Embed Size (px)
Citation preview
DESIGN METHODOLOGY FOR BACKHAULAND DISTRIBUTION NETWORKS
USING TV WHITE SPACES
BY CYRUS GERAMI
A thesis submitted to the
Graduate School—New Brunswick
Rutgers, The State University of New Jersey
in partial fulfillment of the requirements
for the degree of
Master of Science
Graduate Program in Electrical Engineering
Written under the direction of
Narayan Mandayam and Larry Greenstein
and approved by
New Brunswick, New Jersey
May, 2011
c© 2011
Cyrus Gerami
ALL RIGHTS RESERVED
ABSTRACT OF THE THESIS
Design Methodology for Backhaul
and Distribution Networks
Using TV White Spaces
by Cyrus Gerami
Thesis Directors: Narayan Mandayam and Larry Greenstein
Since the FCCs approval of unlicensed use of TV white spaces, the issue of how to
use these white spaces has led to innovative technologies such as cognitive radios as well
as a variety of spectrum policy proposals. There have been proposals to devise alter-
nate rules for spectrum usage citing the overly conservative restrictions on secondary
transmissions to protect incumbents. In this thesis, instead, we propose to utilize white
spaces for a backhaul network for internet traffic based on existing restrictions. Using
the available white spaces and backhaul traffic demands in New Jersey as a case study,
we evaluate the feasibility of such backhauling and present a methodology that can be
used for other areas as well. Using a basic design involving fixed towers and directional
antennas, our results show that the TV white spaces can be an effective medium for
radio backhaul as an alternative to the costly laying of optical fiber. Although the most
recent FCC ruling does not mandate protection of wireless microphones, we show that
meeting the more stringent earlier FCC requirements on sensing and avoiding harm to
wireless microphones would have only a minor impact on capacity. Finally, we study
the aggregating of multiple data traffic flows at the nodes and show that, with proper
engineering, multiple flows have but a slight effect on the need for optical fiber.
ii
Acknowledgements
First and foremost I would like to express my deepest and sincerest gratitude to my
advisors, Prof. Narayan Mandayam and Prof. Larry Greenstein, who supported me
throughout the course of my research with guidance and patience. Their broad knowl-
edge, constructive comments, and logical way of thinking have been of great value to
me. I have learned lots of lessons from them in academics and life, and without their
guidance and persistent help this thesis would not have been possible. I feel fortunate
to have had such supervision to help me grow as an engineer, a thinker, and as a person.
For that I am eternally grateful.
I also wish to express thanks to Prof. Dipankar Raychaudhuri, Prof. Roy Yates,
Richard Frenkiel and Ivan Seskar for their useful inputs and comments. It wouldn’t be
fair to not also mention an extremely important motivating factor for me: WINLAB.
Such a friendly atmosphere and professional research environment is what a student
would hope for and I am honored to have been a “WINLAB Student.”
iii
Dedication
I dedicate this thesis to my family who have supported me throughout my entire life
and are always there for me...
iv
Table of Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Dedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2. TV White Spaces: Regulatory Background . . . . . . . . . . . . . . . . 4
2.1. DTV Transition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2. Some Desirable characteristics of White Spaces . . . . . . . . . . . . . . 5
2.3. Issues Surrounding White Spaces Utilization . . . . . . . . . . . . . . . . 6
2.4. FCC’s Second Report and Order . . . . . . . . . . . . . . . . . . . . . . 6
2.5. FCC ruling on Sep 23, 2010 . . . . . . . . . . . . . . . . . . . . . . . . . 8
3. How Much White Space Is Available? . . . . . . . . . . . . . . . . . . . 10
3.1. Choice of Testbed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2. Database of available channels . . . . . . . . . . . . . . . . . . . . . . . 11
3.2.1. General Approach . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2.2. Portable Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2.3. Fixed Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4. Using White Spaces for Distribution and Backhaul . . . . . . . . . . . 19
4.1. System Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.2. Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
v
4.3. Wireless Microphones (WM) . . . . . . . . . . . . . . . . . . . . . . . . 22
4.4. Achievable Capacity Calculations . . . . . . . . . . . . . . . . . . . . . . 23
4.4.1. Path loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.4.2. SNR and Spectral efficiency . . . . . . . . . . . . . . . . . . . . . 25
4.4.3. Frequency planning . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.5. Demand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.5.1. Usage per household . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.5.2. Active users . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.6. Feasibility of a Backhaul Network in New Jersey . . . . . . . . . . . . . 30
4.7. Aggregating Multiple Flows . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.8. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5. Conclusions and Future Directions . . . . . . . . . . . . . . . . . . . . . 35
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
vi
List of Tables
2.1. Allowable Channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1. Required Separation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.1. Parameters used in calculations . . . . . . . . . . . . . . . . . . . . . . . . 25
4.2. Estimated Population Grades . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.3. Maximum Deliverable Rate . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.4. Usage Per Household . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.5. Demand for Different Hourly Traffic and 30% Active Users (MB/hr) . . . . . 30
4.6. Number of cells that need fiber out of a total 307 cells . . . . . . . . . . . . . 31
vii
List of Figures
3.1. Protected radius, additional separation radius and noise-limited radius of a
primary TV tower and allowed secondary antenna with allowed transmission
coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2. Different grid resolutions . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3. An example of protected radii for co-channel and adjacent-channel trans-
mission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.4. An example of protected areas for transmission using portable devices.
White area is potentially available for transmission. . . . . . . . . . . . . 14
3.5. Map of availability of channel 25 in New Jersey for portable transmission. Dark
green indicates available with max power and light green indicates available
with lower power threshold . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.6. An example of protected areas for transmission using fixed devices. White
area is potentially available for transmission. . . . . . . . . . . . . . . . 16
3.7. Map of availability of channel 25 in New Jersey for fixed transmission
along with grid. Dark green indicates available with max power and light
green indicates available with lower power threshold. . . . . . . . . . . . 17
3.8. Number of available 6-MHz channels within the 5-mile × 5-mile square cells,
where the darkness of the shadings represents the number of channels. For New
Jersey, the number ranges from 7 (lightest shading) to 31 (darkest shading). . 17
viii
3.9. Snap shot of the table for channel availability in each cell where cells are col-
ored depending on availability. Red cells indicate not available, green indicates
available and yellow means partially available. The two right-most columns il-
lustrate the number of available channels and corresponding bandwidth (MHz)
in a particular cell. Note that the last column also illustrates the relative avail-
able bandwidth as a bar chart. . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.1. Illustration of using White Spaces for Distribution and Backhaul . . . . 20
4.2. A plan for using four sector antennas instead of one isotropic antenna at each
base-station . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.3. Received SNR vs. obstruction height on a 5-mile path, for different frequencies.
