PONS

Core optical networks in metropolitan settings are almost exclusively active today. Passive
optical networks (PONs) are mainly used as an access technology, though PON manufacturers
have been trying to promote them for backhaul for years. PONs carry much lower equipment
costs than traditional active optical networks, and they offer aggregate throughput speeds
in the low gigabits per second with the possibility of much higher speeds through a technique
known as wave division multiplexing. For a wireless broadband operator able to obtain
“dark fiber,” explained in detail shortly, a PON provisioned by the broadband operator
may constitute an attractive option for backhaul. As is, the number of firms offering PON
backhaul anywhere in the world probably does not exceed 20. Figure 4-2 shows a passive
optical network. Active Fiber Connections
Active optical networks are based largely on synchronous optical network (SONET) and synchronous
digital hierarchy (SDH), which means that available bandwidth is apportioned in
temporal channels taking the form of fixed time slots. Such networks are inefficient in their use
of network resources, and the services running over them are generally expensive, but reliability
is high. Packet services, which are just beginning to challenge circuit in the marketplace, are
generally less reliable but less expensive per a given throughput speed. Circuit Fiber Connections: SONET/SDH: Circuit time intervals are all multiples of the basic 56 kilobit
per second (Kbps) voice channel. The chief service offering within SONET networks is the T1 at
approximately 1.5Mbps, and the equivalent SDH offering is the E1 at approximately 2Mbps.
As indicated earlier, multiple T1/E1 services may be supported by some of the newer SONET
equipment platforms, but most carriers have not installed such equipment and will not provide
fine increments of bandwidth. In the United States the next level up from the T1 is the DS3
at 45Mbps, and the next increment above that is the OC-3 at 155Mbps. Someone desiring a
20Mbps or 100Mbps backhaul is simply out of luck.
Of course, one can always choose to lease more capacity than is absolutely needed, but the
pricing of DS3s and OC-3s is steep—thousands of dollars a month for the former and well into
the five figures for the latter. For a startup wireless broadband network attempting to build a
business, such large-capacity circuit backhaul connections are a heavy burden, a burden made
heavier by the fact that they are quite inefficient for the transport of packet traffic and carry a
great deal of network overhead.
Packet Fiber: Metro Ethernet or IP: Another option for backhaul is a direct packet connection using
either the Ethernet or IP protocol. Fast packet services over fiber have been offered by a number
of startups such as Cogent and Yipes in the United States and more recently by some of the
incumbents. Quite a number of carriers are now offering such services in East Asia as well, particularly
in mainland China, Hong Kong, and Taiwan. Certainly in the case of the startups, the
pricing has been far more attractive than is the case for circuit services with monthly fees as low
as $1,000 for a 100Mbps burstable connection, but in most cases such services are not delivered
with rigid service-level agreements in respect to jitter, latency, minimum throughput,
bandwidth reservation, and so forth. Direct IP services better support quality of service than do
pure Ethernet services, which is something to consider when evaluating fast packet backhaul
providers.
Where available, packet services over fiber are often preferred to circuit connections
strictly on the basis of price, but the problem is that they are not too generally available, and
in nearly all cases they are offered by entities that would prefer to have the customers whose
transmissions are being backhauled. Then, too, fiber-based access services are expensive to
provision, and the provider of metro Ethernet or direct IP-based services may simply not be
able to lay fiber to the building being served by the broadband wireless operator.
Dark Fiber: Another option, sometimes present and often not, is the lease of so-called dark fiber
from a public utility or public transit system. Gas and electrical utilities, and, to a lesser extent,
transit districts and railroads, own extensive amounts of optical fiber that they use for monitoring
purposes and internal communications. Such private fiber networks are generally grossly
underutilized—largely dark in the parlance of the telecommunications industry—and the
owners are often amenable to leasing capacity, in which case the lessee may be assigned an
individual wavelength or an IP address. Since utilities are not primarily in the business of selling
fiber capacity and do not regard such networks as major profit centers, they can often be
persuaded to lease capacity at reasonable rates. Be forewarned, however, that using leased
dark fiber for backhaul or for any other purpose is a different proposition than purchasing a
T1 or fast packet service from an optical services provider. Public utilities often do nothing
other than provide access to an optical pipe. They do not provide termination or optical
conversion equipment, and they offer no guarantees in respect to network redundancy or provisions for restoration. Consequently, by accepting such an arrangement, broadband wireless
operators are in effect becoming optical carriers as well as wireless access providers since
the burden of operating and maintaining the optical portion of the network will fall squarely on
their shoulders. Stranded Fiber Assets: Yet another possibility is securing what are known as stranded fiber assets.
During the late 1990s and the beginning of this decade, a tremendous amount of fiber was laid,
particularly in the United States, on both the local level and between cities. Such overcapacity
led to falling prices for leased fiber and a wave of bankruptcies among owners of fiber infrastructure.
Quite a bit of fiber laid by bankrupt companies is now available for resale from fiber
brokers such as Fiberloops, often at extremely low prices.
