses
100 GHz channel filters. The upgrade can be implemented gradually
over time so that any port of the ITU channel filter 2010 in FIG.
20 can be used to drop a legacy ITU channel signal.
[0246] The same filtering of subchannels can also be achieved by
placing the ITU channel filters 2110 after the cyclical filter 2120
as shown in the filtering architecture 2100 of FIG. 21.
[0247] Another example of a channel plan 2200 is shown in FIG. 22.
In this case the subchannels 2210 are centered around the ITU
channels 2220 at 50 GHz spacing. In this case the 100 GHz carriers
can be separated from the 50 GHz carriers with (1) Two separate
cyclical filters, one for the 100 GHz grid and one for the 50 GHz
grid, or (2) a cyclical filter that can be shifted between the two
grids by temperature tuning for example, or (3) a cyclical filter
with 8 ports, 4 for the 100 GHz channels, and 4 for the 50 GHz
channels.
[0248] When the subchannels are spaced as in FIG. 22, the outer
channels, i.e. SC-1 and SC-4, can be attenuated by the edges of the
ITU channel filter. This attenuation and its subsequent bit-error
rate penalty can be mitigated, as shown in channel plan 2300 in
FIG. 23, by adding power and/or frequency pre-emphasis on the outer
carriers. This is done by shifting the outer subchannels 2310
toward the ITU channel's center frequency 2320 and boosting their
transmit powers by adjusting the laser transmit powers. This
pre-emphasis can be adjusted by software depending on the penalties
on the edge subchannels.
[0249] The optical network in one embodiment uses a transmitter
module that combines a plurality of subchannels (in the examples
shown herein, 4 subchannels are used, but any number greater than 1
could be employed). Furthermore, the examples shown above are given
for the case where the data mapped to the carrier is approximately
10 Gb/s, but the same methodology could be extended to any
arbitrary rate per carrier.
[0250] 3. Details on the Cyclical Filter
[0251] The cyclical filter can be an interleaver [see, eg, U.S.
Pat. No. 7,257,287], cascaded interleavers, or an array waveguide
grating (AWG) [see, eg, U.S. Pat. Nos. 6,594,049, 4,904,042, and
5,600,742]. For the interleaver case, shown in cyclical filtering
architecture 2400 in FIG. 24, a 4-channel 12.5 GHz cyclical filter
can be made by cascading a 12.5 GHz interleaver 2410 with two 25
GHz interleavers 2420a and 2420b. Note that centering the
subchannels on the ITU grid requires that the subchannels and
interleavers are offset from the ITU grid by half the subchannel
spacing, which is 6.25 GHz in this example. In the case of an odd
number of subchannels (eg, 5 subchannels), the central subchannel
is not offset from the ITU grid.
[0252] Another embodiment of the cyclical filter is based on array
waveguide grating (AWG) technology [see, eg, U.S. Pat. No.
6,594,049]. The principle of operation of cyclical AWG 2500 is
shown in FIG. 25. In this example the AWG 2500 has a channel
spacing of 12.5 GHz. For an application using 100 GHz channel
spacing, the four middle output ports can be used to select the
four subchannels adjacent to the ITU grid.
[0253] The channel spacing of a cyclical AWG is based on wavelength
and the ITU channel spacing is based on frequency. This discrepancy
causes a frequency offset for channels that are further from the
AWG optimum design frequency. This offset is illustrated in graph
2600 in FIG. 26.
[0254] Such use of an AWG may not yield sufficient accuracy for
applications such as the subchannel demultiplexer that need
frequency accuracy within a fraction of the accuracy required by
ITU channel filters. In such cases, a new way of controlling the
temperature of an optical filter provides more accurate operation
of the filter within a wide range of optical frequencies and
ambient temperatures.
[0255] The performance characteristics of the optical filters
employed in DWDM systems change with temperature and over time. In
particular, a central frequency of bandpass optical filters change
due to ambient temperature variations (typically from -5 degC. to
65 degC.), and due to aging. The center frequency of an AWG is
temperature dependent with a shift approximately equal to 1.5
GHz/.degree. C. In applications requiring high accuracy of the
filter spectral response, such as WDM channel filtering, the
temperature of the filter is controlled by mounting it on a heater
or thermoelectric cooler, and using control circuits to maintain
the voltage reading on a thermistor inside the filter package.
[0256] The close spacing of the subchannels as described herein can
be made more accurate with refinement of the AWG design and
controls. First, the change in subchannel offset with frequency can
be compensated by shifting the AWG chip temperature when the ITU
channel is changed (see, eg, top of FIG. 28). Second, the ideal AWG
chip temperature set point depends on the ambient or case
temperature (see, eg, bottom of FIG. 28). Therefore, the AWG
channel spacing can be made more accurate by calibrating the
performance over wavelength and ambient temperature for different
control set points, and then adjusting the control set point in
normal operations depending on the ITU channel and ambient
temperature. Examples of this calibration are shown in both graph
2700 in FIG. 27 and graph 2800 in FIG. 28. The vertical temperature
scale in these figures is expressed in a change of a thermistor
resistance.
[0257] Several examples of the AWG performance under different
operating conditions such as ambient temperature, AWG temperature
setpoint and ITU channel are presented in the graphs shown in FIGS.
29-33. These graphs demonstrate that, by adjusting the setpoint
temperature of the AWG chip according to changes of ambient
temperature and the ITU channel of interest, the center frequency
accuracy of the AWG subchannel spectra can be kept within the
limits required by the system specifications.
[0258] The shift in AWG channel spacing with AWG chip temperature
can also be used to select a different subset of carriers. This is
shown in graph 3400 in FIG. 34, where the operating temperature has
been shifted by approximately 30 degC. to shift the channel spacing
by 50 GHz. Applying this temperature shift enables one to use the
same AWG to demultiplex four subchannels on the 100 GHz grid with
the normal setpoint, or four subchannels on the 50 GHz grid with
the setpoint shifted by approximately 30 degC.
Thermo-Mechanical Considerations
[0259] A high precision and accuracy of the AWG chip temperature
requires a well-designed mechanical enclosure. A detailed design of
such an enclosure is presented in FIGS. 35-37.
[0260] In one embodiment: [0261] An aluminum small top cover for
the AWG chip is employed, such as top cover 3510 in FIG. 35 (a
bottom view of which is illustrated in FIG. 36). This aluminum top
cover 3510 is thermally well attached to the existing bottom
aluminum heat spreader 3520. [0262] A second thermistor 3530 to
monitor the AWG chip temperature is attached directly to the AWG
chip wafer 3540, as close as possible to the optical waveguides at
the wafer center. [0263] The case 3710 (shown in FIG. 37) is filled
with polyurethane foam 3720, including space between the aluminum
wafer cover and plastic top case cover. [0264] A third thermistor
3730 is mounted on the bottom of the case 3710 to measure ambient
temperature.
Electronics Considerations
[0265] The required high precision and accuracy of the AWG chip
temperature also requires accurate control electronics, one
embodiment of which 3800 is presented in FIG. 38, which shows a
Thomson bridge implementation of the chip temperature sensing with
3 Ultra-precise resistors and the AWG chip thermistor. Note that
the ADC is probing the Thomson bridge differentially.
[0266] 4. Subchannel Multiplexer with Polarization Multiplexing and
Demultiplexing
[0267] Many of the embodiments of the present invention discussed
above, including those relating to optical network topologies and
various network elements, are based on wavelength-division
multiplexing--where different frequencies of light represent
different ITU transmission channels and their subchannels.
[0268] Data transport capacity of such networks can be doubled
when, in addition to wavelength multiplexing, polarization
multiplexing is employed. Various polarization multiplexing schemes
can be applied to the wavelength multiplexing systems described
herein, including those implementing subchannel-based
architectures.
[0269] As will be illustrated below, polarization multiplexing can
be used as a means of increasing the subchannel spectral density.
An example of a channel plan 3900 for subchannels in this case is
shown in FIG. 39A. Here the first five subchannels (SC-1 to SC-5)
are aligned along one polarization axis 3910 and the last five
subchannels (SC-6 to SC-10) are aligned along the second
polarization axis 3920.
[0270] In this embodiment, the subchannel transmit laser can be
combined with cascaded polarization combiners, similar to combiners
1511 and 1512 in FIG. 15. At the receive side the subchannels in
orthogonal polarizations can be demultiplexed with the circuit 3950
shown in FIG. 39B. In this circuit, a polarization coupler 3960 is
used to separate the orthogonal polarizations. A polarization
controller 3970 is placed in front of the polarization coupler 3960
to align the subchannel polarization axes to the axes of the
polarization coupler 3960.
[0271] The alignment circuit 3950 shown in FIG. 39B detects the
strength of the subchannel signals in one polarization. This
circuit 3950 assumes that a low-amplitude dither at a fixed
frequency (250 kHz in this embodiment) is superimposed on the
signal transmitters using the first polarization. The control
electronics and software monitor the strength of the received
dither signal 3962 and adjust the polarization controller 3970 to
maximize the signal.
[0272] Key elements of the circuit shown in FIG. 39B include:
[0273] A photodetector 3963 detecting the fraction of light in one
of the outputs of the polarization coupler/splitter; [0274] A band
pass filter (to filter a dither frequency) with an adjustable gain
3964, where the gain adjustment is based on the optical signal
power incoming from a line fiber to the receiver. The
gain-adjusting algorithm assures that the amplitude of the dither
signal at the filter/gain stage output does not change
significantly with significant changes of incoming optical power;
[0275] A clock recovery circuit 3965 to recover the dither
frequency with substantially low time constant; [0276] A lock-in
amplifier 3966 which, synchronously with the recovered clock,
detects the amplitude of the dither signal changing due to
polarization changes of the incoming optical signal; and [0277] A
lock-in amplifier 3967 with substantially smaller time constant of
its output integrator than that of the dither clock recovery
circuitry 3965 to provide a fast feedback signal to the
polarization controller-tracker 3970.
[0278] This circuit can be implemented employing analog electronics
circuitry, or the signal processing can be performed in the digital
domain (eg, by a DSP) after the photodiode analog signal is
converted into the digital domain. Furthermore, both orthogonal
polarizations can be dithered at different frequencies; two
electronics dither processing circuits can be used in parallel,
each optimized for one of the two dither frequencies,
respectively.
[0279] Although it has the added cost of the polarization
controller 3970, this design of a subchannel muxponder has the
advantage that it can double the spectral density by using
polarization as an additional dimension. This embodiment, however,
does not support routing of subchannels from different sites to the
same subchannel receiver since there is only one polarization
demultiplexer per receiver. Polarization-multiplexed subchannels
from different sites could be demultiplexed, though that would
require one polarization controller per subchannel.
[0280] a. Existing Implementation of Polarization Multiplexing
[0281] One embodiment of an implementation 4000a of polarization
multiplexing is presented in FIG. 40A. For a given wavelength,
.lamda..sub.1, of a DWDM system, two independent data channels are
being used: Tx-1 4001a and Tx-2 4002a. The output light of both
lasers is linearly polarized and both polarizations are combined
into a single fiber by a polarization beam combiner (PBC) 4005a in
such a way that the polarizations of Tx-1 4001a and Tx-2 4002a are
linear and orthogonal on the transmit side.
[0282] Both multiplexed polarization channels are added to the
network and propagate through a variety of optical components of
the network such as nodes, wavelength multiplexers and
demultiplexers, fiber, EDFAs, Raman amplifiers, interleavers,
ROADMs, WSSs and so forth. During the propagation, the state of
polarization of each channel Tx-1 4001a and Tx-2 4002a changes due
to birefringent effects of optical network components. Moreover,
since the birefringent effects evolve over time, the state of
polarization changes randomly on the receive end where the
wavelength, .lamda..sub.1 4007a, is dropped.
[0283] Random changes of polarization act on both channels in such
a way that the state of polarization at the receive end of Tx-1
light is still orthogonal to the state of polarization of Tx-2
light. Therefore, polarization demultiplexing can be performed as
long as random polarization of each (or in fact one) channel is
changed to a linear polarization with a known orientation, and the
channels are demultiplexed by a polarization beam splitter/combiner
(PBC) 4015a.
[0284] A change from random polarization to a linear polarization
can be performed by commercially available Polarization Trackers,
such as polarization tracker 4010a. Since dropped polarization
changes over time, a polarization tracker needs to follow these
changes and correct incoming polarization accordingly. This is
typically accomplished by a feedback loop 4020a which detects the
polarization state at the output of the polarization tracker and
provides a control signal 4022a to the tracker 4010a to assure that
the polarization is linear, and a proper polarization channel is
directed to a proper receiver--i.e. the light from Tx-1 reaches the
receiver Rx-1 4011a, and Tx-2 reaches Rx-2 4012a, respectively.
Several known implementations of such feedback mechanisms are
described on the General Photonics website
(http://www.generalphotonics.com/articles.aspx?a=1073).
[0285] One of these implementations presents a scheme 4000a shown
in FIG. 40A in which (on the transmit end) the amplitude of Tx-1
4001a is modulated by a small sine wave signal at frequency
f.sub.dith-1=100 kHz. The modulation depth typically does not
exceed a few percent of the average light intensity. The amplitude
of Tx-2 4002a is not modulated at all.
[0286] At the receive end the polarization tracker dithers
polarization at a frequency f.sub.dith-2 and a photodiode monitors
light intensity in one arm of the polarization beam splitter. The
photodetector detects light from Tx-1 4001a and Tx-2 4002a, both
dithered in intensity by f.sub.dith-2. In addition, the light from
Tx-1 4001a is also dithered in intensity by f.sub.dith-1.
[0287] The signal 4014a from the photodetector 4013a is processed
by feedback control electronics 4020a and input as a feedback
signal 4022a to the polarization tracker 4010a.
[0288] For control purposes: (i) the dithering of the tracker 4010a
at f.sub.dith-2 is used to determine the direction in which to
adjust the polarization in order to track it, if needed, to
accomplish polarization demultiplexing; and (ii) the dithering of
the Tx-1 4001a at f.sub.dith-1 is used to maximize the amplitude of
this dither in the Rx-1 arm of the PBC 4015a at the receive end and
direct a proper transmitter signal to a proper receiver.
[0289] Various novel implementations of this basic approach to
polarization multiplexing are described below.
[0290] b. A Particular Implementation of the Feedback Control
Electronics in FIG. 40A
[0291] FIG. 40B presents a particular implementation 4000b of the
feedback control electronics in the dithering scheme presented in
FIG. 40A. In this embodiment, two electrical circuits are
processing in parallel: (i) a Tx-2 dither at frequency f.sub.dith-1
and (ii) a polarization tracker dither at frequency
f.sub.dith-2.
[0292] The abbreviations in FIG. 40B include the following:
[0293] BPF--band pass filter
[0294] LP--low pass filter
[0295] RMS--root mean square
[0296] TZ--transimpedance amplifier
[0297] ADD--adding two electrical analog voltages
[0298] As shown in FIG. 40B, just before the feedback signal 4022b
is input to the polarization tracker 4010b, it is amplified by a
variable gain amplifier 4025b, where the amplifier gain is adjusted
appropriately to accommodate for changes in optical input power of
a dropped channel while keeping at a constant value the average
voltage of the feedback signal at the input of the polarization
tracker. Alternatively, the variable gain amplifier can be placed
at the output of a TZ 4030b.
