close

Вход

Забыли?

вход по аккаунту

?

Microwave fiber-optic links for microcellular wireless communication networks

код для вставкиСкачать
INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI
films the text directly from the original or copy submitted. Thus, some
thesis and dissertation copies are in typewriter face, while others may be
from any type o f computer printer.
T he quality o f this reproduction is dependent upon th e quality of the
copy subm itted. Broken or indistinct print, colored or poor quality
illustrations and photographs, prim bleedthrough, substandard margins,
and improper alignment can adversely affect reproduction.
In the unlikely event that the author did not send UMI a complete
manuscript and there are missing pages, these will be noted.
Also, if
unauthorized copyright material had to be removed, a note will indicate
the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by
sectioning the original, beginning at the upper left-hand comer and
continuing from left to right in equal sections with small overlaps. Each
original is also photographed in one exposure and is included in reduced
form at the back of the book.
Photographs included in the original manuscript have been reproduced
xerographically in this copy. Higher quality 6” x 9” black and white
photographic prints are available for any photographs or illustrations
appearing in this copy for an additional charge. Contact UMI directly to
order.
UMI
A Bell & Howell Information Company
300 North Zeeb Road, Ann Arbor MI 48106-1346 USA
313/761-4700 800/321-0600
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
NOTE TO USERS
The original manuscript received by UMI contains pages with
slanted print Pages were microfilmed as received.
This reproduction is the best copy available
UMI
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Microwave Fiber-Optic Links for Microcellular
Wireless Communication Networks
by
David Marcel Cutrer
B.S.E.E. (California Institute o f Technology) 1992
M.S.E.E. (University of California, Berkeley) 1994
A dissertation submitted in partial satisfaction o f the
requirements for the degree of
Doctor or Philosophy
in
Engineering-Electrical Engineering
and Computer Sciences
in the
GRADUATE DIVISION
of the
UNIVERSITY o f CALIFORNIA, BERKELEY
Committee in charge:
Professor Kam Y. Lau, Chair
Professor T. K. Gustafson
Professor Peter Y. Yu
1998
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UMI Number: 9902046
UMI Microform 9902046
Copyright 1998, by UMI Company. All rights reserved.
This microform edition is protected against unauthorized
copying under Title 17, United States Code.
UMI
300 North Zeeb Road
Ann Arbor, MI 48103
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Microwave Fiber-Optic Links for Microcellular
Wireless Communication Networks
Copyright© 1998
by
David Marcel Cutrer
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The dissertation of David Marcel Cutrer is approved:
// 9 /
Chair
Date
T&
fPejLu)
Date
Date
University of California, Berkeley
1997
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Abstract
Microwave Fiber-Optic Links for Microcellular
Wireless Communication Networks
by
David Marcel Cutrer
Doctor o f Philosophy in
Engineering-Electrical Engineering and Computer Sciences
University o f California, Berkeley
Professor Kam Y. Lau, Chair
The next generation o f communication systems promises to provide consumers
with a host o f new and improved voice, video, and data services. In particular, the rapid
development o f wireless radio systems has given consumers the luxury o f tetherless
access to telephone conversation, on-line computing, and cable-television service. This
type o f wireless access to high quality information and entertainment services is quickly
becoming a key component o f the information superhighway. New communication
systems must have the bandwidth and spatial diversity to supply wireless users at any
location within the network with the service o f their choice. Optical fiber links, which
have proved to be critical to the development o f high-speed wired networks, may also
provide an excellent connecting infrastructure for providing uniform radio coverage in
wireless networks. This thesis is concerned with device technology and system issues
associated with the implementation o f microwave and millimeter fiber-optic links as the
connecting infrastructure in these wireless networks.
1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The linearity performance of low-cost microwave, and millimeter-wave optical
transmitters will be assessed for the wireless application. Using a statistical model o f user
access in a wireless network, it is shown that by accepting a modest (<0.5%) percentage
of blocked calls, a modest optical link dynamic range o f 91dB (1Hz) is required. By using
multiple fiber-fed antennae per cell and proper network protocol, the required dynamic
range is dramatically reduced to <80dB for the same blocking probability. Also, it is
concluded that the ideal system architecture for providing wireless service is a
combination o f distributed antennas used for radio coverage, and a centralized base
station employed for voice channel capacity.
It is shown that millimeter-wave analog fiber-optic transmission can be
accomplished by resonant modulation of a single contact monolithic semiconductor laser
at the cavity round trip frequency o f 40GHz. Efficient mode coupling is obtained with a
single contact device by utilizing the high attenuation o f the millimeter-wave modulation
signal along the laser stripe. The properties and limitations o f this technique are analyzed
using a distributed circuit model of the laser.
Ultralow threshold (I^lOOuA) lasers can be used in ultralow power optical
interconnect for digital RF transport, preferably in a bias-free digital optical modulation
format. We show that even though optical power requirements for successful
transmission may dictate that the pulse drive current be many times that of I*, reducing
the latter to 10-100pA is still essential in minimizing the total driver power to the laser at
multi-gigabit data rates.
2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Professor Kam Y. Lau
Committee Chairman
3
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Acknowledgements
I have enjoyed working with, and developing a professional relationship with my advisor,
Professor Kam Lau. He is truly an asset to the scientific community, and I feel honored to
have had the opportunity to work with him during my career.
Many thanks to professors Kenneth Gustafson and Peter Yu for their valuable input
during my qualifying exam and dissertation. Thanks also to Professor Joseph Kahn for
his support and input on my qualifying exam.
A special thanks to my fiancee, Beth Mammini, who has seen me though my entire
graduate career. Her unwavering support for all my endeavors has provided me strength
in everything that I do.
During my graduate career, I have had the opportunity to work with, and develop
friendships with several exceptional individuals. Most notable, my work with John
Georges was the most inspired and productive period o f my graduate education. John and
m yself comprise a formidable team that will certainly find greatness. I also had the
pleasure to work with Simon Yeung, whom I have shared an office with for the last three
years. Finally, I have had the pleasure o f working with Adam Schwartz, who has taught
me the true meaning and spirit o f engineering excellence. I see great things ahead for this
group o f people.
iv
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
Thanks to Peter Pepeljugoski, John Park, Lisa Buckman, Meng-Hsiung Kiang, Lenoard
Chen, Bin Wu, Mike Daneman, Ta-Chung Wu, Sydney Kan, Dan Vassilovski, and Olav
Solgaard. My technical discussions and interaction with these individuals contributed
much to my graduate career.
Further thanks to Jeanene Hayes, Heather Brown, Ben Lake, Julie Aquino, Margie
Berger, La Tonia Bolds, Ruth Gjerde, Mary Byrnes, Pat Heman, and many others for
providing the infrastructure to keep Cory Hall running.
Sincere thanks to NDSEG, AT&T, DARPA, and Rome Laboratory for their generous
financial support of my research.
v
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table of Contents
1
2
OVERVIEW_____________________________________________________________________ 1
1.1
In t r o d u c t i o n ......................................................................................................................................................................... 1
1.2
1.3
l .4
1.5
1.6
1.7
F ib e r - f e d M i c r o c e l l s ......................................................................................................................................................3
H is t o r ic a l B a c k g r o u n d ................................................................................................................................................5
F ib e r -O p t ic B a c k b o n e f o r W ir e l e s s N e t w o r k s .............................................................................................6
N e t w o r k B e n e f i t s ..............................................................................................................................................................9
C o m m e r c ia l a n a l o g F ib e r -O ptic D e v ic e T e c h n o l o g y . ...........................................................................12
M il l im e t e r -W a v e O p t ic a l T r a n s m it t e r s ......................................................................................................... 13
1.8
R e q u ir e d D e v ic e P e r f o r m a n c e ...............................................................................................................................16
MOTIVATION AND BACKGROUND______________________________________________ 22
2.1
W h y d is t r ib u t e a n t e n n a s ? ......................................................................................................................................... 2 2
2 . 1.1
Introduction........................................................................................................... 22
2.1.2
Illustrative example................................................................................................. 28
2.1.3
Technical Analysis.................................................................................................. 30
2.2 W h y u s e a n a l o g f i b e r o p t i c s ? ...................................................................................................................................36
2 .3
M il l im e t e r - w a v e F ib e r - o p t i c s ................................................................................................................................. 37
3 DYNAMIC RANGE REQUIREMENTS FOR OPTICAL TRANSMITTERS IN FIBER-FED
MICROCELLULAR NETWORKS_____________________________________________________ 44
3.1
3 .2
I n t r o d u c t i o n ......................................................................................................................................................................4 4
T h e s y s t e m m o d e l ...........................................................................................................................................................4 6
3 .3
M a t l a b S i m u l a t i o n ....................................................................................................................................................... 53
4
MILLIMETER-WAVE OPTICAL TRANSMITTERS________________________________ 58
4.1
4 .2
S e m ic o n d u c t o r L a s e r R F /O p t ic a l P a c k a g in g Is s u e s .............................................................................. 58
D e v i c e S e l e c t i o n , P a c k a g i n g , a n d M M - w a v e T r a n s p o r t a t 4 5 GHz ............................................59
4 .3
R e s o n a n t m o d u l a t io n o f s in g l e c o n t a c t m o n o l it h ic s e m i c o n d u c t o r l a s e r s a t
MILLIMETER WAVE FREQUENCIES................................................................................................................................................ 61
5
DISTRIBUTED ANTENNA SYSTEM IMPLEMENTATION___________________________ 72
5.1
M e a s u r e m e n t b a s e d d e s ig n a n d o p t im a l a r c h it e c t u r e f o r a n in - b u il d in g d is t r ib u t e d
ANTENNA SYSTEM.............................................................................................................................................................................. 72
6
5 .2
R a d io M e a s u r e m e n t s .................................................................................................................................................... 74
5 .3
O p t im a l A r c h i t e c t u r e ................................................................................................................................................ 78
5 .4
5 .5
N o is e f ig u r e a n a l y s i s ...................................................................................................................................................81
C o n c l u s i o n s .......................................................................................................................................................................... 83
LOW POWER, HIGH-SPEED FIBER LINKS FOR DIGITAL RF TRANSPORT__________ 84
6.1
6 .2
D ig it a l R F T r a n s p o r t ..................................................................................................................................................84
U l t r a l o w P o w e r D ig it a l F ib e r - o p t ic L in k s U s in g Z e r o - b ia s e d S e m i c o n d u c t o r L a s e r s 85
7
CONCLUSIONS AND FUTURE DIRECTIONS______________________________________ 94
8
REFERENCES_________________________________________________________________ 96
vi
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
List of Figures
Figure 1.1.1. Comparison of various markets for fiber-optic links.........— ....................— .—.............. 2
Figure 1.2.1. Future wireless personal communication network with fiber-optic backbone
infrastructure.
Figure 1.4.1. Fiber-fed distributed antenna network to provide uniform radio coverage to a given
area............................. .......................................................................................................—................ 8
Figure 1.7.1. Narrow-band, millimeter-wave optical transmitters based on the technique of resonant
modulation..............................................................................................— ....................................... 15
Figure 1.8.1. Illustration of a typical two-tone dynamic range measurement with signal, noise, and
intermodulation distortion (IMD) terms shown...... .................—.......................— .................. 17
Figure 1.8.2. Simulation results for analog link performance in a typical microcellular environment.
______________________________________________________________________________ 19
Figure 2.1.1, In-building coverage and capacity can provided to some, but not all of the floor of the
building by using a single mini-base station
25
Figure 2.1.2. Installing multiple mini-base stations can provide complete coverage at the expense of
reduced local capacity....................................................................................................................... 26
Figure 2.1.3. The ideal performance/cost solution to providing in-building coverage......................... 27
Figure 23.1. Narrowband phased-array antenna network illustrating the versatility of an optical
transmitter capable of mm-wave transport............__ .................—.............................................. 38
Figure 2 3 3 . Millimeter-wave wireless personal communications network illustrating an analog
fiber-optic interconnecting infrastructure.......................— ...........—...........................—...— .... 40
Figure 3.1.1. Fiber-fed distributed antenna wireless network using analog fiber links........................ 45
Figure 3.2.1. Typical spectrum of received signal and IMD powers at antennas A and B shown in
Figure 3.1.1. For this plot, the transmitter dynamic range is 70dB (1Hz)................................... 48
Figure 3.2.2. Percentage of blocked calls as a function of transmitter dynamic range for a single
fiber-fed 300m radius microcell for 5,10 and 20 voice channels................................................... 49
Figure 3 2 3 . Illustration of the maximum signal and maximum C/I protocols............................—.... 50
Figure 3.2.4. Percentage of blocked calls as a function of transmitter dynamic range for 4 and 9
antennas and two different signal protocols. This simulation considers an 1800m square area
with 20 available voice channels............................... ..........................................—..................... 51
Figure 3.2.5. Measured two tone dynamic range of a self pulsating CD laser at 900MHz. The inset
(lOdB/div) shows the two tones centered at 900MHz and the induced intermodulation products.
______________________________________________________________________________ 52
Figure 4.2.1. Illustration of a DBR laser structure used for the research effort. Round-trip resonant
frequency of this device was measured at 45 GHz......................................................................... 59
Figure 4.2.2. Illustration of gold-plated copper carrier used to secure the laser for mm-wave
modulation. The small microstrip board accepts the modulation from the center conductor of a
Wiltron V-connector......................................................................................................................... 60
Figure 4.3.1. The mode-locking phenomenon fundamentally requires an in-homogeneous modulation
of the laser cavity.............................................................................................................................. 62
Figure 4.3.2. Illustration of a single contact mode-locking experiment. A microwave probe is used to
modulate the device at a particular feed point which leads to a partial modulation of the laser
cavity..............___ ............................................................................................................................. 63
Figure 4 3 3 . Experimental set-up used to characterize the single contact resonant modulation. The
streak camera is used for time domain measurements, and the spectrum analyzer is used for
frequency response measurements.................................................................................................. 64
Figure 43.4. Measured streak camera trace of the modulated light output from the single contact
device at 40GHz for two different microwave probe positions along the cavity......................... 65
Figure 4 3 3 . The complete modulation response of the single contact device at low and high
frequencies is shown. The solid curve corresponds to an L/4 fed device, and the dotted curve
corresponds to an L/2 fed device..................................................................................................... 66
Figure 4J.6. Distribution of injected current into the active region of the device as a function of
vii
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
length along the laser stripe at 40 and 90GHz is shown. The inset shows the distributed circuit
model [28] used to calculate the current.
—
— ........ ..— -------- 67
Figure 4.3.7. Calculated mode coupling factor as a function o f probe position for a device with a
round trip cavity frequency of 40GHz. The dotted curve corresponds to the limiting case when
the injected current is a delta
—
...-69
Figure 5.1.1. Floor plan of the radio environment under study. Initial radio measurements were
made in the main corridor with four distributed antennas................. ............ ....—..........
— 73
Figure 5.2.1. Measured received RF power at antennas 1 and 2 as a function of the position along the
hall. The distributions of the received powers are also shown. .............................
75
Figure 5.2.2. Measurement based prediction of the system radio performance with 2 and 4 antennas
used with a diversity selection technique.
