CWDM

CWDM
Technology, Standards, Economics & Applications

Technology, Standards, Economics & Applications

Abstract
Coarse Wavelength Division Multiplexing (CWDM) delivers multiple wavelengths over an
optical fiber at a fraction of the cost of Dense Wavelength Division Multiplexing (DWDM).
It would seem to be a perfect fit for the metro-core and metro-access space, where fiber
optics is increasingly coming into play, but with economics quite different from the long-haul
market, where DWDM has dominated. Rising from the other end of the spectrum is Next
Generation SONET - a multi-protocol transport technology that could be either competitive
or complementary to CWDM. This paper addresses the characteristics of CWDM, and how it
matches up technologically and economically with the requirements of metro applications.

1 Introduction
1.1 Roots of CWDM

1.1.1 Early Generation WDM Technologies & Applications

CWDM technologies have been in use since the early 1980s for transporting multiple
channels with typically 25 nm spacing in the 850 nm window over multimode fiber Local
Area Networks (LANs). However, at that time, these were simply referred to as WDM
technologies. Applications included multi-channel video distribution and bi-directional,
latency-sensitive, telemetry & control information transmitted over a single fiber strand [1].

In the mid 1990s the technology and term Dense Wavelength Division Multiplexing
(DWDM) came into prominence. DWDM enabled over 80 densely spaced wavelengths in the
1550 nm window on a precise grid specified by ITU-T G.6921. Within this window, the
Conventional or C-Band (1530-1565 nm) was the preferred transmission band due the
minimum fiber loss region and due to the availability of Erbium Doped Fiber Amplifiers
(EDFAs) for this band. DWDM and EDFAs were needed to reduce the transmission costs of
high capacity, long haul networks.

With a new generation of DWDM products being developed for the long haul market, an
alternative term was required to differentiate widely spaced wavelengths or bands from
DWDM wavelengths. The term Coarse WDM (CWDM) was born and was increasingly used
from 1996 onwards to differentiate previous WDM technologies from ever-improving
DWDM. However, in the absence of standards, a definitive specification for CWDM
wavelengths and spacing was lacking and this led to some confusion as to its meaning and
application.

1
ITU-T G.692 (dated Oct’98) specified the absolute frequency reference (193.1 THz) and frequency spacings
(50GHz, 100GHz, 200GHz and 400GHz) to be used in different applications.

© Copyright 2002 Page 1

 CWDM Technology, Standards, Economics & Applications

From the mid-late 1990s, CWDM continued to be targeted at 850 nm multimode LAN
applications, but with new Vertical Cavity Surface Emitting Laser (VCSEL) and thin-film
filter technologies to reduce cost and increase packaging density [3]. By the late 1990s,
CWDM was gaining interest within the IEEE 802.3 High Speed Study Group for solving
dispersion and loss problems for 10 Gigabit Ethernet LANs and some 10 x GbE WAN
applications [5, 6]. For the 10GbE LAN applications, 4 wavelengths in the 850 nm or
1310 nm windows were proposed to extend the life of the installed base of multimode fiber in
building and campus environments. To differentiate between the two LAN windows, the
802.3 High Speed Study Group referred to the 850 nm wavelengths as CWDM and the
1310 nm wavelengths as Wide WDM (WWDM)

Throughout the mid-late 90s, there were also references to CWDM in carrier access network
applications such as Passive Optical Networks (PONs) [2, 8], however, by today’s standards,
these were really just band splitters for multiplexing upstream and downstream traffic into the
1310 nm & 1550 nm windows. Such 2-wavelength PONs were possibly referred to as
CWDM since DWDM in the 1550 nm window was the technology of choice for carrier
applications requiring more than 2 wavelengths [7]. With the evolution of narrower, 4-
wavelength CWDM and WWDM LAN technologies, the 1310/1550 nm band splitters were
eventually categorized as simply WDM technology to avoid confusion and to reflect the much
wider spectral passband.

A summary of the early (pre-2000) generation of WDM, CWDM, WWDM and DWDM
technologies and associated transmission windows are shown in Figure 1.

Figure 1 Early Generation WDM, CWDM, WWDM and DWDM Technologies

Page 2 © Copyright 2002

 CWDM Technology, Standards, Economics & Applications

1.1.2 Metro WDM Technologies

Due to the high network capacity growth being sought in the late 1990s, there was a need to
rapidly expand the bandwidth of metro networks with WDM technology. Fast market access
required maximum re-use of existing long-haul DWDM technologies. In some cases, the
product suppliers simply re-marketed their long haul products as metro products.

However, the metro area had an entirely different set of requirements: the distances were
shorter; more fiber was available; in addition to SONET/SDH, more protocols such as Gigabit
Ethernet and Fibre Channel were to be supported; the amounts of information were often
smaller; and the ability to pay for bandwidth was much less. Given these characteristics,
carriers and enterprises requested transmission means quite different from the densely packed,
extraordinarily precise, high bandwidth DWDM technologies so helpful to long distance
transmission.

Since the long haul had been the paradigm, WDM in the metro space in many ways became a
relaxed, far less expensive form of DWDM. The concepts of “wavelength banding” or
“hierarchical WDM” [4, 9] and “more coarsely spaced wavelengths” [9] emerged as
technology solutions, however, these were still fundamentally based on DWDM, with
200 GHz and higher wavelength spacing in the C-Band, on the precise grid specified by ITU-
T G.692.

More coarsely spaced DWDM wavelengths, such as 200 GHz and 400 GHz spacings, reduced
the metro product cost by requiring less stringent temperature and current control. However,
as evident from Figure 1, there was still a large disparity in wavelength spacing and
component cost when compared to the CWDM technologies used in LAN applications and
the WDM technologies used in PON applications. For example, the 850 nm CWDM and
1310 nm WWDM technologies had typically 25 nm wavelength spacing, whereas for
1550 nm DWDM technologies, the 200 GHz and 400 GHz spacings equated to approx.
1.6 nm and. 3.2 nm respectively.2

To further reduce metro product costs, expensive dispersion compensators normally used in
long-haul networks could be removed due to the shorter metro distances required. Expensive
EDFAs were still required however, to compensate for node losses in all-optical metro
networks. The term “Metro DWDM” was thus born to differentiate multi-protocol products
based on relaxed DWDM technologies from the long haul products based on more precise
DWDM technologies.

Notwithstanding the above component changes, there was still a failure on the part of
providers of long-haul DWDM to provide economic packaging and solutions for the metro
space. While Metro DWDM was an attempt to respond to some of these different variables, it
still had many of the undesirable characteristics of the products engineered for the inter-
exchange market.

2
The exact relationship between WDM wavelength and frequency is determined from the equation c = λf where
“c” is the speed of light in a vacuum, “λ” is the wavelength measured in a vacuum and “f” is the frequency.
Frequency is standardized (rather than wavelength), since it is independent of the transmission medium (eg, fiber
or air).

© Copyright 2002 Page 3

 CWDM Technology, Standards, Economics & Applications

For metro applications requiring lower network capacity and cost, solutions to the above
DWDM technology problems required new WDM technologies with simpler, wider tolerance
laser manufacturing practices, less accurate laser temperature control and reduced design
complexity and cost of optical filters. The CWDM technologies developed for the LAN
market were an obvious choice, however, these needed to be re-engineered to provide a range
of wavelengths more suited to the transmission distance requirements of metro applications.