[0299] The RMS detector 4035b (such as LTC1968 from Linear
Technologies) can be replaced by a combination of a clock recovery
circuit 4036b recovering dither frequency at f.sub.dith-1 followed
by a lock-in amplifier 4037b which transforms its AC input signal
into DC.
[0300] In a case when f.sub.dith-2 is substantially lower than
f.sub.dith-1, two parallel paths can be realized as shown in FIG.
40C.
[0301] The processing algorithms shown in FIG. 40B and FIG. 40C
(and all other Figures discussed herein and illustrating
embodiments of polarization multiplexing) can be implemented by
using analog electronics or DSP in the digital domain. In the
latter case, the TZ output signal could be sampled by an ADC, and
DSP processing output could drive a DAC and provide feedback
voltage to the polarization tracker.
[0302] One advantage of such solutions over existing art is that,
in each parallel signal processing path, different processing can
be implemented--e.g. different gain, different spectral transfer
function (e.g. shape of a BPF, LPF), etc.
[0303] c. A Novel Polarization Multiplexed System Based on 3
Dithering Frequencies
[0304] A known method of implementing dithers (see FIG. 40A) can be
significantly improved by adding an additional third dither to Tx-2
at a frequency f.sub.dith-3 4050d as presented in FIG. 40D.
[0305] d. A Particular Implementation of the Feedback Control
Electronics in FIG. 40D
[0306] A particular implementation 4000e of feedback control
electronics for three dithering frequencies is presented in FIG.
40E. In this embodiment, both dithers at the receive end are
detected in two parallel arms of the feedback control electronics,
and after filtering they are subtracted.
[0307] As a result, dithering amplitude at frequency f.sub.dith-1
is maximized and dithering amplitude at f.sub.dith-3 is minimized,
providing substantially better extinction ratio of polarization
tracking. As above with respect to FIG. 40D, parallel paths can be
implemented in two different ways depending on the relationship
between the values of f.sub.dith-2 and f.sub.dith-1--cf FIG. 40E to
FIG. 40F.
[0308] e. Subchannel-Based DWDM Implementations of the above
Polarization Multiplexing Schemes
[0309] The techniques in the above embodiments of dithering and
feedback control electronics can be further enhanced by introducing
subchannels (as discussed above) to the ITU grid of frequencies.
FIG. 40G illustrates one of the possible embodiments (employing
subchannels 4075g) based on the control electronics illustrated in
FIG. 40B.
[0310] While all of the dithering and feedback control electronics
schemes discussed above can be applied to the subchannel
architecture 4000g presented in FIG. 40G, one such example 4000h
(with respect to the feedback control electronics illustrated in
FIG. 40C) is illustrated in FIG. 40H.
[0311] f. A Polarization Multiplexing System without Dithering of
Transmitting Lasers
[0312] In the polarization tracking schemes described above, one or
more dither frequencies are required on the transmit end. In many
applications, introducing dither at the transmit end and detecting
it at the receive end may not be practical. For example, during
propagation of the signal from the transmit end to the receive end,
the dither signal applied to the transmitters can be distorted to
the extent that it is not useful anymore at the receive end as a
valid signal for a feedback loop to track polarization. FIG. 40I
illustrates a novel embodiment 4000i of a polarization tracking
scheme applied to the subchannel architecture of the present
invention which does not require any dithering at the transmit
end.
[0313] In this scheme, the distinction between two orthogonal
polarizations on the transmit end is implemented in the wavelength
domain, i.e. subchannels in orthogonal polarizations are offset by
half of the subchannel spacing, as shown in element 4080i. Note
that the wavelength offset between subchannels in two orthogonal
polarizations does not need to be exactly half of the subchannel
spacing; the offset can be anywhere between half the channel
spacing to zero offset. The selection of the offset in any
particular implementation may depend on the accuracy of the
polarization tracker and the allowed total signal bandwidth. The
feedback control electronics is designed to maximize light
intensity of one or more subchannels (or all subchannels in both
polarizations) after the input signal is demultiplexed into
separate single subchannels. The particular embodiment 4000i
presented in FIG. 40I shows a case of the feedback control
electronics maximizing intensity of two subchannels--one for each
polarization as an example. Note that instead of tapping the output
of each cyclical filter, the signals can be taken from the receiver
bias currents.
[0314] In these polarization tracking schemes, there can be a
requirement for an add/drop node where signals are being added to
the optical network at the same node where signals are optically
passing through. When polarization multiplexing of subchannels is
used as shown in FIG. 39A, there can be a requirement to match the
polarization of the added signals to the polarization of the
passthrough signals. The embodiment 4000j presented in FIG. 40J
shows a polarization recovery module 4010j that monitors the
polarization of signals coming into a node and adjusts a
polarization tracker 4020j so that the polarization of the signal
at the output of the polarization tracker 4020j is aligned with a
linear axis of a polarization maintaining coupler 4030j.
Polarization-maintaining fibers 4040j are used on the add side of
the node to ensure that the polarization of the added signals is
aligned with the passthrough signals. The embodiment shown is for
the case of a 1.times.2 WSS node. Those skilled in the art will
recognize that a similar design can be used for other types of
passthrough nodes where a polarization tracker with a design shown
in one of the embodiments of FIGS. 40A-40I is used to align the
incoming signal polarizations to a linear axis of the add/drop node
and polarization maintaining fibers and optical components are used
in the node to ensure that signals are being added along the same
linear axis.
[0315] 5. Subchannel Mapping of Client Services on the Subchannel
Muxponder
[0316] Client data can be mapped to the subchannels as shown in
FIG. 41, in which data from independent 10 Gb/s sources 4110 is
mapped directly to each subchannel 4120. In each 10G data path an
FEC encoding device 4125 (either an FPGA or ASIC) is used to encode
the data according to an error correction algorithm that improves
the optical performance [G.709 and G.975]. Overhead (OH) data can
also be inserted in the FEC overhead to enable the exchange of
OAM&P (operations, administration, maintenance, and
provisioning) data to be exchanged between the terminals.
[0317] Note that a 10 Gb/s crossconnect 4115 may be inserted
between the input data 4110 and the subchannels 4120. This may be
either a digital or analog electronic crossconnect, or optical
crossconnect [see, eg, U.S. Pat. No. 6,574,386]. It can also be a
protocol-dependent switch such as an ethernet switch. This enables
more flexible functionality as described below.
[0318] The design 4100 shown in FIG. 41 is flexible in that any 10G
protocol (e.g., 10 GE, OC-192, FC-10, etc.) can be connected to any
input port 4110. Each port 4110 has independent clock recovery and
clock multiplication circuits, and can be assigned, via optional
crossconnect switch 4115, to any subchannel 4120. After the data
associated with each subchannel 4120 is encoded (and overhead data
inserted) via FEC encoding device 4125, such data is then sent to
be modulated onto the subchannel's associated subcarrier wavelength
4130.
[0319] It should be noted that tuning a laser to map a client
circuit to a subchannel may require a relative long period of time,
e.g., approximately a minute. Yet, if a subchannel laser has
already been tuned to a particular frequency (subcarrier
wavelength), and a client circuit is being mapped to that
subchannel via crossconnect switch 4115, then the process will
typically require much less time, as the switching time of a switch
such as crossconnect switch 4115 is typically much faster than the
time required to tune or retune a laser.
[0320] FIG. 42 shows an embodiment 4200 where standard digital
multiplexing and switching circuits 4212 are used to multiplex
lower rate data 4210 up to 10 Gb/s. With modular data mapping and
multiplexing, any subchannel can carry either a native 10 Gb/s
service or multiplexed lower-rate services.
[0321] FIG. 43 shows a third mapping method 4300 where data 4310
with a bandwidth greater than the per-carrier bandwidth is
transmitted. In this embodiment, the client data 4310 is inverse
multiplexed (via inverse multiplexer 4313) to divide the data into
4 separate data streams 4314 (each of which can be assigned to a
respective subchannel 4320, e.g., via optional crossconnect 4315).
Inverse multiplexing is well known in the art. For example, IMA
(Inverse Multiplexing for ATM) is a standardized technology used to
transport ATM traffic over a bundle of T1 or E1 cables using
inverse multiplexing.
[0322] In the mapping method 4300 shown in FIG. 43, inverse
multiplexing can be implemented in an Field-Programmable Gate Array
(FPGA) or Application-Specific Integrated Circuit (ASIC). Inverse
multiplexing at the transmit side must be done in such a way that
the original traffic stream can be recovered at the receiver.
Inverse multiplexed data streams have frame markers for the
receiver to re-align the data. The frame markers can be based on
standard protocols, such as SONET, or they can be proprietary. The
receiver also has buffers to hold data before it is realigned. The
buffers typically have enough memory to compensate for any skew in
the network caused by variations in the propagation speed of the
different inverse multiplexed data streams.
[0323] 6. Line Interface Between Subchannel Muxponders and
Lower-Rate Transponders and Muxponders
[0324] In network applications where lower-bandwidth satellite
nodes feed into a hub node, there is a need for a cost effective
solution that supports high bandwidth at the hub and low bandwidth
at the satellite nodes. This can be achieved by an application with
subchannel muxponders at the hub node and lower-rate transponders
or muxponders at the satellite nodes. This requires that the
lower-rate transponders have (1) the same laser tuning capability
as the subchannel muxponder (2) a modulation format that is
compatible with the subchannel muxponder modulation, and (3)
optical filtering to select the subchannels. The optical filtering
may be ITU channel filtering only, as long as the subchannel
muxponders are not using more than one subchannel per ITU
channel.
[0325] FIG. 44 shows an application 4400 with a pair of subchannel
muxponders C1-C2 at Node 1 and pairs of lower-rate transponders
(C3-C4, C5-C6 and C7-C8) at Nodes 3, 4, and 5, respectively. The
subchannel muxponder has its transmit subchannels set as (1)
ITU-210, SC-1, (2) ITU-220, SC-2, (3) ITU-230, SC-3, and (4)
unassigned. A single ITU channel filter (from a FOADM or ROADM) at
the satellite nodes (e.g., drop filter 4410 at Node 2) can be used
to filter out the subchannels from the subchannel muxponder. Each
transponder at the satellite nodes must be set to transmit at one
of the subchannel frequencies being received at Node 1. In this
embodiment, the transponders at Node 1 are set to ITU 200, SC-1;
the transponders at Node 2 are set to ITU-200, SC-2; and the
transponders at Node 3 are set to ITU-200, SC-3. These subchannels
are then dropped by the cyclical filter in the subchannel
muxponders at Node 1. Note that this application 4400 requires that
the lasers in the lower-rate transponders must have the same degree
of accuracy as the subchannel muxponder lasers.
B. Dynamic Network Visibility
Facilitating Network Upgrades and Reuse of Legacy Equipment
[0326] Even when improved functionality is available, such as the
novel subchannel architectures described above, it is still
desirable to minimize the time and expense, as well as disruption
to live networks, associated with upgrading hardware and software,
as well as to reuse legacy equipment whenever feasible. Various
novel techniques along these lines are described below.
[0327] 1. Network Upgrades with Minimal Disruption and No East-West
Ambiguity
[0328] As mentioned above, and shown in FIG. 9, it can be difficult
to ensure proper connections between modular line cards, such as
muxponders and add/drop filters. In ring networks, mistakes can
occur between the east and west connections. The control and
monitoring of added signals can also be facilitated if the fixed or
reconfigurable filter module has monitors on all of its input
ports. These monitors do not prevent misconnections, but they do
provide a means of troubleshooting misconnections.
[0329] We propose two other methods of removing the east-west
ambiguity and selectively routing traffic in either direction or
both directions. The first is shown in FIG. 45 where the muxponder
4510 is connected to a pair of 1.times.2 software-controlled
switches 4520 that control the direction of transmission and
reception.
[0330] If an option for broadcast to both directions is required,
as with protected circuits, then the embodiment shown in FIG. 46
can be used to selectively transmit the traffic to the east, west,
or both directions. The software-controlled switches 4520 in FIG.
45 and 4620 in FIG. 46 remove the east-west ambiguity and enable
network operators to remotely reconfigure the direction of the
traffic.
[0331] This embodiment 4600 is independent of the type of channel
multiplexing and demultiplexing used to add and drop channels from
the network. For example, it can be connected to ROADMs and WSS
networks.
[0332] 2. ITU-Channel Based Network Upgrades to Subchannel
Networks
[0333] Cascaded ITU channel filters and cyclical filters can be
used to upgrade the capacity of WDM networks based on the ITU grid.
FIG. 47 gives an example of a legacy ITU channel-based WDM network
4700 that has been upgraded with subchannels 4710a and 4710b on ITU
channels 192.1 GHz 4720a and 192.2 GHz 4720b, respectively. In this
embodiment, the subchannels increase the capacity of those ITU
channels by a factor of 4. Having the subchannels 4710a and 4710b
at the same bit rate as the legacy ITU services 4720c (on ITU
channel 192.3 GHz 4720c) means that they have similar link budget
rules as the ITU services. This upgrade can therefore be done
without significant changes to the fiber plant and dispersion
compensators since the subchannels 4710a and 4710b are transmitting
at the same date rate (10 Gb/s in this example) as the ITU channels
that are being replaced.
[0334] The network upgrade is further simplified by the high
dynamic range (>20 dB) of the subchannel transmitter (provided
by the VOA 1515 in FIG. 15) and the subchannel receiver (provided
by the EDFA 1522 and VOA 1523 in FIG. 15). These ranges demonstrate
that a legacy network can be upgraded without adding power control
elements to the network such as fixed attenuators.
[0335] Currently installed ITU channel networks can therefore be
upgraded incrementally according to bandwidth demands, with minimal
changes to the installed infrastructure (ITU channel filters,
amplifiers, and dispersion management). Being able to reuse the ITU
hardware provides a distinct cost advantage. Subchannel muxponders
also provide additional functionality as described below.
[0336] The subchannel upgrade described herein supports a very
large network capacity. For example, FIG. 48 shows a network
capacity of 7.0 Tb/s with 704 subchannels. Subchannels are spaced
at 12.5 GHz in both the Conventional (C-band) and Long-wavelength
(L-band) bands. Not only does this subchannel embodiment 4800
provide large capacity, it does so with a small granularity (10
Gb/s), while maintaining client signal synchronization. Subrate
multiplexing can also be combined with subchannels to provide even
finer granularity as shown in FIG. 42.
[0337] Further note that the subchannel implementation based on the
cyclical filter described above provides the means of
deinterleaving channels spaced at 50 GHz so that an external
interleaver is not required in the embodiment shown in FIG. 48.