This plot is based on the ITU path loss model, with the obstruction assumed
to be at the center of the path. . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.4. Population distribution in New Jersey, based on the US Census of 2000. Darker
shadings correspond to higher densities, with the darkest areas concentrated in
the New York City metropolitan area. . . . . . . . . . . . . . . . . . . . . . 27
4.5. (top) five neighboring cells with M common available channels illustrating usage
of different sets of channels with different colors for a reuse factor of 2. (bottom)
a plot of the received SNR from cell 1 vs. distance, illustrating the fading of the
signal power as it reaches cell 3. Each ’X’ on the plot represents the position of
neighboring receivers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.6. Geographical distribution of cells needing fiber, assuming 10% active users. The
usage per house, from left to right, is 18, 90, 126 and 180 MB per hour . . . . 31
4.7. Geographical distribution of cells needing fiber, assuming 30% active users. The
usage per house, from left to right, is 18, 90, 126 and 180 MB per hour . . . . 32
4.8. Geographical distribution of cells needing fiber, assuming 50% active users. The
usage per house, from left to right, is 18, 90, 126 and 180 MB per hour . . . . 33
4.9. An example clustering technique to mitigate aggregation with an initial 8 extra
fiber points. For this example, the average hourly load per household is 126MB,
for 30% of all households. . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
ix
1
Chapter 1
Introduction
The FCC’s opening up of TV White Spaces for unlicensed use has led to innovations in
cognitive radio technology, spectrum sensing as well as novel proposals for dynamic spec-
trum access. “A Survey on Spectrum Management in Cognitive Radio Networks [1],”
talks about how Cognitive Radio networks are being developed to solve current wireless
network problems resulting from the limited available spectrum and the inefficiency in
spectrum usage, by exploiting the existing wireless spectrum opportunistically. This
paper also states that Cognitive Radio networks, equipped with the intrinsic capabilities
of cognitive radio, will provide an ultimate spectrum-aware communication paradigm
in wireless communications. Steve Shellhammer the author of “Spectrum Sensing in
IEEE 802.22 [2]” gives a description of the method used for evaluating specific spec-
trum sensing techniques included in the IEEE 802.22 standard. This evaluation is based
upon various “Blind” as well as “Signal-Specific” sensing techniques. The authors of
“Dynamic Spectrum Access Models: Toward an Engineering Perspective in the Spec-
trum Debate [3]” have taken an engineering perspective toward developing dynamic
spectrum access. The proposed models, promote dynamic access and short-term dedi-
cation of spectrum resources while retaining a bias toward the spectrum property rights
approach. With illustrative results, the authors have proposed the use of a spectrum
policy server to mediate market-based spectrum allocation. The authors of this article
also claim that the spectrum access mechanism and market forces play important roles
in the resulting bandwidth utilization, and that this article presents a foundation for
more realistic engineering models that can shape spectrum policy.
In its Second Report and Order [4], the FCC set forth rules for TVBDs (TV Band
2
Devices) by allowing transmission only in certain channels. To further protect the li-
censed TV towers, FCC has also set forth rules for secondary antennas and transmission
power, which in turn restrict the transmission coverage for a secondary-to-secondary
system. These restrictive requirements have raised several objections and have also
resulted in proposals for alternative rules that can still provide sufficient protection to
the incumbents (e.g. [5]). To date, in spite of these objections, nothing has changed;
the rules remain the same.
In this thesis, we address the following question: How can one use the TV White
Spaces within the existing rules? While proposals have been set forth for use of White
Spaces for local area networks [6], one potential use that might be of interest is a back-
haul network for rural areas and areas with no pre-existing wired infrastructure. We
propose a system using fixed towers and directional antennas that would provide back-
haul and/or distribution for Internet access in places where it is needed and is presently
not available. Using New Jersey as a “testbed”, we calculate the available White Space
and backhaul traffic demands, and evaluate the feasibility of such backhauling and
present a general methodology that can be used for other areas as well.
While optical transport networks are evolving rapidly ([7] and [8] talk about 100
Gigabit Ethernet deployment), having ubiquitous deployment of these, including in
rural areas, is far from being a reality. The backhaul network concept we propose is an
attractive alternative to laying optical fiber, which is costly, while radio towers using
TV White Spaces is far less so. New Jersey is an appropriate testbed for evaluating
this concept. It is the most densely populated state in the US; has large pockets of
urban, suburban and rural areas; and is adjacent to two large metropolitan areas–New
York and Philadelphia.
This thesis is organized as follows: In Chapter 2, we will lay the groundwork for our
studies with some general information about how white spaces came about and some
issues surrounding it. In Chapter 3, the available TV white space spectrum in the UHF
channels for fixed and portable TVBDs is mapped to the geography of New Jersey. In
chapter 4, we present an alternative proposal: to use the white spaces for an Internet
backhaul system using base stations with directional antennas. Also, we describe our
3
methodology and demonstrate its application to an Internet backbone network in New
Jesey. In Chapter 5, we conclude with our main findings.
4
Chapter 2
TV White Spaces: Regulatory Background
2.1 DTV Transition
In 1996, the U.S. Congress authorized the distribution of an additional broadcast chan-
nel to each broadcast TV station so that they could start a digital broadcast channel
while simultaneously continuing their analog broadcast channel [9]. The justification
for this transition was to enable more spectrally efficient broadcasting, thereby allowing
for new services to be offered to the consumer. These included: multicasting; HDTV;
data streaming; better picture resolution, clarity and color; Dolby Surround Sound;
and so on.
Subsequently, Congress set June 12, 2009 as the deadline for full power television
stations to stop broadcasting analog signals. Since June 13, 2009, all full-power U.S.
television stations have broadcast over-the-air signals in digital format only. The switch
from analog to digital broadcast television is referred to as the Digital TV (DTV)
Transition.
Consumers benefited from the DTV Transition because digital broadcasting allows
stations to offer improved picture and sound quality, and digital is much more spectrally
efficient than analog. With digital broadcasting, rather than being limited to providing
one analog program, a broadcaster is able to offer a super sharp High Definition (HD)
digital program or multiple Standard Definition (SD) digital programs simultaneously
through a process called multicasting 1. This means more programming choices for
viewers. Further, DTV provides interactive video and data services that were not
1Multicasting allows broadcast stations to offer several channels of digital programming at the sametime, using the same amount of spectrum required for one analog program. A typical expansion ratiois 4:1.
5
Table 2.1: Allowable Channels
TV Channels Frequency Band Frequency (MHz) Allowed Devices
2 VHF 54-60 Fixed5-6 VHF 76-88 Fixed7-13 VHF 174-216 Fixed
14-20 UHF 470-512 Fixed21-35 UHF 512-602 Fixed and Portable36* UHF 602-608 Portable38* UHF 614-620 Portable
39-51 UHF 620-698 Fixed and Portable
* These channels are set aside for possible Wireless Microphone use (see sec. 2.5).
possible with analog technology.
Another important benefit of the DTV transition is that it freed up parts of the
valuable broadcast spectrum for public safety communications (such as police, fire de-
partments, and rescue squads). A huge chunk of this freed up valuable spectrum has
now been auctioned in part to companies that will be able to provide consumers with
more advanced wireless services (such as wireless broadband). The rest of the new
spectrum was allocated to unlicensed use. This unlicensed spectrum is what is referred
to as White Spaces.