76
Figure 5.3.1. A comparison of flber-to-the-antenna and hybrid fiber-coax architectures for providing
cost effective in-building radio coverage.......................................................................................... 79
Figure 5.3.2. Illustration o f the optimal hybrid fiber-coax architecture implemented in the wireless
field-trial at the University of California at Berkeley.................................................................... 80
Figure 5.4.1. A typical hybrid fiber-coax architecture. The passive coax cable of loss (L) is followed
by an amplifier of gain L and an fiber link of a given noise figure................................................81
Figure 5.4.2. Composite system noise figure as a function coaxial cable length................................... 82
Figure 6.2.1 Illustration of the pseudorandom bit stream, the time evolution of the carrier density,
and the photon density out of the modulated laser. «....•••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 86
Figure 6.2.2. Distribution function of turn-on time for different values of h........................................ 89
Figure 6.23. BER as a function of drive current for various device thresholds................................... 90
Figure 6.2.4. Electrical drive power to the laser as a function of the device threshold current, for
various bit rates, and carrier lifetimes............................................................................................ 92
viii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
List of Tables
Table 1.8.1. Comparison of the performance of various analog optical transmitters..— ...—........... 20
Table 2.1.1. Summary of illustrative comparison between distributing radios vs distributing antennas
and centralizing capacity
29
Table 2J.1. FCC licenses that have been granted to corporations interested in operating and testing
at mm-wave frequencies...........................................................................................................—.... 42
Table S2.1. Comparison of the antenna dynamic range requirements and radio coverage
characteristics of the hallway under study..........................................................—..........— ....... 77
ix
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1 Overview
1.1
introduction
The next generation o f communication systems promises to provide consumers with a
host of new and improved voice, video, and data services. In particular, the rapid
development of wireless radio systems has given consumers the luxury o f tetherless
access to telephone conversation, on-line computing, and cable-television service. This
type of wireless access to high quality information and entertainment services is quickly
becoming a key component of the information superhighway. New communication
systems must have the bandwidth and spatial diversity to supply wireless users at any
location within the network with the service o f their choice. Optical fiber links, which
have proved to be critical to the development o f high-speed wired networks, may also
provide an excellent connecting infrastructure for providing uniform radio coverage in
wireless networks. A comparison of markets for fiber-optics is shown in Figure 1.1.1.
The markets are segmented into four quadrants selected by high-end applications vs
high volume / cost sensitive applications, and analog vs digital applications. High-end
digital applications include long distance telecommunications, while high-end analog
applications include cable television signal delivery and satellite receiver applications.
Fiber-optics also finds substantial deployment for the transport o f high speed digital data
in local area computer networks. In this case, the cost of the optical transmitters and
receivers must be kept low since every computer on the network must deploy these opto­
electronic devices. The new emerging area in which fiber-optics promises to play a major
role, is the use o f fiber for the delivery of commercial wireless services.
1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Applications of Fiber-Optics
Digital
- T elecom
Shared
- Long Haul (>50km)
Medium
(high cost) -AT& T
Mass
Deployment
(low cost)
-L A N s
- 850nm mm-fiber
- HP, IBM
Analog
-CA TV
- Satellite/ R ad ar
- High Linearity
Fiber for Wireless
Emerging Area
Figure 1.1.1. Com parison of various m arkets for fiber-optic links.
2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
This so called “fiber for wireless” application is an analog application of fiber-optics
since the wireless signals are subcarrier multiplexed signals at microwave frequencies,
but is also very cost / volume sensitive since a large number o f fiber-optic antennas must
be deployed to provide coverage in a wireless network. This new application of fiberoptics presents a number o f unique challenges, some o f which are addressed in the
sections below.
1.2
Fiber-fed Microcells
This thesis is concerned with the need and implementation o f analog fiber-optic
communication links as the connecting infrastructure in wireless networks, providing
uniform radio coverage in present and future networks as illustrated in Figure 1.2.1. This
optical fiber ‘backbone’ for wireless radio systems has many unique performance
advantages over conventional solutions. Since this is a new application of fiber-optics,
several new device and network issues need to be addressed. Such as, in what way can
present wireless systems benefit from a fiber-optic infrastructure, and what is the required
opto-electronic device performance for satisfactory operation? Also, given that fiberoptics is playing an increasingly major role in wireless networks, what are the prospects
for future devices that provide greater performance and/or a reduced cost over presently
available devices? These issues will be addressed as follows: First, a brief historical
background on the evolution o f fiber-optic links will be presented, followed by a
discussion of the motivations for using these links as the connecting infrastructure in
wireless networks.
3
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
Optical Fiber Backbone
Figure 1.2.1. Future wireless personal communication network with fiber-optic
backbone infrastructure.
4
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Second, an overview o f the cost and commercial availability o f analog optical
transmitters for the present low-frequency (<3GHz), and new devices for future highfrequency wireless systems will be presented. Next, an analysis o f the required fiber-optic
link performance for the wireless application will be addressed using a simulation of a
fiber-fed radio system. Given this required level o f performance, the feasibility of using
present commercial devices and future devices in wireless radio systems can be assessed.
Finally, the future commercial and research directions o f this area will be discussed.
1.3
Historical Background
Since the development o f optical fiber and the semiconductor laser, fiber-optic links
have found application as both digital and analog communication channels. A natural use
o f these low-loss, high bandwidth optical links is for long-haul (>10km) and/or high
speed (>lGb/s) transmission o f digital signals. In this application, the light output from a
laser is simply modulated on and off such that each light pulse represents a bit of
information. These types of links are used in telecommunication systems and for high­
speed computer interconnects. Digital fiber-optic links have been commercially available
for over 15 years, and during this time researchers have continually improved the
reliability, speed, and cost o f these links.
A more recent application of fiber-optics is the use of linear optical devices for the
transmission of analog signals. In this scenario, a high-frequency analog signal is
transmitted in the time varying intensity of the light output from the optical transmitter.
In characterizing the suitability of optical devices for analog signal transmission, the
5
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
primary issue is the linearity o f the light output with respect to the modulation signal, and
the signal-to-noise ratio after transmission through the link. Analog fiber links have found
commercial application for cable television (CATV) video signal distribution [1]. Analog
CATV transmission using fiber-optics was actually proposed in the early 1970’s, but the
device linearity performance at that time was so poor that the idea was ruled out
immediately. Recent advances in linear fiber-optics technology have made this
application feasible. Many analog video channels (in the frequency range 50-550MHz)
are sent over a fiber link to a head-end distribution point where they are distributed to
subscribers over coaxial drops. The low-loss optical fiber allows for transmission o f the
video signals over a long distance (>5km) without signal repeaters. Analog fiber links
have also found application in military systems for the delivery o f signals to and from
remote satellite locations, and as feeders for phased-array radar systems [2], where many
microwave antennas must be driven with synchronous signals in the GHz frequency
range. Both the CATV and the military applications o f analog fiber links are very
demanding in terms of device performance (i.e. optical transmitter linearity and noise).
Hence, analog fiber links are historically high cost products for these applications. This is
in sharp contrast to the cost and device performance required for devices used in the
emerging area o f fiber-fed wireless networks to be discussed below.
1.4
Fiber-Optic Backbone for Wireless Networks
The goal of a wireless radio system is to provide acceptable communications to
mobile and spatially distributed users. A primary difficulty in achieving this goal is
creating a network with uniform radio coverage. This problem can be severe in urban
6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
environments where buildings and other obstructions shadow the radio signals. This is
unfortunate because urban settings and buildings are areas o f high consumer demand for
wireless services. The classic solution to this radio coverage problem is simply to build
more antennae sites, or base stations, to cover a given area. For example, this approach
has been implemented by cellular telephone service providers, thus creating more
coverage and capacity in many popular urban areas. Although the addition of new base
stations does improve coverage, the installation o f the expensive radio hardware, and the
required real estate for the cell site may be prohibited by cost. An alternative, and more
cost effective solution, is to implement the analog fiber-optic links discussed in the
previous section as feeders to remote antenna sites (called ‘microcells’) within the
coverage area o f a single cell site [3-6]. If we consider a cell covered by a single base
station, as illustrated in Figure 1.4.1, the radio coverage within the cell can be improved
by deploying several remote antenna sites within the cell that are linked to the base
station via analog fiber optic links. In this way, each antenna site simply consists o f an
antenna, an amplifier, and an optical transmitter.
The use o f optical fiber for this application provides advantages that have been
exploited in digital links for many years, including low-loss, wide bandwidth, and
immunity to electromagnetic interference. The resulting fiber-fed distributed antenna
network provides blanket radio coverage to the area without the construction o f any new
base stations. The same concept can be applied to solve in-building radio coverage
problems, where radio signals cannot adequately reach areas deep into the building. In
this case, fiber-fed antennas can be placed at strategic locations within the building, and
the fiber is routed to the roof where the radio coverage is excellent.
7
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Analog
Fiber
Links
Base
Station
Mobile
Users
RF signals from
usar handsets
G
Remote Antenna
Optical Fiber
Q
* W ill
Base Station
Figure 1.4.1. Fiber-fed distributed antenna network to provide uniform radio
coverage to a given area.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Also, in-building wireless LAN and PBX services can be supported by the fiber-fed
distributed network without interfacing with the outside base station. For both the
outdoor fiber-fed microcellular systems, and the indoor ‘picocellular’ systems, the
feasibility of the fiber-optic approach is very sensitive to the cost of the fiber links. In
fact, the entire appeal o f using fiber-fed antennas versus construction of new base stations
is a potential savings in cost. Due to the cost-sensitive nature o f the wireless application,
emphasis will be placed on the tradeoff between device cost and performance in the
discussion below. This issue is typically not emphasized in traditional fiber-optic
applications such as the telephone and CATV industries because in these systems, high
performance fiber links are implemented such that the cost o f the link is shared among
many consumers. For fiber-fed microcells, cost effective deployment of many fiber-links
to provide blanket radio coverage is desired, and hence the device cost/performance
tradeoff must be addressed.
1.5
Network Benefits
Many o f the problems in the existing cellular systems can be improved considerably
by using multiple antennae within each microcell. In this case, a mobile user within the
cell is, on average, closer to an antenna, allowing a reduction in the transmission power.
With reduced power, the network has a shorter frequency re-use distance (which
accommodates more users per unit area), and the user gains a substantial increase in
battery lifetime (which reduces the size and recharge time o f the user's portable unit).
Although these advantages are very attractive, covering a cell with multiple antennae,
each with a full set o f radio and data multiplexing and demultiplexing gears, is very
9
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
costly. An alternative is to use "dumb" antennae which simply detect, amplify and
transmit the entire radio band, verbatim, via fiber optics to a central station within the cell
where all the channel multiplexing and demultiplexing are done. In this case, the cost of
setting up each antenna is greatly reduced, and one can afford to install more of them
within a cell. In fact, there is no reason for each unit cell to have its own central station with fiber links, multiple cells can share a central station where all the multiplexing and
demultiplexing gears are located. The verbatim transmission o f a radio signal is precisely
the subject matter o f high frequency optoelectronics and analog fiber-optic links - an area
which has seen much action in recent years in connection with CATV systems and
microwave phased-array antennae.
In addition to the above improvements, the fiber-linked distributed array concept
reduces the dynamic range requirements on the network, and facilitates centralized
network control. As an example of this dynamic range improvement, consider that one
omnidirectional antenna is located at each cell comer, with an additional omnidirectional
antenna at the center. Each omnidirectional antenna at the comer serves three cells which
intersect at that comer, and the antenna at the center serves only that cell. The cell has a
central station (CS) for processing the RF signals, which is linked by analog fiber-optics
to the 7 antennae serving the cell. Assume that a user is located somewhere in the cell and
is transmitting. In an urban environment, it has been shown that the received power at
each antenna is inversely proportional to the cube o f the distance between the user and the
antenna. The CS will receive all of these signals via fiber, pick the one with the strongest
signal for each user, and discard the others. The received power is reasonably uniform
(within 20 dB) in a large portion of area inside the cell, except when the user is very close
10
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
to an antenna site. When this happens, the close-in user (or users) "monopolizes" that
particular antenna, making that antenna unavailable for reception for users that are farther
away due to a finite dynamic range for the receiver electronics as well as the fiber-link.
However, since there are multiple antennae covering the cell, reception is not impeded
except in certain unlikely, unfortunate circumstances where all antennae are
simultaneously "monopolized" by close-in users. This results in a "lock-out" situation for
certain users. The higher the dynamic range of the antennae receiver and fiber-optic link,
the less likely for lock-out to occur, whereas its likelihood of occurrence increases with
the number of users in the cell. Therefore, based on the component and fiber-link
parameters as well as radio propagation characteristics, certain optimal cell size and radio
frequency allocations exist.
Another beneficial feature of the distributed-antenna concept is power-savings for the
mobile user. Power consumption is a paramount consideration in any mobile/portable
unit relying on battery power. The transmission power is reduced in the distributedantenna cellular network due to the closer proximity o f the users to the antennae. For a
cell with unit radius, the longest distance from any user to an antenna is 0.577. Since the
received power is inversely proportional to the cube o f the distance, we have a power
saving factor of 5 compared to a cell serviced with a single antenna.
In addition to the
dynamic range improvement, the central remoting o f the antennae allows flexible control
of the network from the central station (CS). For example, all of the "hand-offs" o f users
from one cell to another are easily facilitated from a central location. Also, dynamic
channel assignment (DCA) to the users from the central station can relieve the "hot"
traffic spots within a cell and reduce interference between users. Also, the CS can borrow
11
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
unused radio channels from low traffic cells and redistribute them to high traffic areas.
1.6
Commercial Analog Fiber-Optic Device Technology
The discussion of device technology will concentrate on the transmission o f analog
signals by the direct current modulation o f semiconductor lasers. The competing
technology is a system where a constant light source is sent into an external device where
the light transmission through the device is modulated. This external modulation has the
advantage that the optical light source can be located at a central location (like the base
station), and the remote antenna site can simply consist of the external modulator [7].
Direct current modulation, although it requires a semiconductor laser transmitter at each
antenna site, is typically a more simple and cost-effective technique.
There are a variety of commercially available analog optical transmitters based on
direct current modulation o f semiconductor lasers in terms of both price and performance.
The highest bandwidth commercially available device has a modulation bandwidth of
~20GHz and costs approximately $20,000. On the other hand, typical low bandwidth (12GHz) Fabry-Perot lasers can cost below $100 each in quantity. Although modulation
bandwidth is only one of several performance criteria, the point is that the
cost/performance range of commercially available devices is large, and even the lowest
cost devices have modulation bandwidths in the GHz frequency range. Given that most
present wireless services operate in frequency bands less that about 3GHz, there is a
strong motivation for the implementation o f low-cost, commercially available devices in
present wireless systems. This topic will be discussed below where the required device
12
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
performance for wireless systems is addressed.
1.7 Millimeter-Wave Optical Transmitters
The low-cost transmitters discussed above may be acceptable for the existing wireless
cellular systems that operate in the 800MHz-2.4GHz frequency range; however, as
demands for wireless services increase, the need for larger radio bandwidth will
inevitably drive wireless systems to operate at higher carrier frequencies, where the radio
bandwidth is more plentiful [8]. The detailed motivations and applications o f millimeterwave fiber optics will be addressed in a later chapter o f this thesis. The development of
low-cost, high-speed analog optical transmitters is a major research challenge. This
problem can be simplified by noting that high frequency radio communications can take
place in a relatively narrow bandwidth around a high frequency carrier. For example, a
100MHz radio band centered at 50GHz only occupies a 2% fractional bandwidth. With
this insight, it is promising to investigate narrow-band optical transmitters that operate at
high (>20GHz) frequencies for wireless communications.