[0338] 3. Estimate of OSNR at a Receiver
[0339] Optical amplifiers, such as erbium-doped fiber amplifiers
(EDFAs), can be deployed in a network to compensate for the optical
fiber loss. But optical amplifiers add amplified spontaneous
emission (ASE) noise to the signal. This diminishes the optical
signal to noise ratio (OSNR). OSNR is a significant factor that
affects the performance of an amplified optical network. Therefore,
when a network is being deployed and upgraded, it is desirable to
have a measurement of the OSNR of a signal. This can be measured
with an instrument such as an optical spectrum analyzer (OSA).
However, adding full-spectrum channel monitors at all network
points can add significant cost to the network.
[0340] OSNR is given by the ratio between the optical signal power
to the ASE noise power in a given noise bandwidth. Typically 0.1 nm
is used in the industry as the noise bandwidth. For example,
consider the following 2001 IEEE article ("OSNR Monitoring
Technique Using Polarization-Nulling Method," IEEE Photonics
Technology Letters, vol. 13, p 88 (2001)), which presents a method
of measuring OSNR of a link by measuring the ASE noise that is
orthogonal to the signal. This method requires additional equipment
to be deployed in the field. We disclose here a lower-cost method
of using existing optical networking equipment to measure the OSNR
of the signals. This method uses the signal transmitter and
receiver hardware to measure the OSNR, so it does not require any
additional equipment.
[0341] Measuring OSNR with a high dynamic range requires that the
optical receiver be able to monitor power with a high dynamic
range. This can be accomplished by placing a resistor in series
with the photodiode bias current and measuring the voltage across
the receiver with a log amp connected to an analog to digital
converter (ADC).
[0342] First, the receiver and filter in front of the receiver,
e.g. the cyclical filter in the case of a subchannel muxponder, are
calibrated in the factory. The receiver can be calibrated by
inserting light with a known power level and measuring the response
of the log amp. The filter can be calibrated with a wideband source
(such as ASE from an EDFA) that is calibrated with a commercial
OSA. Then, assuming that the source has noise spectral density
S(W/Hz), and the noise power measured on the receiver is P(W), the
equivalent noise bandwidth, Bf, of the cyclical filter is given by
Bf=P/S.
[0343] Then the OSNR of a signal can be measured each time the
signal is enabled or tuned to a new channel or subchannel. This
assumes that the noise level is independent of the signal level,
which is typically the case in a multichannel DWDM network with
gain-controlled and gain-flattened amplifiers. The receiver first
records the power before the channel is enabled, so that it is
measuring the noise power, Pn (W).
[0344] After the channel is enabled, the receiver records the
signal plus noise power, given by:
Pt=Pn(W)+Ps(W)
Then the signal to noise ratio is given by
OSNR(dB)=10*log 10(((Pt-Pn)*Bf)/(Pn*Br))
[0345] where Br is the reference noise bandwidth.
[0346] This method requires that an idle receiver constantly update
and store its noise power in memory and use that power level to
calculate the OSNR after the signal is enabled. It has the drawback
that it cannot track network changes that alter the OSNR after the
signal is added. But, it gives an indication when the channel is
first enabled if the OSNR is too low. After the channel is added,
the BER can be tracked to indicate network degradations. This
method has the advantage of providing software with only the added
cost of placing the log amp and ADC after the receiver, calibrating
the receiver and filter, and recording the OSNR.
[0347] In a multichannel DWDM network with gain-controlled and
gain-flattened amplifiers, the OSNR depends on the transmitter
launch power to first order. Using this method to monitor OSNR when
channels are added enables the management software to equalize the
OSNR of the channels at the receiver. After channels are enabled,
the measured OSNRs provide an estimate as to how much the transmit
powers need to be adjusted. For example, if one channel has OSNR 1
dB lower than the other channels, it can be equalized by increasing
its launch power by approximately 1 dB.
[0348] Although this method can be used by other WDM equipment, the
requirement of calibrating the effective bandwidth of the optical
filter means that the accuracy is diminished unless the filter is
located on the same circuit pack as the receiver, as in the case of
the subchannel muxponders described herein.
[0349] 4. Measurement of EDFA OSNR Contributions
[0350] When networks are being upgraded, it is desirable to have
visibility of the added channel at all network sites to aid in
ensuring good performance of the added channel and debugging any
issues. Adding full-spectrum channel monitors at all network points
can add significant cost to the network. We disclose here a
lower-cost method for out-of-band and out-of-service monitoring of
an optical path. When a channel is out-of-service, it can be tuned
to the monitor wavelength and the power levels at each node can be
measured and communicated by the management software.
[0351] FIG. 49 shows an embodiment 4900 of a method to track added
channels 4910 and monitor out-of-service channels. An unused
channel on the edge of the amplifier gain spectrum is used for this
purpose, e.g. at 1528 nm. Using a channel on the edge of the gain
spectrum will not interfere with the usable channels, but there
will still be some gain so that the edge channel propagates through
the amplifier chain.
[0352] Placement of the monitors depends on the network
configuration but it is ideally before 4920 and after 4930 each
amplifier 4925 as shown in FIG. 49. The monitor 4920 before each
amplifier 4925 can be used to estimate the OSNR contribution from
each amplifier 4925 [see, eg, U.S. Pat. No. 6,040,933] and the
difference between the input and output powers can be used to
estimate the gain of each amplifier 4925. Note that the gain at the
monitor wavelength may be lower than the gain in the amplifier's
signal band; yet this difference can be calibrated in the factory
for each amplifier 4925.
[0353] Therefore, a signal laser can be tuned to the monitor
channel and its power levels can be verified at each network point.
The power levels can be compared to the existing channels to
balance the power of the added channel relative to the live
channels. After the performance has been verified at the
out-of-band monitor frequency, the channel can be tuned to its
designated frequency. Standby channels and protection channels can
also be periodically tuned to the monitor frequency for
verification.
[0354] The noise of the amplifier 4925 can also be estimated in
this configuration by disabling the monitor transmitter. In that
case the output monitor photodiode (e.g., 4930b) will measure the
noise at the monitor wavelength integrated over the passband of the
filter (e.g., 4930a). The passband of this filter can be calibrated
so that the noise measured by the photodiode 4930b can be given
relative to a reference noise bandwidth, such as 0.1 nm.
[0355] If the amplifier 4925 has its gain and noise vs. wavelength
stored in a calibration table, then the gain and noise at other
signal wavelengths can be determined from the gain at the monitor
wavelength.
[0356] 5. Determining the Net Chromatic Dispersion and
Polarization-Mode Dispersion of a Fiber Link with Subchannel Delay
Times
[0357] The performance of an optical link depends on the net
dispersion of the fiber and components in the path such as
amplifiers and filters. System operators often do not have a record
of the precise values of the dispersion of installed fibers. If the
net dispersion of the link is outside of the allowed range of the
transmitters being used, then dispersion compensators may have to
be installed. If the dispersion of a link is not known precisely,
then whether or not dispersion compensation is required may not be
known a priori. Deploying a network with incorrect dispersion
compensation can cause bit errors that are difficult to debug.
[0358] To be precise, before installing or upgrading an optical
network, field technicians may have to measure the dispersion of
each fiber in the network. Measuring the fiber dispersion can be
done with commercially available dispersion testers [eg, Exfo
FTB-5700]. Portable dispersion measurement equipment, and travel
and labor costs to perform these measurements, can be quite
expensive.
[0359] We therefore propose a solution to this problem in which the
transmission equipment is used to measure the net dispersion of the
fiber link. This measurement requires a subchannel muxponder like
the one shown in FIG. 15. Measuring the fiber dispersion requires
that at least two of the subchannels be in a maintenance mode
because the measurement will disrupt traffic on the subchannels
that are used. This is the case when a service is being
installed.
[0360] The dispersion of a fiber link is given in units of ps/nm.
This determines the delay in ps per nm spacing between two
carriers. The delay can be measured on a subchannel muxponder line
card (employing a circuit such as circuit 5000 shown in FIG. 50) as
follows:
(1) Set the subchannels to use the same clock source such as
reference oscillator 5010 on the board normally used as the
reference clock for maintenance signals such as ODU-AIS. (2) The
same reference signal must be transmitted on the subchannels. For
this purpose a user-defined test sequence 5020 can be programmed
into most commercially available SERDES or FEC devices. (3) The
SERDES or FEC devices for the two subchannels can be synchronized
by simultaneously releasing them from reset. (4) At the receive
side the phase difference between the two subchannels can be
measured with a commercially available phase detector 5030 such as
the AD8302 from Analog Devices. This assumes that the voltage V
from the phase detector 5030 has been calibrated to provide a
constant C in (ps/V) to be used in the calculation. (5) From these
measurements, the net dispersion of the fiber link (in the measured
phase difference) is given by:
D=V*C/.DELTA..lamda. where .DELTA..lamda. is the wavelength spacing
between carriers.
(6) To get high resolution and wide range, adjacent subchannels can
be used to measure larger dispersion values, and the outside
subchannels can be used to measure smaller dispersion values. For
example, with a device like the AD8302 that provides an output of
roughly 10 mV/ps of phase difference, if the fiber dispersion is
2000 ps/nm, then the output of the phase detector 5030 will be 6 V
for subchannel spacing of 37.5 GHz (or 0.3 nm) and 2 V for a
subchannel spacing of 12.5 GHz (or 0.1 nm). In this case the 6 V
theoretical value may saturate the electronics so the measurement
from the closer spaced channels would be used.
[0361] Note that measuring dispersion requires software
communication between the transmitter and receiver. We assume that
an OSC link or in-band communication channel are available for this
purpose. Also note that this measurement can only be done
out-of-service, e.g. when the service is being installed. The
measurement process can be as follows: [0362] Network management
sends a command to the link (transmitter and receiver) to measure
dispersion [0363] Transmitter uses one subchannel for overhead and
signals over that subchannel to the receiver that it is switching
to dispersion measurement condition [0364] Transmitter switches the
other 3 subchannels to the common reference clock and inserts the
test sequence on those channels [0365] Receiver measures the phase
differences between two of the subchannels [0366] The receiver can
repeat the measurement for other channels [0367] Software
calculates the dispersion seen by the signals and raise an alarm if
it is out of range
[0368] Note that there may be an unknown delay between channels,
for example from delays in the first-in first-out (FIFO) buffers in
the serializer. The error from these unknown delays can be
eliminated by measuring the phase differences versus the channel
spacing. The subchannels to be measured can be tuned toward each
other, and away from each other, while measuring the phase
difference. Several points measured in this fashion can then be fit
with a least-squares fit to get the slope, and hence dispersion.
The tuning range depends on the passband of the cyclical filter. An
example of a phase difference versus channel spacing measurement is
shown in graph 5100 in FIG. 51. Note that extrapolating the
difference to zero frequency difference leads to a fixed delay
caused by the electronics. The slope of this curve gives the factor
V/.DELTA..lamda. in the equation above, and the fixed offset does
not affect the measurement.
[0369] Alternatively, we can eliminate the unknown transmitter
phase difference by making a measurement with transmitter 1 sending
.lamda.1 and transmitter 2 sending .lamda.2, and then swapping the
lambdas and subtracting the results. Assuming that transmitter 1
has fixed delay Td1 and transmitter 2 has fixed delay Td2, and the
transmitter delay sending .lamda.1 (T.lamda.1) is the same as the
transmitter delay sending .lamda.2 (T.lamda.2), then:
Tdiff1=(Tx1+T.lamda.1)-(Tx2+T.lamda.2)
Tdiff2=(Tx1+T.lamda.2)-(Tx2+T.lamda.1)
Tdiff1-Tdiff2=2(T.lamda.1-T.lamda.2)
[0370] Then the measured dispersion (or delay between the two
wavelengths) is given by the following expression that does not
contain the unknown transmitter delays:
T.lamda.1-T.lamda.2=(Tdiff1-Tdiff2)/2
[0371] Furthermore, this method can be used to estimate the
polarization-mode dispersion (PMD) seen by the signals. PMD is a
form of modal dispersion where two different polarizations of light
in an optical fiber propagate at different speeds due to random
imperfections and asymmetries. PMD is a statistical effect, and it
depends on alignment of the launched state of polarization (SOP).
When PMD is measured by a dedicated instrument according to
standard FOTP-124, the polarization of the measurement light source
is scrambled so that the instrument can average the measured value
over all polarization states.
[0372] In one embodiment of the current invention, the relative
dispersion between two subchannels can be measured as described
above. If the subchannel muxponder in FIG. 15 is used such that the
carriers have orthogonal polarizations, then the delay between two
wavelengths can be measured for parallel polarizations and
orthogonal polarizations.
[0373] If subchannels 1 and 3 have the same polarization, and
subchannels 2 and 4 have the same polarization, and the
polarization of subchannels 1 and 3 is orthogonal to the
polarization of subchannels 2 and 4, then the method above can be
used to measure the following dispersions (as shown in graph 5200
in FIG. 52):
D12=dispersion measured with subchannels 1 and 3 in ps/nm
D24=dispersion measured with subchannels 2 and 4 in ps/nm
[0374] The difference in those delays is then equal to the PMD for
the SOP held by the subchannels. This is of more relevance than the
PMD averaged over all SOPs, because it will be the PMD that affects
the actual signal transmissions. Statistical averaging over time
can be used with this technique to get the statistical distribution
of PMD.
[0375] A distinct advantage of these measurement techniques is that
they measure the cumulative CD and PMD seen by the signal in a
single measurement. Another advantage is that these methods can
measure the link without requiring a technician to break the link
to insert external test equipment. Other methods may only measure
the characteristics of the transmission fiber in several steps, and
not measure any contributions from the optical modules used for
transmission. A further advantage of these methods is that the
results can easily be displayed by the network management
software.
[0376] All transmitters have an acceptable dispersion window, i.e.
range of dispersion values for which the transmit signal will have
an allowed dispersion penalty. If the measured network dispersion
or PMD is out of this range, then the software can raise an alarm
to the network operator to signal that dispersion compensation is
required in the network.
[0377] 6. Network Upgrades with Minimal Disruption of Live
Traffic
[0378] The traffic demands on an optical network evolve over time.
As the demand changes, wavelengths and subchannels may have to be
added or removed from service. These changes are typically done
during scheduled maintenance windows. If human, hardware, or
software errors occur during a network change, then some or all of
the static traffic may be adversely affected. We therefore disclose
a method for software to control a network upgrade to make the
traffic changes as non-disruptive as possible.
[0379] When a channel is added, or the transmitter wavelength is
tuned to a different channel, the output power must be disabled to
avoid interference on the other channels. Then, when the
transmitter power is enabled, it should be turned on gradually to
avoid any adverse effects on the live channels. While it is being
enabled, the BER of the other channels should be monitored for
increased BER. One embodiment of a sequence for tuning a channel is
shown in graph 5300 in FIG. 53. This requires software
communicating with all nodes to monitor the BER (e.g. at 5310) of
all the live channels while turning on the transmit power (e.g.
starting at 5320). A similar procedure can be followed when a
channel is first enabled.
[0380] A procedure for adding a channel with the techniques
described above is as follows: [0381] 1 Install the subchannel
muxponders and connect the fiber jumpers (transmitter is disabled
until the traffic is assigned) [0382] 2 Use the bandwidth map to
assign the subchannel frequencies [0383] 3 Assign the protection
bandwidth [0384] 4 Assign the traffic direction [0385] 5 Assign the
services (protocols) [0386] 6 Tune the transmitter to the EDFA
monitor wavelength, turn on the power to a level that equals the
other channels and check the OSNR contributions of the amplifiers.