As shown in Figure 2, in addition to the C-Band, there are the lesser-known wavelengths
bands, called the Original (1310 nm window), Extended, Short & Long or more simply, the
O, E, S & L bands. These bands can potentially be used to provide 10x more bandwidth than
the C-band or 10x wider wavelength spacing for the same bandwidth. The latter option was a
logical solution to the metro WDM technology and cost problems.

From the year 2000, papers started to appear promoting CWDM technologies and standards
for metro applications with 20-25 nm filter spacing and “uncooled” Distributed Feed-Back
(DFB) lasers [10, 11]. Recognition by carriers and the ITU of the need for low-cost CWDM
and the technology differences between CWDM and DWDM resulted in two separate WDM
standards:
- ITU-T G.694.1 “Spectral Grids for WDM applications: DWDM Frequency Grid”
(May’02)
- ITU-T G.694.2 “Spectral Grids for WDM Applications: CWDM Wavelength Grid”
(Jun’02-draft).

Figure 2 illustrates the ITU CWDM wavelength grid. These events led to the birth of “Metro-
CWDM” technologies as we define them today for the carrier market.

Figure 2 Metro CWDM Wavelength Grid as specified by ITU-T G.694.2

Page 4 © Copyright 2002

 CWDM Technology, Standards, Economics & Applications

1.2 CWDM Technologies & Standards

1.2.1 ITU CWDM Wavelength Grid

Metro CWDM technologies now comprise optical filters and uncooled lasers with 20 nm
spacing. There are 18 wavelengths currently specified with nominal wavelengths ranging
from 1270 nm to 1610 nm inclusive.

Figure 2 shows a mapping of the ITU-T G.694.2 CWDM wavelength grid into the O,E,S,C &
L bands. A typical attenuation curve for the installed base of ITU-T G.652 fiber (eg, Corning
SMF-28 and OFS Conventional fiber) is also shown. The mapping of CWDM wavelengths
onto the fiber attenuation curve has been done for greater clarity and to highlight the higher
loss incurred by some wavelengths (eg, in the E-band) for this type of fiber.

1.2.2 First Metro CWDM Implementations

As evident from Figure 2, five of the CWDM wavelengths fall within the E-Band. This band
is normally not used on standard G.652 type fiber due to the water peak. The loss due to the
water peak is typically 0.5dB/km which is not large, however, the maximum loss can be
2dB/km or greater. Carriers are not willing to take the risk that purchased equipment may not
operate on some or all of their metro G.652 type fiber, consequently, the first products to
implement and deploy CWDM in metro applications focused on the O, S, C & L bands only.

For all-optical metro applications where the O-Band has already been provisioned for a
1310 nm transport service, there is a preference to leave this band untouched and add new
capacity to the existing fiber via eight CWDM wavelengths in the S, C & L bands. For this
reason, the first CWDM lasers and filters developed to meet the emerging ITU-T G.694.2
standard were designed for eight CWDM wavelengths between 1470 nm & 1610 nm
inclusive. The fiber attenuation in this region is at or near its minimum, which is needed to
maximize the number of Optical Add/Drop Multiplexer (OADM) nodes and/or the
transmission distance between optical regenerators. CWDM transmitters in this region can
also use higher power lasers than the O-band while meeting Class 1 eye safety requirements.

Within the O-Band, various 1310 nm transport services are specified, such as existing
SONET/SDH or new IEEE 802.3ae 10 Gigabit Ethernet (10 GbE) services.

1.2.3 IEEE 10GbE Standard

The 10GbE standard supports several Physical-Layer (PHY) Media Dependent sub-layer
options in the O-Band. Of particular relevance to this paper is:

10GBaseLX-4 - Wide WDM (WWDM) PHY as detailed in section 1.2.4 below.

The 10GBaseLX-4 standard for the 1310 nm window resulted from the CWDM LAN studies
undertaken in the mid-late 1990s - as mentioned previously.

© Copyright 2002 Page 5

 CWDM Technology, Standards, Economics & Applications

Initially a 10GBaseSX-4 CWDM option was proposed for the 850 nm window and there are
proprietary products available that implement this option. The 10GBaseLX-4 standard is
outlined in detail below.

1.2.4 10GBaseLX-4

This standard is very similar to the ITU CWDM standard, except that it uses 24.5 nm spacing
between wavelengths and is differentiated by the term Wide WDM (WWDM). This standard
originated to allow the installed base of fiber-optic cabling in buildings and campuses to be
used for 10GbE. It is therefore intended for both multimode and singlemode fiber cables for
which excessive dispersion would be a problem at 10 Gbit/s serial line rates. Instead, 4
wavelength channels or “lanes” in the O-Band are defined (as listed below), each at a rate of
3.125 Gbit/s. Transmission distances are specified to 10 km but some commercial WWDM
devices can reach up to 20 km.

Lane 0 1269.0 - 1282.4 nm (1275.7 nm nominal)

Lane 1 1293.5 - 1306.9 nm (1300.2 nm nominal)
Lane 2 1318.0 - 1331.4 nm (1324.7 nm nominal)
Lane 3 1342.5 - 1355.9 nm (1349.2 nm nominal)

1.2.5 OIF VSR-5

In January 2002, the Optical Internetworking Forum (OIF) commenced the specification of a
Very Short Reach (VSR) Level 5 interface for Intra Office (CO) and client interconnects at
OC-768 (40 Gbit/s) rates. Several options are planned, including 4 x 10 Gbit/s CWDM over
singlemode fiber in the O-Band (1310 nm) [12]. The IEEE 10GBaseLX-4 wavelength plan is
proposed for this.

1.2.6 CWDM & WWDM Standards Convergence

Submissions have been made to the ITU for the IEEE 10GBaseLX-4 and OIF VSR-5
wavelength plan to be adopted for the ITU CWDM standard. One option could mean
replacing the existing 5 wavelengths spaced 20 nm apart in the O-Band (as defined by ITU-T
G.694.2), with 4 wavelengths spaced 24.5 nm apart (as defined by IEEE 10GBaseLX-4). This
would result in 17 rather than 18 CWDM wavelengths. These 17 wavelengths could for
example, be used for 16 user-channels plus a dedicated, high-speed Optical Supervisory
Channel (OSC).

Convergence of the Metro CWDM and LAN WWDM standards would further reduce
component and systems costs. Apart from the loss of one (almost useless) CWDM channel,
there do not appear to be any negative aspects associated with a single, converged CWDM
standard.

Page 6 © Copyright 2002

 CWDM Technology, Standards, Economics & Applications

1.3 Metro CWDM System Components

1.3.1 CWDM-Optimized Optical Fiber

For metro fiber network upgrades and greenfields applications, the opportunity now exists to
install the latest ITU-T G.652.C fiber technology which substantially eliminates the water
peak at 1383 nm and thus releases the E-Band for further capacity expansion. Both OFS and
Corning have fiber products on the market that conform to this standard. Multiplexer products
with typically 16 CWDM wavelengths are now emerging for broadband metro applications.

1.3.2 Existing CWDM Lasers

Direct modulated CWDM lasers with bit-rates up to 2.5 Gbit/s are optimized for low cost.
Their design is based on tried and proven DFB laser technology. The DFB technology has the
benefits of a narrow line-width with highly suppressed side-modes, thus providing similar low
dispersion performance to direct modulated DWDM lasers. As a result, CWDM lasers are
capable of transmitting 2.5 Gbit/s over distances of 80 km on ITU G.652 fiber.