A potentially low cost, narrow-band, high-frequency optical transmitter is based on
the technique o f resonant modulation [9]. A semiconductor laser, cleaved to a given
length (L), has a set of allowed optical modes that are separated in frequency by c/2nL,
where n is the effective index of refraction of the laser, and c is the speed o f light. By
properly choosing the length of the laser, this mode separation can be chosen to lie within
the millimeter-wave frequency range. For example, if n=3.5 and L = 850pm, then the
optical mode spacing is ~50GHz. By modulating the gain o f the laser (i.e. the current) at
13
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
this intermodai frequency, one observes a significant modulation o f the light output from
the laser. This phenomenon results from the coupling of the optical modes of the laser,
and is historically referred to as ‘mode-locking’. In terms o f modulation response, the
effect appears as a narrow-band resonance centered at the intermodai frequency. Figure
1.7.1 shows an illustration of the technique o f resonant modulation and a typical
measured modulation response at low and high frequencies. Note that for this device the
conventional low-frequency response is limited to approximately 1GHz, but the resonant
modulation effect produces a -100MHz bandwidth response at a center frequency of
~45GHz. There are several important points to be observed. First, this technique is a way
o f utilizing ordinary devices, with no special processing or design for high frequency
performance, to transmit high-frequency analog signals. Second, by cleaving the device
to the appropriate length, a narrow-band optical transmitter can be produced at an
arbitrary center frequency. These devices show good potential for the support of future
high frequency wireless systems.
14
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Laser
fcenter = *—T
2nL ®
center
03
-o
C/3
c
A
U
o
w"
0>
C*
-10
c
o
-20
Frequency (GHz)
Figure 1.7.1. Narrow-band, millimeter-wave optical transmitters based on the
technique of resonant modulation.
15
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1.8
Required Device Performance
A major factor in the cost of an analog optical transmitter is the required linearity o f
the optical modulation signal. This linearity requirement is typically specified by a twotone dynamic range measurement as shown in Figure 1.8.1. Two equal amplitude
sinusoidal signals at frequencies f, and f2 are used to modulate the link. The third order
nonlinearity o f the link introduces intermodulation distortion (IMD) resulting in output
signals at 2f,-f, and 2f,-f„ as shown in the inset o f the figure. This type o f distortion is o f
particular interest since the impairment lies ‘in-band’ near the signal channels of interest.
Plotting the output signal, IMD, and noise powers as a function o f the input signal power
to the link results in a two-tone dynamic range plot as shown in Figure 1.8.1.
Note that there is a range of input powers over which the output signal is above the
noise floor, and the distortion below the noise floor. When specified in a 1Hz noise
bandwidth, this range o f powers is called the two-tone dynamic range o f the optical link,
and is a figure of merit for the linearity o f the link. Also, the input power for which the
IMD equals the signal power is called the IP3 point, and is simply related to the dynamic
range discussed above. With this understanding o f the standard characterization of analog
optical transmitters, we can address the question: What is the required dynamic range o f a
fiber-optic antenna remoting link in a practical microcellular application?
16
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
DVD
S.-100
s
X**x xxx x xx~x
XX
-120
NiseFlcxr
IP3
-140
-80
-60
-20
Input RFRwer((Bi$
Figure 1.8.1. Illustration of a typical two-tone dynamic range measurement with
signal, noise, and intermodulation distortion (IMD) terms shown.
17
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In fact, for wireless systems, the two-tone specification o f link linearity greatly over­
estimates the required link performance [10]. This is because the two-tone specification
corresponds to the rare and unfortunate situation where two high power wireless users
happen to be assigned to adjacent frequency channels. More practical values of dynamic
range requirements can be obtained by performing a statistical study o f wireless access in
a microcellular environment. Figure 1.8.2 shows the results o f such a simulation based on
the AMPS wireless standard which allocates 30kHz o f bandwidth per user and requires
an 18dB carrier-to-interference ratio (C/I) for acceptable service. The details of this
simulation are discussed in a section later in this thesis. The model assumes that users are
randomly located between 5 and 300 meters from the microcell antenna site. Assuming a
standard (1/d)4 path loss results in a range of received powers at the antenna o f
approximately 70dB [11]. If the system designer were to stick to the ‘spur-free’ condition
imposed by the two-tone specification, the parameters above imply a required dynamic
range of ~120dB. By considering many configurations of users within the microcell, and
calculation of the resulting IMD introduced by the fiber link, one can determine the
percentage of calls that do not maintain the required 18dB o f C/I after the link. This
blocking probability is plotted as a function of the link two-tone dynamic range in Figure
1.8.2 for 5, 10 and 20 users within the microcell. Increasing the dynamic range of the link
clearly reduces the percentage of blocked calls. Also, increasing the number o f users
within the cell increases the required dynamic range to reach a given blocking probability
due to the increased number of intermodulation products.
18
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Fiber-fed Mcrocell Sundation Results
20 Channels
Two-ToneEXnanic Range (dB)
Figure 1.8.2. Simulation results for analog link performance in a typical
microcellular environment
19
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
We see that by accepting a very small probability o f blocked calls (0.5%) that the
required link dynamic range is significantly reduced from 120dB to the 90dB range. This
concept of allowing a small degradation in network performance in exchange for a 30dB
reduction of dynamic range requirement has substantial implications for the use o f lowcost optical transmitters in this application.
With this insight into the required dynamic range o f analog optical transmitters for
wireless applications, it is interesting to compare the performance o f the various analog
optical transmitters discussed above. Table 1.8.1 shows a comparison of the noise floor,
IP3 point, and two-tone dynamic range o f a typical CATV Distributed Feedback (DFB)
laser, a low-cost Fabry-Perot laser, and a millimeter wave optical transmitter based on
resonant modulation. Note the wide range of performance in noise, linearity, and dynamic
range between the devices.
High Performance CATV
Distributed Feedback (DFB) Laser
-155
+25
120
Low-Cost Fabry-Perot Laser
-115
+20
90
Millimeter-Wave Optical
Transmitter
Based on Resonant Modulation
-100
+5
70
Table 1.8.1. Comparison of the performance of various analog optical transmitters.
20
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
To achieve the absolute spur-free operation, the conventional CATV optical transmitter is
required. This conclusion is consistent with the approach taken by cellular service
providers in the past. By accepting the slight degradation in network service (0.5%
blocking probability, which is acceptable even for wired networks), the low cost
commercially available Fabry-Perot lasers may be appropriate for use in both outdoor
microcellular and indoor picocellular systems. For the millimeter-wave optical
transmitters based on resonant modulation, the dynamic range is very modest by
conventional CATV measures. Device structures and techniques to improve the dynamic
range o f these transmitters is a significant future research challenge.
In summary, low cost analog fiber-optic networks must be developed to meet the
rapidly expanding demand for wireless communication services. This effort involves the
design o f networks that can be supported by low cost, commercially available devices,
and the development narrow-band, high-frequency optical transmitters for future wireless
systems. These technologies are expected to have a significant impact on the realization
o f future wireless personal communication networks.
21
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2
Motivation and background
2.1
Why distribute antennas ?
This section examines from a broad perspective the role o f a low-cost antenna system
for providing wireless service. The optimal solution for supplying both coverage and
capacity is found to be a distributed antenna system - mini-base station hybrid
architecture. In this scheme, the network inherently allocates channel capacity as needed
when user traffic patterns change over the course o f a day. Other benefits include:
completely uniform radio coverage, the need for handoff is eliminated, network control
and maintenance are centralized, and the network is easily upgraded.
2.1.1
Introduction
As consumer demand for wireless communication services continues to grow, service
providers are quickly developing and deploying wireless networks with more capacity
and functionality than ever before. The users o f new wireless systems demand several
features from the network: (1) high quality of service, (2) ubiquitous coverage, and (3)
small handsets with a long talk time. As discussed in [12], the current wireless systems do
not meet all of these requirements, and are in the process of evolving to meet consumer
demands. For example, cordless telephone systems provide good quality voice service
with small, low-power handsets, but do NOT provide ubiquitous coverage. Typical
systems allow the user to roam within the vicinity o f a localized base unit, with no
handoff between base units. Some cordless systems (e.g. the SpectraLink PCS 200) allow
the user to roam within a building or a convention center using a portable phone that is
22
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
only functional within that particular building. Again, this system falls short o f ubiquitous
coverage where each individual has a single “one person - one number” communicator
that can be used anytime at any location.
In contrast, the cellular telephone service providers have a wide scale network,
capable of providing wireless telephony over urban and residential areas. Although they
provide wide area coverage, the current cellular systems suffer from poor local radio
coverage or “dead spots.” Typical dead spots include: office buildings, shopping malls,
hotels, hospitals, and tunnels. Furthermore, even in areas that are partially covered, the
mobile is required to transmit high power levels resulting in a reduced talk time per
battery charge. An additional problem with cellular is that many calls are blocked during
peak hours when “all circuits are busy” due to limited voice channel capacity. It is clear
that to meet the new demands o f wireless consumers, cellular providers and the new PCS
providers must build out a “low-tier” network that provides both radio coverage and
channel capacity on a local level. The most important area where this low-tier service
must be provided is inside buildings, since this is where people spend most of their time.
Although most in-building environments have wired telephone service, wireless
customers demand the ability to send and receive calls while they are away from their
desk, in a meeting, or visiting another building. Only the in-building wireless systems
discussed below can provide this type of service. Unfortunately, the in-building
environment is also the most challenging area to provide radio coverage due to severe
attenuation and multi-path effects. Also, the potentially high density of users within a
typical office building require in-building cellular systems to provide channel capacity to
these users.
23
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
From the above discussion, it is clear that present and future cellular and PCS systems
will need methods of providing both coverage and capacity for in-building environments.
To understand the various methods for providing this in-building service, the first floor of
the electrical engineering building at the University o f California at Berkeley will be used
as an example. For this floor, given a number o f anticipated subscribers, and by assuming
standard traffic and usage statistics, one can calculate the number o f voice channels (N)
required to serve this entire floor with a low call blocking probability. Once this is done,
one proposed solution to the in-building coverage problem is to simply install a “mini­
base station” with N cellular channels to cover the floor. This solution is illustrated below
in Figure 2.1.1. Note that since cellular bandwidth is scarce, the decision to allocate N
channels to a particular building is not taken lightly by the cellular provider, and in many
instances the system designer may need to reduce N due to frequency allocation issues.
The term “mini” is used to distinguish this base station from the large outdoor
base stations that are used to cover many square kilometers. As shown in Figure 2.1.1,
although the mini-base station provides the N channels of capacity that are required for
the users on the floor, the radio coverage range is limited by the portable transmit power
and the thermal noise of the front-end receiver o f the base station. Note that the
fundamental issue in solving the radio coverage problem is dealing with the wide
dynamic range of received powers at both the portable and the base station. Simply
increasing the transmit power does NOT solve this problem since far away users are
covered at the expense of saturating users that are close to the base station. Extensive
radio measurements have been made on this particular floor, and it has been determined
that for typical mobile transmit powers (+20dBm), one antenna site is NOT sufficient to
24
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
cover the floor. In this system, a minimum o f one base station per floor is required to
cover the entire building, and more may be required to completely provide radio coverage
on each floor.
Areas of poor
or non-existent
radio coverage
Mini-base station
N Channels
Good coverage
and capacity
Figure 2.1.1. In-building coverage and capacity can provided to some, but not ail of
the floor of the building by using a single mini-base station.
To circumvent the limitations of the single mini-base station approach discussed
above, another approach is to install multiple mini-base stations, each having capacity
N/4 as illustrated in Figure 2.1.2. Note that in this architecture, the mini-base stations are
adjacent to one another, and therefore cannot share the same frequency spectrum due to
interference between the base stations. In fact, it has been found experimentally that there
should be a minimum o f two floors between areas where frequencies are re-used [13].
Also, for cellular base stations, the cost o f the station is approximately proportional to the
number of channels that the station provides [14] since each channel requires a separate
radio frequency transceiver. Therefore, if multiple base stations are installed per floor, the
25
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
number o f voice channels per base station must be reduced to maintain a cost-effective
solution. Therefore, if the number o f frequency channels allocated to the building is fixed
(N), each o f the four base stations must have an N/4 voice channel capacity.
D isadvantage of approach:
R educed local capacity
Mini-base
station '
N/4 C hannels
Good —
coverage
Figure 2.1.2. Installing multiple mini-base stations can provide complete coverage at
the expense of reduced local capacity.
This architecture has complete radio coverage and the same total channel capacity (N) as
the single base station; however, the drawback to this scheme is that the local capacity of
the network is compromised. Although N channels are still used in the network, the
traffic-handling capacity is much less, as will be seen in the example below. Consider the
consequence when most of the people on the floor move to one section o f the floor. For
example, during lunch many employees in an office building may go to the cafeteria, or
during a conference break many attendees may go to the main hotel lobby. In this case,
the base station serving this portion o f the floor will experience call congestion and call
blockage, while the capacity o f the other base stations on the floor go unused. Also,
26
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
although the mini-base stations are capable o f handoff from one base station to another,
the additional required network control increases the cost and may be something that the
system designer would rather avoid.
A new trend for in-building wireless systems is to solve both o f the above mentioned
problems by using a hybrid architecture that implements both a mini-base station and
distributed antennas. Consider the advantages of using a single base station architecture
in conjunction with a distributed antenna network as shown in Figure 2.1.3. This
architecture provides the ideal solution to the coverage/capacity problem. The base
station provides the voice channel capacity, and the distributed antenna network provides
the radio coverage. In this architecture, the capacity of the centralized base station is
automatically allocated to various locations in the building as user traffic patterns change
over the course o f a day. There are no handoffs required since the entire floor is served by
one base station, the services of the network are simple to upgrade, and the network
control and maintenance are centrally located.
Ideal solution: Install distributed antennas
Provides coverage and dynamic capacity allocation
Lowcost
distributed
antennas
Figure 2.1.3. The ideal performance/cost solution to providing in-building coverage.
27
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
For example, the Motorola INReach™ system implements this architecture, and typically
implements a single 24 channel mini-base station, and a distributed antenna network to
distribute the radio signals to various locations in the building. Given the advantages of
this architecture, it is likely that both distributed antenna networks and mini-base stations
will play a major role in future in-building cellular systems. The advantages o f the
distributed antenna - mini-base station approach are summarized below:
Advantages of the distributed antenna approach:
•
Base station capacity is dynamically allocated as needed.
•
Provides completely uniform radio coverage at the lowest cost.
•
No channel handoff is required as the user roams within the coverage area.
•
The base station services are easily modified and upgraded.
•
Maintenance and control o f the network are centralized.
2.1.2 Illustrative example
In order to illustrate the significance and performance advantages associated with
centralizing radio capacity, and using a distributed antenna system, the following
example is considered. Assume that a 10 story building requires wireless service for a
total o f 100 subscribers. A traffic level o f 0.1 Erlangs per user is assumed. We wish to
compare the system requirements under a variety o f scenarios.
28
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(1) Distribute capacity.
Assume that one mini-base station is required to cover each floor o f the building, for a
total of 10 mini-base stations. The traffic that each mini-base station must handle depends
on the assumed user mobility. If users are uniformly distributed at all times (i.e. 0%
mobility) then each base station must handle 10 users, or 1.0 Erlangs of traffic, which
requires 5 radio channels per base station. However, if users are 50% mobile, then each
base station must handle 55 users or 5.5 Erlangs o f traffic, which requires 12 radio
channels per base station. Finally, if users are 100% mobile, each base station must
handle all 100 users, or 10 Erlangs of traffic, which requires 18 radio channels per base
station.
(2) Centralize capacity and use a distributed antenna system.
In this case the single, centralized base station must handle the traffic o f all 100 users,
or 10 Erlangs which requires 18 radio channels. Note that this solution inherently
provides 100% mobility for the users in the network. The results are summarized in Table
2.1.1 below.