Record the OSNR at the receiver and raise an alarm if it is too
low. [0387] 7 Turn off the transmitters and tune the subchannels to
the designated frequencies [0388] 8 Gradually turn on the transmit
power as described below. [0389] 9 Establish an overhead channel
between the two end points with the line side at a predetermined
rate. This is easier to do with OTN framing, where the overhead
channel is independent of the protocol being used. It is also
easier if the same rate is used for the different protocols. Use
the overhead channel to exchange path trace information (raise an
alarm if there is a path trace mismatch), IP and subchannel
addresses. [0390] 10 Measure the dispersion (CD and PMD) seen by
the subchannels as described above, and raise an alarm if the
dispersion is out of spec [0391] 11 Enable the traffic and monitor
the BER, raise an alarm if it is too high [0392] 12 Traffic is
established, start the traffic monitors and update the performance
status
[0393] 7. Diagnostic Tool to Capture all Digital Data and Compare
to Known Values
[0394] Electronic equipment, such as the optical equipment
described herein, can have several digital devices with registers
containing configuration and status data. These registers may be
implemented in custom-designed or off-the-shelf Application
Specific Integrated Circuits (ASICs), field programmable gate
arrays (FPGAs), and/or complex programmable logic devices (PLDs).
Registers may hold provisioning information (such as the bit rate),
loopback condition, or data protocol, and contain alarm conditions
and performance monitoring data. Typically, real-time operating
system (RTOS) software configures the registers depending on
instructions from the network operator, and reads the registers to
discern status.
[0395] With the development of complex ASICs and FPGAS, there may
hundreds or thousands of registers on a line card. The RTOS
software may provide commands to users to read certain registers,
but typically it is difficult to get full visibility of all the
digital data on a line card and to debug issues on the line
card.
[0396] We propose here a diagnostic tool that reads all ASIC, FPGA,
and CPLD registers on a line card, compares the read values to
expected values and reports the differences to a user for
debugging, and/or uses the differences to provide debugging advice.
The interface is a spreadsheet, created for example with Microsoft
Excel, that has been programmed to contain a list of all the
devices on a line card with one spreadsheet tab per device.
[0397] FIG. 54 shows an example of such a spreadsheet 5400. The
first column 5410 lists the register numbers of a device, the
second column 5420 lists the address, the third column 5430 lists a
name for the register, the fourth column 5440 lists the type of
register (RW=read/write), the fifth column 5450 lists the expected
value of the data read from each register, the sixth column 5460
reads the actual value, the seventh column 5470 lists the
difference between the expected and actual columns, and the eighth
column 5480 gives a description of each register.
[0398] The spreadsheet 5400 is first generated by a person that
transposes the functional specification for each device to the
spreadsheet. Reading the data from the line card requires that the
user have a communications link from their laptop to the line card.
This can be done with a serial debug port or standard protocol such
as telnet. The expected values can be generated from the functional
specification, or by reading a device in a known good state and
copying the read values from the "Actual" column to the "Expected"
column. Differences between read and expected values can be
selectively highlighted, as shown in element 5490. Several standard
configurations can be stored in the spreadsheet so that the user
can compare values depending on the expected card configuration.
The spreadsheet has an "Update" button 5495 that triggers a macro
to read the registers on the line card (user can select which
devices are updated), copy the data to the appropriate "Actual"
column in the spreadsheet, and highlight any differences between
the actual and expected values. The spreadsheet can also be
programmed to hide data when the actual and expected values are
equal.
[0399] This diagnostic spreadsheet 5400 therefore provides a
network operator, field support engineer, or design engineer with a
quick method to determine if a line card is correctly configured,
or if there are any current alarms or hardware faults.
[0400] 8. Network Upgrades with Minimal Software Upgrades to Legacy
Equipment
[0401] As telecommunications products evolve, there can be a need
to add new products that reuse the shelf, backplane and/or
management card of a legacy product. Furthermore, the new product
release may have software that is not necessarily backwards
compatible to the legacy product. This may be the case because the
new product is enhanced with a new software architecture, and
making the new software backwards compatible may require a lot of
time and investment in engineering. To get a new product to market
sooner, the new product may be deployed alongside the legacy
product in two separate shelves 5510 and 5520 as shown in FIG. 55.
This implementation 5500 has the disadvantage that it requires
extra shelf space for the new product even though there may be
empty slots in the chassis of the legacy product.
[0402] We therefore disclose a method of combining the new product
and legacy product in a single shelf 5610 as shown in
implementation 5600 in FIG. 56. Rather than using the two ethernet
busses to provide redundant management, one ethernet bus 5620 is
used by MGT-1 5625 to manage the legacy cards 5627 and the second
ethernet bus 5630 is used by MGT-2 5635 to manage the new cards
5637. Each MGT has a separate IP address and is connected
separately to the EMS server 5650 via an IP network. In this
embodiment 5600, the new line cards 5637 use the second ethernet
plane and MGT software (1) disables the redundancy feature, (2)
disables handshaking between MGT cards, and (3) disables shared
control lines on MGT-2 5635.
[0403] The EMS 5650 managing shelf 5610 will display it as two
separate shelves (with separate IP addresses). Over time, as the
software in the legacy equipment is upgraded, the legacy line cards
5627 can be upgraded remotely with software downloads to be managed
by MGT-2 5635. Eventually all the legacy cards 5627 will be
upgraded so that all cards are managed by the new software on MGT-2
5635, and MGT-1 5625 is not being used. At that point the software
on MGT-1 5625 can be upgraded with redundancy enabled so that the
chassis is managed with MGT-2 5635 and MGT-1 5635 is on standby to
provide redundant management.
C. Subchannel Routing, Switching, Concatenation and Protection
[0404] Having described the core hardware elements of a
subchannel-based architecture, as well as various techniques (which
can be applied to ITU channel-based, as well as subchannel-based
systems) for facilitating network visibility generally and in the
context of network upgrades (where legacy equipment is reused
whenever feasible), we now turn our attention to various methods
for implementing, on a subchannel-based architecture, the routing,
switching, concatenation and protection of client circuits across
nodes of an optical WDM network. It should be noted that the extent
to which these methods are implemented in software and/or
general-purpose or dedicated hardware is generally a matter of
design choice.
[0405] 1. OSC Options and Routing Protocols
[0406] WDM equipment often requires that the EMS have a management
connection to all remote nodes for functions such as provisioning
equipment, reporting faults, downloading software upgrades, and
retrieving and reporting performance metrics. The node management
card (MGT) also needs a management connection to remote nodes for
end-to-end provisioning, controlling protection switching, and
reporting remote performance and faults. For these functions,
current WDM equipment deploys an optical service channel (OSC) that
is outside of the ITU-T G.692 spectral window, i.e. at 1510 nm or
1620 nm.
[0407] We disclose here an alternative implementation that enables
remote management information to be transmitted over (1) a separate
unamplified wavelength 5710 such as 1510 nm or 1620 nm, (2) a
separate amplified wavelength 5720, (3) the overhead channel 5730
of a subchannel 5735, or (4) an unused portion of the payload (not
shown). In one embodiment, the management software uses OSPF
routing to select the overhead channel with the highest bandwidth.
Other routing protocols such as RIP may also be used. A general
embodiment 5700 of this alternative routing of management traffic
is shown in FIG. 57, and the advantages and limitations of various
different approaches are illustrated in Table 1 below.
[0408] The alternative path selected for any given implementation
is a design choice made after balancing the various advantages and
limitations of each approach, such as those shown in Table 1. As
will become apparent below, use of any of the alternative paths, as
opposed to a dedicated OSC, enhances remote management
functionality by enabling greater network visibility of information
at a lower level of granularity, which in turn facilitates the
detection and repair of problems, as well as the modification and
upgrading of network functionality.
TABLE-US-00001 TABLE 1 Remote Management Path Advantages
Limitations OSC - non amplified wavelength Highest bandwidth Higher
cost to add and drop a special wavelength Not affected by amplifier
Not enough reach for failure extended links OSC - amplified
wavelength Highest bandwidth Affected by amplifier failure Good for
extended links Out-of band overhead (GCC) No extra cost Lower
bandwidth Affected by amplifier and line card failures Unused VLAN
or VCG on muxponder No extra cost Affected by amplifier and line
card failures Higher bandwidth than GCC Needs to be re-routed when
port is assigned to traffic
[0409] Similar to the standard OSPF protocol, management traffic is
routed between nodes based on a metric that is inversely
proportional to the bandwidth of each path. This requires each
independent link to first establish end-to-end connectivity by
handshaking with the remote end. After the connection is
established, each link must publish its availability, end points,
and bandwidth measure. OSPF routing tables on the MGT then select a
route between each node. If required, load balancing can also be
implemented on the management channels. In general, the routing
algorithm and updates follow a standard OSPF implementation.
[0410] 2. Security Application with Subchannel Hopping
[0411] In addition to remote management applications, subchannels
also facilitate secure communications that rely upon existing
optical infrastructure, such as the subchannel muxponder shown in
FIG. 17. The embodiment shown in FIG. 17 can be used to provide
secure communications by constantly redistributing the traffic
among the carriers. This requires handshaking between the terminals
to synchronize the mapping and remapping between the client traffic
and the subchannels. The channel overhead can be used to signal
between the subchannel transmitters and receivers. One embodiment
of an algorithm for such subchannel distribution includes the
following steps: [0412] 1. Assign traffic to the carriers [0413] 2.
Use channel overhead to signal the carrier distribution to the
receiver [0414] 3. Receiver sends acknowledgement [0415] 4. Send
the traffic and start reconfiguration timer [0416] 5. After timer
expires, transmitter generates new random carrier distribution and
signals that distribution to the receiver [0417] 6. Transmitter
buffers starts buffering the traffic or hold off with Pause signals
[0418] 7. When receiver sends acknowledgement for the new channel
acknowledgement, start transmitting with new carrier
distribution
[0419] 3. Optical Routing and Switching at the Subchannel Layer
[0420] Having described details of the design and control of
subchannel muxponders, we now disclose novel applications of
subchannel muxponders for optical routing and switching across
network nodes. In one embodiment, data mapped onto the subchannels
can be routed to any end terminal by selectively and independently
tuning the wavelength of each subchannel, i.e. software tuning a
laser's frequency to select a subchannel.
[0421] As noted above, this is distinguished from prior
implementations using multiple lasers that were constrained to be
(a) fixed within the same ITU window, and (b) transmitted to the
same receive node. This current embodiment uses ITU-T G.692
compliant add/drop multiplexers to select WDM channels incident on
a network element to be dropped at that node or to pass through to
the next node.
[0422] To appreciate the detailed implementation of the
applications described herein, it is helpful to review some general
background information regarding the routing functions and software
used to manage these subchannel routing and switching applications.
Managing a large number of ITU channels, subchannels, and client
services requires multi-layer routing software that uses subchannel
mapping to direct services between endpoints.
[0423] In this regard, FIG. 58 shows the addition of a new
management sublayer, the Optical Subchannel Layer 5810, between
existing FEC Encode/Decode and Wavelength Assignment layers. This
layer 5810 manages the subchannels within each ITU channel.
[0424] To maintain the degree of flexibility necessary to manage
the vast array of options afforded by the use of subchannels, the
network management system responsible for maintaining the
assignments of client services to subchannels throughout an optical
WDM network employs (in one embodiment) the following set of rules:
[0425] Multiplexing lower-rate services into a subchannel is
constrained by hardware capabilities and bandwidth per subchannel
[0426] Inverse multiplexing of higher-rate services to subchannels
should be done on the same ITU channel (although this network
design does not preclude a distribution over different ITU
channels, it is impractical since it requires dedicated hardware at
the WDM receive side to recombine the subchannels) [0427] Routing
tables must provide overall network visibility [0428] The total
number of lower-rate services per subchannel is fixed by the
deployed hardware, but software should be flexible to allow future
upgrades [0429] The total number of subchannels per ITU channel is
fixed by the deployed hardware, but software should be flexible to
allow future upgrades, e.g. from 4 to 10 subchannels [0430]
East-west ambiguity must be resolved, software must control and
track the traffic to fiber mapping [0431] Optional user-defined
labels should be supported for users to refer to services and nodes
with meaningful labels [0432] User is given choices of available
paths [0433] Routing is distance and OSNR aware so that a path with
better optical performance is preferred [0434] Protection traffic
is placed on the side of the ring with worse optical performance
[0435] Reuse of bandwidth at all layers in a network is supported.
[0436] The number of nodes connected to a given node is as low as 1
for point-to-point applications, and as high as N-1, for mesh
applications with an N-port WSS [0437] Interworking between
subchannel muxponders and legacy ITU channel equipment is
supported
[0438] We now describe various methods for managing optical
bandwidth in accordance with the above requirements. In the
examples below, we describe the routing tables for a ring network,
but the tables do not limit the number of nodes that can be
connected to each node.
[0439] When a new service or node is being added to the network,
software-assisted routing either selects the lowest-cost available
subchannel(s) or, if no subchannels are available, requests the
network operator to add subchannel muxponders as required. Because
of the flexibility of this design, the traffic assignments to
subchannels and assignment of subchannels to ITU G.692 channels is
not known a priori. These assignments depend on the network's
real-time traffic demand and evolution of the network over
time.
[0440] An example of a simple optical network is shown in FIG. 59.
At each node the routing table numbers the node ports from P1 to
PN. A ring node has 8 ports (East add, East drop, West add, West
drop, East line in, East line out, West line in, and West line
out). Although we give examples here of ring nodes of degree 2
(where a node is connected to 2 other nodes), this methodology can
easily be extended to higher degree nodes by adding more ports to
the routing description. Also note that the routing description and
tables are independent of the particular hardware used for adding
and dropping channels. In FIG. 59 we show generic nodes that
selectively add, drop, and passthrough traffic from and to the line
fibers without restrictions on the hardware.
[0441] In FIG. 59 each node has 8 ports that can selectively direct
the traffic in each ITU channel as follows:
[0442] P1) Adds channels from the node to the output port P7
[0443] P2) Adds channels from the node to the output port P8
[0444] P3) Line Input port that can drop channels to P5 or
passthrough channels to P8
[0445] P4) Line Input port that can drop channels to P6 or
passthrough channels to P7
[0446] P5) Drop port that can drop channels from P3
[0447] P6) Drop port that can drop channels from P4
[0448] P7) Line output port
[0449] P8) Line output port
[0450] In one embodiment, bandwidth is managed with routing tables
exchanged between network nodes. A simple routing table that
describes the possible connections is shown in FIG. 60. A "1" in
the table between input ports and output ports indicates that a
connection between those ports is possible with a single hop, and a
"0" indicates that a connection between those ports is not
possible.
[0451] After a node is commissioned and connected via the line
fibers to a second node, the OSC connection between the nodes is
used to exchange the node connections. The management cards at each
node then exchange the routing information to build up a route
connection table as shown in FIG. 61.