The low cost, low power and small space benefits of CWDM lasers results from their
uncooled design. This means that they do not have bulky heat sinks, control circuits and
Thermo-Electric Coolers (TECs) coupled close to the laser chip, which saves electrical power
and space. A typical optical output power of 1mW (0dBm) is achieved with low cost CWDM
lasers.

The CWDM transmission system has been designed such that the CWDM laser wavelength
manufacturing accuracy and drift with temperature is less than the CWDM filter bandwidth.

More specifically, G694.2 recommends that the CWDM filters have a guard-band equal to
one third of the channel spacing. This means that for 20 nm spacing, the useable filter
bandwidth should be no more than 13 nm and hence the CWDM laser wavelength cannot vary
by more than 6.5 nm from the nominal center wavelength. This is illustrated in Figure 3 for
the case of 1550 nm.

DFB laser wavelengths vary from their nominal manufactured wavelength at a batch-
dependent rate of between 0.08 nm/°C and 0.12 nm/°C. For telecommunications-grade
equipment, lasers should as a minimum, be specified to operate over the temperature range:
0 °C to +70 °C. Assuming that a laser is selected from a bin to be at its nominal wavelength
(eg, 1550 nm) at the mid-point of the specified temperature range (+35 °C), then over a 0°C to
+70 °C variation, its wavelength can vary by as much as ± 4.2 nm. This means that to stay
within the 13 nm bandwidth of the filter, the maximum manufacturing tolerance on a CWDM
laser’s nominal wavelength is ± 2.3 nm. This is relatively small, but is not as tight as the
tolerance required for a DWDM laser.

© Copyright 2002 Page 7

 CWDM Technology, Standards, Economics & Applications

Figure 3 CWDM vs DWDM - Manufacturing and Temperature Tolerances

1.3.3 CWDM vs DWDM Laser Technology Comparison

As evident from Figure 3, 50 GHz, 100 GHz and 200 GHz DWDM lasers must be
manufactured with a tight tolerance on their nominal wavelength. Additionally, DWDM lasers
must be cooled to isolate the laser from the ambient temperature changes that would cause its
wavelength to drift outside the DWDM filter pass-band. Each laser includes a thermister to
monitor the laser temperature and in the case of 200 GHz fixed wavelength lasers, it is often
the thermister that is calibrated during manufacture to set the precise ITU wavelength of a
laser. The calibrated thermister increases the manufacturing yield by allowing a wider
variation in the nominal wavelength of the laser than the 200 GHz DWDM filter pass-band
would suggest (approaching that of a CWDM laser). In the case of 50 GHz lasers, a thermister
enables coarse temperature control but an external wavelength locker is required for fine
wavelength control.

1.3.4 Vertical Cavity Surface Emitting Lasers (VCSELs)

Vertical Cavity Surface Emitting Lasers (VCSELs) are a type of semiconductor diode laser
whose cavity is perpendicular to the wafer plane. A VCSEL cavity comprises distributed
Bragg gratings that stabilize the wavelength and suppress the side-modes. This design
provides the advantages of low-cost manufacturing and high spectral performance. Since light
is emitted vertically with respect to the wafer plane, VCSELS are ideal for laser array
applications, such as parallel optical interconnects. They are now manufactured in volume for
GbE and 10 GbE WWDM LAN applications with 850 nm / 1310 nm and
singlemode/multimode options.

Page 8 © Copyright 2002

 CWDM Technology, Standards, Economics & Applications

As mentioned in section 1.2.5, the OIF VSR-5 study group is now specifying 4 x 10 Gbit/s
O-Band CWDM lasers which can be used for very short-reach LANs and Central-Office (CO)
interconnects. These CWDM devices will most likely be based on low cost 1310 nm
VCSELs.

Success has recently occurred in the development and manufacture of Long Wave (1500-
1610 nm) VCSELs (LW-VCSELs) for singlemode fiber and WDM applications [13]. Both
fixed wavelength and tunable variants using integrated MEMS technologies have been
demonstrated. Transmission distances of 80 km have been achieved at 2.5 Gbit/s with small
power penalty and the technology is heading towards 10 Gbit/s. Wavelengths suitable for
CWDM & 100 GHz DWDM applications have been achieved.

Sample LW-VCSEL devices have been tested across the entire industrial temperature range
and have performed successfully without the need for thermal compensation, thus eliminating
associated packaging costs. Such devices will be suitable for low cost, single and multi-
channel Intermediate Reach and Long Reach SONET/SDH applications. As an extension to
OIF VSR-5, it is likely that 10 Gbit/s S, C & L-Band CWDM lasers will be commercialized
and as discussed later in section 2.3.1, legacy Dispersion Shifted Fiber (DSF) that has been
abandoned could be re-lit to enhance this development and extend the reach of 40 Gbit/s
CWDM systems.

1.3.5 Receivers

In contrast to standard single-protocol receivers, the receivers used in multi-channel CWDM

(and DWDM) systems often require larger bandwidths that can capture all the specified bit-
rates and protocols. The front-ends of these receivers use wavelength agnostic PIN or
Avalanche Photo-Diode (APD) detectors that cover the entire ITU CWDM band. It is the
CWDM filters that provide the wavelength selectivity. The benefit of the PIN detectors is a
lower-cost, simpler receiver design. The benefit of the APD detectors is a 9-10dB
improvement in receiver sensitivity.

Wideband Trans-Impedance Amplifiers (TIAs) are integrated with the detectors for maximum
receiver sensitivity. Regeneration is then provided using either 2R (reshape) or 3R (retime)
techniques. The latter is nearly always used for the higher data rates to maximize receiver
sensitivity and transmission distance and to meet the customer-interface jitter specifications.
Programmable, multi-rate Clock/Data Recovery (CDR) devices are invariably used as 3R
regenerators in WDM systems. Simple, low cost limiting amplifiers are used for 2R
regenerators.

Typical 2.5 Gbit/s receiver sensitivities and associated optical power margins are –24dBm for
a PIN/TIA with limiting-amp and CDR and –33dBm for an APD/TIA with limiting-amp and
CDR (assuming 223-1 PRBS & 10-10 BER). Semiconductor Optical Amplifiers (SOAs) will
soon be available to extend the end-end optical power margin. However, these do not solve
dispersion problems, so 2.5 Gbit/s metro CWDM networks extending beyond the nominal
dispersion limit of CWDM lasers and G.652 fiber will need 3R regeneration at typically, 60-
80 km intervals.

© Copyright 2002 Page 9

 CWDM Technology, Standards, Economics & Applications

1.3.6 CWDM Filters

CWDM filters are implemented using thin film technology. As shown in Figure 4, they are
available as discrete single-channel filter devices and as integrated multiplexer / demultiplexer
devices with typically 4 or 8 wavelength ports. Various configurations of these devices can be
used to implement a multi-channel Optical Add/Drop Multiplexer (OADM) product. CWDM
filters can be specified for uni-directional transmission on 2-fiber networks or for bi-
directional transmission on single fiber networks. The latter option has the advantages of
lower first-in cost for leased-fiber applications and reduced fiber-count for fiber-exhaust
applications.

Figure 4 CWDM Filter Options: 1-Channel and 8-Channel Mux/Demux Schematics

As evident from Figure 3, individual CWDM filters are required to have at least 13 nm wide
flat-top passband with small passband ripple and a sharp spectral skirt. Typical insertion
losses are around 1 dB for single-channel filters and around 4 dB for 8-channel mux/demux
filters. Typical adjacent channel isolation is >30 dB – aimed at reducing crosstalk between
channels.