Scheme
Mobility (%)
# of base
stations
# o f channels per
base station
# of total radio
channels
Distributed Base Stations
0
10
5
50
Distributed Base Stations
50
10
12
120
Distributed Base Stations
100
10
18
180
Centralized Base Station +
distributed antennas
100
1
18
18
Table 2.1.1. Summary of illustrative comparison between distributing radios vs
distributing antennas and centralizing capacity.
29
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
Clearly, the distributed antenna scheme is the superior solution, since it provides 100%
mobility with less than half the channels required to provide 0% mobility in a distributed
radio system.
2.1.3 Technical Analysis
This section provides a technical analysis, comparing the relative merits of
distributing mini-base stations, versus centralizing capacity and distribution antennas.
The distributed mini-base station approach has become feasible due to advances in IC
fabrication, which enables complex switching and channel-handoff functions to be
performed in a relatively compact and inexpensive module which can be distributed in a
building without excessive demand in real-estate or physical obstruction. Which o f the
two approaches, the “Distributed Antenna” (D.A.) approach, or the microcellular minibasestation approach (B.S.), yields the greater cost-performance benefit, will be
determined both by fundamental issues in traffic engineering, and (obviously) the relative
costs in implementing these systems. The purpose o f this section is to examine and
quantify some of these issues, and to show that in almost all practical circumstances, a
hybrid architecture consisting o f a D.A. system and a small number o f mini-basestations,
will yield significant performance advantages over a pure distributed B.S. microcellular
architecture.
Referring back to Figure 2.1.2 and Figure 2.1.3, to quantify the cost/performance
advantages of the distributed antenna network, we consider the ideal situation in which
traffic is evenly distributed over the floor at all times. The advantages will be even greater
30
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
when traffic patterns change over the course o f a day, since the capacity of the centralized
B.S. can be automatically allocated to various locations as need, an option NOT available
to distributed mini-B.S. We assume that the radio coverage is identical for the antenna
located on a mini-B.S. and an antenna o f a D.A. system. The need of radio-coverage
adequacy requires subdivision o f the floor into M such coverage areas (in the example in
Figure 2.1.3, M=4). Let the total traffic load be A Erlangs for each o f the M coverage
areas. The number o f channels N(A) required to handle this traffic with, say, a 1% grade
o f service, is given by the well-known Erlang-B formula. This is the number o f channels
required o f each mini-B.S. in the distributed mini-B.S. microcellular approach o f Figure
2.1.3. On the other hand, if each coverage area is served by an antenna in a D.A. system
which then routes the traffic of the entire floor to a central basestation, the traffic load
presented will then be M*A, and the number o f channels required, N(M*A) is again
given by the Erlang-B formula, for the same grade of service of 1%. We make the
simplifying assumption that the cost o f the mini-B.S. is proportional to the number of
channels equipped: the amortized cost per channel is Cchanncl , while the amortized cost
(including installation) per antenna in a D.A. is Cantenna . The total cost of the distributed
mini-B.S. system is then M*N(A)*Cchannel while that o f the D.A. system is
N(M*A)*Cchannd + M*CinlCTna . The cost ratio o f these two architectures, which provide
identical performance under ideal uniform traffic situation, can be expressed as:
Cost(D.A.) / Cost(mini-B.S.) = (N(M*A) + q*M)/(M*N(A))
where (q) is the cost ratio Cantenna / Cchannel . The total system cost-ratio is plotted in the
figures on the following two pages for number of coverage areas 5, 10, 20 and 40
31
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
respectively. All plots are against traffic (A) per coverage area, for different values o f r\.
Values for t| o f 1, 0.5, 0.2, and 0 are illustrated by the colors blue, green, black, and red
respectively.
As an example, consider a 10-story building with occupancy o f 100 persons per
floor. Using a cellular user penetration rate of 20%, and a standard traffic o f 20
mErlangs/user, the total traffic per floor is 0.4 Erlangs. Assume that each floor constitutes
a coverage area, so that M=10. From the figure, the cost advantage o f the D.A.
architecture over that of distributed mini-B.S. is a factor o f 3 for tj = 0 (i.e., the antenna
costs nothing) and a factor of 1.6 for r\ =1 (an antenna costs as much as a mini-B.S.
channel). At present, a 4-voice channel mini-B.S. costs approximately S15K, while the
amortized cost per antenna of a D.A. system is around $750, resulting in
tj
=0.2. The
present cost saving o f implementing a D.A. system over that o f a distributed mini-B.S.
system is a significant factor of 2.8.
32
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
Cost(DA) / Cost(mini-B.S.)
1.0
0.8
r
0.4
0.2
M=5
1.5
0.5
Traffic / coverage area (Erlangs)
2.0
1.5
0.5
Traffic / coverage area (Erlangs)
2.0
Cost(DA) / Cost(mini-B.S.)
1.0
0.4
0.2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0.5
1.0
1.5
Traffic / coverage area (Erlangs)
0.5
1.0
1.5
Traffic / coverage area (Erlangs)
34
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
Several interesting facts can be gleamed from these simple calculations:
(1) The cost savings o f D.A. tends to be higher for larger buildings (large M) and for
systems implemented for lower traffic scenarios, with the implication that D.A.’s are
particularly attractive for buildings that has a high priority on coverage but which
phone conversations tend to be short - hospitals and factories are prime examples;
(2) The cost o f mini-B.S.’s is expected to drop in the next few years with the volume
buildup o f IC manufacturing capacities, but according to Fig. 4, even with a 5-fold
drop in price of mini-B.S. (and assuming the cost of D.A. stays put, an unlikely
scenario), D.A. systems still hold a cost advantage of between 1.5 to 2 for medium
and large size buildings (M>10) and for a majority of traffic conditions.
The “uniform-traffic” assumptions used above actually gave undue advantage to the
distributed mini-B.S. approach, since dynamic allocation o f channels is not possible. In
addition, there are no handoffs required since the base station is centrally-located. The
services of the network are simple to upgrade, and the network control and maintenance
are centrally located.
In summary, for in-building radio environments that require cellular radio
coverage and capacity, the optimal architecture is a combination o f a mini-base station
and a low-cost distributed antenna system. This architecture provides dynamic allocation
o f radio channels, low-cost radio coverage, requires no channel handoff, and facilitates
the centralized maintenance and control of the network.
35
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2.2
Why use analog fiber optics?
There are multiple advantages o f using analog fiber optics in wireless systems.
Optical fiber is light-weight, low-cost, immune to RF interference, and can transmit
signals over many kilometers with minimal loss. The real question is whether the wireless
signal should be transmitted over the fiber in analog or digital format. O f course the
benefit o f analog signal transmission is that the RF signal is transmitted “verbatim” from
point A to point B using the optical fiber. This type of system is signal frequency and
format independent in that any wireless signal, modulated in any way, can be transmitted
over the analog fiber link. Furthermore, the remote antenna site is very simple since it
consists only o f the physical antenna, an amplifier and the analog fiber-optic transceiver.
In comparison, there are two methods by which the signals could be transmitted in
baseband digital format. First, the voice signals could be de-modulated from the RF
carriers at the antenna site and the resulting baseband signal could be sent over the fiber.
This technique has two disadvantages: (1) the down-conversion and de-modulation
electronics at the antenna site are costly and consume electrical power, and (2) the de­
modulation process is format specific, hence separate electronics are required for each
and every wireless protocol that are needed to be sent over the fiber link. Second, the
entire wireless bandwidth can be digitized using a high speed A-D converter, and the
resulting digital bit stream sent over the optical fiber. This scheme also has two
drawbacks: (1) Again the high speed A-D converters are costly and consume much
electrical power, and (2) The resulting digital bit stream is very high speed, and may
exceed the direct modulation bandwidth o f the fiber-optic transmitter. Consider the
36
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
following example:
The PCS wireless band in the U.S. is 60MHz wide. To prevent aliasing with perfect
sampling one needs to sample at least two times the bandwidth. In practice, one needs to
sample 4-5 times the bandwidth due to non-ideal filtering and sampling. Furthermore,
most wireless antenna applications require at least 70dB o f dynamic range. For the A-D
conversion, this corresponds to a resolution of 12 bits per sample. Hence, the total serial
digital bit stream is 60*5*12 Mb/s = 3.6Gb/s. Although there are digital fiber-optic links
that can handle this bit rate, one would do just as well to implement the analog link and
avoid the need for all of the complex filtering, down-conversion, sampling, and digital
signal processing required in order to transmit the signals in digital form.
In summary, the use of analog fiber optics in wireless systems offers a number of
performance and practical advantages over competing schemes. The next section
discusses the motivation for transmitting millimeter-wave signals over the fiber-optic
cable.
2.3
Millimeter-wave Fiber-optics
The advantages of using analog optical fiber links for the remote transport of microwave
and millimeter-wave (mm-wave) signals in military phased-array antenna systems are
well-established [15-16]. The low loss, light weight, immunity to EMI, high bandwidth,
and ease o f installation make optical fiber a preferable alternative to metallic waveguidebased media for applications such as phased-array antennas, remote synchronization of
antenna stations, and as a connecting infrastructure for indoor microwave and mm-wave
37
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
picocellular and cable-television wireless networks. Figure 2.3.1 illustrates a typical
application in which many remote antenna sites are fed via low-loss optical fiber links
from a central location.
Continuous RF
Phase Control
> 1 km
✓ 1T
Tx
±
Tx
mm-wave
source
±
TTD n o t req u ired
for n arro w b an d
Figure 2.3.1. Narrowband phased-array antenna network illustrating the versatility
of an optical transmitter capable of mm-wave transport
The use of fiber-optic delay lines and integrated photonic devices allow for compact,
true-time delay o f the RF signals, and hence a means for accurate beam forming and
steering. One area of significant research is exploring the use o f these millimeter wave
frequency band opto-electronic devices for wireless applications. In particular, the
development of low-cost analog fiber-optic links capable of efficient transmission in the
millimeter-wave frequency range has direct application in both commercial and military
markets as will be discussed below.
A particularly exciting aspect o f this work is its direct application in commercial
markets such as wireless cellular communication systems [17-19], and wireless mm-wave
cable-television distribution [20,21]. Existing cellular and personal communication
38
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
systems at 900 MHz and 1.8 GHz are severely limited in bandwidth, however the demand
for high bandwidth wireless services continues to increase. Technological advances must
be developed to meet the demands o f this growing market for personal communications.
Moving to the under-utilized mm-wave region provides the bandwidth needed to support
a large number of users and high-bandwidth services. Due to the limited propagation
distance of mm-waves, however, uniform radio coverage can only be provided by
employing a network of strategically placed distributed antennas. Each antenna extends
radio coverage over a small area, or pico-cell. Frequency reuse is increased due to the
naturally confined picocells created by the limited propagation distance o f mm-waves.
The mobile user’s battery power is also conserved due to the small coverage areas. It is
envisioned that a mm-wave distributed antenna network can lead the way towards the
proliferation of low power, hand held transmitters capable o f wireless voice, video, and
data communications anywhere. Fiber will play a key role in these systems, providing an
interconnecting infrastructure for the antennas. As an example, consider the in-building
mm-wave distributed antenna network illustrated in Figure 2.3.2 below.
antennas can be placed anywhere —in corridors, in offices and elevator shafts.
39
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
MM-wave
Palmtop computer
or p h o n e ^ ^
Optical fiber
backbone for
----mm-wave transport
Figure 2.3.2. Millimeter-wave wireless personal communications network
illustrating an analog fiber-optic interconnecting infrastructure.
Signals coming from user handsets are received by the antennas and transported to a
centralized location (e.g. basement) using the narrowband mm-wave fiber-optic links.
Use o f mm-wave fiber-optic links for the remote delivery o f signals to and from these
antenna sites allows centralization o f the mm-wave hardware and demultiplexing
electronics. Centralized processing simplifies the network control and allows for
implementation of protocol to improve the network performance. Note that since a large
number o f antennas are employed, the cost o f the mm-wave fiber-Iinks must be low. The
resonant modulation technique, in which an ordinary low-bandwidth device is used to
transmit mm-waves, is an excellent candidate for the practical implementation o f such
systems. The ability to perform mm-wave RF phase control in the system illustrated in
Figure 2.3.2 above also yields significant advantages. By phasing the signals from
various antennas, the fading characteristics of the radio environment in which the
40
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
antennas are deployed can be controlled. Thus, the presence o f a deep fade in the middle
o f the corridor, for example, can be “pushed” to the wall by adjusting the phase of the
mm-wave optical transmitters.
Commercial interest in use of the mm-wave band for personal communications is
growing in the world wide market. Production o f short-haul mm-wave transceivers is
expected to reach 78,000 units in 1997, with a compound annual growth rate o f 29%.
Millimeter-wave fiber links are expected to play a crucial role as a connecting
infrastructure for these transceivers. In Japan, the 59.0-60.0 GHz has been allocated as an
"experimental band" for personal communications. A Japanese company, ATR Optical
and Radio Communications Research Laboratories, is aggressively developing fiber-optic
technologies based on external modulators in the mm-wave frequency band for wireless
communication applications [19]. In England, British Telecom is exploring the use of the
38 GHz frequency band. Growing interest in the United States in mm-wave wireless
systems has led to several requests for FCC licenses. Table 2.3.1 lists a representative
sample of several companies that have requested a license to operate and test at mm-wave
frequencies.
41
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Corporation Granted FCC
license
Telesis Technology Lab
(formerly Pacific Telesis Group)
LDH International
Time Warner Cable Group
Goeken Custom Communications,
Inc.
Personal Communications
Network Services o f NY Inc.
SM Tek
Vicinity
San Francisco, CA
Chicago, IL
Dallas, TX
New York, NY
Los Angeles, CA
Greenville, NC
Denver, CO
Atlanta, GA
New York, NY
St. Petersburg, FL.
Cincinnati, OH
Columbus, OH
Milwaukee, WI
Chicago, IL
New York, NY
Milwaukee, WI
Chicago, IL
Frequency Band
Requested
37 - 39.6 GHz
27.5-28.14 GHz and
28.5-29.14 GHz
17.7-19.7 GHz
27.5-28.14 GHz and
28.5-29.14 GHz
17.7-19.7 GHz
38.7-38.9 GHz and
58.0-58.3 GHz
Table 2.3.1. FCC licenses that have been granted to corporations interested in
operating and testing at mm-wave frequencies.
The goal of the optical transmitter is to transport the signal over distances > 1 km
with high fidelity. The use of optical fiber for the transport o f narrowband mm-wave
signals provides significant advantages over metallic waveguide-based media. The low
loss, light weight, immunity to electromagnetic interference, high bandwidth and ease o f
installation make fiber an excellent medium for applications such as phased-array
antennas, remote synchronization of antenna stations, and as an interconnecting
infrastructure for indoor mm-wave picocellular and cable-television wireless systems. In
42
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the narrowband mm-wave phased-array antenna system illustrated above, the ability o f
the optical transmitter to transport mm-wave signals eliminates the need for bulky
waveguides, profoundly impacting system cost and complexity. It is therefore o f
immense interest to develop novel optical transmitters capable o f efficient optical
modulation at mm-wave frequencies.
43
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3
3.1
Dynamic range requirements for optical transmitters in fiberfed microcellular networks.
Introduction
The use o f analog fiber-optic links as the connecting infrastructure in wireless
networks has recently been proposed [22]. Future wireless systems must provide uniform
radio coverage to spatially distributed mobile users in a cost effective manner. Small
(radius~300meters) radio microcells can serve a high density o f users, and require low
user handset transmit power compared to large (r~lkm ) macrocells in existing systems. A
microcellular network can be implemented by using a fiber-fed distributed antenna
network as shown in Figure 3.1.1.