[0452] The first table 6110 in FIG. 61 only contains the intra-node
connections for the 3 nodes shown in FIG. 59. In the second table
6120, we assume that the connection of Node1, Port 8 is made to
Node 2, Port 3. The table 6120 then gets filled in with new values
which indicate that, since channels from Node 1 Ports 2 and 3 can
be connected to Node 1, Port 8, they can also be connected to the
same outputs as Node 2, Port 3. These connections are given the
value of "2" in the table since they involve 2 hops. The next
version of the routing table 6130 shows the additions for the
connection of Node 1, Port 7 to Node 3, Port 4. The other cases
shown in FIG. 61 demonstrate how the routing table is filled out.
The net results, in table 6140 in the lower right-hand corner of
FIG. 61, has all the connections filled out.
[0453] The fiber connections shown in this 2-dimensional table can
also be listed in a linear representation as shown in FIG. 62, has
tables 6200 that list the action of each port as follows:
[0454] A) To indicate an add port
[0455] D) To indicate a drop port
[0456] PI) To indicate a passthrough input port (line input)
[0457] PO) To indicate a passthrough output port (line output)
[0458] Each table proceeds from top to bottom with the propagation
of light around the ring. For example the left-hand table 6210 of
FIG. 62 represents the inner fiber of FIG. 59. Proceeding from the
first line of the table 6210, channels input to Port 4 of node 1
can be dropped to port 6 of node 1, and channels added at Port 1 of
Node 1 can be added to the passthrough channels. These channels are
then passed to the output Port 7 that is connected to the Node 5
Port 4 input port. The sequence then repeats for each node around
the ring. The table 6220 on the right-hand side of FIG. 62 shows
the progression for the outer fiber in FIG. 59. Note that these
tables "wrap around", i.e. the bottom port of each table is
connected to the top port. For example, in the left-hand table 6210
Node 2, Port 7 at the bottom is connected to Node 1, Port 4 at the
top.
[0459] This 2-dimensional routing table 6200 contains the possible
connections only for each channel, or subchannel, but does not
contain the additional dimension required to specify the actual
configuration of each subchannel. To demonstrate that function,
consider the example of a ring network with subchannel routing.
[0460] Each ITU channel is divided into N subchannels (N>=2); in
our examples we assume that N=4. In a network with a plurality of
nodes, each node can demultiplex at least one ITU channel. The
control plane can map a plurality of client signals to tunable
lasers at each node, and can route any client signal to any other
node by tuning the transmitter laser of that signal to a subchannel
within the ITU channel associated with the destination node.
[0461] FIG. 63 illustrates the concept with a simple network 6300
where a pair of single ITU-T G.692 channel filters are used at each
node to drop the ITU channel indicated. As noted above, the ITU
filter function can be realized by a variety of technologies, such
as fixed thin-film filters or a ROADM. Since the carriers are
tunable, the simplest means to add the signals to the ring is to
use a wavelength-independent coupler (CPLR) as shown.
[0462] In this example, the traffic map has two 10 Gb/s connections
between every node in a protected full-mesh configuration. The
mapping of subchannels is given in the legend in the middle of the
diagram. Note that each node has an unassigned subchannel available
to carry more traffic. Also note that subchannels are reused at
different ITU channels, e.g. SC-2 is used to connect Node 200 to
Node 220, as well as Node 210 to Node 230.
[0463] FIG. 64 shows the simplified connection diagram 6400 for the
4-node ring example with the same numbering as the 3-node example
in FIG. 59. After a subchannel is added at a node and connected to
one direction, the OSC connection communicates the channel add to
all the downstream nodes. Each node then classifies that subchannel
as being passthrough, or dropped if a channel filter is used to
drop that subchannel. When the subchannel is dropped and connected
to a subchannel receiver, the connection between transmitter and
receiver is updated in the routing table.
[0464] In FIG. 65 and FIG. 66, we show two different means of
displaying the connection map for the 4-node ring example shown in
FIG. 64.
[0465] FIG. 65 shows a connection map 6500 displaying the
connection between the end points of the subchannel muxponders at
each node and FIG. 66 lists the state of each subchannel at each
point in the network.
[0466] FIG. 65 lists the network hierarchy from nodes (first row
and first column) to cards (second row and second column) to
channels (third row and third column) to client ports (fourth row
and fourth column). Note that an ITU channel is assigned to each
card, which is the ITU channel being received by that card. A
circuit is displayed in the table as a highlighted square linking
two client ports. To illustrate an example connection, the table
6500 has highlighted the connection between Client Port 3 on Card 2
at Node 1 (transmitting at Subchannel 3 of Channel 230) and Client
Port 3 of Card 8 at Node 4 (transmitting at Subchannel 3 of Channel
200). The rest of the connections in the network of FIG. 63 have
been entered in table 6500 of FIG. 65 as dark rectangles.
[0467] Available bandwidth in FIG. 65 is indicated by empty rows
and columns. For example, the row associated with Node 1, Card 1,
and Port 4 does not have a rectangle linking that port to a
circuit, so that port is available for network upgrades.
[0468] FIG. 66 lists the state of each subchannel at each
connection point for the network of FIG. 64. The first three
columns of FIG. 66 list the fiber connections according to the
convention illustrated in FIG. 62. There is one column in this
table 6600 for each subchannel. The state of a subchannel is listed
with the following nomenclature: [0469] UEQ (Unequipped)--the
subchannel is not present [0470] IS-A (In Service Add)--the
subchannel is being added at that port [0471] IS-P (In Service
Passthrough)--the subchannel is present and being passed through to
the next port [0472] IS-D (In Service Drop)--the subchannel is
being dropped at that port [0473] UAS-D (Unassigned Drop)--the
subchannel is not present, but there is a filter present that would
drop that subchannel
[0474] Note that table 6600 is divided into two halves. The top
half 6610 is for the counterclockwise fiber connection in FIG. 63
and the bottom half 6620 is for the clockwise fiber connection in
FIG. 63. When the first node in a network is commissioned, a
routing table is populated with all available bandwidth listed as
unequipped (UEQ). The available bandwidth is determined by the
management software. For the purpose of simplicity in this example,
we assume that it starts with 4 ITU channels and 4 subchannels per
ITU channel. Furthermore, we assume that east and west drop filters
are deployed at the first site that drops a subset of ITU channels,
which then constricts a subset of the UEQ channels to be dropped at
that node.
[0475] To illustrate how this table 6600 works, consider the
example of the connection, highlighted in FIG. 65, between Client
Port 3 on Card 2 at Node 1 (transmitting at Subchannel 3 of Channel
230) and Client Port 3 of Card 8 at Node 4 (transmitting at
Subchannel 3 of Channel 200). This connection is listed in the
second to last column of FIG. 66 for Subchannel 3 of Channel 230.
"IS-A" is listed in the third row of that column to indicate the
start of the connection at Port 1 of Node 1. The subchannel passes
through to Port 7 of Node 1, across the line fiber to Port 4 of
Node 4, and is then dropped at Port 6 of Node 4. In the reverse
direction (clockwise fiber), the Subchannel is added at Port 4 of
Node 2 in the third column, second-to-last row. It then passes to
Port 8 of Node 4, across the line fiber to Port 3 of Node 1, and is
then dropped at Port 5 of Node 1. Note that this connection wraps
around to the top of the second half 6620 of the table 6600.
[0476] This circuit has a corresponding protected connection that
is highlighted in FIG. 67. This connection uses the same
subchannel, but it propagates around the other side of the ring,
passing through Nodes 2 and 3 between the Node 1 and Node 3
terminals.
[0477] FIG. 66 indicates the assignment and use of subchannels at
any point in a fiber connection. For example, the subchannels
present at the output of Port 7 from Node 1 is listed in the fourth
row, where the assignments are in bold characters. That row
indicates that Subchannel 1 of Channel 210, Subchannels 2 and 3 of
Channel 220, and Subchannels 2,3 and 4 of Channel 230 are In
Service, and the other subchannels are unequipped.
[0478] To determine if there is a subchannel available for a new
circuit, consider the request for a new circuit between Node (added
at Port 2) and Node 2 (dropped at Port 5). The subchannels
available for such a circuit are highlighted in FIG. 68. The table
6800 lists 9 subchannels as being unused on that link. However, it
is preferable to use a subchannel within an ITU channel that is
already dropped at that node. Subchannel 4 of Channel 210 is
available at Node 2, as indicated by the "UAS-D" entry in table
6800.
[0479] As shown in these examples, the routing table indicates to
the network operator which subchannels and channels are in use at
each location of the network. With the fully-tunable subchannel
lasers, any unassigned laser can then be mapped to any unused
subchannel to provide the requested circuit. When a new circuit is
requested, the table can indicate which subchannels are available,
and which have the lowest cost of deployment. The routing
information can also be passed to higher layers of software that
monitor and control the subchannels.
[0480] The software can therefore provide to the network operator
capacity lists and/or maps of in-service capacity, present but not
deployed capacity, and unused capacity. The tables can also have
options to group subchannels by those that are (1) deployed and in
service (2) installed, but not in service, and (3) available to be
deployed.
[0481] This architecture also supports sub-rate multiplexing within
a subchannel. Various mappings 6900 to subchannels are shown in
FIG. 69. Subchannel mapping supports optical concatenation, e.g.,
four 10 Gb/s subchannels can be concatenated to carry a 40 Gb/s
signal.
[0482] The same routing tables can be used to support an overlay of
lower-layer protocol routing with subchannel optical routing.
[0483] For example, suppose the subchannel muxponder supports
standard 10 Gigabit Ethernet data on the client side, and the 10
Gigabit Ethernet data comes from a 10.times.1 Gigabit Ethernet
multiplexer.
[0484] An example of the same routing approach can be applied to
the overlay 7000 shown in FIG. 70, which shows the overlay of
10.times.1 GE services on one of the available subchannels of the
4-node network shown in FIG. 63. The subrate muxponder in this case
has 10 client SFP ports and an ethernet switch that maps the client
ports to VLANs on the line side. The traffic in the VLANs is mapped
to one of the available 10 Gb/s subchannels transmitted between the
three nodes.
[0485] In FIG. 70, the following services are provisioned: [0486] 3
circuits from Node 1 to Node 2 over VLANS 1-3 [0487] 4 circuits
from Node 1 to Node 3 over VLANS 4-7 [0488] 3 circuits from Node 2
to Node 1 over VLANs 8-10 [0489] 6 circuits from Node 3 to Node 1
over VLANS 1-6
[0490] FIG. 71 shows the VLAN routing map 7100 for the subrate
overlays displayed in FIG. 70. Similar to the tables above, the
following notation is used: [0491] IS-A (In Service Add)--the VLAN
is being added at that port [0492] IS-D (In Service Drop)--the VLAN
is being dropped at that port [0493] IS-DP (In Service
Drop-Passthrough)--the VLAN is being dropped to the adjacent card
to form a passthrough connection to the next node [0494] IS-DP (In
Service Add-Passthrough)--the VLAN is being added from the adjacent
card to form a passthrough connection to the next node
[0495] The last column 7110 of FIG. 71 lists a possible mapping of
the sublayer service to the subchannels in the network shown in
FIG. 63.
[0496] This architecture and routing method provides a means of
interconnecting the rings and spurs shown in FIG. 10. For
inter-ring traffic, a subset of the total number of ITU channels
can be assigned to the inter-ring traffic and remaining ITU
channels can be assigned to the intra-ring traffic. Fixed or
reconfigurable filters can then be used to direct the inter-ring
traffic and intra-ring traffic. The routing tables can be extended
to cases where there are spur nodes, and interconnected rings.
[0497] The subchannel routing software provides multilayer routing
where the first layer manages client services, the second layer
manages subchannels, the third layer manages ITU channels, and the
fourth layer manages fiber connections. Additional tables can
indicate the status of these services and connections. The tables
also provide route discovery for subchannels and services.
[0498] FIG. 72 shows an example of a status table 7200 that can be
displayed by the network management software to the network
operator. The techniques described above can be used to determine
the optical data listed. The first table 7210 in FIG. 72 lists
status of the subchannels transmitted from Node 1 in the example of
FIG. 63. The columns list, in order from left to right, the local
IP address of each line card, the transmitted subchannel, the
service on each subchannel, a unique label used to identify the
service, the destination node, the transmit port connecting the
subchannel to the line fiber, identification of any subchannel that
is protecting the traffic, the remote received power, OSNR,
dispersion, PMD, and bit error rate. Similar entries are provided
in the bottom table 7220 for the received subchannels.
[0499] In these tables, visual cues can be used to alert the
operator about network problems. For examples, metrics that are
failing a requirement can be colored red, and metrics that are
close to failing can be colored yellow. Moreover, in another
embodiment, such cues could trigger automated actions including
notifications of a problem or predefined corrective measures such
as provisioning or de-provisioning a circuit.
[0500] 4. Directionless Subchannel Muxponder
[0501] There is a need in WDM optical networking for directionless
transponders and muxponders. In this context, "directionless" means
that the circuit provided by the transponder or muxponder can be
remotely switched by software to be on either side of a ring, i.e.
the transmitter can switch between transmitting in the clockwise
direction or counter-clockwise direction; and the receiver can
select a circuit from the clockwise direction or counter-clockwise
direction.
[0502] FIG. 73 shows an example of a directionless subchannel
muxponder implementation 7300. The switching is accomplished by
placing a red/blue filter 7310 at the transmitter output and a
red/blue filter 7320 at the receiver input. The red/blue filter
splits the C-band ITU channel bandwidth into two halves, "red"
channels are ITU channels 200-390 and "blue" channels are ITU
channels 400-600. Note that this division is arbitrary and
dependent on the quality of the red/blue filter. Current filters
may not have adequate isolation at the splitting frequency, so that
1 or more channels at the middle, e.g. channels 390-410, may be
unusable.
[0503] With the red/blue filters inserted as shown in FIG. 73, a
subchannel muxponder can set a subchannel transmitter to a blue
frequency to transmit to one direction or to a red frequency to
transmit to the other direction. Similarly, red or blue channels
can be selected from either direction to be received.
[0504] The configuration 7400 of FIG. 74 enables a subchannel
muxponder to be deployed as a switchable subchannel crossponder. A
"crossponder" in this context is a muxponder that can transmit and
receive on two line ports. One advantage of a subchannel
crossponder is that it can redirect traffic away from a span for
node insertion as shown in FIG. 74. For example, traffic can
transmit and receive on both fiber spans 7410 or be redirected to
only one of the spans 7420, or a new node can be inserted,
effectively replacing one of the fiber spans 7430. A subchannel
crossponder can also be used to bridge traffic on two diverse spans
for protection switching applications as described below.
Subchannels can be selectively dropped or added using the
crossconnect switch 4115 shown in FIG. 41.