Due to the thin film processes and materials used, the thermal stability of the CWDM filter
center-wavelength is excellent – typically resulting in less than 0.002 nm/°C drift with
temperature. This equates to less than ±0.07 nm wavelength change over a ±35°C temperature
variation. Manufacturing tolerances can be an issue however, with filter devices exhibiting up
to ±0.3 nm center-wavelength variance. When included with the CWDM laser temperature
variance, the maximum manufacturing tolerance on a CWDM laser’s nominal wavelength is
reduced to ± 2.0 nm. CWDM lasers are readily available with this manufacturing tolerance.

Notwithstanding the specification by ITU-T G694.2 of CWDM wavelengths centered on

nominal values (1550 nm etc), an interesting quirk has occurred in the laser and filter device
specifications. The lasers are specified by the manufacturers to be on the nominal ITU
CWDM grid at 25 °C (rather than 35 °C), which would be in the center of the filter passband
if the maximum case temperature was to be 50 °C (eg, for a commercial-grade product).
However, the 0-70 °C range needed for carrier-class applications means that at 35 °C, the
lasers would nominally have drifted to a wavelength equal to the ITU grid plus 1 nm. This has
caused several filter suppliers to re-specify all the CWDM filter center wavelengths to be off-
center by 1 nm to 1551 nm etc!

Page 10 © Copyright 2002

 CWDM Technology, Standards, Economics & Applications

1.3.7 OADMs in Metro CWDM Networks

Figure 5 illustrates a linear (bus) network comprising an 8-wavelength CWDM mux/demux

filter at the head-end (eg, Central Office) and four OADM filters distributed between remote
customer nodes, spaced 15 km apart. The total network size is 60 km, which in this example
is loss-limited – assuming an APD detector, 1 dB OADM insertion loss (express & add/drop),
4 dB mux/demux insertion loss, and 0.4 dB/km fiber + splice loss at the 1470 nm and/or
1610 nm wavelengths.

For 2.488 Gbit/s OC-48/STM-16 applications, this OADM network is also approaching the
80 km dispersion limit of many direct-modulated CWDM lasers. This is due to chromatic
dispersion of the modulated 1610 nm wavelength in standard G.652 fiber. Such fiber is
optimized for zero dispersion at 1310 nm and thus causes significant dispersion at 1610 nm.

As shown in Figure 5, calculating the maximum size of the downstream segment of an un-
amplified CWDM bus network requires simple math – not dissimilar to the tap-loss
calculations used for the coax. segment of HFC networks. The flip side to this is that un-
amplified CWDM networks are intrinsically small in size and thus confined to metro-access
applications having a small number of customer nodes.

1.3.8 Repeaters and Amplifiers

Expansion of CWDM networks to greater distances and/or more nodes than shown in Figure
5 requires the insertion of either a repeater or an amplifier. Repeaters can provide 3R
regeneration (re-amplify, reshaping & retiming) to overcome loss and dispersion limitations
whilst amplifiers provide 1R regeneration to over come optical power loss limitations only.

Single wavelength regeneration is an incremental solution that is sometimes used to overcome

loss or dispersion limitations on an "as required" basis. For example, if a OC-48 service
shown in Figure 5 is required at a node well downstream of the GbE node, this can be made
possible by inserting a 3R repeater in place of the existing OC-48 interface as shown in
Figure 6. This is potentially a low cost solution, but only if other services are required at the
repeater node.

As evident from Figure 6, one of the problems with all-optical WDM networks with single
wavelength repeaters is a disparity in optical signal levels for different wavelengths, which
can lead to optical crosstalk problems between adjacent channels. Even without single-
wavelength repeaters, such problems also occur in the upstream path close to the head-end
where large amplitude signals are added to heavily attenuated signals from remote nodes. To
solve this problem, all-optical CWDM systems are often supplied with a range of optical
attenuators or different OADM filter tap-ratios and customized optical network design
software.

An alternative solution to the loss/dispersion limited network expansion problem is to

regenerate all wavelengths at appropriate points in the metro network before the optical signal
degrades too much. Such a multi-wavelength repeater might be implemented as shown in
Figure 7, where 3R regenerator bit-rates are programmed for each channel at the time of
connection establishment.

© Copyright 2002 Page 11

 CWDM Technology, Standards, Economics & Applications

Figure 5 Example Network of 2-Channel OADMs and a 8-Channel Mux/Demux

Figure 6 Single-Wavelength 3R Repeater

Figure 7 Multi-Wavelength 3R Repeater

Page 12 © Copyright 2002

 CWDM Technology, Standards, Economics & Applications

Note that unless integrated with a major optical switching center, a multi-wavelength 3R
repeater is not practical in a high capacity DWDM network due to the larger number of
wavelengths to regenerate and/or the size, power and cost associated with each wavelength
regenerated. This is why C & L band optical amplifiers are preferred for regional and long
haul DWDM networks.

For loss-limited, CWDM metro networks having a large number of nodes, an alternative,
lower cost solution to the multi-wavelength 3R repeater is a simple optical amplifier (1R
regenerator) as used in the regional and long-haul DWDM networks. The expected
commercial availability from 2003 onwards of a range of small, low cost Semiconductor
Optical Amplifiers (SOAs) that together cover the O, E, S, C & L bands will enable this
solution to be implemented.

Figure 8 shows a schematic of such a SOA device.

Figure 8 SOA – 1R Regenerator for CWDM Networks

2 Metro CWDM Benefits

2.1 Component Related Benefits

2.1.1 CWDM vs DWDM Filters

Dielectric Thin Film Filter (TFF) technology is used to manufacture low cost CWDM filters.
The desired filter characteristics (center-wavelength, channel bandwidth, ripple height,
insertion loss, skirt-width and adjacent channel isolation) are all achieved through controlled
deposition of optical layers of different refractive index dielectric material on a glass
substrate. For DWDM applications, the same TFF technology has been refined and has
become an attractive, low cost alternative to fiber Bragg gratings. As a result, the cost
differences between CWDM and DWDM filters is now dominated by the number of optical
layers required to implement them. Basically, the narrower the wavelength spacing, the
greater the number of optical layers and the tighter the accuracy needed to meet the thickness
requirements of each layer.

Consequently, CWDM filters are inherently less expensive to make than the DWDM filters
due to the fewer number of layers in the filter design. Typically there are over 100 layers
required for a 200 GHz filter design as used in Metro DWDM products, whereas there are
only 50 layers in a 20 nm filter as used in Metro CWDM products. The result is shorter
manufacturing time, less materials and higher manufacturing yields for CWDM filters. As a
result, CWDM filter costs are generally less than 50% of the cost of comparable DWDM
filters [14].

© Copyright 2002 Page 13

 CWDM Technology, Standards, Economics & Applications

2.1.2 CWDM vs DWDM Transmitters

As mentioned in section 1.3.2, Distributed Feedback (DFB) lasers are currently used for both
CWDM and DWDM transmitters. Laser technology is therefore not a major cost
differentiator.

For metro applications, direct (current) modulation is preferred for both CWDM and
(200+GHz) DWDM lasers to minimize the cost of the transmitter. Consequently, laser driver
technology is not a major cost differentiator. This is in contrast to regional and long haul
applications, where higher-cost externally modulated DWDM lasers are often used to avoid
frequency chirping effects (cause by direct modulation), associated dispersion and
transmission distance limitations.