Specifically, White Spaces refers to the unused television spectrum that traditionally
existed between channels as buffers or empty spectrum left over or vacated by TV sta-
tions through the DTV transition. This TV band spectrum can be used for Secondary
services in coexistence with Primary licensed services, Only certain TV channels have
been freed up as White Spaces. These channels are in the high VHF and UHF TV
bands and their frequencies range from 54 to 698 MHz. Table 2.1 shows the channels
allowed for unlicensed use by the FCC, subject to certain restrictions and regulations.
2.2 Some Desirable characteristics of White Spaces
Propagation: Because of the lower frequencies of TV channels (i.e., lower than those
for the cellular and PCS bands at 900 MHz and 1.9 GHz, respectively, and the Wi-Fi
bands at 2.4 and 5.0 GHz), signals in White Spaces can propagate to larger distances,
permitting larger coverage regions.
6
Penetration: Also because of the lower frequencies, White Space signals can better
penetrate walls of rooms and buildings, further extending coverage.
Bandwidth: White Spaces also offer the promise of faster speeds, up to 100 megabits
per second. That can be used for end users or to connect local Wi-Fi hotspots. We will
discuss the details of exact bandwidth availability in Chapter 3.
2.3 Issues Surrounding White Spaces Utilization
Technical: Since White Spaces would remain unlicensed, the use of it could interfere
with local broadcasters. On top of that, interference among unlicensed users (secondary
co-existence) would be an issue. The use of wireless of microphones could also be
compromised by interference from these White Space devices, as discussed in Chapter
4.
Political: Technology heavyweights like Google, Dell, Microsoft and others have
rallied on behalf of White Spaces. The National Association of Broadcasters has filed
a lawsuit to halt the use of White Spaces because of fears of interference Theatrical
groups and sports franchises have also raised concerns about interference for wireless
microphones Consumer groups such as Free Press, Public Knowledge, Consumers Union
and others have pushed for the use of White Spaces.
Applications: The added range and performance of VHF and UHF could help con-
nect rural communities, allow schools to light up entire campuses, help service providers
relieve burdened cellular networks and could help with things like in-home video stream-
ing and smart meter monitoring. Systems providing Super Wi-Fi or even mobile com-
munication are also possibilities.
2.4 FCC’s Second Report and Order
In 2008, the FCC released its Second Report and Order [4] on the TV unlicensed Spec-
trum or the TV White Spaces. According to the regulations set forth in this document,
TVBDs are required to be equipped with both spectrum sensing and geo-detecting
capabilities in sync with database systems to protect the TV signals from interference.
7
Under the rules, TV bands devices have to check on analog broadcasting activity
in their area every 60 seconds. But the Order also required an automatic sensing test
for wireless microphones and other “low power auxiliary signals” streaming at a certain
threshold every 30 seconds. If a device sensed one, it would have to vacate that channel
within two seconds flat.
The FCC has defined two classes of devices: Fixed devices (height of above 10
meters); and personal/portable devices (height of 3 -10 meters). It has also defined two
modes under portable devices: Mode I under control of a device that employs geo-
location/database access; and Mode II employs geo-location/database access itself.
Fixed use: The rules allow fixed devices to operate on any channel between 2 to
51, except for channels 3, 4 and 37. In addition, fixed devices will be excluded from
operating on channels 14 through 20 in certain geographic areas where private and
commercial wireless facilities are licensed to operate. The FCC prohibits operation of
fixed devices on channels 3 and 4 in order to prevent direct pick-up interference when
poorly shielded TVs are connected to VCRs, DVRs and cable set-top boxes that output
on these channels. Fixed devices may operate at power levels of up to 4 watts EIRP,
equivalent to 30 dBm (6dBi antenna gain). Also, fixed devices cannot operate in busy
channels (co-channel transmission) or channels which have busy neighboring channels
(adjacent-channel transmission).
Portable use: Portable wireless devices will be allowed to operate on an unlicensed
basis in the higher portion of this band only, from channels 21 to 51, with the exception
of channel 37. Channels 2 to 21 are excluded to protect licensed wireless microphone
users. Due to their greater risk of interference to licensed services, the power level of
portable devices will be limited to up to 100 milliwatts EIRP (20 dBm), although the
maximum power limit will be lowered if the antenna gain of the portable device exceeds
0 dBi. Also portable devices cannot operate in busy channels (co-channel transmission)
and should use less power (40 mWatts, or 16 dBm) to transmit in channels which have
busy neighboring channels (adjacent-channel transmission).
Some more rules are as follows:
- Devices require receive sensitivities of -114 dBm.
8
- Devices must use adaptive power control.
- Channels 36 and 38 must have strict out-of-band emission control and fixed devices
are not allowed to transmit in these channels.
- The sensing antennas must be mounted outside and must be at least 10 meters
above ground. The transmit antennas must be outdoors and be mounted no more than
30 meters above ground.
- Before utilizing any channel, TVBDs are required to access a TV band database
operated by a third party in order to determine permissible operating channels. This
must be followed by spectrum sensing to confirm the emptiness of these channels.
- TV database should contain the following information: Transmitter coordinates;
effective radiated power (ERP); height above average terrain of the transmitter (HAAT);
horizontal transmit antenna pattern (if the antenna is directional); channel number; and
station call sign.
Device makers, Google, and various public interest groups strongly objected to the
spectrum-sensing requirement. Dell, Microsoft, Adaptrum and other companies argued
that wireless mics would be protected by the TV bands database and the presence of an
array of “safe harbor” channels where White Space devices won’t be allowed to stream.
Other groups argued that trying to sense these signals at various low levels would pump
up time and development costs.
Another point that Google and the Public Interest Safety Coalition brought up was
that spectrum sensing would in many instances protect many wireless microphones
systems that go to unlicensed.
2.5 FCC ruling on Sep 23, 2010
As part of the new rules adopted on Sep 23, 2010, the FCC agreed to set aside two
channels for wireless microphone use to mitigate potential interference issues. Based
on this rule, the 1st vacant channel on either side of 37 (i.e., 36 or 38) would be
the first candidates for wireless microphones. If neither 36 nor 38 are vacant in a
region, channels 35 or 39 would be considered and also, it goes on to say that if no
9
channel on one side of 37 is free then it will be the 1st two vacant channels on the
other side. But the commission said it would not require device makers to include
geolocation spectrum sensing technology in new devices to ensure that these products
don’t interfere with existing services already using the spectrum. This is a relaxation
of a potentially difficult and expensive problem for any White Space system. Instead,
devices will query a special geolocation database that makes sure no one is using that
spectrum before it transmits. This database check is largely to prevent White Space
services from interfering with broadcast TV signals (Primary users).
The FCC’s Office of Engineering and Technology will select companies that will
manage the database going forward. Companies such as Google and Spectrum Bridge
have submitted proposals for this role.
10
Chapter 3
How Much White Space Is Available?