Such stranded assets can be an enormous bargain for the startup broadband access provider,
but their usefulness is subject to several qualifications. Usually the optical fiber that was
installed did not form a comprehensive or pervasive network but more commonly took the
form of a core ring that served to anchor what are known as laterals, fiber runs to individual
buildings. Laterals, the access portion of a fiber network, are by far the sparsest fiber deployments
and reach less than 10 percent of business locations in the United States. From any
given base station a fiber lateral may simply be unavailable.
One can, of course, situate one’s base stations with a view to exploiting extant fiber
resources, if one can determine their whereabouts. A number of consultancies such as the
aforementioned Fiberloops and TeleGeography offer databases of fiber resources, but no one
pretends that any such database is exhaustive. Because fiber is an inherently valuable resource
in and of itself, requires no licensing, and has proven itself to be long lasting and reliable, and
thus far has supported steady increases in throughput speeds through upgrades and improvements
in the terminal devices, acquiring fiber is generally a wise decision. But unless the fiber
extends the entire distance between the points to be linked, it is fairly useless for backhaul.
The network operator who can obtain only some portion of the fiber required to complete
a backhaul connection may choose to construct the remaining portion required. This is not an
operation to be undertaken lightly and in most cases will be prohibitively expensive. Trenching
costs for fiber builds can run as high as several hundred thousand dollars a mile in a large metropolitan
area and several tens of thousands in more rural areas. In some cases fiber may be
hung from utility poles or even snaked through sewer or gas lines, both installation methods
being roughly half the cost of trenching. Still, installation in the best of circumstances is seldom
cheap and, moreover, usually entails a long and involved permitting process. Fiber builds are
also time consuming as a rule.

Other Physical Media for Backhaul
While fiber is certainly the preferred medium for backhaul in respect to speed and availability,
it is far from the only option, and frequently it is far from the most cost effective.
Free-Space Optics for Backhaul
A further possibility is obtaining a fraction of the fiber needed on a lease or ownership basis
and utilizing free-space optics to provide fill-in. One manufacturer, LightPointe, makes a freespace
optical system where the transceiver can link fiber sections transparently with no need
for optical-electronic-optical (OEO) conversion, always an expensive proposition. Hybrid Fiber Coax
In a few locations, mostly in the United States, a new type of hybrid fiber-cable system operator
known as an overbuilder may offer packet services capable of speeds in the 10–30Mbps range,
sufficient for backhaul in some instances. Few such operators guarantee quality of service,
however, and here again the cable operator is likely to be competing for the same access customer
as the broadband wireless operator.
VDSL
VDSL, with speeds in excess of 50Mbps in some cases, might also be considered for backhaul
where available, but availability is limited as yet. Even more limited opportunities exist for utilizing
a power line carrier for backhaul, a technology where data is transmitted over AC power
lines at throughput speeds as high as 30Mbps. One New England–based company named
Amperion is pursuing a strategy of combining powerline carrier backhaul with wireless broadband
access, but to date it has achieved few commercial deployments. Bear in mind that any
such arrangement requires the active participation of the local electrical utility, and while such
entities have long evinced interest in gaining a stake in the communications business, few have
made major commitments to doing so.
Wireless Bridge Connections
Wireless broadband operators, can, if they so choose, pursue a wireless pure play in regard
to backhaul and utilize wireless point-to-point “bridge” connections from the base station to
a central office. They can do so using their own spectrum, or they may elect to use another
wireless service provider, preferably one utilizing millimeter microwave equipment or freespace
optics.
Such bridge connections can utilize either low microwave or millimeter wave frequencies
and will normally employ the full available spectrum in the one airlink. Very high-gain,
narrow-beam antennas are the rule here, and maximum transmission distances are multiples
of the radii of the cells being backhauled. Wireless bridge backhaul connections exceeding 30
miles are feasible in some instances.
The same spectrum can be used for backhaul as is used for access, but the prevailing practice
is to use a dedicated band to avoid interference since the backhaul has the potential to
interfere with every subscriber in the cell inasmuch as it reuses all of the spectrum. The 5.8
U-NII band is frequently assigned to backhaul and offers a good combination of bandwidth
and range—100 megahertz (MHz) and more than 20 miles, best case. Other frequencies
favored for backhaul purposes include the 18GHz and 23GHz licensed bands where licenses
are fairly easily obtainable in the United States, the 24GHz unlicensed band, the 28GHz–31GHz
Local Multipoint Distribution Service (LMDS) bands, the 38GHz and 39GHz bands, and the
unlicensed band at 60GHz. In all of these bands at least 100MHz of spectrum is available and
interference is minimal. Range is an issue, however, especially as one goes higher in frequency,
and in the highest band, centered at 60GHz, distance should not exceed a kilometer.
The cost of radios for the millimeter wave regions is significantly higher than for the lower
microwave bands, and bridge links are minimally several thousand dollars apiece. In the case
of LMDS, 38GHz, and 39GHz, the cost is apt to be much higher, approaching $100,000, though
that must be balanced against the generous allocations of bandwidth and resultant high
throughputs and the fact that the bridge equipment represents a one-time capital cost along
with the generally fairly minimal recurrent costs associated with roof rights.