The dominant factor that differentiates CWDM transmitter costs from DWDM transmitter
costs is the WDM channel spacing. The channel spacing determines how far the associated
laser that lights the channel can drift from the nominal wavelength due to manufacturing
tolerances, temperature range and modulation current. This was graphically illustrated in
Figure 3.

The manufacturing, packaging and thermal control technologies needed to constrain the
wavelength variances are what differentiate the CWDM from DWDM transmitters used in
metro applications. For example, the manufacturing wavelength tolerance of a fixed-
wavelength DWDM laser die is of the order of ±0.2 nm; whereas the manufacturing
wavelength tolerance for CWDM laser die is of the order of ±2.0 nm. Lower die yields thus
drive up the costs of DWDM lasers relative to CWDM lasers. Using wider-spaced DWDM
lasers, such as 400 GHz – 800 GHz wavelength spacing (on ITU G.694.1 grid) lessens
manufacturing-yield as a cost differentiator.

Uncooled CWDM lasers are typically packaged with a laser die and monitor photodiode
mounted in a hermetically sealed coaxial container with a glass window. These containers are
aligned with either a fiber pigtail or an alignment sleeve that accepts a connector. The
container plus sleeve forms a coaxial package called a Transmitter Optical Sub-Assembly
(TOSA). A typical TOSA is approximately 2 cm long and 0.5 cm in diameter. The laser driver
is normally separate to the laser and is typically mounted on a PCB close to the laser. The size
of the laser driver is approx. 0.5 cm x 0.5 cm x 0.1 cm and the entire CWDM laser transmitter
(laser + laser-driver) consume about 0.25 watts of electrical power per wavelength. This
equates to an electrical-optical conversion efficiency of approx. 0.4%.

Cooled DWDM lasers are offered in either a butterfly or dual in-line package and contain the
laser die, monitor photodiode, calibrated-thermister, Peltier cooler (TEC) and heatsink. These
devices are almost always pigtailed, requiring fiber management. The laser transmitters vary
in size from manufacturer to manufacturer. Some are about 5cm long, 5cm wide and 1cm high
and include the laser driver and laser temperature control circuitry. A DWDM transmitter,
associated laser, laser driver, TEC, monitoring and control circuitry consume around 5 watts
of electrical power per wavelength. This equates to an electrical-optical conversion efficiency
of 0.02%.

Page 14 © Copyright 2002

 CWDM Technology, Standards, Economics & Applications

The following summarizes the differences between DWDM and CWDM transmitters:
The volume occupied by a DWDM laser transmitter is about 8 times the volume of a
coaxial CWDM laser transmitter.
The power consumed by a DWDM transmitter is about 20 times the power consumed
by a CWDM transmitter. For a sixteen-channel WDM system, the CWDM
transmitters consume approx. 4 watts, while the same functionality in a DWDM
system can consume over 80 watts.
Due to the above issues, packaging a DWDM laser transmitter is more expensive than
an uncooled CWDM laser transmitter. As a result, DWDM transmitter components are
typically 4-5 times the cost of their CWDM counterparts [14].

2.1.3 LW-VCSELs vs Uncooled CWDM DFB Lasers

Fixed wavelength LW-VCSELs promise to achieve lower cost-points than existing DFB
CWDM lasers and tunable LW-VCSELs will be attractive for OADM applications [13].
However, the greatest cost benefits will occur for point-point applications if multiple
uncooled CWDM lasers having different wavelengths can be manufactured on the same die
and packaged as a single multi-channel device.

In addition to the above direct cost benefits, there are indirect cost benefits of LW-VCSELs.
For example, LW-VCSELs exhibit very low threshold currents (≈1-2 mA) which simplifies
and reduces the cost of transmitter driver designs. However, they do not produce as high
optical power output as uncooled CWDM DFB lasers. If higher output powers ≈1 mW are
required, then an external SOA can be used to amplifier the output of a single LW-VCSEL or
can be shared to boost the output of multiple
LW-VCSELs.

LW-VCSEL arrays also offer significant space savings and associated direct and indirect cost
benefits. Future devices could extend the transmission distance of the OIF proposed 40 Gbit/s
systems using 4 x 10 Gbit/s CWDM channels as discussed in section 1.2.5. If such devices are
developed and commercialized, 40 Gbit/s TDM transmitters may find it extremely difficult to
compete in terms of cost, power, space and transmission distance.

2.1.4 Future Mass Produced Components

Significant manufacturing cost reductions in CWDM components are expected over the next
2 - 3 years due to automated manufacturing processes and increased component integration.
TFF filter costs are projected to drop by a factor of three during this time. LW-VCSEL
transmitter and APD/TIA receiver arrays with integral thin-film mux/demux filters are likely
to be developed for multi-channel point-point Metro CWDM applications. This will further
reduce the cost per channel. As evidence of this trend, 10GbE CWDM VCSEL transceivers
and mux/demux filters are already available with this level of integration and associated mass
production cost benefits.

© Copyright 2002 Page 15

 CWDM Technology, Standards, Economics & Applications

2.2 Product Related Benefits

The smaller size and lower power consumption of CWDM vs DWDM components translates
into smaller CWDM multiplexer product dimensions, fewer or lower-consumption power
supplies and reduced thermal management equipment to remove the heat that both the
transmitters and power supplies generate. These differences are graphically illustrated in
Figure 9 using data from section 2.1 for representative 16 wavelength DWDM and CWDM
multiplexer products.

Figure 9 Representative DWDM vs CWDM Products – Space & Power Differences

The multiplexer architectures shown in Figure 9 are based only on the most significant
component differences between 200 GHz DWDM and CWDM technologies, being thin-film
filters and directly modulated, DFB laser transmitters. This approach provides an
unadulterated comparison, which is not diluted by other product features and component
technologies (eg, sub-lambda multiplexing) that have little to do with the DWDM vs CWDM
technology comparison.

In both cases, pigtailed lasers are assumed, which impact on the printed circuit board sizes
due to fiber management issues (minimum bend-radius). Note that in contrast to DWDM,
CWDM lasers are becoming available in connectorized versions that enable even more
architecture options for CWDM products. DC-DC power supplies are assumed to have 80%
power conversion efficiency.

Page 16 © Copyright 2002

 CWDM Technology, Standards, Economics & Applications

The representative product space and power differences are significant – but of little surprise
to equipment manufacturers who have experience in both DWDM and CWDM product
developments. As shown in Figure 9, the combination of fiber management and power (heat)
density drive the DWDM products towards a large chassis with many vertical card types. In
contrast, the low power density of the CWDM technologies permits a low-profile chassis with
fewer module types, or even a single motherboard in the case of a commercial-grade product.

The additional space requirements of DWDM products result in considerable extra cost –
especially in co-location situations where the space is leased from another carrier. Additional
space is also required for larger AC-DC power supplies and/or backup batteries (not shown in
Figure 9) – which are a major consideration in the operation of telecommunications
equipment in COs and Outside-Plant (OSP) / Fiber to the Cabinet (FTTC) applications.

In addition to the power supply and thermal management requirements and associated capex
costs of DWDM systems, opex costs also increase due to the power and air-conditioning load.
Due to the above considerations, the life-cycle cost of a DWDM solution compared to a
CWDM solution is greater than the basic component cost comparison would indicate.