[0505] 5. Subchannel Optical Protection
[0506] The subchannel network design offers a flexible means of
protecting services. Traffic that propagates in one direction in a
subchannel can be protected by the same or a different subchannel
propagating around the ring in the opposite direction. This
architecture also supports shared optical protection. Regardless of
the protection architecture deployed, a protected circuit requires
two basic functions at the terminals--the bridge function and
switch function. The signal to be protected has to be bridged onto
two redundant paths at the transmit end, and one of the two signals
from the redundant paths must be switched at the receive end to be
selected as the working circuit.
[0507] In the example shown in FIG. 63, there are two duplex
connections between every node--one in the clockwise direction, and
one in the counter-clockwise direction. The two subchannels can be
used for independent services or for protecting against fiber
cuts.
[0508] An example of a simplified dedicated protection
implementation 7500 is shown in FIG. 75 where the traffic between
two nodes, Node 1 and Node 2, is protected with a working fiber
pair 7510 and a protection fiber pair 7520. There are two protected
circuits, one between Clients 1 (C1) and 3 (C3), and one between
Clients 2 (C2) and 4 (C4). At the transmitter, the traffic from
Clients 1 and 3 is bridged onto I-200, SC-1 and I-400, SC-3; and
the traffic from Clients 2 and 4 is bridged onto I-200, SC-2 and
I-400, SC-4. This traffic is transmitted onto both working fibers
with the I-200/I-400 channel filters.
[0509] At each receiver, a 1.times.2 switch selects traffic from
one of the designated subchannels. For example, the switch 7530 at
Node 1 for Client 1 selects the I-200, SC-1 received signal in
normal mode, and selects the I-400, SC-3 received signal when the
working fiber pair is cut. The selection of subchannel can be done
with the integrated crossconnect switch shown in FIG. 17, or with
the use of external optical or electrical switches.
[0510] Although the crossconnect shown in FIG. 17 provides more
connectivity, it also adds to the cost of the terminal equipment.
If the transceiver module does not have a crossconnect, there are
other means for performing the bridge and switch. For example,
equipment 7600 in FIG. 76 includes two optical splitters on the
client side. The top splitter 7610 is used to bridge the client
data 7605 onto the first and third subchannels. At the receive
side, the signals from the first and third subchannels are
connected to a second optical splitter 7620. The software on the
transport module selects which subchannel signal is sent back to
the client 7605 by enabling one of the client transmit lasers. This
function could also be accomplished with a 1.times.2 optical switch
7630 instead of the second optical splitter 7620. The first option
may be preferred because (1) cost of an optical splitter is lower
than that of an optical switch and (2) the software controlling the
switch is self-contained on the transport module--the second option
requires external software communication between the transport
module and the switch, which makes the switch slower and less
reliable.
[0511] Another embodiment for performing the bridge and switch is
shown in FIG. 77. In this embodiment 7700, the bridge function is
done with an optical loopback 7710 on the client side between CL-1
and CL-3.
[0512] The dedicated protection architecture shown in FIG. 75 can
be deployed in much larger applications with many more nodes and
many more subchannel circuits. In these cases it may be desirable
to use shared protection to free up more subchannels for working
traffic.
[0513] FIG. 78 shows an embodiment of a subchannel crossponder 7800
that can be used to support dedicated or shared protection. On the
transmit side, the first two carriers 7810a and 7810b are
transmitted to one line fiber 7825 and the other two carriers 7810c
and 7810d are transmitted to the second line fiber 7835. The
receive side has a coupler 7840 to receive subchannels from both
line fibers 7825 and 7835. The integrated crossconnect switch 7850
can be used for dedicated protection (1) at the transmit side to
bridge client traffic onto two subchannels, and (2) at the receive
side to select traffic from one direction. The switch 7850 can also
be used for a shared protection application to regenerate the
protection subchannel. These cases are discussed below.
[0514] FIG. 79 shows an application 7900 that uses a subchannel
crossponder at two distinct nodes connected by two fiber pairs. The
top fiber pair 7910 is designated as the working pair and the
bottom pair 7920 is designated as the protection pair. Subchannels
SC1 and SC2 are transmitted in the clockwise direction and SC3 and
SC4 are transmitted in the counter-clockwise direction. There are
two working circuits on the top working fiber, (1) SC1 transmitting
from Node 1 to Node 2 with SC4 transmitting from Node 2 to Node 1;
and (2) SC2 transmitting from Node 1 to Node 2 with SC3
transmitting from Node 2 to Node 1.
[0515] The integrated crossconnects 7930a and 7930b perform the
bridge and switch functions for Nodes 1 and 2, respectively.
Traffic on the first client port at Node 1 is bridged to SC1
(transmitted over the working fiber) and SC3 (transmitted over the
protection fiber). At the receive side the switch function in the
crossconnect 7930a selects the traffic from SC1 or SC3 to be
transmitted back to the first client port at Node 2. This design of
the crossponder therefore provides dedicated protection of two
working circuits with Subchannel 3 protecting the traffic in
Subchannel 1 and Subchannel 4 protecting the traffic in Subchannel
2.
[0516] FIG. 80 shows an embodiment 8000 where the subchannel
crossponder is used in a shared protection application. In this
example, there is a single working circuit between each pair of
neighboring nodes (N1 to N2, N1 to N3 and N2 to N3). The
subchannels assigned for the working circuits are SC1 for the
clockwise direction and SC3 for the counter-clockwise direction.
SC4 is allocated as a shared protection subchannel for failures in
SC1; and SC2 is allocated as a shared protection subchannel for
failures in SC3. Therefore at each node where client traffic is
being added, client traffic that is mapped to SC1 is also bridged
to SC4, and client traffic that is mapped to SC3 is also bridged to
SC2. At the drop side, the switch function in the crossconnect
normally selects SC1 and SC3.
[0517] FIG. 81 shows how the network shown in FIG. 80 is protected
by a fiber cut 8130. In this embodiment 8100, the fiber pair 8110
between Nodes 1 and 2 has been cut. After the control software
determines that the fiber pair 8110 between Node 1 and Node 2 has
been cut, e.g. by Loss of Signal alarms, it sends messages over the
OSC (optical service channel) to each node in the ring. The nodes
adjacent to the cut are instructed to switch to protection so that
the first client circuit at Node 1 switches from selecting the
traffic from SC3 to selecting the traffic from SC2 on the other
side of the ring, and the first client circuit at Node 2 switches
from selecting the traffic from SC1 to selecting the traffic from
SC4 on the other side of the ring.
[0518] At the intermediate nodes (Node 3 in this example), the
protection subchannels are switched to passthrough mode. In this
example that means that SC4 is regenerated from Node 1, passing
through Node 3, to Node 2, and SC2 is regenerated from Node 2,
passing through Node 3, to Node 1.
[0519] The shared protection switching with subchannels requires
real-time messaging software between the nodes over an overhead
channel to coordinate the protection switching. It should be noted
that, because one embodiment of the subchannel crossponder supports
different protocols on each port, the protocol of the protection
subchannel may have to be switched as well during the protection
switch.
[0520] Also note that the protection bandwidth (SC2 and SC4 in this
example) is normally unused so that it is available for
low-priority traffic. For this example, a low-priority circuit
could be established between the second client port at Node 1 and
the second client port at Node 2 transmitting on SC2 from Node 1 to
Node 2, and transmitting on SC4 from Node 2 to Node 1. Similar
circuits could be established between Node 1 and Node 3, and
between Node 3 and Node 2. Those circuits would be dropped in the
event of a protection switch of high-priority traffic, since the
protection switch then uses the protection bandwidth to maintain
the high-priority circuit.
[0521] This type of shared protection falls under the category of
OSPR (Optical Shared Protection Ring), that use a division between
working and protection bandwidth similar to that used by a SONET
Bidirectional Line Switched Ring (BLSR) defined in the Telcordia
Standard GR-1230-CORE. [See also "Transparent Optical Protection
Ring Architectures and Applications", by M. J. Li et al, IEEE
Journal of Lightwave Technology, Vol 0.23, No. 10, p. 3388
(2005).]
[0522] In this protection architecture, when a failure occurs that
causes a loss of signal on a span between two OADM nodes, a
protection switch is initiated by the control software. Signaling
messages from the nodes on either side of the failure are
transmitted around the ring on the side away from the point of
failure. Upon receiving the request to effect a protection switch,
all the intermediate nodes discard the extra traffic, if used, to
set up a protection link for the failed traffic.
[0523] It should be noted that the designs described above
represent not only a novel device--i.e., the first subchannel
crossponder--but a novel application of this device--i.e., the
first application of a subchannel crossponder in a shared
protection ring. This type of shared protection can also be applied
to subrate overlay channels, such as the examples shown in FIG.
70.
[0524] Also note that the crossconnect switch 1750 in the
subchannel muxponder 1700 (shown in FIG. 17) or crossponder 7800
(shown in FIG. 78) can be used to provide 1:N shared protection
against laser failures. For example, 3 client services can be
mapped to the first 3 subchannels, and if any of the hardware used
by one of the first 3 subchannels fails, the affected traffic can
be bridged and switched to the fourth subchannel.
[0525] 6. Distributed WDM Switching Network
[0526] By mapping client data to N subchannels on M ITU channels
with tunable lasers, this architecture can support ring and mesh
topologies with up to up N.times.M strict-sense nonblocking
connections with low-cost fixed optical filters. In the examples
described above, M=4 for 4 subchannels at 10 Gb/s, while N=on the
order of 40 for C-band ITU channels spaced at 100 GHz, and N=on the
order of 80 for C-band ITU channels spaced at 50 GHz, and N=on the
order of 160 for C and L band channels spaced at 50 GHz. With 4
subchannels per ITU channel, this embodiment can therefore support
a 640.times.640 full logical mesh of 10 Gb/s services, in a
strict-sense nonblocking switching architecture.
[0527] Furthermore, if J subrate services are mapped to each
subchannel, then J.times.M.times.N circuits can be supported by
this network.
[0528] When 10 subchannels are mapped into an ITU channel,
providing 100 Gb/s, the switching network size increases to a
1600.times.1600 full logical mesh of 10 Gb/s services.
[0529] FIG. 82 shows the case of a large distributed optical
switching network 8200. It is on a physical ring, but provides a
logical mesh connection between all circuits as shown in network
8300 in FIG. 83. A circuit can be established between a pair of any
client ports at any nodes by tuning the subchannel lasers. Spurs or
other rings can be connected as well with ITU channel filters used
to direct the traffic across the ring. As discussed previously, the
ITU channel filter function can done with fixed or reconfigurable
filters in a broadcast and select configuration or configuration
with selective add and drop filters.
[0530] The subchannel routing software described in this document
can be used to determine which subchannels are available on each
network segment, establish and monitor subchannel circuits, and
reconfigure subchannels for network upgrades such as node
insertions. The subchannel muxponder can be used to upgrade an
existing ITU network to provide M times the capacity while
providing the additional functionality of subchannel routing,
reconfiguration and restoration, switching, and physical layer
monitoring. The network can also be operated in a hybrid manner,
where subchannels coexist with standard ITU channels as ITU
subchannels are subdivided into subchannels as the network grows.
This provides a "pay as you grow" cost advantage.
D. Shared Wavelocker for Controlling Subchannel Frequencies
[0531] The above descriptions of various subchannel-related
embodiments assume tunable lasers of sufficient precision to
reliably distinguish multiple subchannels (eg, subcarrier
frequencies) within a single ITU channel. As noted above, tunable
lasers for such subchannel-related applications require even
greater accuracy than for ITU channel-based applications.
[0532] As is the case with all electronic and optical components,
the performance characteristics of the lasers employed in DWDM
systems change with temperature and with time. In particular, the
frequency of emitted laser light changes due to ambient temperature
variations (typically from -5 degC. to 65 degC.) and due to
aging.
[0533] If the WDM system requirements call for better absolute
laser frequency stability than that of the DWDM (Dense
Wavelength-Division Multiplexing) laser itself, an external
wavelocking scheme is preferred. In this scheme: (i) each laser has
a set-point for its target frequency; (ii) the laser absolute
frequency is measured by a measurement means that has the required
absolute frequency accuracy; (iii) the control electronics and/or
software calculate the offset between each laser's actual frequency
and its target frequency, and communicates the error to each laser
controller; and (iv) the laser set-point is adjusted appropriately
to reduce the frequency error to a value within the system
requirements. In this context, the terms "laser frequency" and
"laser wavelength" may be used interchangeably.
[0534] U.S. Pat. No. 6,282,215 discloses a wavelocking scheme that
employs one or more Fabry-Perot etalons placed inside each laser
cavity. Yet, using at least one integrated etalon per laser adds
cost to the system, and the locking error on each laser does not
ensure that the lasers are all on the same frequency grid.
Moreover, current integrated wavelocker designs offer absolute
frequency accuracy on the order of +/-1 GHz or more. This level of
absolute accuracy is adequate for WDM systems where channels are
located on the ITU grid with spacing of 50 or 100 GHz, but this
level of accuracy is not sufficient for DWDM systems such as those
based on subchannels within an ITU with channel spacing on the
order of 10 GHz.
[0535] Existing architectures partially address these drawbacks
with designs that can lock multiple lasers to a grid using a shared
external etalon. See, for example, U.S. Pat. No. 6,369,923, which
uses an etalon with spacing between resonant frequencies equal to
the channel spacing. It applies the same dither on each laser, with
the dither being activated on one laser at a time while each laser
is being monitored. U.S. Pat. No. 7,068,949 also uses an etalon
with spacing between resonant frequencies equal to the channel
spacing. It applies different pilot tones or low-amplitude
frequency dithers (i.e. dithering) to each laser to be locked. The
wavelocker detects the pilot tones with a Fast Fourier Transform
(FFT) signal analysis.
[0536] However, these methods do not scale to very large channel
counts because the signal-to-noise ratio (SNR) of the signal
required for locking decreases as more channels are added. These
methods lock the laser array to the frequency grid of the
Fabry-Perot etalon with an error on the order of +/-1 GHz. These
methods also require a dither or pilot tone to be applied to each
transmitter. The dither can add a bit-error penalty, and it also
requires specialized hardware to be added to each laser.
[0537] We therefore disclose a method that offers significant
performance and cost advantages over state-of-the-art wavelocking
designs. In particular, this disclosed solution:
1 does not require any modifications to the
transmitters--specifically, it does not require dithering the
lasers, which can add a bit-error penalty; 2 scales to a very large
DWDM channel count without a decrease in Signal-to-Noise-Ratio
(SNR) of the signal required to wavelock the lasers; 3
independently measures and locks the frequency of each DWDM laser;
4 references the frequencies of each DWDM laser directly to an
absorption frequency given by the fundamental physical quantum
properties of molecules in gaseous state. Thus, it is predominantly
independent of engineering and design tolerances, inaccuracies,
manufacturing defects, etc., that are inherent, for example, in
Fabry-Perot based wavelockers; 5 provides absolute frequency
accuracy of wavelocking or referencing DWDM laser frequencies at
least 20 times better (e.g. +/-50 MHz) than that of existing
solutions with minimal errors over time and temperature; 6 provides
locking without requiring hardware calibration; 7 provides an
absolute accuracy that is determined by a National Institute of
Standards and Technology (NIST) certified reference that is
directly linked to the fundamental frequency standards used by
NIST; and 8 provides wavelength accuracy comparable to that
obtained by a NIST reference, since the frequency reference is
based on absorption lines that have been identified by NIST.
[0538] See the following references for background in this context:
(i) W. C. Swann and S. L. Gilbert, "Line centers, pressure shift,
and pressure broadening of 1530-1560 nm hydrogen cyanide wavelength
calibration lines", J. Opt. Soc. Am. B, vol. 22, no. 8, pp.
1749-1759, August 2005; (ii) S. L. Gilbert, W. C. Swann and C-M
Wang, "Hydrogen Cyanide H.sup.13C.sup.14N absorption reference for
1530 nm to 1565 nm wavelength Calibration--SRM 2519a, Standard
Reference Materials, National Institute of Standards and Technology
Special Publication 260-137, 2005 Edition; (iii) W. C. Swann and S.
L Gilbert, "Pressure-induced shift and broadening of 1510-1540 nm
acetylene wavelength calibration lines", J. Opt. Soc. Am B 17, pp.
1263-1270 (2000); and (iv) S. L. Gilbert and W. C. Swann, "Carbon
monoxide absorption references for 1560 nm to 1630 nm wavelength
Calibration--SRM 2514 (.sup.12C.sup.16O) and SRM 2515
(.sup.1C.sup.16O), Standard Reference Materials, National Institute
of Standards and Technology Special Publication 260-142, 2002
Edition.
[0539] In one embodiment, a measured DWDM signal laser is
referenced to a tunable laser (oscillator) that has a well-known
and stable absolute frequency by measuring the beat frequency
between the two lasers. This gives a frequency error equal to the
frequency accuracy of the reference laser. If improved accuracy is
required, the accuracy of the reference laser is improved by
calibrating it in real-time to the absorption lines of a gas
cell.
[0540] To minimize the system cost, this embodiment employs means
of locking multiple DWDM lasers with a shared measurement system.
These same measurement techniques could, however, be used with a
single laser.
[0541] The strength of the beat frequency signal between two lasers
depends on the polarization alignment of the two lasers.
Polarization alignment between fiber-pigtailed lasers can be
maintained with polarization-maintaining fibers (PMF). However,
most DWDM components have single-mode fiber (SMF), which does not
preserve the polarization. Some embodiments described below use PMF
to control the polarization. Other embodiments use a polarization
scrambler for cases where PMF is not available.
CASE A
Oscillator and Signal Lasers Linearly Polarized and all
Polarizations Aligned
No Absorption Cell
Embodiment 1
Beating of Two Lasers, Both at Fixed Frequencies
[0542] The simplest example of a frequency beating phenomenon may
be described as follows. Assume that two laser beams of equal
optical power and aligned polarizations are launched into the
Polarization Maintaining (PM) optical fiber. Assume that the first
laser is tuned to an optical frequency f.sub.o (oscillator) and the
second laser is tuned to an optical frequency f.sub.s (signal) so
that the frequencies differ by:
f.sub.RF=|f.sub.s-f.sub.o, (e.g. f.sub.RF=500 MHz);
[0543] Also assume that the linewidth of both lasers
(.DELTA.f.sub.o-FWHM for the oscillator and .DELTA.f.sub.s-FWHM for
the signal laser) are approximately the same and equal to:
.DELTA.f.sub.FWHM=.DELTA.f.sub.o-FWHM=.DELTA.f.sub.s-FWHM;
[0544] and substantially smaller than f.sub.RF (e.g.
.DELTA.f.sub.FWHM=1 MHz).
N.B. sub-index "FWHM" stands for a "Full Width at Half Maximum"
linewidth definition.
[0545] When the combined light of both lasers is detected by a
photodetector with an electrical bandwidth larger than f.sub.RF,
but much smaller than the frequencies of optical light f.sub.s or
f.sub.o, the photocurrent of the photodetector would have a
sinusoidal electrical signal oscillating at the beat frequency
f.sub.RF. This phenomenon is known as frequency beating, in this
example between f.sub.s and f.sub.o.
[0546] The photocurrent of the photodetector detecting the laser
frequency beating can be amplified by a trans-impedance amplifier
(TZ) and filtered by an electrical bandpass filter (BPF) centered
at a fixed or tunable frequency f.sub.RF with a fixed or tunable
electrical bandwidth .DELTA.f.sub.RF-FWHM (.DELTA.f.sub.RF-FWHM
being smaller than 2*f.sub.RF, but larger than .DELTA.f.sub.FWHM).
If required, the beat signal can be amplified at the input or
output of the BPF. The BPF can act also as an amplifier. The beat
signal can be detected as follows. The BPF output is rectified, for
example, by using a zero offset rectifier (i.e. a rectifier without
a typical 0.7V voltage drop (offset) of a simple semiconductor
silicon diode rectification); averaged, e.g., by an
operational-amplifier-based integrator; digitized by an analog to
digital converter (ADC); and processed in the digital domain by,
e.g., a digital signal processing (DSP) circuit. Note that the band
pass filtering, rectification, amplification and any other signal
processing required can be performed after direct ADC conversion of
the TZ output in the digital domain, e.g. within a microprocessor
or a DSP.
[0547] For the purpose of this description, we call the electronic
detection and processing system that measures the beat frequency
the "RF Detection of the Beat Signal" (RFDBS), and its Direct
Current (DC) (and/or direct voltage) signal at the input of the ADC
after analog processing described above as the "RFDBS Output."
Embodiment 2
Beating a DWDM Signal Laser at Fixed Frequency (with a Narrow
Optical Carrier) with an Oscillator Laser with a Scanned
Frequency
[0548] We now expand the first example to a more practical beating
system suitable for DWDM applications. We now assume that in
addition to all the conditions described above, the frequency of
the reference laser, f.sub.o, is tuned from
f.sub.s-f.sub.RF-.epsilon. to f.sub.s+f.sub.RF+.epsilon., where
.epsilon. is much larger than .DELTA.f.sub.FWHM. See graph 8400 in
FIG. 84 which illustrates this case.
[0549] While scanning the oscillator laser within the said range,
the amplitude of the RFDBS output would show two strong peaks
(maxima substantially higher than zero) at frequencies of the
oscillator laser approximately equal to f.sub.o=f.sub.s-f.sub.RF
and f.sub.o=f.sub.s+f.sub.RF, where the equal sign means an
equality within +/-.DELTA.f.sub.RF-FWHM or a few times
+/-.DELTA.f.sub.RF-FWHM.
[0550] In this example the RFDBS output is maximized when the
oscillator frequencies are set to f.sub.o-left and
f.sub.o-right.
[0551] By analyzing the frequency positions of two RFDBS output
peaks one can determine that the frequency of the signal laser to
measured is given by:
f.sub.o-set=(f.sub.o-left+f.sub.o-right)/2=f.sub.s.
[0552] The described example is valid not only for DWDM signal
lasers which are narrow in spectrum, as in FIG. 84, but also for
modulated DWDM lasers at high bit rates (e.g. 10-40 Gb/s) with
modulation formats such as Non-Return-to-Zero (NRZ) or
Carrier-Suppressed-Return-to-Zero (CS-RZ) that have significant
spectral energy at the optical carrier. In the case of CS-RZ
modulation format, the RFDBS output will have four distinct peaks
and the formula to determine f.sub.o-set is slightly more complex
than the prior formula given above for the NRZ modulation
format.
Embodiment 3
Beating a DWDM Signal Laser at a Fixed Frequency (with a Broad
Spectrum) with a Scanned in Frequency Oscillator Laser
[0553] For DWDM modulation formats such as duo-binary (DB),
Quadrature-Phase-Shifted-Keying (QPSK) or
Differential-Quadrature-Phase-Shifted-Keying(DQPSK), the optical
carrier is substantially or totally suppressed and the signal laser
spectrum is broad (e.g. several or several tens of GHz) as shown in
graph 8500 in FIG. 85.
[0554] In this case, the RFDBS Output signal would in general show
two broad spectral lines or more. They can be distinct and
separated in spectrum as in FIG. 85 or overlap to some or large
extent depending on the choice of detection parameters such as
f.sub.RF, .DELTA.f.sub.RF-FWHM, .DELTA.f.sub.o-FWHM, bandwidth of
the photodetector, gain and integration constant of the processing
circuit, intensity of both signal and oscillator lasers and other
design parameters for a particular implementation.
[0555] In one embodiment, the RFDBS signal spectrum, namely the
RFDBS output versus the oscillator laser frequency, can be
processed and analyzed, and the absolute frequency of the signal
laser, f, (e.g. defined at the center of its broad spectrum) can be
determined.
General Approach
[0556] In general, for each modulation format of DWDM signal lasers
and specific DWDM system requirements, the optimum set of system
parameters such as f.sub.RF, .DELTA.f.sub.RF-FWHM,
.DELTA.f.sub.o-FWHM, bandwidth of the photodetector, gain and
integration constant of the processing circuit, intensity of both
signal and oscillator lasers and so on, exist. Optimization of
these parameters and the overall design is well known in the art of
heterodyne radio receivers design.
Implementation Type 1
[0557] FIG. 86 illustrates one possible implementation 8600 of this
approach. The linear polarization of the lasers is preserved in
polarization maintaining fibers (PMF) 8610a and 8610b, and both
lasers (Oscillator Laser 8615 and Signal Laser 8625) are combined
in such a way that, at the output of PM Coupler 1 8630, their
polarizations are aligned.
[0558] An algorithm implemented by the control unit 8650 in FIG. 86
includes the following steps:
1 Set the frequency of the signal laser 8625 to an initial setpoint
value f.sub.s-set-ini, which may or may not be equal (within
required system tolerance) to the absolute frequency of the signal
laser required, f.sub.s-target; 2 Scan the frequency of the
oscillator laser 8615 from f.sub.s-target-f.sub.RF-.epsilon. to
f.sub.s-target+f.sub.RF+.epsilon., recording at each scanning step
the frequency setpoint of the oscillator laser 8615. Make sure that
.epsilon. is large enough, e.g. much larger than the anticipated
signal laser absolute frequency error
f.sub.error=f.sub.s-set-ini-f.sub.s-target; and the bandwidth of
the signal .DELTA.f.sub.s-FWHM; 3 Record at each scanning point an
amplitude of RFDBS output, thereby obtaining at the end of the scan
a table of data with beat signal versus the oscillator laser
frequency; 4 By processing and analyzing the RDFBS Output spectrum,
determine the current absolute frequency of the signal laser,
f.sub.s-current; e.g. by using equation (1) for NRZ modulation
format; 5 Calculate correction to the signal laser setpoint as
f.sub.s-set-corr=f.sub.s-target-f.sub.s-current; 6 Modify the
signal laser frequency setpoint as
f.sub.s-set-final=f.sub.s-set-ini+f.sub.s-set-torr; 7 Set signal
laser frequency to f.sub.s-set-final and call it from then on
f.sub.s-set-ini; 8 Repeat steps 1 to 7 as fast and as often as
needed to keep the signal laser frequency within the accuracy
required by the DWDM system.
With Absorption Cell
[0559] In DWDM systems which require higher or substantially higher
absolute signal laser frequency stability and accuracy than that
given by the oscillator laser, the oscillator laser can be
calibrated during each scan by a more precise external frequency
reference.
Description of a Gas Cell as Wavelength Reference
[0560] A very stable and accurate frequency reference is available
in the form of absorption cells filled with specific molecules in a
gaseous state. A gas chosen for this application has many narrow
absorption lines within the spectral region of interest. For
example, the following molecules can be employed in different
spectral regions:
TABLE-US-00002 1255-1355 nm HF Hydrogen Fluoride O-band 1510-1540
nm .sup.12C.sub.2H.sub.2 Acetylene S-band 1529-1564 nm
H.sup.13C.sup.14N Hydrogen Cyanide C-band 1560-1600 nm .sup.12CO
Carbon Monoxide L-band 1590-1640 nm .sup.13CO Carbon Monoxide
L-band+
[0561] To assure absolute accuracy of the location of the
absorption lines in the optical spectrum, the molecules in the gas
cell must contain only particular atomic isotopes with well-defined
reference frequencies.
[0562] A mixture of molecular gases can be used to cover a much
wider spectrum range than listed above. For example, a mixture of
Hydrogen Cyanide (H.sup.13C.sup.14N), Carbon Monoxide (.sup.12CO)
and Carbon Monoxide (.sup.13CO) covers the spectral range 1520 nm
to 1640 nm.
[0563] A typical absorption spectrum 8700 of Hydrogen Cyanide
(H.sup.13C.sup.14N) is shown in graph 8700 in FIG. 87.
[0564] The spectral positions of all absorption lines change only
very slightly with temperature. Since the temperature dependence of
line positions is well known, this dependence can be calibrated out
if the cell temperature is being measured and processed by the
control unit (such as control unit 8650 in FIG. 86).
[0565] The spectral positions of all absorption lines are
predominantly dependent on a gas pressure in a cell as illustrated
in graph 8800 in FIG. 88, which does not change in time and is
known and determined in the cell manufacturing process.
[0566] The linewidths of the absorption lines also change with gas
pressure as illustrated in graph 8900 in FIG. 89. When lower
pressure is chosen, the lines are narrower in spectrum and their
positions can be measured more accurately. The gas pressure can be
determined by measuring and analyzing the spectral widths of the
absorption lines.
[0567] As mentioned above, several molecular gasses (including
Hydrogen Cyanide (H.sup.13C.sup.14N)) have been fully characterized
by NIST and provide excellent absolute frequency references.
[0568] Graph 9000 in FIG. 90 shows the absolute frequency accuracy
of Hydrogen Cyanide (H.sup.13C.sup.14N) absorption line positions
over a typical operating temperature range of -5.degree. C. to
+70.degree. C. The vertical axis is given in ITU channel numbers
where, for example, channel 20 corresponds to the optical frequency
192.0 THz and channel 60 corresponds to 196.0 THz.
[0569] Graph 9100 in FIG. 91 shows the FWHM linewidths of Hydrogen
Cyanide (H.sup.13C.sup.14N) absorption lines. It is apparent that
lower gas pressures offer narrower absorption peaks, which provides
more measurement accuracy.
[0570] A summary of the absolute errors in absorption peak
positions for Hydrogen Cyanide (H.sup.13C.sup.14N) at three
different gas pressures is presented in Table 2 and Table 3
(assuming no corrections for temperature dependence).
TABLE-US-00003 TABLE 2 Parameter Units MIN MAX Operating
Temperature Range degC. -5 70 Gas Pressure Uncertainty % -25 25
Fitting Error MHz 2.5 Line Spacing GHz 58 172
TABLE-US-00004 TABLE 3 Cell Pressure (kPa) Units 1.0 3.3 13.3 Max.
Frequency Error (2 * Sigma) MHz 11 32 57 Max. Linewidth FWHM GHz
0.8 2.3 9.0
[0571] Table 3 shows that when a cell with 1.0 kPa gas pressure is
used, the maximum (worst line) uncertainty of the absorption peak
spectral position is +/-11 MHz (within 2 standard deviations). The
average error among all lines under these conditions is +/-7.6 MHz.