44
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
O utdoor
Fiber-fed
A ntenna
Network
Analog
Fiber
Links
Base
Station
Mobile
Users
Rem ote Antenna
Optical Fiber
To B ase Station
Figure 3.1.1. Fiber-fed distributed antenna wireless network using analog fiber
links.
The received RF signals at each antenna site are transmitted over an analog fiber-optic
link to a central base station where all demultiplexing and signal processing are done. In
this way, each remote antenna site simply consists o f a linear analog optical transmitter,
an amplifier, and the antenna. The cost of the microcell antenna sites must be greatly
reduced before the deployment of these networks is practical. The required dynamic
range o f the analog optical transmitter is a major factor in cost. Previous analysis [3] on
dynamic range requirements assumed an absolute spur-free condition for each FDM
channel, and resulted in a link dynamic range o f >100dB (1Hz). This chapter investigates
the dependence of this dynamic range on the number of voice channels, the density of
45
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
antenna coverage, and the network protocol. For a single antenna serving a cell, it is
shown that by accepting a small (<0.5%) call blocking probability, a modest dynamic
range of 91dB for 20 FDM voice channels is required. Furthermore, it is shown that by
using multiple fiber-fed antennae per cell, and employing an appropriate network
protocol, a 73dB dynamic range link can provide acceptable network performance.
3.2
The system model
The dynamic range of fiber-optic links is commonly specified by the range o f input
powers over which the output signal is spur-free for a two-tone input. In a wireless
system, this two-tone specification under-estimates the link performance since it
corresponds to the rare, unfortunate situation when two high power users are assigned to
adjacent frequency channels. To determine practical values for the required dynamic
range of optical links, a statistical simulation of user access in a wireless microcell is
performed. The model is based on the AMPS cellular system which uses FDM for
multiple access, requires an 18dB carrier-to-interference (C/I) ratio, and allocates 30kHz
of bandwidth per channel. A standard model for multi-path environments [23] is
employed in which the received RF power varies a (1/d)4 where d is the distance from
the antenna to the user. Experimental measurements in a typical line-of-sight urban
microcellular system [5] have demonstrated that this model empirically provides a lower
bound for the received power including local fading effects o f the signal. It is also
assumed that the user are only allowed to within 5 meters o f each antenna site, which
leads to a maximum RF power variation of ~70dB for a 300meter microcell. Although
46
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
the effects o f shadowing are not included in the model, existing mobile handsets typically
have some limited power control capability which can compensate for shadowed
environments. A series o f simulations are performed in which users appear randomly in
the cell, and the spectrum of received RF powers at each antenna is amplified by an
automatic gain control (AGC) amplifier that maintains the weakest signal to 18dB above
the noise floor o f the link. Next, the intermodulation distortion (IMD) terms from the
optical transmitter at each channel are calculated as:
r
IMD = y
\
Z
? pi +
\2 a > l - 0 ) j
Z4W*
Q)t + Q ) j - Q ) k
Where the summations are over all the combinations of ©;, ©j, and ©k that fall on the
given channel and the proportionality constant is determined by the two-tone dynamic
range o f the link. Figure 3.2.1 shows a typical spectrum of output signal powers and the
resulting intermodulation distortion (IMD) terms from two of the antennas shown in
Figure 3.1.1.
47
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
o
1
'
1
□
Signal Pow er
EZI IMD Pow er
-10
E
CD -20
ID
-30
IO
■“
•40
CL -50
■Q.
+*
3
o
-60
-70
-80
-90
5
D15
10
Channel Num ber
D
20
Figure 3.2.1. Typical spectrum of received signal and IMD powers at antennas A
and B shown in Figure 3.1.1. For this plot, the transmitter dynamic range is 70dB
(1Hz).
After the link, any channel that does not meet the minimum 18dB C/I ratio is counted as a
blocked call. This process is repeated until the product o f the number of runs and the
number of voice channels equals 10s. The average percentage o f blocked calls is then
calculated as a function o f the link two-tone dynamic range.
First, we consider a circular microcell (r=300m) with one fiber-fed antenna covering
the cell. Figure 3.2.2 shows the average blocking probability as a function of the link
two-tone dynamic range for 5, 10, and 20 available FDM voice channels.
48
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5
0s
>»
a
(0
20 Channels
4
n
3
O)
2
o
c
u
o
Spur Free
-120 dB
1
CD
0
70
75
80
85
90
95
Two-Tone Dynamic Range (dB)
Figure 3.2.2. Percentage of blocked calls as a function of transmitter dynamic range
for a single fiber-fed 300m radius microcell for 5,10 and 20 voice channels.
As expected, the number of blocked calls decreases with increasing link performance.
The dotted line in the figure corresponds to a relatively small blocking probability of
0.5%. The link performace required to achieve this blocking level increases with the
number of channels due to the increase in the number o f intermodulation products. For 20
channels, the criteria is met for a modest link dynamic range o f 91dB. Next, we consider
multiple antennae within an 1800m square cell. When using these multiple antennae, the
base station must decide how to assign users to a particular antenna. For example,
referring to Figure 3.2.3, channel 1 has a higher C/I at antenna A while channel 2 has a
higher C/I at antenna B.
49
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Antenna A
O
B ase
Station
Signal
Antenna B
IMD
i Maximum signal protocol: C hoose A
i Maximum C/I protocol: C hoose B
Figure 3.2.3. Illustration of the maximum signal and maximum C/I protocols.
We consider two different protocols for processing the signals from the multiple
antennae. First, we consider that each channel is assigned to a particular antenna based on
the strength of its received signal power. Second, we consider that each channel is
assigned based on the C/I o f that channel. This C/I protocol is more difficult to
implement than the maximum power protocol since it requires a measurement of both the
signal and the interference power on each channel. Figure 3.2.4 shows the blocking
probability with 4 and 9 antennas per cell for both o f these protocols.
50
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
— 4 an ten n as
—9 a n ten n as
Maximum C/I
Protocol \
§!5
Maximum Signal
Power Protocol
(0
.a
2
o.
o>
e
u
0.5%
o
CD
70
75
80
85
90
95
Two-Tone Dynamic R ange (dB)
Figure 3.2.4. Percentage of blocked calls as a function of transmitter dynamic range
for 4 and 9 antennas and two different signal protocols. This simulation considers
an 1800m square area with 20 available voice channels.
Notice the dramatic influence o f the network protocol on the optical transmitter
requirements. For the maximum signal protocol, there is not a significant difference
between using 4 and 9 antennas, and the link dynamic range requirement remains in the
~90dB range. This can be understood by realizing that the maximum signal protocol is
equivalent to dividing the cell into several smaller cells. In these smaller cells, the
variation in received RF power is reduced; however, the probability that a user will
saturate an antenna is increased since there are more antennae in the cell. In order to take
advantage o f the spatial diversity o f the distributed antenna network, the maximum C/I
protocol must be implemented. These simulations indicate that the required dynamic
range is reduced to 78dB for 4 antennae and to 73dB for the 9 antennae case. With this
protocol implemented, when a close-in user monopolizes a particular antenna, the other
51
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
users in the vicinity of that antenna can be “picked-up” by one of the other antennas that
is not saturated. In this case, the network performance is only very poor for the rare
situation when all of the distributed antennas are saturated by close in users. Even further
reductions in the required link dynamic range can be obtained by taking advantage of
dynamic channel assignment (DCA) strategies which allocate user channels such that the
generated intermodulaiton products discussed above are minimized.
To put in perspective the modesty of these dynamic range requirements determined
from the above analysis, Figure 3.2.5 shows the experimentally measured two-tone
dynamic range of a Sharp compact disk (CD) laser.
JJJL L
900MHz
66dB (10 kHz) - 92.7dB (1 Hz)
-60
-40
-20
0
Input RF P ow er (dBm)
Figure 3.2.5. Measured two tone dynamic range of a self pulsating CD laser at
900MHz. The inset (lOdB/div) shows the two tones centered at 900MHz and the
induced intermodulation products.
Operating at an optical power o f 6.9mW without an optical isolator, the dynamic range at
900MHz is 92.7dB. The inset shows the two tones centered at 900MHz and the induced
52
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
intermodulation products. This measurement may have significant impact in the use of
very low cost optical transmitters in the fiber-fed microcellular networks discussed above.
In conclusion, the dynamic range requirements of optical transmitters used in fiberfed distributed antenna networks are determined using a statistical model for the number
o f dropped calls due to the generated IMD products. A single antenna microcell fed with
a 91dB link can support 20 FDM channels with a dropped call probability of 0.5%. By
covering a cell with multiple antennas, and implementing a optimum protocol at the base
station, the link linearity requirement can be reduced to <80dB. These results may have
significant implications in the practical implementation o f next-generation microcell
personal communication networks. A related problem involving the distortion of power
amplifiers was considered in [24] for two users in a single antenna cell.
3.3
Matlab Simulation
%Program to simulate mobile user access in a fiber-fed wireless network
function y = npowerl6
global freq %number of available frequency channels
freq = 20;
load twenty %load the file containing the IMD products
numrun = 1000;
%number of runs to consider
%antenna data
numant = 4; %number o f antennas in the given area
antrad = 5; %antenna radius (in meters)
si = size(IM l,l); %number o f 2W1-W2 IMD terms
s4 = size(IM4,l); %number o f W1+W2-W3 IMD terms
cell = 2400;
%square outer cell size (in meters)
53
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
%power and blocking calculation data
cnr =18; %required carrier-to-interference ratio
bw = 30000; %per channel bandwidth (Hz)
DR = 90;
%link two-tone dynamic range (dB-Hz2/3)
t = 4;
%power law dependence o f fields
%generate locations o f antennas
xant(l) = cell/8; yant(l) = cell/8;
xant(2) = 3*celI/8;
yant(2) = cell/8;
xant(3) = 5*cell/8;yant(3) = cell/8;
xant(4) = 7*cell/8;yant(4) = cell/8;
xant(5) = cell/8; yant(5) = 3*cell/8;
xant(6) = 3*cell/8;yant(6) = 3*cell/8;
xant(7) = 5*cell/8;
yant(7) = 3*cell/8;
xant(8) = 7*cell/8;yant(8) = 3*cell/8;
xant(9) = cell/8; yant(9) = 5*cell/8;
xant(10) = 3*cell/8; yant(10) = 5*cell/8;
xant(l 1) = 5*cell/8; yant(l 1) = 5*cell/8;
xant(12) = 7*cell/8; yant(12) = 5*cell/8;
xant(13) = cell/8; yant(13) = 7*cell/8;
xant(14) = 3*cell/8; yant(14) = 7*cell/8;
xant(15) = 5*cell/8; yant(15) = 7*cell/8;
xant(16) = 7*cell/8; yant(16) = 7*cell/8;
subcell_width = cell/4;
%center of a subcell
xcenter = subcell_width/2;
ycenter = subcell_width/2;
alpha = 2*(((3/2)*DR) - 10*logl0(bw) - cnr);
dropl = 0; %number o f dropped calls
drop2 = 0;
drop3 = 0;
A = zeros(freq, numant); %matrix that holds received powers
H 1 = zeros(freq+1,1);
%matrix that holds data for bar graph
H2 = zeros(freq+l,l);
H3 = zeros(freq+l,l);
for run = 1:numrun
%generate matrix of received powers
54
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
for i = lrfreq
z = 0;
while z < antrad
xx = subcell_width*rand;
yy = subcell_width*rand;
z = sqrt((xx-xcenter)A2+(yy-ycenter)A2);
end
subcell = ceil(numant*rand);
if subcell = 2
xx = xx + subcellw idth;
elseif subcell = 3
xx = xx + 2*subcell_width;
elseif subcell = 4
xx = xx + 3*subcell_width;
elseif subcell = 5
yy = yy + subcell_width;
elseif subcell = 6
xx = xx + subcell_width;
yy = yy + subcellw idth;
elseif subcell = 7
xx = xx + 2*subcell_width;
yy = yy + subcellw idth;
elseif subcell = 8
xx = xx + 3*subcell_width;
yy = yy + subcell_width;
elseif subcell = 9
yy = yy + 2*subcell_width;
elseif subcell = 10
xx = xx + subcell_width;
yy = yy + 2*subcell_width;
elseif subcell = 1 1
xx = xx + 2*subcell_width;
yy = yy + 2’"subcell_width;
elseif subcell = 1 2
xx = xx + 3 *subcell_width;
yy = yy + 2*subcell_width;
elseif subcell = 1 3
yy = yy + 3*subcell_width;
elseif subcell = 1 4
xx = xx + subcell_width;
yy = yy + 3 *subcell_width;
elseif subcell = 1 5
xx = xx + 2*subcell_width;
yy = yy + 3*subceII_width;
55
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
elseif subcell = 1 6
xx = xx + 3*subcell_width;
yy = yy + 3*subcell_width;
end
%calculate powers
forp = 1inumant
A(i,p) = (1/(sqrt((xx-xant(p))A2+(yy-y ant(p))A2)))At;
end
end
%perform AGC
g = min(A);
for p = 1rnumant
A(:,p) = A(:,p)./g(p);
end
B = zeros(freq, nuraant);
%matrix that holds imd powers
for p = 1:numant
for i = l:sl
B(IM 1(i, 1),p)=B(IM 1(i, 1),p)+A(IM 1(i,2),p)* A(IM 1(i,3),p)* A(IM 1(i,4),p);
end
for i = l:s4
B(IM4(i,l),p)=B(IM4(i,l),p)+4*A(IM4(i,2),p)*A(IM4(i,3),p)*A(IM4(i,4),p);
end
end
%power combine protocol.. matrix HI .. blockl
P = sum(A')';
Q = sum(B')';
C = 10*logl0(P./Q)+alpha;
j =0;
for i = 1:freq
if C(i) < cnr, j = j+1;, end
end
dropl = dropl+j;
H l(j+ l)= H l(j+ l)+ l;
%max power protocol.. matrix H2 .. block2
[Y,I] = max(A');
P = V;
for i=l:freq
Z(i) = B(i,I(i));
56
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
end
Q = Z';
C = 10*logl0(P./Q)+alpha;
j=0;
fori = l:freq
if C(i) < cnr, j = j + 1 e n d
end
drop2 =drop2+j;
H2G+1)=H2G+1)+1;
%max ctoi protocol.. matrix H3 .. block3
W = zeros(freq, numant);
for j=l:numant
W(:j) = A(:J)./B(:j);
end
V = max(W')';
C = 10*log 10(V)+alpha;
j =0;
fori = l:freq
if C(i) < cnr, j = j + 1 end
end
drop3 = drop3+j;
H3G+i)=H3G+i)+i;
if rem(run,10) = 0, run, end
end
block 1 = (100*dropl)/(freq*numrun)
save al6d90hl HI
block2 = (100*drop2)/(freq*numrun)
save al6d90h2 H2
block3 = (100*drop3)/(freq*numrun)
save al6d90h3 H3
end
57
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4
Millimeter-wave optical transmitters
This chapter deals with the topic o f millimeter-wave signal distribution over optical
fiber for application in future wireless networks. The novel concept under investigation is
the use of resonant modulation, by which efficient millimeter-wave modulation of an
optical signal can be achieved by direct current modulation of a semiconductor laser.
There are two challenges that must be addressed to accomplish this goal (1) Packaging of
the opto-electronic device (the laser) must be carefully considered such that the
millimeter-wave signal can actually be delivered to the device, and (2) the limitations and
the influence of the device properties o f the resonant modulation phenomenon are critical
to the development of transmitters based on this technique. Both of these topics are
addressed below.