3.1 Choice of Testbed
In addition to limiting secondary transmission to a list of allowed channels, the FCC has
set a protected radius around each licensed TV tower in which no secondary transmis-
sion is allowed. Also, to compensate for the interference introduced by the secondary
radio towers, FCC has set an extension to the protected radius for co-channel trans-
mission as well as adjacent-channel transmission. Table 3.1 shows this extended radius
for different device types and different use cases.
To further illustrate this issue, figure 3.1 shows the protected radius (rp), the
additional separation radius (ras)and the noise limited radius (rnl)around a TV tower
in comparison with a secondary antenna and its presumed coverage.
Figure 3.1: Protected radius, additional separation radius and noise-limited radius of a primaryTV tower and allowed secondary antenna with allowed transmission coverage
On top of the above restrictions, for fixed devices (base-stations or towers), co-
channel or even adjacent-channel transmission is not allowed and for portable devices,
co-channel transmission is not allowed and adjacent-channel transmission is allowed
11
Table 3.1: Required Separation
Required Separation (km) From Digital or Analog TV(Full Service or Low Power) Protected Contour
Unlicensed Device AntennaHeight
Co-channel Adjacent-channel
Less than 3 meters 6.0 km 0.1 km3 - less than 10 meters 8.0 km 0.1 km10 -30 meters 14.4 km 0.74 km
with lower transmission power.
A major step towards determining whether White Spaces can be used for a system
is to map the allowed White Space channels to the geography of the area under study,
in order to calculate the available bandwidth per square cell. For this purpose we chose
New Jersey as our test-bed.
To quantify White Space availability in New Jersey, we first divided the state into
a grid of 10-mile × 10-mile square cells, as shown in Figure 3.2a. At later stages of
our study, each square cell was divided into four 5-mile × 5-mile square cells for better
channel availability resolution and to better support our fixed antenna proposal, details
of which we will show in chapter 4. This gave us about 300 cells instead of the 80 cells
we previously had. Figure 3.2b shows the new grid.
3.2 Database of available channels
3.2.1 General Approach
In order to build up a database of available channels per square cell in New Jersey, we
used Google maps [10] and the FCC database of TV tower locations [11] along with
their coverage radii and additional extensions. As an example, Figure 3.3a and 3.3b
show the TV tower coverage for channel 33 in the state of New Jersey considering co-
channel and adjacent-channel transmission. This example tower is located in the state
of New York with coverage spanning over New Jersey. We consider not only towers
in New Jersey, but also towers in New York, Pennsylvania and Delaware that have
coverage extending into New Jersey. We include these three states because a given TV
12
(a) grid of 10mi x 10misquare cells
(b) grid of 5mi x 5mi squarecells
Figure 3.2: Different grid resolutions
towers coverage can span up to 100 km, which can limit use of the same channel in a
neighboring state.
We use our grid of 5-mile × 5-mile square cells and implement all FCC restrictions
and TV tower coverage areas on to the map of New Jersey. For each channel we are
able to observe cells with availability. Repeating this process for each allowable channel
gives us the total channels available in each area. Keep in mind that in this practice we
have only considered UHF channels (i.e. channels available from channel 14 to 51, as
shown in Table 2.1). Since there are different sets of FCC rules for fixed and portable
devices, we would need to build two separate channel databases.
3.2.2 Portable Devices
For portable TVBDs, co-channel transmission is prohibited but adjacent-channel trans-
mission is allowed with less transmission power (40 mW instead of the maximum of 100
mW). To understand channel availability in each cell of our grid, we first need to map
the channel coverage and appropriate extra separation to our grid on top of New Jersey.
Potentially available area for each channel is determined by subtracting the TV tower
prohibited zone from the total area of New Jersey.
13
(a) Channel 33 TV tower protectedarea for co-channel transmission
(b) Channel 33 TV tower protected areafor adjacent-channel transmission
Figure 3.3: An example of protected radii for co-channel and adjacent-channel trans-mission
Now based on the FCC rules, in order to actually be able to transmit in any chan-
nel we need to make sure adjacent-channel conditions are also satisfied. So if towers
transmitting in adjacent channels have coverage spanning over the potentially available
areas, transmission power must be lowered down. To be able to transmit in full power,
not only the very channel of transmission but also both adjacent channels should be
fully available. As an example let us assume we want to transmit in channel 25. Figure
3.4b shows coverage of channel 25 with the extra radius for co-channel transmission.
The dark and light blue show the coverage area and the extra radius. Now to actu-
ally transmit in this channel we also consider channels 24 and 26. Figures 3.4a and
3.4c show the coverage area with the extra radius for adjacent-channel transmission
for these adjacent channels. Now Areas that are available (white) in Figure 3.4b but
not available (blue) in any of figures 3.4a or 3.4c, are allowed for transmission with
lower power. Areas that are available in all three figures (i.e. channels) are available
for full power transmission. To comprehend the geography of channel 25 availability,
14
(a) Potential area ofchannel 24 for portableuse
(b) Potential area ofchannel 25 for portableuse
(c) Potential area ofchannel 26 for portableuse
Figure 3.4: An example of protected areas for transmission using portable devices.White area is potentially available for transmission.
we superimpose all three figures. Figure 3.5 shows the superimposed map, in which
the color green means available for full power transmission, light green means available
for low power transmission.
We can superimpose our grid on top of this map and repeat this modeling for each
allowable channel and have a database of how many channels are available in each
square cell. Note that based on FCC allowed channels, only channels 21 - 51 (512 - 698
MHz) are available for portable devices.
3.2.3 Fixed Devices
For fixed TVBDs, co-channel transmission is prohibited as well as adjacent-channel
transmission. To understand how many channels are available in each cell, we again
need to map the channel coverage and additional separation to New Jersey and super-
impose our grid. The potentially available area for each channel is as before, determined
by subtracting the prohibited area from the area of New Jersey. Now for fixed devices,
to be able to transmit in full power, not only the very channel of transmission but also
both adjacent channels should be completely available. Coming back to the example
of transmission in channel 25, transmission is allowed only if not only channel 25, but
15
Figure 3.5: Map of availability of channel 25 in New Jersey for portable transmission. Darkgreen indicates available with max power and light green indicates available with lower powerthreshold
also channels 24 and 26 are available. Figures 3.6a, 3.6b and 3.6c respectively show
potentially available areas for fixed transmission for channels 24,25 and 26. Note that
the dark and light blue colors show the coverage area and the extra radius. Also note
that since channel 25 is intended for transmission, the extra separation ring for that
channel is larger than that of 24 and 26.
Now the actual allowed area for transmit in channel 25 is the area that is white in
all three figures (i.e. available in all three channels). In other words, superimposing all
three figures, whatever left that is not blue, is available for transmission. Figure 3.7a
shows the availability of channel 25 for fixed devices, in which the color green means
available for full power transmission and the color blue means not allowed.