2.3 System Related Benefits

2.3.1 Dispersion Shifted Fiber (DSF)

ITU-T G.653 dated October 2000 defines single-mode fiber that has been dispersion shifted to
have both chromatic dispersion and attenuation nulls at 1550 nm. This type of Dispersion
Shifted Fiber (DS fiber) was optimized for use in very high bit rate TDM (40 Gbit/s OC-768)
applications and has been widely deployed by countries that were early leaders in fiber
technology. Since carriers often deploy the same type of fiber throughout their long haul,
regional and metro networks to reduce inventory costs, there is likely to be a large investment
in DS fiber in the metro network.

Unfortunately, DS fiber is not suitable for use with C-Band DWDM technologies due to
“4-wave mixing”. It can be used with new L-Band DWDM technologies for long-haul
applications, but L-Band DWDM is far too expensive for metro network applications.

For metro applications, CWDM laser and filter technologies are more suited to DS fiber, since
they are low cost and do not suffer from optical crosstalk induced noise due to 4-wave
mixing. This is because even if one of the CWDM wavelengths is at the dispersion null, the
next adjacent wavelength is 20 nm away, which means that extremely high optical powers and
long-haul transmission distances would be required before there would be any evidence of
interference.

Due to the above issues, many carrier networks that were deployed early have been upgraded
with Non-Zero DS fiber (NZDSF) as defined by ITU-T G.654. As a result, there may be a
large investment in DS fiber lying dormant in the ground. CWDM products enable existing
DS fiber investments to be reactivated and/or new investments in NZDSF infrastructure to be
deferred – thus preventing the unnecessary tearing up of an entire section of DS fiber cable.

© Copyright 2002 Page 17

 CWDM Technology, Standards, Economics & Applications

Furthermore, as mentioned in section 1.2.5, legacy DS fiber that has been abandoned could
also be fired-up again to extend the reach of future 4 x 10 Gbit/s CWDM systems operating in
the S, C & L Bands. In such situations, Metro CWDM products will pay for themselves many
times over.

2.3.2 Regenerative OADMs

Unless DS fiber is available with a dispersion null near the center of the CWDM band, an all-
optical CWDM network will be severely limited in size. In the case of standard ITU-T G.652
fiber, dispersion limits the end-end transmission distance to approx. 80 km when 2.5 Gbit/s
data is modulated onto a 1610 nm-wavelength DFB laser source. The availability of SOAs
that amplify CWDM wavelengths in the S, C & L bands will not overcome this dispersion
limitation.

For regional and long-haul DWDM applications, a combination of external rather than direct
laser modulation, fiber dispersion compensators and optical amplifiers are used to achieve all-
optical transmission distances in excess of 3000 km. Needless to say, such solutions are far
too expensive for extending the size of Metro CWDM networks.

However, the small size, low power and low cost of CWDM transmitters, receivers and
filters, and the smaller number of wavelengths transported, have caused conventional WDM
network design wisdom to be “turned on its head”. These technologies have enabled new
Metro WDM product design options that until now would have been considered impractical
and/or too costly.

For example, Figure 10 illustrates that a multi-wavelength 3R repeater as previously shown

in Figure 7, can be enhanced to include full OEO Add/Drop Multiplexing (ADM) features
normally found in major optical switching centers, thus creating a Regenerative OADM (or
R-OADM). Multiple R-OADMs can be connected to form a regenerative CWDM network,
which may be configured as a bus or ring. When configured as a ring, full logical mesh
connectivity and multiple (BLSR & UPSR) protection options are possible. This is referred to
as a “true ring” topology, as against the looped-bus topology used by many all-optical OADM
& PON networks.

Figure 10 Regenerative CWDM Network

Page 18 © Copyright 2002

 CWDM Technology, Standards, Economics & Applications

As shown in Figure 10, a CWDM networks built with R-OADMs simplify network design
and long-term planning. This is because each node is like a SONET ADM and the length of
each link can be calculated independently of other links – using simple SONET-like
Intermediate Reach (IR) or Long Reach (LR) design rules. The choice of IR or LR depends on
the CWDM laser power and the receiver type (PIN or APD). Like SONET & SDH,
regenerative OADM networks can cover large metro areas, exceeding 160 km (100 miles).
For the carrier, these benefits result in significantly lower opex costs and depending on the
network configuration, can result in equal or lower capex costs compared to similar OADM
networks that provide ad-hoc regeneration.

Regenerative OADMs enable more flexible network solutions that provide the multi-protocol
transparency and low latency of WDM and the network design simplicity of SONET & SDH
(TDM). New TDM technologies, such as Next Generation SONET (NG-SONET) will have
an important part to play above the CWDM layer, by enabling smaller (sub-lambda) channel
granularities and hence greater bandwidth efficiency. However, NG-SONET imposes greater
processing overheads, latency, power consumption and cost when it tries to compete head-on
with the multi-protocol simplicity of regenerative CWDM networks.

In conclusion, CWDM based R-OADMs with 10 Gbit/s capacity per fiber-strand today
(40 Gbit/s in the future) offer functional and economic advantages compared to conventional
CWDM based OADMs and OC-192 NG-SONET based ADMs in metro-access and metro-
core networks.

3 Metro CWDM Market Segmentation

3.1 Definitions

Metro-access and metro-core applications for CWDM networking products define the
following horizontal (Layer-0) market segments:

H1: Fiber to the Building (FTTB): low cost, commercial or carrier-grade metro-access
products
H2: Remote Terminals (FTTC): roadside hardened carrier-grade products for OSP
applications
H3: Inter-Office (CO-CO): carrier-grade products that solving immediate metro-core needs

For each of these horizontal market segments, there are at least two and generally more of the
following vertical (Layer 1-7) market segments:

V1: Basic Services: POTS, ISDN, Internet & Pay-TV (DS3, E3, OC-3 etc & HFC);
V2: Broadband Services: High-speed Internet, Video-on-Demand (OC-12 – OC-48 or GbE);
V3: Transparent Services: Leased wavelengths (λ) or leased channels (DS3/E3/100BaseT);
&
V4: Enterprise/SAN: GbE LAN extension, ESCON/FICON Storage Area Networks (SANs);

The horizontal CWDM market segments are outlined in the following sections, with reference
to the vertical market applications.

© Copyright 2002 Page 19

 CWDM Technology, Standards, Economics & Applications

3.2 Fiber to the Building (FTTB)

Building equipment rooms are popular locations for co-locating carrier’s metro-access
equipment and private enterprise’s LAN/WAN gateway equipment. Enterprise customers and
carriers both require that Fiber to the Building (FTTB) networks support multiple services
(eg, OC-n, FICON, GbE) and multiple path protection options. As shown in Figure 11, ring
topologies are often used as an efficient means of providing both protected and unprotected
services to different customers.

Metro CWDM networks are ideal for these

applications since space is a premium in
equipment rooms and high power-
consuming equipment is not
desirable where there is
inadequate air conditioning.
The low capex and opex
cost benefits of CWDM are
essential for justifying new
higher bandwidth services
that the enterprise customers
demand but are often reluctant
to pay a premium for.

Examples include GbE

LAN extension, ESCON
& FICON (Fibre Channel)
Storage Area Networks (SANs),
as well as existing broadband services
transported via ATM or Packet over SONET.

As shown in Figure 11, metro CWDM rings

with the full logical mesh connectivity offered by
R-OADMs are ideal for vertical market segments
requiring transparent service connections between
buildings in a Central Business District (CBD).

R-OADMs also support the ability to re-use the limited

number of wavelengths and channels available, which
further improves the economics of deploying 10Gbit/s
CWDM equipment in FTTB applications.