4.1
Semiconductor Laser RF/Optical Packaging Issues
Packaging a semiconductor laser for millimeter-wave signal transport presents new
and significant challenges in opto-electronics. On rare occasions has a semiconductor
laser been prepared for signal transport beyond 30 GHz. The use of simple narrowband
reactive matching circuits are essential for the efficient delivery of the mm-wave signal
into the contact o f the laser. Monolithic millimeter-wave narrowband integrated circuit
(MMIC) amplifiers placed before the laser can increase the minimum detectable RF
power, and are required to modulate the laser at the cavity round-trip frequency. Highfrequency bias tees are necessary to couple the mm-wave signal, bias current, control
current into the laser. RF shielding o f the package and spurious radiation at millimeter-
58
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
wave frequencies must be addressed for the optical transmitter. Circuit-board real-estate
may be conserved by collocating the electronic high-speed drive with the laser and mmwave components. Effects o f cross-talk between these circuits must be ascertained.
Standard issues such as optical isolation and laser-to-fiber coupling must be dealt with.
Elimination of the optical isolator can significantly reduce the cost of the transmitter.
Under this scenario, the coupling efficiency (into a single-mode angle-polished fiber)
must be reduced to eliminate the deleterious effects of back reflections. Reducing optical
coupling further drives down cost due to decreased manufacturing labor, at the expense o f
reduced RF link efficiency.
4.2
Device Selection, Packaging, and MM-wave Transport
45 GHz
Figure 4.2.1. Illustration of a DBR laser structure used for the research effort
Round-trip resonant frequency of this device was measured at 45 GHz.
59
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
DBR lasers emitting at 1.5 pm were cleaved to various lengths corresponding to
cavity round-trip frequencies in the range from 36-45 GHz. Details o f one of the DBR
lasers used to perform RF measurements is provided in Figure 4.2.1 above.
The length of each section o f the DBR laser is also shown in Figure 4.2.1. The device
was cleaved to a total length of ~1.2mm, corresponding to a cavity round-trip resonant
frequency o f 45 GHz. For maximum flexibility, an independent bias current is provided
to each contact, and modulation is delivered to the gain section of the laser. Control of the
phase o f the mm-wave signal is accomplished by varying the bias current into the phase
and grating section o f the device. Prior to performing modulation response and RF phase
measurements, the device must be packaged for high-frequency operation. As illustrated
in Figure 4.2.2, the laser is first mounted on a standard gold-plated copper carrier.
Bias Current Contacts
for RF Phase Control
To Wiltron
V-connector
S
'
L
Figure 4.2.2. Illustration of gold-plated copper carrier used to secure the laser for
mm-wave modulation. The small microstrip board accepts the modulation from the
center conductor of a Wiltron V-connector.
60
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The copper carrier was modified slightly to account for the multiple contacts of the laser.
Connection of each laser contact to the copper carrier is accomplished via a bond wire
approximately 20-30 mils in length.
Another bond wire makes contact to the small
microstrip board situated adjacent to the laser on the copper carrier as shown in Figure
4.2.2. The microstrip board subsequently makes contact to the center conductor of a
Wiltron V-connector. The impedance o f the entire assembly is measured at 45 GHz using
a high-frequency network analyzer. Swept frequency Smith chart measurements are made
from 44-46 GHz. The impedance data obtained with the analyzer are fed into Touchstone
microwave simulation software to obtain the physical dimensions o f the microstrip board
required for a 50 Q impedance match at 45 GHz. Based on the simulation results, a
single-section microstrip board is fabricated on a standard Duroid microwave substrate
using conventional circuit board etching techniques. Trimming o f the microstrip board is
done with a razor blade to obtain a narrowband match at 45 GHz.
4.3
Resonant modulation of single contact monolithic semiconductor
lasers at millimeter wave frequencies
There is a growing interest in the use of optical fiber for the remote transport of
millimeter wave (>30GHz) signals for phased-array antenna systems [15] and wireless
personal communication networks [3]. Given that the direct current modulation
bandwidth of semiconductor laser is currently limited to ~30GHz [25], the technique of
resonant modulation o f monolithic semiconductor lasers at millimeter-wave frequencies
is a promising technology for the above applications [26] [27]. Due to the fact that
61
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
efficient longitudinal mode-coupling demands inhomogeneous modulation o f the laser
cavity, previous demonstrations o f monolithic resonant modulation were performed using
spilt-contact lasers. This concept is illustrated in Figure 4.3.1 below.
Demands partial cavity modulation
{\ ) modulation
•<
mirror
c/2nL
I "A/w
-------------- L
►
optical out
Modulation m u st be applied away from cavity cen ter
©
optical out
L/2
Figure 4.3.1. The mode-locking phenomenon fundamentally requires an in­
homogeneous modulation of the laser cavity.
Although the availability o f multiple contacts increases the range of bias conditions of
the laser, the increased device complexity is undesirable. Recent measurements of
millimeter-wave propagation along a semiconductor laser showed a high attenuation
(~60dB/mm at 40GHz) o f the signal along the laser stripe [28]. In this chapter, it is
demonstrated that the confinement o f the modulation current resulting from the high
signal attenuation can be utilized to achieve resonant modulation of a single contact
monolithic semiconductor laser at 40GHz. This concept is illustrated schematically in
Figure 4.3.2 for a ridge waveguide structure.
62
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
Modulation
Feed
Partial Cavity
Modulation
Attenuation ~ 500 dB/cm @ 40 GHz
Figure 4.3.2. Illustration of a single contact mode-locking experiment. A microwave
probe is used to modulate the device at a particular feed point which leads to a
partial modulation of the laser cavity.
The injected modulation current spreads out in a localized region near the feed point,
leading to a partial modulation o f the laser cavity. The modulated light and the small
signal response o f the device are experimentally measured at two different points along
the stripe of a semiconductor laser. A modulation efficiency of -20dB relative to that at
low frequencies is obtained. Also, using a simple distributed circuit model o f the laser in
conjunction with conventional mode-locking theory, the characteristics and limitations of
this technique are investigated.
The device used for the measurements is an InGaAs ridge waveguide laser with three
quantum wells. The ridge structure is 4pm wide, and the ground plane contacts are
63
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
located 80pm on either side o f the laser stripe. Although the device geometry is not
perfectly coplanar, the 6pm difference in height between the ridge and ground plane is
small enough that a coplanar (ground-signal-ground) 40GHz bandwidth probe can be
used to inject signals into the device. For a cleaved length o f approximately 1000pm, the
laser has a threshold current of 37mA, and a lasing wavelength o f 1.0pm. The
experimental set-up for all o f the measurements is illustrated in Figure 4.3.3 below.
60 dB
waveguide
mbcer
40 GHz
frwnTx
Figure 4.3.3. Experimental set-up used to characterize the single contact resonant
modulation. The streak camera is used for time domain measurements, and the
spectrum analyzer is used for frequency response measurements.
In a first set of experiments, the optical output was operated in synchroscan mode, and
triggered with a phase locked 100MHz output from the synthesizer. Figure 4.3.4 shows
the optical modulation measured with the probe positioned at L/4 and L/2 away from the
edge of the laser.
64
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
'</>
ffl
-
-L /2
/V/\AjAy^lKv\A#
100
150
200
250
300
Time (ps)
Figure 4.3.4. Measured streak camera trace of the modulated light output from the
single contact device at 40GHz for two different microwave probe positions along
the cavity.
For both curves, the RF drive power and the bias current to the laser are the same.
The streak trace clearly demonstrates millimeter-wave modulation of the light output at
the intermodal frequency of 40.4GHz. As expected, the mode-locking efficiency is
substantially reduced when the laser is modulated near the center of the cavity (-L/2).
Next, the small signal modulation response o f the device at millimeter-wave frequencies
was acertained by sweeping the synthesizer over the frequency range of interest. At the
receiver, the millimeter-wave optical signal was detected with a high-speed detector
followed by a down-converting mixer driven at 39.0GHz. The output of the mixer was
65
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
amplified by 45dB and observed on an RF spectrum analyzer. Figure 4.3.5 shows the
small signal modulation response o f the device at low frequencies and near the cavity
round-trip frequency.
00
2
<D
<0
C
o
a
(0
-10
-20
O'
-30
<D
c
o
3
■o
O
5
L/4
-40
-50
L/2
-60
0.6
1.2
Frequency (GHz)
Figure 4.3.5. The complete m odulation response of the single contact device at low
and high frequencies is shown. T he solid curve corresponds to an L/4 fed device,
and the dotted curve corresponds to an L/2 fed device.
For the quarter length (~L/4) fed device, the peak of the millimeter-wave response is
20dB below the DC modulation efficiency, and the width o f the passband is
approximately 160MHz. These measurements are comparable to previous results
obtained with split-contact lasers under homogeneous bias [27]. Again, note that the
response o f the center fed device is small compared to the ~L/4 case. These streak camera
and RF response measurements substantiate the claim that efficient mode-coupling is
66
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
made possible due to non-uniform current injection into the laser at millimeter-wave
frequencies.
We investigate the modulation position dependence o f the observed mode coupling in
a single contact device using the distributed circuit model used in reference [28] and
shown in the inset of Figure 4.3.6. The calculated amplitude of the normalized injection
current into the laser is shown as a function o f position along the device away from the
feed point at 40 and 90GHz.
1.0
3
(0
■+*
0.8
£
0.6
c
3
o
C
0.4
40GHz
o
3
o
0)
0.2
90GHz
c
0
200
100
300
Position Along Laser Contact (|im)
Figure 4.3.6. Distribution of injected current into the active region of the device as a
function of length along the laser stripe at 40 and 90GHz is shown. The inset shows
the distributed circuit model [28] used to calculate the current
Notice that the amplitude of the injected current decreases rapidly with position, and
drops is insignificant levels beyond 200pm away from the feed point. For an intermodal
frequency of 40GHz, the laser is typically 1000pm long, resulting in a fractional cavity
67
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
modulation o f approximately 20%. The modulation efficiency is addressed using a
conventional mode-locking analysis along with the current distribution obtained from the
distributed circuit model discussed above. The self consistent solution for active modelocking [29] shows that the optical modulation (p) is expressed as:
n
i
1
. n<h.
r 1l
M
p = 24
Where n is the amplitude o f the time variation in the electron density, n* is the threshold
electron density, and the mode discrimination factor (b) is the ratio of the optical gain
bandwidth to the round trip cavity frequency. The parameter (Q is the overlap integral
between adjacent longitudinal modes and the spatial variation of the gain modulation
along the cavity. Since the modes are orthogonal, this amplitude is zero if the modulation
is uniform over the cavity. If we assume that the modulation in the photon density is
small, such that the spatial dependence of the gain modulation can be approximated by
that of the injection current, and that the cavity modes are given by solutions to the one
dimensional Helmholtz equation, 4 can be expressed as:
where L is the laser length, and I(z) is the normalized spatial distribution of the
modulation current. To determine I(z), we model the laser as a transmission line with two
68
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
open ends. The impedance o f an open ended transmission line is:
Z - 2 0 coth(y /)
where Z0 is the characteristic impedance of the line, y is the complex propagation
constant, and / is the distance from the probe to the open end [30]. Figure 4.3.7 shows this
calculated mode-coupling amplitude as a function o f position along the laser stripe for an
intermodal frequency of 40GHz.
0.07
Optim um
0.06
0.05
£
0.04
0.03
8 current
confinement
0.02
0.01
0.00
0
L/4
L/8
3U8
L/2
Cavity Probe Position
Figure 4.3.7. Calculated mode coupling factor as a function of probe position for a
device with a round trip cavity frequency of 40GHz. The dotted curve corresponds
to the limiting case when the injected current is a delta function.
The maximum value of 0.06 is approximately a factor of 5 smaller than the maximum
achievable value o f £ obtained when exactly one half o f the cavity is uniformly
69
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
modulated. For comparison, if we consider that the actual injection current profile is a
delta function of amplitude 1/y (i.e. no spreading o f the modulation current), then the
above equation reduces to:
which is shown as the dotted curve in Figure 4.3.7. The similarity of this curve to the
exact numerical calculation highlights the fact that the extent o f the modulation current is
very localized, and that 4 is proportional to (yL)1. Hence, the value of 4 can be optimized
by designing the transmission line for a particular value o f y. Note that the optimum
probe position is approximately at L/8 away from the laser facet. In fact, the conclusion
that ~L/8 is the optimum feed point remains true over the entire millimeter-wave range
(30-100GHz). The maximum value o f 4 also remains approximately constant over this
range. This is understood by realizing that although the signal propagation is reduced at
higher frequencies, the resonant device length is shorter, resulting in a comparable
fractional cavity modulation. These results show that modulation of single contact
monolithic semiconductor lasers is possible over the entire millimeter-wave range up to
100GHz.
In summary, it is demonstrated for the first time the efficient modulation o f a single
contact monolithic semiconductor laser at the intermodal frequency o f 40GHz. Both
temporal and RF measurements show clear evidence o f millimeter wave modulated light
output from the laser. Our claim that the single contact mode-locking is due to the limited
propagation of the millimeter-wave signal is substantiated by the observed feed point
70
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
dependence of the modulation efficiency. This technique was studied using a simple
distributed circuit model o f the laser in conjunction with conventional mode locking
theory. The optimum feed-point and mode coupling factor over the millimeter-wave
frequency range are found to be approximately L/8 and (yL)"‘ respectively. These results
are a very important step toward the realization o f practical millimeter-wave optical
transmitters based on direct modulation o f monolithic semiconductor lasers.
71
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5
5.1
Distributed antenna system implementation
Measurement based design and optimal architecture for an in-building
distributed antenna system
As the consumer demand for ubiquitous wireless coverage continues to grow, service
providers are quickly developing and deploying networks with greater utility than ever
before. It is clear that to meet the new demands of wireless customers, cellular and PCS
service providers must upgrade their networks to provide complete radio coverage. This
evolution has motivated the need for low-cost systems that transport radio signals to and
from areas o f poor signal coverage such as: office buildings, shopping malls, hotels,
hospitals, and tunnels. The most important area where this problem must be solved is
inside of buildings, since this is where people spend most o f their time. Unfortunately, the
in-building environment is also the most challenging area to provide radio coverage due
to severe attenuation and multi-path effects. Wireless providers seek solutions to this
problem that optimize the performance/ cost ratio of the network.
In-building radio coverage can be provided by deploying a network of distributed
antennas, thus providing high-quality uniform wireless service. The use of analog fiberoptics as a connecting infrastructure for such a distributed antenna network has been
proposed [3]. Optical fiber may be ideal in some applications due to its low loss, light
weight, high bandwidth, and immunity to electromagnetic interference. In the design o f a
distributed antenna system, there are two major performance/cost design issues: (1)
Where (if at all) should fiber-optic links be used?, and (2) What dynamic range is
required for each antenna in the network? This chapter presents a design approach to
72
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
answer these questions in the context o f a real in-building environment illustrated in
Figure 5.1.1 below.
UCB Cory Hall Floor Plan (1st Floor)
^ Antenna 2
Antenna 1
Measurement Path
Figure 5.1.1. Floor plan of the radio environment under study. Initial radio
measurements were made in the main corridor with four distributed antennas.
It is found that by proper antenna placement, the dynamic range requirement of the
antenna sites can be relatively low (~90dB-Hz2/3). Given that fiber-optic links with this
performance typically have a high noise figure (~30dB), the fiber link can be used to feed
multiple antennas through -5 0 meters o f coaxial cable with minimal performance
degradation. By feeding multiple antenna sites per fiber link, this hybrid fiber-coax
architecture achieves a performance nearly equivalent to a completely fiber-fed antenna
73
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
system at a much lower cost.