We superimpose our grid on top of this map (figure 3.7b) and repeat this modeling
for each allowable channel, and this gives us a database of how many channels are
available in each square cell. Note again that only UHF channels where used in this
exercise. Figure 3.8 illustrates a map of available channels for fixed TVBDs, where the
degree of shading indicates the number of channels available. In this case, the minimum
number of available channels is 7 and the maximum is 31. This would mean that at every
point in New Jersey as least 42 MHz of bandwidth (7 channels of 6-MHz bandwidth)
16
(a) Area of channel 24for fixed use
(b) Area of channel 25for fixed use
(c) Area of channel 26for fixed use
Figure 3.6: An example of protected areas for transmission using fixed devices. Whitearea is potentially available for transmission.
is available for unlicensed use. Also note that for this exercise, only full-power TV
stations have been considered and inclusion of low-power TV, PLRMS/CMRS and
other licensed services would slightly reduce the channel availability in some localities.
Nevertheless, the number of available channels from our calculations is comparable to
those obtained using online databases such as [12] and [13].
To be able to further analyze the White Space availability, we have quantified the
channel availability for each cell. Table 3.9 shows an example of this quantification
for fixed transmission. The vertical axis of this table is cell index and the horizontal
axis is channel number. In other words, each unit in this table shows the availability
of a certain channel in a certain cell. The color green in each unit illustrates that the
vertical reference channel is available in the horizontal reference cell. Accordingly, red
cells show unavailability and yellow cells show partial availability. The last columns of
this table show the number of channels available as well as the total bandwidth available
in each cell. Note that the percentage of channel availability in each cell with partial
availability, has been calculated and to be on the conservative side, we have considered
cells with above 90% availability as available and all the rest as not available.
17
(a) availability of channel 25for fixed use
(b) superimposition of gridon availability map
Figure 3.7: Map of availability of channel 25 in New Jersey for fixed transmission alongwith grid. Dark green indicates available with max power and light green indicatesavailable with lower power threshold.
Figure 3.8: Number of available 6-MHz channels within the 5-mile × 5-mile square cells, wherethe darkness of the shadings represents the number of channels. For New Jersey, the numberranges from 7 (lightest shading) to 31 (darkest shading).
18
Figure 3.9: Snap shot of the table for channel availability in each cell where cells are coloreddepending on availability. Red cells indicate not available, green indicates available and yellowmeans partially available. The two right-most columns illustrate the number of available chan-nels and corresponding bandwidth (MHz) in a particular cell. Note that the last column alsoillustrates the relative available bandwidth as a bar chart.
19
Chapter 4
Using White Spaces for Distribution and Backhaul
In this day and age, connectivity to high-speed Internet is becoming not just a com-
modity but a necessity. Yet a major part of the country is still using dial-up Internet
connections. Many people in rural areas and small cities currently cannot not have
access to broadband Internet; on the other hand, many others in large metropolitan
cities can have super high speed Internet service provided via Fiber Optical cables to
their doorsteps! Helping to provide high-speed Internet access in areas with no existing
network or infrastructure, leads to another attractive use case for White Spaces.
In this section we propose a system using fixed towers and directional antennas
which would provide backhaul and/or distribution for Internet access in places where
it is needed and is presently not available. Using New Jersey as a testbed, we calculate
the available White Space and backhaul traffic demands and evaluate the feasibility of
such backhauling. In the process, we present a general methodology that can be used
for other areas as well.
4.1 System Description
This proposed system can be used to route an area’s aggregate Internet data to another
area through a network of antennas with extremely high speed. This would not mean
using such a system for client-to-server connections, but rather using it as an extension
for high-speed Internet connections to reach areas where it does not currently exist.
According to our proposal, we would have a network of base stations and connect
them through White Spaces (figure 4.1). Keep in mind that these base stations will
have a maximum height of 30 meters (which is the max allowed by FCC) and their
transmit power will be limited to 4 W (again, FCC limitations). And also, we will
20
implement FCC rules for fixed TVBDs. Note that since the purpose of this system is to
aid existing networks, there will not be base stations in all cells, but only in cells that
do not already have high speed Internet available and qualify based on our criteria.
Interestingly, we shall see that the system will end up having base stations in rural
areas that are the intended beneficiaries for this system.
Figure 4.1: Illustration of using White Spaces for Distribution and Backhaul
Base stations in each cell will be used to wirelessly connect local traffic to wired
(e.g., fiber-connected) stations using our network. From there, Internet data can be
routed to their final destinations. In other words, if a local client connects to the grid
by any means (e.g. wifi or wire), the traffic will be routed through the radio network to
the nearest fiber point which is best located to route the traffic to its final destination.
To quantify White Space availability for such a system in New Jersey, we used
our grid of 5 mi x 5 mi square cells. Now assuming there is a radio tower in (or
near) the middle of each of these cells, the distance to transmit data from one cell to
another would be 5 miles. In order to increase efficiency and achieve more concentrated
21
transmission, we assume that each radio tower has 4 sector antennas covering all four
directions instead of one isotropic antenna, as shown in Fig 4.2. This allows for more
concentrated line-of-sight (LOS) transmission and less interference.
Figure 4.2: A plan for using four sector antennas instead of one isotropic antenna at eachbase-station
4.2 Methodology
To determine whether TV White Spaces can be used for backhaul, we need to com-
pare the achievable capacity for each link with the demand for service in each square
cell. Achievable capacity will be derived using FCC power limits and widely accepted
propagation models. Traffic demand per cell will be derived using a Cisco data traffic
survey [14].
If the deliverable capacity is less than the demand in any area, connectivity cannot
be provided by utilizing the White Spaces and fiber would be needed instead. Inter-
estingly, the areas in which fiber would be needed are found here to be those already
provided with wired service. We also find, as we will show, that in the vast majority
of cells in New Jersey, the deliverable capacity is greater than the likely demand. This
will be important in showing the feasibility of such a system. Later in the study, we
use the difference in demand and deliverable capacity to justify the routing abilities of
our system.
22
4.3 Wireless Microphones (WM)
Although the latest set of rules devised by the FCC does not require TVBDs to protect
Wireless Microphones, it is interesting to show that this system will not even violate the
previous, more stringent set of rules. Under FCC’s previous sets of rules, any scheme
for using White Spaces, including the one proposed here, must be able to sense WMs
and shut down transmission in TV channels where they are detected. The operative
definition of ’detected’ is that the received power is measured to be at or above -114 dBm
when using an isotropic antenna mounted 10 m or higher above ground. In this section
we show that although in the last, more relaxed, rulings of FCC, this requirement is
omitted, the performance of our system remains unharmed while implementing WM
protection rules. For our system in New Jersey, we assume an antenna height 30 m above
ground and make a first-order (conservative) calculation of the number of detectable
WMs per radio tower.
A typical WM transmits at a power of 50 mW (17 dBm), so a path loss above
131 dB would yield received WM signals below threshold. To estimate path loss vs.
distance, we use Hata’s Suburban model [15] in the UHF band, and we add 10 dB for
transmission loss into buildings, a very conservative increment according to most reports
(cf. [16]). The result is that ground-level WMs inside buildings would be detectable
within a 1-mile radius around our radio towers. The radius would be even smaller using
Hata’s Urban model ([ [15]]) and/or a higher (and more likely) transmission loss into
buildings.
Next, we estimate the expected number of WMs in a 1-mile radius, i.e., within
an area of 3.14 square miles, in New Jersey. According to [17] and [18], there are
about 2500 schools and 4500 churches in the state, and these are assumed here to
be the dominant WM users. Given a total area of 7836 square miles, this translates
to averages of 1 school and 1.8 churches per 1-mile radius. Thus, at most three 6-
MHz channels might require shutdown in a typical cell at a given time, which is small
compared to the number of channels available per cell (7-31, as noted previously).
Several additional ameliorating factors can be noted. First, the affected WMs in
23
a given cell may congregate in one or two channels rather than three; second, data
on school and church locations could be used to advantageously tailor the positions of
radio towers; third, the focus of our study is rural areas, where the densities of schools
and churches may be below the average; fourth, peak-hour usages of WMs in schools
and churches (daytimes and evenings, respectively) differs from those for Internet users
(9 pm - 1 am [14]), thus reducing the ’duty cycle’ of channel shutdowns needed to
protect WMs. We conclude that the unavailability of channels due to WMs would have
a minor impact on the performance of our proposed system.
Aside from meeting FCC requirements, we believe that our system would pose little
potential danger to WM usage. This is because of the height, directionality and fixed
positions of the radio tower antennas, and the fact that most WMs would be near the
ground. A more detailed study would be needed to confirm this result or to engineer
around potential problems. At this stage, this is no longer a requirement and this study
is just for information purposes.
4.4 Achievable Capacity Calculations
In order to compute the achievable capacity, we assume that all TV White Space trans-
mitters (towers) have the ability to execute perfect spectrum sensing and avoid causing
any interference to any of the primary (protected) devices using this spectrum. We
calculate the achievable capacity based on the available bandwidth and spectrum ef-
ficiency achieved by these TV White Space transmissions. The available bandwidth
was described in Chapter 3. For determining the spectral efficiency, we need the re-
ceived signal-to-noise ratio (SNR) of the TV White Space transmission. The SNR at
the receiving antenna depends on the transmission power, path loss and noise. The
transmission power for fixed devices has been limited by the FCC to 4 watts (6 dBW)
and noise consists of thermal noise augmented by the receiver noise figure, assumed
here to be 10 dB.
24
4.4.1 Path loss
To determine the path loss in our 5-mile LOS links, we use the ITU Terrain Propagation
Model [19]. This model is ideal for LOS transmissions with regular obstructions on the
path. Under this model the median path loss (PL50%) is given by
PL50% = FSL+Ad dB, (4.1)
where FSL is the free-space path loss given as
FSL = 20 log10 d+ 20 log10 1000f + 32.44 dB. (4.2)
The inter-terminal distance d is in km, the frequency f is in GHz, and Ad is the excess
path loss above and beyond free-space loss. It is based on the antenna height, the
obstruction height and the link distance, and is given by
Ad = −20h/F1 + 10 dB, (4.3)
where h is the height difference between the most significant path blockage and the
line-of-sight path between the transmitter and the receiver; and F1 is the radius of the
first Fresnel zone given by
F1 = 17.3
√d
4fm, (4.4)
where d is in km and f is in GHz. Note that the obstruction is considered to be at the
mid-point between the two antennas.
To be on the conservative side, we add a dB term to the path loss to account for
random variations about the ITU prediction; and we model this term as a Gaussian
variate of zero mean and standard deviation σ . Also, to be conservative, we assume
the random variation term is at its 99% value on all links (only 1% of all links have a
greater value), in which case we can write the path loss on every link as
PLtotal = PL50% + 2.3× σ dB. (4.5)
25
4.4.2 SNR and Spectral efficiency
The received path loss can be written as
SNR = PT − 10× log10(kTB)−NF − PLtotal dB, (4.6)
where PT is transmit power, B is bandwidth (6 MHz), kT is thermal noise density, and
NF is the receiver noise figure (10 dB).
The theoretical upperbound, η, on spectral efficiency is given by
η = log2(1 + 10SNR10 ) bps/Hz. (4.7)
The parameters used for our analysis are given in table 4.1. Note that we consider
two different values for obstruction height; this is due to different structure heights in
cities and is based on population. In our model, we have assumed 15 m obstruction
heights (5-story buildings) for sparsely populated areas, and 30 m heights (10-story
buildings) for densely populated areas. Also, note that some parameters in table 4.1
are specified by the FCC and others are typical numbers used in similar studies.
Table 4.1: Parameters used in calculations
Link height 30 meters
Obstruction height 30 meters (for densely populated areas)15 meters (for sparsely populated areas)
Link distance (d) 5 miles (8 kilometers)
Frequency (f) 580 MHz (average White Space frequency)
Transmission power (PT ) 6 dBW (4 watts)
Thermal noise (kTB) -136 dBW
Noise Figure (NF ) 10 dB
Shadowing margin (σ) 3 dB
The effect of obstruction height on path loss in the ITU model is illustrated in
Figure 4.3, which plots SNR vs. obstruction height for different frequencies.
Now, for the two different obstruction heights, we can compute the SNRs and the
corresponding spectral efficiencies as follows:
26
Figure 4.3: Received SNR vs. obstruction height on a 5-mile path, for different frequencies.This plot is based on the ITU path loss model, with the obstruction assumed to be at the centerof the path.
ho = 15 m⇒ SNR = 18.7 dB⇒ η = 6.23 bps/Hz
ho = 30 m⇒ SNR = 9.3 dB⇒ η = 3.25 bps/Hz
To determine where to assume a sparse population and where to assume a dense
one, we did the following: Based on US Census 2000 data for population density in
New Jersey [20], we estimated the population for each cell. We modeled this with 10
population grades. The first two high population grades where considered as densely
populated with obstructions heights of 30 meters where the rest were considered as
sparse with 15-meter high obstructions. Table 4.2 shows this estimation with finer
detail. Also, the shadings of Figure 4.4 indicate the population distribution of New
27
Jersey based on our grid.
Table 4.2: Estimated Population Grades
Grade 1 2 3 4 5 6 7 8 9 10
PopulationDensityBracket
Max 14000 8000 5000 3000 1400 800 500 200 100 40
Min 8000 5000 3000 1400 800 500 200 100 40 0
Number of Cells 10 10 35 15 20 38 19 59 63 38
Figure 4.4: Population distribution in New Jersey, based on the US Census of 2000. Darkershadings correspond to higher densities, with the darkest areas concentrated in the New YorkCity metropolitan area.
4.4.3 Frequency planning
Most available channels are similar in neighboring cells. Thus, when using same chan-
nels in neighboring cells, we will have co-channel interference problems. In an array of
cells, frequency planning would insure that the receiving terminal in a link receives a
relatively higher signal power from its corresponding transmitter rather than any other
neighboring transmitter. In other words, in Figure 4.5, if cell 1 is transmitting to cell 2
in a certain channel, cell 2 cannot use that channel to transmit to cell 3. To guarantee
28
this, cells 1 and 2 should transmit on different channels. For example, with a frequency
reuse factor of 2, cells 1 and 3 would both use frequency fa but not fb, and cells 2 and 4
would both use frequency fb but not fa. Note that cell 4 will receive signals from cells
1 and 3, though the latter signal will be stronger by about 14.5 dB (see lower part of
Figure 4.5, which shows the decay of the received SNR with distance, using our path
loss model).
Figure 4.5: (top) five neighboring cells with M common available channels illustrating usageof different sets of channels with different colors for a reuse factor of 2. (bottom) a plot of thereceived SNR from cell 1 vs. distance, illustrating the fading of the signal power as it reachescell 3. Each ’X’ on the plot represents the position of neighboring receivers.
To alleviate co-channel interference (e.g., the interference to the cell 4 receiver from
the cell 1 transmitter), we can assume that each tower has a second receive antenna,
used to effect interference cancellation. Alternatively, we can assume a larger reuse
factor (3 or higher), at a cost in capacity. In our calculations here, we will assume that
a second receive antenna is used to cancel interference, and that the reuse factor is 2.
Under these assumptions, we can construct a database for the maximum deliverable
rate from each cell. Table 4.3 shows available bandwidth before and after frequency
planning, cell spectral efficiency, maximum deliverable rates in Mbps and MB/hr for
29
some example cells.
Table 4.3: Maximum Deliverable Rate
CellAvailableBandwidth
AvailableBandwidth(frequencyplanning)
Cell SpectralEfficiency(bps/Hz)
Max DeliverableRate (Mbps)
Max Deliverable(MB per hour)
113 66 33 3.25 107.25 48262.5
114 66 33 6.23 205.59 92515.5
121 84 42 6.23 261.66 117747
122 72 36 6.23 224.28 100926
123 72 36 6.23 224.28 100926
4.5 Demand
To characterize the Internet traffic demand in each location in New Jersey, we use
a Cisco survey on Internet traffic [9]. This survey says that an average broadband
connection generates 11.4 GB traffic per month (375 MB per day per connection). In
peak hours generated traffic is 18 MB per connection per hour in comparison to nonpeak
hours in which generated traffic in 15 MB per connection per hour. In addition, Cisco
has estimated that peak Internet traffic may grow seven-fold by 2013, compared to a
five-fold increase of average Internet traffic.
In our model, we used this information to estimate the future demand. We take
into consideration two parameters: usage per household and simultaneous active users.
4.5.1 Usage per household
We have the Year 2000 population data for each cell of the grid shown for New Jersey.
The average population per household in the US is 2.48 [21], which we round up to 3.
Also, 74.2% of people in the US have Internet access [22]. From these estimates, we
can estimate the number of households (Internet clients) in each cell. In making this
calculation, we assume a current peak traffic of 18 MB/hour, with a potential future
growth in traffic by factors of 5, 7 and 10 (corresponding to 90, 126 and 180 MB/hour in
the future). Table 4.4 shows usage per household calculations for some example cells.
In table 4.4, Cell, is the cell number; followed by Pop. Grade, which is the population
30
grade where that cell resides; population per square mile; total cell population; number
of household in cell; and number of Internet clients per cell.
Table 4.4: Usage Per Household
CellPop.Grade
Pop persq. mi.
Cell Pop.Householdper Cell
Internet Usersper Cell
113 2 6500 162500 54166.67 40191.67
114 3 4000 100000 33333.33 24733.33
121 10 20 500 166.67 123.67
122 9 60 1500 500 371
123 10 20 500 166.67 123.67
4.5.2 Active users
We parameterize the percentage of active households for any given time and take 10%,
30% and 50% active households as three different scenarios.
Based on these two parameters–peak demand and percent active households, we can
construct a database, consisting of the demand for different hourly traffics and different
active user percentages for each cell. Table 4.5 shows demand for the same set of cells
with different hourly traffics for 30% active users.
Table 4.5: Demand for Different Hourly Traffic and 30% Active Users (MB/hr)
Cell30%ActiveUsers
Cell Demandfor 18 MB perhour
Cell Demandfor 90 MB perhour
Cell Demandfor 126 MBper hour
Cell Demandfor 180 MBper hour
113 12057.5 217035 1085175 1519245 2170350
114 7420 133560 667800 934920 1335600
121 37.1 667.8 3339 4674.6 6678
122 111.3 2003.4 10017 14023.8 20034
123 37.1 667.8 3339 4674.6 6678
4.6 Feasibility of a Backhaul Network in New Jersey
A comparison of achievable rates and demand per cell lead us to some findings on the
feasibility of our proposed backhaul system. If the achievable capacity is less than
demand in a given cell, fiber would be needed. Otherwise, radio towers and TV White
Spaces could be used to transfer traffic between neighboring cells.
31
Table 4.6 shows the number of cells (out of a total of 307 cells) that would need
fiber connectivity to its neighbors, under our study assumptions. Also, Figures 4.6,
4.7 and 4.8 illustrate the geographical distribution of these cells on a New Jersey map,
assuming 10%, 30% and 50% simultaneous active users, respectively.
Table 4.6: Number of cells that need fiber out of a total 307 cells
Per Household Usage 10% 30% 50%
18MB 20 63 70
90MB 70 117 137
126MB 80 135 141
180MB 91 139 155
Figure 4.6: Geographical distribution of cells needing fiber, assuming 10% active users. Theusage per house, from left to right, is 18, 90, 126 and 180 MB per hour
4.7 Aggregating Multiple Flows
The method we have used so far for calculating the feasibility of the backhaul network
is sufficient to guarantee the traffic can be transferred from one cell to its neighbor
(i.e. we can guarantee that a link between two neighboring cells has sufficient transfer
capacity). In most use-cases though, traffic will be routed through a number of nodes
to reach the intended destination and the data traffic carried through these links is
more likely to be an aggregate of a number of flows. Thus our methodology conditions
are necessary but not sufficient to prove the feasibility of this system.
32
Figure 4.7: Geographical distribution of cells needing fiber, assuming 30% active users. Theusage per house, from left to right, is 18, 90, 126 and 180 MB per hour
The issue of aggregating flows will not have a major effect on such a system because
this network is meant to aid existing fiber infrastructure in connecting to areas where
there is no current service. The purpose of this network would be to retrieve local traffic
and transfer it to the nearest fiber point, so that from there the traffic can be routed
to the final destination. In other words, since traffic will not be routed through a large
number of hops, with some minor provisions, this aggregation can be dealt with.
Our proposed solution for this problem is to use the Excess Capacity for this aggre-
gation. In the database of available capacity vs. demand, we observe that in almost
all cells where the achievable capacity is greater than the demand, this difference is
a significant amount. This extra capacity above the demand is what we refer to as
“Excess Capacity”. This Excess Capacity can be used for routing traffic.
On top of using Excess Capacity we can use clustering techniques, plant extra fiber
points at cluster heads, and route traffic through these fiber points. This would allow
traffic to reach fiber with fewer hops thus ameliorating the aggregation effect. A suitable
choice of cluster head within a cluster is that cell for which the capacity is highest, so
that more Excess Capacity is available for high traffic loads. Although a detailed routing
study would be appropriate to mitigate this issue, the current level of detail is sufficient
for use with our general methodology.
An example of such a solution is illustrated in Figure 4.9 where all New Jersey
33
Figure 4.8: Geographical distribution of cells needing fiber, assuming 50% active users. Theusage per house, from left to right, is 18, 90, 126 and 180 MB per hour
cells are grouped into 8 clusters and each cluster has a fiber planted at its cluster
head. In this situation, the number of hops to reach a fiber point from any cell is 2
or 3. By comparing the Excess Capacity with twice (or three times) the demand we
can determine whether these 8 extra fiber points are sufficient. In clusters where the
Excess Capacity is less, we add another fiber point. The particular example in Figure
4.9 refers to 30% active clients using 126 MB/hr (an average case in our study). An
extra 10 fiber points (2 in addition to the 8 already marked in the figure) are required
here to guarantee enough capacity for routing. Note that adding this to our previous
calculations still yields a reasonable number of fiber points and does not diminish the
feasibility of our system.
4.8 Conclusion
We observe that most places that cannot be supported by using radio towers and White
Spaces are in the New York or Philadelphia metropolitan areas. It is safe to assume,
therefore, that these areas already have a fiber infrastructure in place. Conversely, all
of the areas which are somewhat remotely located, basically places that may not have
access to wired infrastructure, do have ample TV White Space capacity. The issue of
multiple flows can be dealt with by provisions such as clustering and adding a couple
of extra fiber points. This makes wireless backhauling a viable proposition.
34
Figure 4.9: An example clustering technique to mitigate aggregation with an initial 8 extrafiber points. For this example, the average hourly load per household is 126MB, for 30% of allhouseholds.
35
Chapter 5
Conclusions and Future Directions
In this thesis, we have derived a database of available White Spaces for fixed devices
taking into account transmitters in New York, Pennsylvania and Delaware, and pro-
posed to utilize White Spaces for a backhaul network for Internet traffic based on
existing FCC restrictions. Using the available White Spaces in the UHF channels for
fixed TVBDs and backhaul traffic demands in New Jersey for our case study, we have
evaluated the feasibility of such backhauling and have presented a methodology that
can be used for other areas as well. Our results show that the TV White Spaces can be
an effective medium for radio backhaul as an alternative to the costly laying of optical
fiber. A first order analysis of the traffic flow in this system enabled us to tackle the
issue of aggregation with minimal provisions such as clustering and having more fiber
points. This augmentation to our approach also adds to the reliability of such a system,
and in turn, to the justification for such methodology.
Most parameters used in our calculations are conservative, with results that are
correspondingly pessimistic. Under these conservative assumptions, and for the most
extreme conditions considered, only 50% of New Jersey cannot be covered using radio
towers and TV White Spaces. This is a very satisfying result, leading to a positive
finding for the proposed backhaul concept. In addition, the methodology used here
can be transported to other cases (e.g., other states) to see if the same concept can be
beneficially applied elsewhere.
This work can be continued by conducting detailed studies on the most recent ruling
(September 2010) regarding wireless microphone provisions; incorporating LPTV and
other local low power transmission stations to White Space availability calculations
(for more accuracy); and studying in detail appropriate routing schemes for mitigating
36
aggregation. To further expand such a study one could use the same methodology for
a national use-case and in doing that, more detailed path loss modeling can be done
based on different terrain and population conditions.
37
References
[1] I.F. Akyildiz, Won-Yeol Lee, M.C. Vuran, and S. Mohanty. A survey on spectrummanagement in cognitive radio networks. Communications Magazine, IEEE, 46(4):40 –48, april 2008. ISSN 0163-6804.
[2] S. J. Shellhammer. Spectrum sensing in ieee 802.22. IAPR Wksp. Cognitive Info.Processing, Santorini, Greece, Jun. 2008.
[3] O. lleri and N.B. Mandayam. Dynamic spectrum access models: toward an engi-neering perspective in the spectrum debate. Communications Magazine, IEEE, 46(1):153 –160, january 2008. ISSN 0163-6804.
[4] In the matter of unlicensed operation in the tv broadcast bands: Second report andorder and memorandum opinion and order. Federal Communications Commision,Tech. Rep. 08-260, Nov. 2008. URL http://hraunfoss.fcc.gov/edocspublic/
attachmatch/FCC-08-260A1.pdf.
[5] S. M. Mishra and A. Sahai. How much white space is there? Tech.Rep. EECS-2009-3, Jan. 2009. URL http://www.eecs.berkeley.edu/Pubs/
TechRpts/2009/EECS-2009-3.html.
[6] Paramvir Bahl, Ranveer Chandra, Thomas Moscibroda, Rohan Murty, and MattWelsh. White space networking with wi-fi like connectivity. In SIGCOMM ’09:Proceedings of the ACM SIGCOMM 2009 conference on Data communication,pages 27–38. ACM, New York, NY, USA, 2009. ISBN 978-1-60558-594-9.
[7] J. D’Ambrosia. 100 gigabit ethernet and beyond [commentary]. CommunicationsMagazine, IEEE, 48(3):S6 –S13, march 2010. ISSN 0163-6804.
[8] G. Wellbrock and T.J. Xia. The road to 100g deployment [commentary]. Commu-nications Magazine, IEEE, 48(3):S14 –S18, march 2010. ISSN 0163-6804.
[9] The digital tv transition. URL http://www.dtv.gov/.
[10] Google maps. URL http://www.maps.google.com/.
[11] List of all class a, lptv, and tv translator stations. Federal Communications Com-mision, Tech. Rep., 2008. URL http://www.dtv.gov/MasterLowPowerList.xls.
[12] List of licensed tv channels available at each location. URL www.tvfool.com.
[13] List of available channels for unlicensed use. URL www.showmywhitespace.com.
[14] Cisco visual networking index: Usage study. 2009. URL http:
//www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/ns705/
Cisco_VNI_Usage_WP.pdf.
38
[15] M. Hata. Empirical formula for propagation loss in land mobile radio services.Vehicular Technology, IEEE Transactions on, 29(3):317 – 325, aug 1980. ISSN0018-9545.
[16] T. S. Rappaprt. Wireless communications: Principles and practices. IEEE Pressand Prentice Hall, page 124, 1996.
[17] New jersey public schools fact sheet. URL http://www.state.nj.us/education/
data/fact.htm.
[18] New jersey christian church directory. URL http://www.churchangel.com/
newjersy.htm.
[19] Itu-r recommendations. URL http://www.itu.int/publ/R-REC/en.
[20] Us census 2000 gazetteer files. URL http://www.census.gov/geo/www/
gazetteer/places2k.html.
[21] State and county quickfacts. URL http://quickfacts.census.gov/qfd/meta/
longHSD310200.htm.
[22] Usage and population statistics. URL http://www.internetworldstats.com/.