Figure 11 FTTB Application

Page 20 © Copyright 2002

 CWDM Technology, Standards, Economics & Applications

3.3 Remote Terminals (FTTC)

Due to population growth and resultant urban sprawl, typically 25% of residential customers
in the US today are now greater than 33 kft (10 km) from a Central Office (CO) and 15% are
now greater than 50 kft (15 km) from a CO. This trend is at odds with the demand for greater
bandwidth to the home and business since Asymmetric Digital Subscriber Line (ADSL)
technology is generally specified for a serving area of less than 12 kft (3.7 km) from the CO.

For these reasons, Digital Loop Carrier (DLC) access networks previously
deployed to deliver basic POTS services
to outlying areas have now become the
platform of choice for delivering the
new range of broadband DSL services.

The DLC access networks comprise

roadside-hardened Remote Terminal (RT)
equipment installed in Outside Plant (OSP)
cabinets. With the installation of these
Fiber to the Cabinet (FTTC) networks, the
take-up of ADSL has increased due to the
higher bandwidths offered by shorter last-mile
transmission distances. As a result, the DLC
fiber loop is now the final bandwidth
bottleneck in the broadband services
delivery chain. There is now the need
for a high capacity backhaul solution
to this problem.

Due to the low cost, power and space

benefits of CWDM, this is an
ideal “pair-gain” technology
for installing in Outside Plant
applications - close to the customer.

In contrast to NG-SONET upgrades, a

regenerative CWDM network (Figure 12)
provides a cost-effective “no fork-lift upgrade”
solution for adding a new OC-12 or OC-48
backhaul for the new broadband DSL services,
without affecting an existing OC-1 or OC-3 ring
previously installed for the basic POTS services.
Figure 12 FTTC Application

© Copyright 2002 Page 21

 CWDM Technology, Standards, Economics & Applications

3.4 Inter-Office (CO-CO)

There is little doubt that the demand for inter-office network bandwidth will grow at a rate of
at least 20% per annum for the next few years. Excess fiber capacity is being consumed and
wavelengths are needed now in those pockets of the metro network where fiber exhaust is
occurring first. The issue for the carriers today is whether to run more fiber or to install a
WDM overlay. If only metro DWDM overlays are considered, then the economics and
logistics issues may favor not doing anything at all – which will delay new service incomes.

However, the emergence of carrier-grade metro CWDM products,

which conform to the ITU-T G.694.2 wavelength grid, has
changed the economics of providing a WDM overlay.
CWDM requires considerably less equipment
space and power, and offers typically 8x pair-
gain on existing ITU-T G.652 and G.653 (DSF)
fiber infrastructure. At a conservative 20% yearly
growth rate, one may get another ten years life out
of the installed fiber base. At a very optimistic
50% yearly growth rate, this equates to another
five years of fiber life – which these days, is
longer than most carrier’s business plans.
Furthermore, for situations where DS
fiber is widely deployed (eg, Japan),
CWDM enables a much greater
return on investment of this
fiber infrastructure than
previously thought possible.

Compared to metro-access
networks, Inter-Office networks
are much larger and may not be
hubbed. Instead, all COs in the metro-core network
may be peers. Regenerative CWDM products that
support longer transmission distances (at least 160 km)
and “true ring” capabilities (such as full logical-mesh
connectivity) are therefore ideal for inter-connecting
COs within the metro-core network. One existing and
five private network applications (A - E) of CWDM in
Inter-Office networks are illustrated in Figure 13. These
6 services require only 4 wavelength channels per fiber
due to the wavelength re-use capabilities of R-OADMs.

Figure 13 Inter-Office Application

Page 22 © Copyright 2002

 CWDM Technology, Standards, Economics & Applications

3.5 NG-SONET, CWDM & DWDM Product Positioning

NG-SONET, CWDM and DWDM all have technical and economic benefits in different
segments of the telecommunications market. As a result, carrier transport networks are
evolving into a hierarchical bandwidth structure, with increasing network capacity from NG-
SONET to CWDM to DWDM. It is expected that most service providers and corporate
customers will interface to the Metro network at the NG-SONET and CWDM layers. The
choice will depend on the total bandwidth, bandwidth-utilization and latency requirements of
the customer’s applications.

Customers with high bandwidth and high bandwidth-utilization needs or low latency
requirements (eg, FICON / Fibre Channel) will generally interface at the CWDM layer.
Applications include Local Area Network (LAN) extension and Networked Storage (SANs).
Such customers might have their own Gigabit Ethernet switches or FICON directors as their
interface to the CWDM layer. These switches and directors are optimized for single-protocol
applications and provide more economic use of a wavelength with statistical multiplexing and
finer bandwidth granularity for the multitude of computers and peripherals connected to them.

Like CWDM ADM products, NG-SONET ADM products support multiple protocols and
rates. However, NG-SONET products are also like LAN switches and SAN directors, in that
they provide finer bandwidth granularity to many more customers. NG-SONET ADM
products are often referred to as Sub-Lambda multiplexers.

NG-SONET ADM products may segment the bandwidth of a CWDM wavelength using
either time division multiplexing (TDM) and switching (TDS) or packet multiplexing and
switching techniques.

The ITU-T G.7041/Y.1303 “Generic Framing Procedure” dated Dec’01 is an example of the
TDM/TDS option. This is most suited to transparent or channelized services with high
bandwidth utilization of the channel so provisioned. Furthermore, “true-ring” NG-SONET
networks with STS-1 (51.84 Mbit/s) space/time switching granularity at each ADM node
enable channel re-use and hence offer more economic use of the CWDM wavelength
capacity.

The IEEE 802.17 Resilient Packet Ring (RPR) standard – due for ratification in 1Q’03, is an
example of the packet multiplexing and switching option. This is most suited to packet
services for which the bandwidth utilization per customer is low. Statistical multiplexing of
packets enables many more customers to share the total bandwidth of a CWDM wavelength.
The “true-ring” benefits of the RPR standard enables full logical mesh connectivity,
bandwidth re-use and more economic use of the CWDM wavelength capacity.

The ITU and IEEE standards are complementary and for hybrid channelized and packet
services, the two Sub-Lambda multiplexing and switching options can co-exist on the same
CWDM wavelength. Either fixed or dynamic sharing of the total wavelength capacity is
possible using one or more STS-3c (155.52 Mbit/s) virtual tributaries per service with virtual
concatenation.

© Copyright 2002 Page 23

 CWDM Technology, Standards, Economics & Applications

4 Summary
4.1 CWDM Technology in the Metro space

Coarse WDM (CWDM) technology has been available since the early 1980’s. It has now
matured and is migrating from the LAN space to the Metro space. ITU standards have been
produced which secure the acceptance of CWDM in the eyes of the telecommunications
carriers. CWDM will consequently play a complementary role to Dense WDM (DWDM) and
Next Generation SONET (NG-SONET) in meeting the network capacity and operational
requirements of existing and new broadband services in the metro-area.

The key technical benefits of CWDM that make it an attractive alternative to DWDM and
NG-SONET for transporting multiple broadband services & protocols in the Metro space are:
a) Smaller component size (CWDM transmitters use approx. 12.5% of the volume of
DWDM);
b) Low power (CWDM transmitters consume approx. 5% of the power of DWDM
transmitters);
c) Multi-protocol support (CWDM provides this with much less overhead than NG-SONET);
d) Low end-end latency (in contrast to NG-SONET, CWDM latency is comparable to bare
fiber);
e) Less rack space (smaller components, less cooling and smaller power supplies);
f) Less air-conditioning (lower power dissipation of components and power supplies);
g) Longer backup battery life (lower power consumption of components & supplies).

Until now, the deficiencies of CWDM included:

• Lack of support for 10 Gbit/s transport services
• No optical amplification
• The limited number of CWDM wavelengths available
These deficiencies are rapidly evaporating, as summarized below under Future Directions.

4.2 CWDM Benefits in the Metro space

If we focus on components designed for Metro applications, CWDM transmitters are about
20-25% of the cost of equivalent DWDM transmitters and CWDM filters are about 50% of
the cost of equivalent DWDM filters. These are primary capex benefits. Secondary capex
benefits – which in some cases dominate – include savings in expensive co-location rack
space, savings in thermal management equipment, power supplies, backup batteries and air
conditioning plant. Opex benefits include reduced electrical power and air-conditioning costs.

The above are generic economic benefits of CWDM. There are also implementation-specific
economic benefits. As outlined in this paper, CWDM networks comprising Regenerative
Optical Add/Drop Multiplexers (R-OADMs) offer the operational benefits of NG-SONET
(such as simple network design rules) and the low latency, multi-protocol benefits of all-
optical CWDM networks (due to reduced data processing). The result is a “best-of-both-
worlds solution”. The operational and reduced data processing benefits equate to Opex and
Capex cost savings.

Page 24 © Copyright 2002

 CWDM Technology, Standards, Economics & Applications

4.3 CWDM Applications in the Metro space

Metro CWDM networks plug the gap between the high capacity, regional and long haul
DWDM networks and the emerging NG-SONET networks. In particular, regenerative
CWDM networks and NG-SONET networks fulfill complementary needs in the metro space.

Regenerative CWDM networks in particular, enable wavelength services to be provisioned

over a large metro area, with the functional and economic benefits of full logical mesh
connectivity, wavelength re-use and low end-end latency. These features are applicable to the
Inter-Office (CO-CO) and Fiber to the Building (FTTB) segments of the metro network.
Vertical application markets include support for basic services such as POTS, plus: new
broadband services such as high speed Internet & Video-on-Demand; transparent wavelength
services; Enterprise LANs & Storage Area Networks (SANs). The low latency benefits of
CWDM are especially attractive in ESCON and FICON / Fibre Channel based SAN
applications, where excessive end-end latency seriously reduces the data throughput
efficiency due to high-layer protocol issues.

The low space, power and cost benefits of CWDM also enable its deployment in the Outside
Plant (OSP) or Remote Terminal (RT) segments of the metro market. This is the final
bottleneck in the bandwidth chain from the service provider to the end-customer. CWDM
pair-gain technology enables new broadband services to be deployed from remote OSP
cabinets without forklift upgrades to the existing equipment that provides the basic (POTS)
services.

NG-SONET ADM networks and products complement CWDM networks. They provide
access to many customers, each requiring smaller bandwidth granularity than a whole
wavelength. As such, NG-SONET products are often referred to as Sub-lambda multiplexers.
The combination of regenerative, true-ring CWDM networks with NG-SONET ADM
multiplexers, provides a very cost-efficient network solution with full logical mesh
connectivity between many customers.

4.4 Future directions

The forthcoming emergence of 10 Gbit/s Long Wave VCSEL technology will enable 40
Gbit/s (OC-768) CWDM networks to be implemented on a single fiber strand, using 8
wavelengths in the S, C & L bands. The wider deployment of low water-peak fiber as defined
by ITU-T G.652.C will extend the useable fiber capacity to 16 CWDM wavelengths. The
addition of Semiconductor Optical Amplifiers (SOAs) will increase the number of OADM
nodes that can be deployed in an all-optical metro CWDM network. Integration and mass
production of a single connectorized package comprising thin film CWDM filters, Long
Wave VCSEL lasers, PIN/APD receivers, 2R and 3R regenerators will further reduce the
space, power and cost of CWDM technologies. This level of integration is less likely to occur
with DWDM technologies due to intrinsic space and power issues. Regenerative OADM
products employing these future technologies will enable 160 Gbit/s per fiber-pair metro
networks spanning over 160 km, at much lower cost than DWDM.

© Copyright 2002 Page 25

 CWDM Technology, Standards, Economics & Applications

5 References
1) M. Miller, N. Lewis, J. Ravita, R. Winfrey, M. Page, W. Lewis, “A Four Channel
Bi-Directional Data Link Using Wavelength Division Multiplexing”, Proceedings
9th International Fiber Optic Communications and Local Area Networks
Exposition, San Francisco USA, pp 107-111, 18-20 Sep’85
2) P.P. Iannone, K.C. Reichmann, N.J. Frigo, “Broadcast Digital Video Delivered
Over WDM Passive Optical Networks”, IEEE Photonics Technology Letters,
Vol.8, No.7, pp 930, Jul’96
3) G.G. Ortiz et al, “Monolithic, Multiple-Wavelength Vertical-Cavity Surface-
Emitting Laser Arrays by Surface-Controlled MOCVD Growth Rate Enhancement
and Reduction, IEEE Photonics Technology Letters, Vol.9, No.8, pp1069, Aug’97
4) K.Harada et al, “Hierarchical Optical Path Cross-Connect Systems for Large Scale
WDM Networks”, OFC’99, San Diego USA, pp 356-358, Feb’99
5) N. Chitica, J. Daleiden, M. Strassner, K. Streubal, “Monolithic InP-Based
Tuneable Filter with 10 nm Bandwidth for Optical Data Interconnects in the
1550 nm Band”, IEEE Photonics Technology Letters, Vol.11, No.5, pp584,
May’99
6) W. St.Arnaud, “1310 nm vs 1550 nm window for 10GbE”, IEEE 10G Study
Group, 6May’99,
http://grouper.ieee.org/groups/802/3/10G_study/email/msg00022.html
7) E. Warren, “When WDM Makes Sense in the Access Network: An Economic
Analysis”, JDS Uniphase White Paper, Aug’99
8) Chang-Joon Chae et al, “A PON System Suitable for Internetworking Optical
Network Units Using a Fiber Bragg Grating on the Feeder Fiber”, IEEE Photonics
Technology Letters, Vol.11, No.12, pp 1686-1688, Dec’99
9) Adel A.M. Saleh, Jane M. Simmons, “Architectural Principles of Optical Regional
and Metropolitan Access Networks”, Jnl of Lightwave Technology, Vol.17,
No.12, pp 2431-2448, Dec’99
10) James Campbell, “Coarse WDM makes waves in metro/access markets”, Laser
Focus World supplement, Nov’2000
11) B.R. Eichenbaum, S.K. Das, “Economics for Choosing a Coarse WDM
Wavelength Grid”, National Fiber Optic Engineers Conference, 2001, pp 1444-
1448, Jun’01
12) Optical Internetworking Forum (OIF) – Very Short Reach (VSR) Level 5 Project
for 40 Gbit/s optical interfaces, Feb’02
13) P. Kner, D. Sun, J. Boucart, P. Floyd, R. Nabiev, D. Davis, W. Yuen, M. Jansen,
and C. J. Chang-Hasnain, “VCSELs Go the Distance”, Optics & Photonics News,
pp 44-47, Mar’02
14) James Campbell, “Tsunami Optics – riding the CWDM wave”, Optical Keyhole,
( http://www.opticalkeyhole.com/keyhole/html/tsunami.asp ) 18Jan’02

746-000-001/1

Page 26 © Copyright 2002