5.2
Radio Measurements
An in-building wireless field-trial is currently in progress at the electrical engineering
building at the University o f California at Berkeley. The floor plan o f the first floor of
this building is shown in Figure 5.1.1. To perform an initial system design, radio
measurements were taken along the main hallway as illustrated by the dotted line in the
figure. Dipole antennas were positioned at the four comers o f the hallway (as shown), and
used to measure the received signal power from a mobile dipole antenna driven with a
+15dBm oscillator at 900MHz. The mobile antenna and remote antennas were positioned
1 and 3 meters above the ground respectively. The entire length around the square
corridor is approximately 140 meters. The signal strength at each of the 4 antennas was
measured independently. Figure 5.2.1 below shows the measurement results for two of
the antennas. On the left, the received radio signal strength is plotted as a function of the
mobile distance along the hallway. On the right, the distributions of the received signal
powers are shown.
Note the wide range (~70dB) o f received powers as the mobile travels around the
corridor. If one were to use only one fiber-fed antenna to cover the entire floor, the link
dynamic range would need to be ~1 lSdB-Hz273 to maintain adequate voice quality (C/N =
18dB) and spur-free operation. Given the high cost of such fiber-optic links, a more
optimal architecture is needed.
74
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
^-20u
S
i » ‘l-j i ■ i
.
'W '
£
|- a *
l/M, ■
if fki
(if
n
o
a
I
A i i .t r
,-J * ^ u -
a
0+
to
:i I, .
In! i <r l 1J i
to o — m
3■JP
o ~ Jtr TO
ny
|-K»
o ft 4a ft anooiarjao
s
»*i ' i • i
€ 4
| -60.
£
■O 40.
Si
j .m
M
12
10
8
6
4
2
0-
Antema2
rkk
? 'k f
o a am anwiaia
n
e
a
a
to
Tttr
... . I '1'1kliTlflilO
■r
4)—T )
AoeiwdBmerW
Ration (meteis)
Figure 5.2.1. Measured received RF power at antennas 1 and 2 as a function of the
position along the hail. The distributions of the received powers are also shown.
75
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
-20
I
-40
!a. -60
■a -80
A
^
i/vx h .i
rt'.y ap.A .
y -i
'*li\
v-j|i
n
I
j
,
Pkc.yf
Ttoo Antennas
W' 100
£
0 20 40 60 80 100120140'
-20
-40
-60
%
£ -80
KhitAntennas
■o
%
Z -100
8
0 ' 20 4 0 ' 60 u'o iooi2in4o
£
Position (meteis)
14
12
10
8
6
4
2
0
14
12
10
8
6
4
2
0
n
o
s
s**
^lOo -su
-6u -46 -io
o
©
s
s
-1*)0 _-$ti— -60
-tU
0
Received Bower (dBm)
Figure 5.2.2. Measurement based prediction of the system radio performance with 2
and 4 antennas used with a diversity selection technique.
76
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
By noting the complementary nature of the signals received at antennas 1 and 2, we
observe that an architecture that implements both antennas, and selects the stronger of the
two signals would provide better radio coverage and a reduced range o f powers at each
remote antenna site. Figure 5.2.2 shows the calculated consequence o f such a diversity
combination method using antennas 1 and 2 (two antennas) and by using all o f the
antennas (4 antennas). Note that by using two (four) antennas, the range o f received
powers is reduced to ~35 (25) dB, resulting in an antenna dynamic range requirement of
~85 (70) dB-Hz273. Given that very low-cost fiber-optic transmitters can have an analog
dynamic range performance o f ~90dB-Hz2/3 [10], the use o f two antennas for this
particular environment is ideal. A comparison o f the required antenna dynamic range as
function of the number of antennas is shown in Table 5.2.1 below.
H
I
Range of Selected
Powers (dB)
68.0
35.7
23.0
Required SFDR
(dB-Hz273)
115.8
83.5
70.8
0
2.8
39.5
Percent of
Hallway
‘Over-Covered’
(%)
Table 5.2.1. Comparison of the antenna dynamic range requirements and radio
coverage characteristics of the hallway under study.
77
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The fact that two antennas are ideal for this system is further emphasized by
calculating the fraction o f the hallway that is ‘over-covered’ which we define to be a
region where the signals received by any two active antennas are within 3dB o f each
other. Table 5.2.1 shows that the use o f four antennas over-covers the hallway by ~40%,
as compared to -3% if only two antennas are used. Also, this overlap between antenna
coverage areas leads to severe fading if simple power combination is used to connect the
antennas. In this simple environment, the use of two antennas provides the optimum
tradeoff between radio coverage, antenna dynamic range requirements, and antenna
isolation.
5.3
Optimal Architecture
Given the results of the previous section, the next step in the system design is to
determine the optimal architecture for connecting the distributed antennas to the RF
distribution point. This distribution point is typically located either on the roof or in the
basement o f the building. It is well known that the use o f optical fiber is ideal for the
transport o f RF signals over a relatively long distance; however, its use is not justified for
short distance connections since coaxial cable links are significantly lower in cost than
analog fiber-optic links. With this consideration, the most cost-effective scheme for
connecting distributed antennas is to use a hybrid fiber-coax (HFC) network as illustrated
in Figure 5.3.1 below.
78
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Fiber-to-the-Antenna
Hybrid Fiber-Coax Network
L ow -cost fiber link
(-9 0 dB-Hz273)
To/From Roof
Low-cost fiber link
(-90 dB-Hz273)
To/From Roof
▲
-SO m eters
coaxial cable
Com biner
A
Antenna
Coverage Area
Antenna
Coverage Area
15,000 sq-ft
> 60,000 sq-ft
Figure 5.3.1. A comparison of fiber-to-the-antenna and hybrid fiber-coax
architectures for providing cost effective in-building radio coverage.
This conclusion is based on the fact that QOdB-Hz273 analog fiber-optic links typically
have a high RF noise figure (~30dB) [32]. In this case, one can add up to ~30dB o f
coaxial cable loss before the fiber link with no degradation in system performance. For
typical coaxial repeater specifications, this corresponds to a distance of approximately 50
meters. In this way, the advantages of both low-cost fiber links and coaxial cable are both
exploited to achieve optimal network performance at the lowest cost. This point is clearly
illustrated in Figure 5.3.1 since the HFC system covers a much larger area than the fiberfed antenna at a comparable cost.
For the radio environment under study, the two antennas on the first floor are
79
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
separated from each other by only ~50 meters, and hence are a perfect candidate for
hybrid fiber-coax interconnect. The optimal design in this case is to feed the two first
floor antennas with coaxial cables, and to trunk the combined signals from both antennas
to/from the roof of the building using a low-cost, 90dB-Hz2/3 analog fiber-optic link. This
design has been implemented and tested in the field-trial at U.C. Berkeley, and is
illustrated in Figure 5.3.2 below. The network provides superior radio coverage on the
first floor o f the building under multi-user test conditions.
Optimal Hybrid Fiber-Coax Architecture
Fiber Link To/From
RF Distribution
Coaxial
Antenna Feed
Figure 5.3.2. Illustration of the optimal hybrid fiber-coax architecture implemented
in the wireless field-trial at the University of California at Berkeley.
80
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.4
Noise figure analysis
The conclusion o f the above section can be understood by considering a noise figure
analysis of the system shown in Figure 5.4.1 below.
Fiber Link
NF = Flink
Figure 5.4.1. A typical hybrid fiber-coax architecture. The passive coax cable of loss
(L) is followed by an amplifier of gain L and an fiber link of a given noise figure.
In the architecture shown, a remote antenna is connected via a passive coaxial cable of
loss L. After this cable, the loss o f the cable is compensated by an amplifier with Gain =
L, and a noise figure Famp. The output of this amplifier is used to modulate a fiber-optic
link o f noise figure Flink. The composite noise figure o f this system (F^,) is given by the
following formula:
Ftol- L + {Famp- \ ) L + {FUnk- \ )
Now, 100 meters of low loss coaxial cable has a loss o f approximately 25dB at 1GHz.
Typical microwave amplifiers have a noise figure o f 2dB and typical analog fiber-optic
links have a noise figure o f 30dB. Using these numbers, the composite noise figure is
81
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
plotted as a function of coaxial cable length, in Figure S.4.2 below.
L = 25dB (100m)
F = 2dB
34
HFC A ntenna
link
1 dB penalty
00
■o
O)
LL.
(0
O
z
0
20
80
40
60
Cable Length (m eters)
100
Figure 5.4.2. Composite system noise figure as a function coaxial cable length.
The lower, flat curve is the noise figure o f the fiber-optic link. A system that
incorporates fiber-to-the-antenna will have a 30dB noise figure, independent o f cable
length. The upper curve corresponds to the hybrid-fiber-coax solution discussed above.
Notice that up to 90 meters o f passive coaxial cable can be placed between the fiber link
and the antenna site before a ldB penalty in noise figure is observed. This illustrates that
claims made in the previous sections in that multiple passive antenna sites can be fed
using a single fiber-optic link with a minimal (1 dB) degradation in system performance.
The hybrid fiber-coax solution has significant cost advantages over the fiber-to-theantenna approach.
82
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.5
Conclusions
In summary, a real in-building radio environment is currently under study to evaluate
the architecture and dynamic range requirements o f distributed antenna networks. It is
found that two antennas implemented in a hybrid fiber-coax architecture will provide
good radio coverage of the first floor at the lowest cost. By proper antenna placement, a
modest (~90dB-Hz273) dynamic range analog fiber-optic link can be used for transmitting
the radio signals to and from the roof of the building.
83
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6
6.1
Low power, high-speed fiber links for digital RF transport.
Digital R F Transport
The previous chapters in this thesis have focused on the analog transport of
microwave and millimeter wave signals using fiber-optics. The benefits of this technique
are mainly in its simplicity, since the analog fiber link is basically a radio “extension
cord”, delivering signals from point A to point B. An alternative approach to distributing
RF signals is the use o f digital RF transport. In this scheme, the entire RF spectrum is
digitized using an A-to-D converter, and sent over a digital fiber-optic link. There are
several advantages o f this scheme: (1) digital signals can be re-generated, and can be sent
over much longer cable lengths before signal degradation is observed, and (2) once the
signals are in digital form, off-the-shelf digital signal processors can be used to perform
advanced features on the radio signals such as diversity, signal combination, channel
demodulation, and digital filtering. The main disadvantage o f such a scheme is the power
consumption o f the A-to-D converter and the digital fiber-optic transmitter required.
Research is currently under-way to produce low power A-to-D converters. This chapter
will address the development and requirements for high-speed, low-power digital fiber
optic links. The challenge here is the simultaneous high-speed and low-power
requirement. The U.S. PCS block is 60MHz wide. With 5x over sampling, and 12bits per
sample, a serial bit rate o f 3.6Gb/s is required. The following section addressed the topic
o f ultra-low power digital fiber-optic links operating at multi-gigabit rates.
84
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6.2
Ultralow Power Digital Fiber-optic Links Using Zero-biased
Semiconductor Lasers
Intense interest in recent years in ultralow threshold lasers, through a combination o f
quantum confined and/or microcavity VCSEL structures, was motivated mainly by their
potential for optoelectronic integrated circuits (OEICs) due to their low electrical power
requirements. Lasers with a threshold in the lOOpA range already exist [33] [34]. This
threshold is small compared to the drive current above threshold needed to generate the
typical required optical power for high data rate communication. It thus appears that
continued research to further lower the threshold will have little impact on the laser driver
requirements o f the OEIC’s. It is shown in this section that this is not true when one fully
considers the switching dynamics o f semiconductor lasers in an OEIC environment,
particularly at high data rates.
The analysis is based on the presumption that the laser is digitally modulated and the
zero-bias modulation format [35][36] is used. This modulation format is highly favorable
in OEIC’s since it eliminates the need for optical monitoring and feedback control of the
bias point of each individual laser. This bias circuitry is a complication, particularly for
VCSEL’s, that consumes both power and on-chip real estate. It is well known that zerobias on/off switching produces data dependent tum-on delays that result in a degradation
o f the data. This is illustrated in Figure 6.2.1 for the time evolution o f the electron density
and the optical output under a pseudorandom pulse modulation.
85
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
I(t)
N (t)
P (t)
Figure 6.2.1 Illustration of the pseudorandom bit stream, the time evolution of the
carrier density, and the photon density out of the modulated laser.
When the optical data stream is fed into a typical digital receiver, the resulting timing
jitter introduces a bit-error-rate (BER) degradation. This degradation depends on the
relative magnitude of the time delay compared to the bit rate. For lasers biased below
threshold, the tum-on delay consists of the time for the drive current to fill the electron
density up to the threshold level, plus an additional time for the photon density to build
up from spontaneous emission. The variance o f the time for the photon density to build
up has been studied [37] and shown to be <10 ps for a wide range of drive currents. In
comparison, a laser modulated in the zero-bias modulation format typically requires a
much longer time (~ lOOps) to reach threshold. To model the pattern dependent switching
dynamics o f the laser, we assume that the tum-on delay is given simply by the time for
86
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
the carrier density to reach threshold. With this assumption, it is clear that a laser with a
lower threshold will require less drive current in order to achieve a certain tum-on delay,
and thus reducing the power requirement for the laser driver overall.
Since the tum-on delay is data dependent and is thus random, we must describe it in
terms of a statistical distribution.
The tum-on delay can be determined by considering the rate equation for the electron
number (n) in the active region of the laser:
where t is the carrier recombination time, I(t) is the drive current, e is the electronic
charge, a is the optical gain coefficient, and p is the photon number in the active region.
If the laser is biased below threshold, we can assume that the photon number in the cavity
is approximately zero (p ~ 0).
With this assumption, (1) can be solved for a step input current o f amplitude Imat t=0.
Where the approximation is valid for times small compared to (t), and
number at t = 0. The initial carrier number in the laser
is the carrier
depends on the number of “0”
87
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
bits which precede the “ 1” bit. We see from (1) that at the end o f a bit, the carrier number
decays exponentially with time constant (tau) such that:
Hi
= nth ex
Where N is the number of “0” bits preceding the considered bit, n^ is the threshold carrier
number, and T=l/B where B is the bit rate. If the modulation current pulses follow a truly
random digital pattern, the statistics of N are simply that of a geometric distribution: p(N)
= y(2'N) where y = ln(2) if we take p(N) to be a geometric distribution. Noting from Eq.(l)
that IA= nth*e/(x), defining n(ton) = n*, and t0 =
we can write:
Now, defining r\ = ln(2)B(T), we can use the statistics o f N to determine the distribution
function of ton:
This distribution function is plotted in Figure 6.2.2 for different values o f t].
88
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
t
It
Figure 6.2.2. Distribution function of turn-on time for different values of h.
A noteworthy feature of this plot, is that for r\ <1, P(ton) is peaked around t = t0, and for tj
> 1 it is peaked around t=0. This is expected since at low bit rates the average charge in
the laser is small, and hence ton ~ t0. In comparison, at high bit rates ton ~ 0 since the
charge does not have much decay time, and therefore remains close to the threshold
value.
In order to quantify the requirements for drive current for satisfactory laser
modulation performance, we assume that the optical pulse stream is fed into a thermal
noise dominated simple photodiode receiver (50 ohm load) through 6dB o f optical loss (a
typical number in short distance optical interconnect). We assume that the received
photocurrent after equalization at the receiver has the simple form [38].