These uncertainties can be improved by 10-20% by calibrating out
the temperature dependence.
[0572] Note that the absolute wavelength accuracy of the best
commercially available laboratory wavemeters is +/-30 MHz. These
instruments are very costly (tens of thousands of dollars), bulky,
operate only at a laboratory temperature range, and are not
suitable for DWDM system applications. Due to technical, physical,
size and cost limitations, their absolute accuracy has not been
improved at all over the last 15 years. At present, and at least
the near future, only national standards laboratories are able to
measure optical frequencies with better absolute accuracy than
+/-30 MHz.
[0573] We describe herein methods to measure the absolute laser
frequency with an accuracy of approximately +/-10 MHz in a cost
effective way, and with a design that can easily be implemented in
a DWDM system.
Calibration of the Oscillator Laser Frequency Setpoints by
Reference Frequencies of the Absorption Cell.
[0574] This section describes how we can use the absorption cell
described above as an ultra-stable optical frequency (wavelength)
reference.
[0575] A part of the oscillator laser light is transmitted through
the absorption cell and the resulting optical output from the cell
is detected by a photodiode, followed by a TZ, a low-pass filter
(LPF) (possibly with amplification) and converted to the digital
domain by an ADC. We refer to the DC voltage at the input of the
ADC as the "CELL Output". Again as before, amplification, LPF and
other processing functions can be implemented in the digital
domain, e.g. in a DSP, when the ADC is installed just after the
TZ.
[0576] While the oscillator laser is being scanned in frequency,
the CELL Output signal would stay mostly constant (corresponding to
a fully transparent cell without any gas) except for the spectral
regions which have gas absorption lines.
[0577] A typical spectral shape of an absorption line of a gas cell
is presented in graph 9200 in FIG. 92. In this particular example,
the absorption line P16 for Hydrogen Cyanide (H.sup.13C.sup.14N) is
measured at a quite high pressure of 13 kPa (100 Torr).
Consequently the line is quite broad, which reduces the accuracy
with which the peak position can be determined.
[0578] The oscillator frequency in FIG. 92 was scanned in 0.1 pm
steps (74 GHz).
[0579] After accumulating all the absorption line points during the
scan (i.e. the CELL Output versus the oscillator set frequency),
the collected data can be analyzed by variety of methods in order
to find an absorption dip position f.sub.CELL-i, where index "i"
numbers consecutive dips in the whole absorption spectrum (e.g. 50
strong lines in Hydrogen Cyanide).
[0580] The simplest method and the least accurate method would be
to select a sampling point of the scan with the lowest value. The
most sophisticated and most accurate method would involve:
1 fitting a straight line or a parabola to the relatively flat
background signal around the absorption line and subtracting this
fit from the measured sampling points of the absorption line; 2
fitting to the full (and background free) absorption shape a
theoretical Galatry function which offers a perfect fit to the
shape of the line; 3 determining from the Galatry function an
optical frequency of the dip position, f.sub.CELL-i 4 finding the
oscillator frequency setpoint, f.sub.o-near-CELL-i, which is the
closest to dip position; 5 fitting a straight line or a parabola to
several oscillator optical frequency setpoints, f.sub.o-set-j,
(where index "j" numbers consecutive sampling setpoints) recorded
during the scan for each sampling point and located approximately
symmetrically around f.sub.o-near-CELL-i 6 interpolating the
function determined in point (v) to calculate the oscillator
frequency setpoint, f.sub.o-set-CELL-i, which corresponds to tuning
the oscillator to the minimum of the absorption line with index
"i", f.sub.CELL-i. 7 assigning to the oscillator setpoint
f.sub.o-set-CELL-i an optical frequency value determined by NIST
for this absorption line, f.sub.CELL-i.
[0581] Simpler methods can be applied such as:
1 fitting a Voigt or Lorentzian profile function instead of Galatry
function; 2 not subtracting the background; 3 fitting a Gaussian
function or a parabola only in the neighborhood of the absorption
dip;
[0582] Regardless of the method used in determining the spectral
dip position, the end result of this process is to find the
setpoint of the oscillator, f.sub.o-set-CELL-i, which would tune it
as accurately as possible to the dip of the absorption line and to
assign to this setpoint the NIST-determined optical frequency,
f.sub.CELL-i.
Embodiment 4
Beating of a DWDM Signal Laser at Fixed Frequency (with a Broad
Spectrum) with an Oscillator Laser Scanned in Frequency and a
Frequency Reference Cell
[0583] Calibration of the oscillator frequency setpoints against
the known positions of the cell absorption peaks can be used to
significantly improve the absolute frequency accuracy of the
oscillator and consequently the signal laser.
[0584] The oscillator frequency setpoints can be calibrated with a
very high absolute accuracy only at the absorption peak
frequencies, f.sub.o-set-CELL-i=f.sub.CELL-i.
[0585] Any other optical frequency of the oscillator setpoints
needs to be calibrated by interpolation or extrapolation of two or
more f.sub.o-set-CELL-i determined setpoints.
[0586] FIG. 93 shows an example of a calibration of the oscillator
laser frequency setpoints by the reference frequencies of the
absorption cell. In this example, the oscillator laser frequency is
being scanned over a wider range of optical frequencies, such that
during scanning at least two absorption lines are measured,
resulting in calibration of at least two frequency points of the
oscillator during the scan. Preferably, the absorption lines
measured should be placed on both sides of the measured signal
wavelength.
Implementation Type 2
[0587] A practical implementation 9400 of calibration of the
oscillator setpoints and the measurement of beat frequency spectrum
is shown on FIG. 94.
[0588] A temperature sensor 9410 measuring the temperature of the
absorption cell 9420 is optional and required only if a slight
temperature dependence of the absorption line positions needs to be
calibrated out.
[0589] An algorithm implemented by the control unit 9450 in FIG. 94
includes the following steps:
1 Select a particular signal laser frequency to be measured and/or
controlled f.sub.s-target; 2 Set the frequency of the selected
signal laser to an initial setpoint value f.sub.s-set-ini, which
would be close but may or may not be equal (within required system
tolerance) to the absolute frequency of the signal laser 9425
required, f.sub.s-target; 3 Find from a theoretical look-up table
frequencies of (e.g. two) the closest cell absorption peaks,
f.sub.CELL-i1 and f.sub.CELL-i2 to the left and right of
f.sub.s-target, respectively (f.sub.CELL-i1<f.sub.CELL-i2); 4
Scan the frequency of the oscillator laser 9415 from
f.sub.CELL-i1-.epsilon. to f.sub.CELL-i2+.epsilon., recording at
each scanning step the frequency setpoint of the oscillator laser
9415. Make sure that .epsilon. is large enough, e.g. large enough
that both absorption peaks are fully measured including the flat
spectral regions to the left of peak i1 and to the right of peak i2
for background subtraction, if required; 5 Record at each scanning
step the amplitude of the CELL Output signal, which results in a
record of the cell absorption spectrum versus the oscillator
frequency (given by its setpoints at each scanning step); 6 Record
at each scanning point the amplitude of the RFDBS Output signal,
which results in a record of the RFDBS Output spectrum versus
oscillator frequency; 7 In a processing unit (e.g. DSP), analyze
all measured cell absorption lines and find the oscillator laser
9415 frequency setpoints corresponding to the minima of the
absorption lines, f.sub.o-set-CELL-i1 and f.sub.o-set-CELLi2. By
interpolation or extrapolation, calibrate the oscillator setpoints
for all scanning steps and use the calibrated frequencies for each
scanning step in the analysis of the RFDBS Output; 8 By processing
and analyzing the RFDBS Output spectrum, determine the current
absolute frequency of the signal laser 9425, f.sub.s-current; e.g.
by using equation (1) for NRZ modulation format and numerical
techniques (similar to those described for finding the spectral
position of the cell absorption dip) to find a center frequency of
the signal laser 9425, f.sub.s-current; 9 Calculate a correction to
the signal laser 9425 setpoint as
f.sub.s-set-corr=f.sub.s-target-f.sub.s-current; 10 Modify the
signal laser 9425 frequency setpoint as
f.sub.s-set-final=f.sub.s-set-ini+f.sub.s-set-corr; 11 Set signal
laser 9425 frequency to f.sub.s-set-final and call it from now on
f.sub.s-set-ini; 12 Repeat steps 1 to 11 as fast and as often as
needed at frequency f.sub.s-target as required by DWDM system
accuracy. 13 Repeat steps 1 to 12 for another DWDM signal laser or
lasers.
Implementation Type 3
[0590] The implementations described above assumed that the optical
output power of the oscillator laser does not change during the
frequency scan. Generally, this assumption may not be correct and
power variability of the oscillator potentially distorts both the
CELL Output and RFDBS spectra measurements.
[0591] FIG. 95 illustrates an implementation 9500 in which the
variations of the oscillator optical output power are measured and
used as a reference signal to correct (normalize) both CELL Output
and RFDBS Output spectra.
Implementation Type 4
[0592] For some DWDM system applications, the implementations
described above may not be accurate when the signal laser is being
modulated. This occurs when the signal laser light is modulated by
an RF signal, e.g. at 10 Gb/s, resulting in a signal optical
bandwidth, .DELTA.f.sub.s-FWHM, of several GHz, e.g. 16 GHz. Such a
wide optical bandwidth would produce an approximately equally wide
RFDBS Output spectrum, thereby reducing the accuracy of the
measurement by an equivalent amount.
[0593] FIG. 96 illustrates an implementation 9600 in which this
problem is avoided. In this embodiment, the signal laser 9625 light
is split (tapped at 9660) before the RF modulation is applied (by
data modulator 9670) and the frequency of the tapped light is
determined by one of the methods described above.
CASE B
Oscillator Laser Output is Linearly Polarized but the Signal Lasers
Polarization Randomly Changes in Time
[0594] In the above embodiments, both signal and oscillator lasers
light are polarized linearly and both polarizations are aligned on
the photodetector. In some applications, the implementation of
these designs would be too costly or not practical.
[0595] The strength of the beat frequency (RFDBS Output) depends on
the polarization alignment between the two lasers. When the lasers
are polarized orthogonally to each other the beat frequency would
not be generated at all. This issue is addressed below in a
practical design for random polarizations.
Implementation Type 5
[0596] FIG. 97 illustrates an implementation 9700 in which there
are no restrictions for the state of polarization of the signal
laser.
[0597] In this implementation, the oscillator laser 9715 light is
transmitted through a polarization scrambler 9720 with preferably
high scrambling speed (e.g. over 1 kHz). At the output of the
polarization scrambler 9720, the polarization of the oscillator
changes rapidly, and after some time reaches all or almost all
polarization states possible.
[0598] The scrambled oscillator light 9730 will beat with the
randomly polarized and time-dependent signal light for half of the
time on average, providing that:
1 the integration time constant of the integrator placed before the
ADC converter which provides RFDBS Output is substantially longer
than the scrambling speed; and 2 the scanning speed of the
oscillator laser is slow enough that, at each sampling point, both
RFDSB and CELL Outputs would reach a steady state.
[0599] The implementation 9700 illustrated in FIG. 97 works equally
well as all the previously described implementations in which both
lasers polarizations were preserved and aligned. The polarization
scrambler 9720 can be introduced to any of the four implementations
described above, and PM fibers in these implementations can be
replaced by non-PM fibers, e.g. SMF-28.
[0600] Thus, this novel approach applies to both types of fiber
systems: (i) based on polarization maintaining fiber; and (ii)
standard non-polarization maintaining fiber such as SMF-28; or
hybrid. In both systems, the absolute accuracy of stabilizing a
DWDM signal laser or lasers remains the same.
Stabilization of Multiple Signal Lasers in DWDM System
[0601] This approach of stabilizing DWDM signal laser wavelength
can be applied in many different architectures to stabilize
multiple signal lasers.
Implementation Type 6
[0602] FIG. 98 and FIG. 99 show possible architectures which share
the frequency stabilization hardware and control circuits among
many signal lasers (with signals propagating along multiple fibers
in embodiment 9800 or a single fiber in embodiment 9900.
[0603] This approach can also be used as a very accurate, high
resolution optical spectrum analyzer or fiber monitor. FIG. 100
shows a typical DWDM network node 10000 where all incoming and
outgoing optical spectra in all fibers are monitored.
[0604] FIG. 101 shows an electro-optic circuit 10100 that can be
used to measure the beat frequency for the case where there are
polarization-maintaining fibers connecting the signal lasers 10125
and the reference laser 10115. Note that the detection circuits in
FIGS. 101 to 103 can be implemented with the same blocks in a
different order, for example the AC RF bandpass filter with gain
10110 can be placed before the Zero offset rectifier 10120.
[0605] The absorption frequencies are characterized, and do not
vary significantly with temperature. Measuring the absorption of
the gas cell 10150 versus the frequency of the reference tunable
laser 10115 enables the software to apply a calibration factor to
the reference laser 10115 to increase its accuracy to the order of
.+-.0.1 GHz or better.
[0606] A microprocessor algorithm for the circuit shown in FIG. 101
includes the following steps:
1 Tune the reference laser 10151 to an absorption band; 2 Measure
the offset between the laser set point and expected frequency of
the absorption band (see 10152). Record the difference, f.sub.C,
and apply the offset of the following measurements; 3 Tune the
reference laser 10151 to the frequency of the signal laser 10125 to
be measured and scan the reference laser 10151 across the signal
bandwidth while measuring the strength of the beat signal (see
10153). Fit a curve to the measured beat signal and interpolate to
get the value of the reference laser 10151 frequency f.sub.R that
maximizes the beat signal f.sub.D. Calculate the frequency of the
signal laser 10125 as fs=f.sub.R+f.sub.D+f.sub.C; and 4 Calculate
the error in the signal laser 10125 and send a message to the laser
source to correct its setpoint (see 10154). Raise an alarm if the
laser 10125 is not responding to the tuning messages. Repeat until
the error is within an acceptable bound and then repeat for the
next laser 10125.
[0607] This method can also be used to measure the spectrum of
signals in a transmission fiber to provide a monitoring and alarm
function.
[0608] There are often cases in an optical network where it is
difficult to maintain the polarization between the reference laser
and the signal lasers. In such cases, the beat signal can be
measured by using a polarization controller 10250 (as shown in
embodiment 10200 in FIG. 102) to align the reference laser 10215
with the signal to be measured. As each signal laser 10225 is
scanned, the polarization controller 10250 is adjusted to maximize
the beat signal.
[0609] An alternative embodiment 10300 is shown in FIG. 103 where a
polarization scrambler 10320 is used instead of a polarization
controller.
[0610] Another embodiment of a subchannel muxponder 10400 is shown
in FIG. 104. In this embodiment, integrated transceivers 10410,
such as DWDDM XFPs, are tuned to the correct subchannel frequency.
The shared wavelength locker 10413 can be used at the transmitter
to improve the nominal channel spacing of the integrated optical
devices.
[0611] For example, 50 GHz XFPs can be locked to a 25 GHz or 12.5
GHz grid with the improved frequency accuracy given by the locker
10413.
* * * * *
References