89
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Where the link efficiency (Tilink) includes the laser differential quantum efficiency (0.3
W/A), the link loss (6dB), and the detector responsivity (0.5 A/W). Standard calculations
o f the error rate in the presence o f Gaussian noise are performed. The noise variance of
the “ 1” bits and “0” bits are equal; however, the “ 1” bits are degraded by the tum-on jitter
effect discussed above. Following the approach in reference [39], the bit error rate at the
receiver can be expressed as:
Where D is the decision level, sigma is the variance o f the thermal noise, and erfc(x) is
the complementary error function. For each data point, the decision level (D) is
numerically chosen to minimize the corresponding BER.
-3
100 mA\
10 mA
-13
50
100
150
200
50 mA
250
300
350
400
450
Figure 6.2.3. BER as a function of drive current for various device thresholds.
90
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 6.2.3 shows the calculated BER as a function o f the drive current to the laser (Im)
for various values o f 1^, with a carrier lifetime o f 2 ns and a bit rate o f lGb/s. Clearly,
increasing
improves the BER. Note that for I* < lOpA the power penalty is not severe;
however, for I^ lO p A the threshold current has a large influence on the error rate.
We compute the required drive current to maintain a BER o f 10'9 as a function of the
laser threshold current. This required drive current is used to calculate the average
electrical power consumption for the laser given by: Pe = l/2Itn(V0+ImZo) where we
assume V0 = 1.5V and Z0 = 100 ohms. For drive currents in the milliampere range, the
above expression is dominated by the term involving the tum-on voltage (V0). Hence, in
these low power links, the tum-on voltage o f the laser is much more critical than the
series resistance in determining the overall power consumption. Figure 6.2.4 shows the
required electrical power as a function of threshold current for lOOMb/s, 1 Gb/s, and
5Gb/s.
91
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Electrical Power (mW)
20
1.5
1.0
1 Gb/s
0.5
0.0
100
1000
Threshold Current (uA)
Figure 6.2.4. Electrical drive power to the laser as a function of the device threshold
current, for various bit rates, and carrier lifetimes.
92
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
The solid and dashed curves are for carrier lifetimes o f 2 ns and 4 ns, respectively.
From the curves, we see that the lowest threshold current which still yields a drive power
advantage is approximately lOOpA for lOOMb/s, 50 pA for lGb/s, and lOpA for 5Gb/s
data rate. The desired threshold current clearly decreases as the data rate is increased. For
larger data rates, the carrier density has less relaxation time between bits resulting in
more significant pattern dependent switching delays. A lower threshold device is then
required to maintain satisfactory error performance for low electrical drive powers. Also,
notice that for the giga-bit data rates, the carrier lifetime has a significant effect on the
required drive power. This is expected since at high bit rates (>1 Gb/s) the carrier lifetime
is long compared to the bit period, and hence has a large influence on the carrier density
relaxation between bits and the corresponding time to reach threshold. It should be noted
that the above results are for a specific link loss and receiver configuration.
In summary, we have examined the extent to which electrical drive power
requirements of optical transmitters can be lowered by lowering the laser threshold to the
sub-lOOuA level. A simple model for the BER experienced in a digital optical link
employing the zero bias modulation format was used. At data rates below 1 Gb/s, there is
no significant advantage to lower the laser threshold to much below lOOuA, while at
multigigabit rates, there is a significant advantage in reducing the laser threshold current
down to the 10-uA range.
93
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
7
Conclusions and Future Directions
The majority o f consumer wireless services at the present time are concentrated in
radio bands below 3GHz. The above analysis has argued that low-cost commercially
available semiconductor lasers have sufficient performance for the support of wireless
systems in these bands. Given that these relatively low frequency radio systems will
continue to dominate in the near future, the existing challenge is the implementation o f
practical fiber-fed radio systems based on low-cost devices. These fiber-fed radio
networks introduce a host of new research issues such as optimum network protocols for
the assignment of mobile users to a particular antenna or group of antennas. Also,
techniques for interfacing fiber radio systems with existing wired networks is expected to
be an important issue.
As demand for higher bandwidth services such as wireless cable television and
multimedia data communications grow, the need for a high frequency infrastructure will
be unavoidable. Given the commercial availability o f low cost semiconductor lasers
capable o f supporting low-frequency wireless services, the challenge in device
technology is the development o f the narrow-band, high frequency optical transmitters
based on resonant modulation for future high frequency applications. New device
structures are needed that are optimized for efficiency and linearity under resonant
modulation. Another major research issue is the necessary microwave packaging of these
devices with driver electronics such that efficient delivery of the millimeter-wave signals
to the laser is accomplished.
94
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In summary, low-cost analog fiber-optic networks must be developed to meet the
rapidly expanding demand for wireless communication services. This effort involves the
design o f networks that can be supported by low-cost, commercially available devices,
and the development of narrow-band, high-frequency optical transmitters for future
wireless systems. These technologies are expected to have a significant impact on the
realization o f future wireless personal communication networks.
95
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
8
References
[1] J.A. Chiddix, H. Laor, D.M. Pangrac, L.D. Williamson, and R.W. Wolfe, “AM Video
on Fiber in CATV Systems: Need and Implementation,” IEEE Journal on Selected Areas
in Communications, vol. 8, no. 7, pp. 1229-1239, September 1990.
[2] W. Ng, A.A. Walston, G.L. Tangonan, J.J. Lee, I.L. Newberg, and N. Bernstein, “The
First Demonstration o f an Optically Steered Microwave Phased Array Antenna Using
True-Time-Delay,” Journal o f Lightwave Technology, vol. 9, no. 9, pp. 1124-1131,
September 1991.
[3] W.I. Way, “Optical Fiber-Based Microcellular Systems: An Overview,” IEICE
Transactions on Communications, vol. E76-B, no. 9, pp. 1091-1102, September 1993.
[4] T.S. Chu, and M.J. Gans, “Fiber Optic Microcellular Radio,” IEEE Transactions on
Vehicular Technology, vol. 40, no. 3, pp. 599-606, August 1991.
[5] L.J. Greenstein, N. Amitay, T.S. Chu, L.J. Cimini, G.J. Foschini, M.J. Gans, Chih-Lin
I, A.J. Rustako, R.A. Valenzuela, and G. Vannucci, “Microcells in Personal
Communications Systems,” IEEE Communications Magazine, pp. 76-88, December
1992.
[6] M. Shibutani, T. Kanai, W. Domom, K. Emura, and J. Namiki, “Optical Fiber Feeder
for Microcellular Mobile Communication Systems,” IEEE Journal on Selected Areas in
Communications, vol. 11, no. 7, pp. 1118-1126, September 1993.
[7] H. Ogawa, K. Horikawa, and Y. Nakasuka, “Cascaded External Optical Modulation
Link for Radio Distribution,” 1994 MTT-S International Microwave Symposium Digest,
pp. 487-490, San Diego, CA, 1994.
96
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[8] H. Ogawa, D. Polifko, and S. Banba, “Millimeter-Wave Fiber Optics Systems for
Personal Radio Communication,” IEEE Transactions on Microwave Theory and
Techniques, vol. 40, no. 12, pp. 2285-2293, December 1992.
[9] J.B. Georges, M.-H. Kiang, K. Heppell, M. Sayed, and K.Y. Lau, “Optical
Transmission of Narrow-Band Millimeter-Wave Signals by Resonant Modulation of
Monolithic Semiconductor Lasers,” Photonics Technology Letters, vol. 6, no. 4, pp. 568570, April 1994.
[10] D.M. Cutrer, J.B. Georges, T.H. Le, and K.Y. Lau, “Dynamic Range Requirements
for Optical Transmitters in Fiber-Fed Microcellular Networks,” Photonics Technology
Letters, vol. 7, no. 5, May 1995.
[11] A.J. Rustako, N. Amitay, G.J. Owens, and R.S. Roman, “Radio Propagation at
Microwave Frequencies
for Line-of-Sight
Microcellular Mobile
and
Personal
Communications,” IEEE Transactions on Vehicular Technology, vol. 40, no. I, pp. 203210, February 1991.
[12] Donald Cox, “Wireless Personal Communications: What is it?” IEEE Personal
Communications, Vol. 2, No. 2, April (1995), pp. 20-35.
[13] G.K. Chan, and F.A. Razaqpur, “Spectrum Requirements of an Indoor Pico-Cell
Radio System,” IEEE Transactions on Vehicular Technology, Vol. 44, No. 1, February
(1995).
[14] Donald Cox. Stanford University professor o f electrical engineering. Private
communication. February 21,1996, 1:00pm.
97
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[15] A.S. Daryoush, “Optical Synchronization of Millimeter-Wave Oscillators for
Distributed Architectures,” IEEE Transactions on Microwave Theory and Techniques,
Vol. 38, No. 5, pp. 467-476, May 1990.
[16] E. Ackerman, D. Kasemset, S. Wanuga, R. Boudreau, J. Schlafer, and R Lauer, “A
Low-Loss Ku-Band Directly Modulated Fiber-Optic Link,” IEEE Photonics Technology
Letters, Vol. 5, No. 11, Nov. 1993.
[17] K. Morita and H. Ohtsuka, “The New Generation of Wireless Communications
Based on Fiber-Radio Technologies,” IEICE Trans. Commum., Vol. E76-B, No. 9, pp.
1061-1068, September 1993.
[18] Y. Takimoto and T. Ihara, "Research Activities on Millimeter Wave Indoor
Communication Systems in Japan," IEEE MTT-SDigest, V-l, pp. 673-676, (1993).
[19] H. Thomas, N. Imai, ATR Optical and Radio Communications Research
Laboratories,
"Millimeter
waves
over
optical
fiber
for
broadband
wireless
communication," Optical Fiber Conference (OFC), San Jose, Feb. 1994.
[20] J. O'Reilly and P. Lane, “Remote Delivery of Video Services Using mm-Waves and
Optics,” J. o f Lightwave Technology, Vol. 12, No. 2, pp. 369-375, February 1994.
[21] R. Olshansky, “Options for video delivery,” Optical Fiber Conference (OFC), San
Jose, Feb. 1994.
[22] D.M. Fye, “Design of Fiber Optic Antenna Remoting Links for Cellular Radio
Applications,” 40th IEEE Vehicular Technology Conference, Orlando, FL, May (1990),
pp. 622-625.
98
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[23] D. Parsons, “The Mobile Radio Propagation Channel,” Halsted Press: New York
(1992).
[24] M Li, T.A. Wilkinson, M. Beach, S.K. Barton, H. Xue, I.R. Johnson, and A. Nix,
“Analysis o f Intermodulation Distortion Specificatios for Radio LANs Using Multicarrier
Schemes,” Electronic Letters, Vol. 29, No. 13, June (1993), pp. 1229-1231.
[25] S.D. Offsey, L.F. Lester, W.J. SchafF, L.F. Eastman, “High-speed Modulation of
Strained-Layer InGaAs-GaAs-AlGaAs Ridge Waveguide Multiple Quantum Well
Lasers,” Appl. Phys. Lett., Vol. 58, no. 21, pp. 2336-2338, 27 May 1991.
[26] J.B. Georges, D. Cutrer, and K.Y. Lau,
"Theory of Resonant Modulation at
Millimeter-Wave Frequencies o f Inhomogenously Biased Monolithic Semiconductor
Quantum-Well Lasers ", IEEE Photonics Technology Letters, Vol. 7, No. 3, pp. 263-265,
March 1995
[27] J.B. Georges, D. Cutrer, Meng-Hsuing Kiang, and Kam Y. Lau, "Multichannel
Millimeter Wave Subcarrier Transmission by Resonant Modulation of Monolithic
Semiconductor Lasers," IEEE Photon. Tech. Lett., Vol. 7, no. 4, pp. 431-433, April 1995.
[28] D.A. Tauber, R. Spickermann, R. Nagarajan, T. Reynolds, A L. Holmes, Jr., and
J.E. Bowers,
"Inherent bandwidth limits in semiconductor lasers due to distributed
microwave effects," Appl Phys. Lett., Vol. 64, No. 13, pp. 1610-1612, March 1993.
[29] K.Y. Lau “Narrow-band Modulation of Semiconductor Lasers at Millimeter Wave
Frequencies (>100GHz) by Mode Locking,” IEEE Journal of Quantum Electronics, vol.
26, no. 2, pp. 250-261, February, 1990.
[30] S. Ramo, J.R. Whinnery, and T. Van Duzer, “Fields and Waves in Communication
99
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
Electronics,” John Wiley and Sons, New York (1965).
[31] B. Sartorius, M. Mohrle, and U. Feiste, ”12-64 GHz Continuous Frequency Tuning
in Self-Pulsating 1.55 mm Multiquantum-Well DFB Lasers," IEEE Joum. Selected
Topics in Quantum Electronics, Vol. 1, No. 2, pp. 535-538, June 1995
[32] ActiveLite™ Antenna Data Sheet, Allen Telecom Group, Systems Division, 1994.
[33] T.R. Chen, L.E. Eng, B. Zhao, Y.H. Zhuang, and A. Yariv, “Strained single
quantum-well InGaAs lasers with a threshold current of 0.25 mA,” Appl. Physics Lett.,
vol 63, no. 19, pp. 2621-2623,1993.
[34] T. Numai, T. Kawakami, T. Yoshikawa, M. Sugimoto, Y. Sugimoto, H. Yokoyama,
K. Kasahara, and K. Asakwaw, “Record low threshold current in microcavity surface
emitting laser,” Jap. J. Appl. Phys., Part 2, vol. 32, no. 10B, pp. L1533-L1534, 1993.
[35] K.Y. Lau, N. Bar-Chaim, P.L. Deny, and A. Yariv, “High-speed digital modulaiton
o f ultralow threshold (<lmA) GaAs single quantum well lasers without bias,” Appl.
Phys. Lett., vol. 51, no. 2, pp. 69-71, 1987.
[36] T. Odagawa, K. Nakajima, K. Tanaka, H. Nobuhara, T. Inoue, N. Okazaki, and K.
Wakao, “High-speed operation of strained InGaAs/InGaAsP MQW lasers under zero-bias
condition,” IEEE J. Quantum Electronics, vol. 29, no. 6, pp. 1682-1686,1993.
[37] P.K. Pepeljugoski, D.M. Cutrer, and K.Y. Lau, “Parametric dependence of timing
jitter in gain-swithed semiconductor lasers,” Applied Physics Letters, vol. 63, no. 26, pp.
3556-3558, 1993.
[38] G.P. Agrawal, and T.M. Shen, “Power penalty due to decision time jitter in optical
communication systems,” Electronics Letters, vol. 22, pp. 450-451, 1986.
[39] K. Schumacher, and J.J. O’Reilly, “Power penalty due to jitter on optical
100
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
communication systems,” Electronics Letters, vol. 23, no. 14, pp. 718-719, July 1987.
[40] D.Wake, N.G. Walker and I. C. Smith, "Zero-bias edge-coupled InGaAs photodiodes
in millimetre-wave radio-fibre systems,” Electron. Lett., Vol. 29, no. 21, pp. 1879-1881,
Oct. 1993.
[41] Lee, William, C.Y., “Mobile Cellular Telecommunications,” McGraw-Hill, Inc,
Second Edition, (1995).
101
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
IMAGE EVALUATION
TEST TARGET (Q A -3 )
<0
<;5“V
'/.■
150mm
IIVM GE. I n c
1653 East Main Street
Rochester. NY 14609 USA
Phone: 716/482-0300
Fax: 716/288-5969
0 1993, Applied Image, Inc., All Rights Reserved
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
*
<
< £ \
Документ
Категория
Без категории
Просмотров
0
Размер файла
4 100 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа