WP-ISA-031502-2.

0

IntelliCell

: A Fully Adaptive Approach

to Smart Antennas








ArrayComm, Incorporated
2480 North First Street
San Jose, CA 95131 USA
Main Number: +1 408.428.9080
[email protected]
www.arraycomm.com


IntelliCell: A Fully Adaptive Approach to Smart Antennas

Contents

Introduction.................................................................. 3
1. Basic Cellular Architecture ........................................ 3
2. Coverage................................................................... 5
3. Spectral Efficiency..................................................... 5
4. The Quest for better Coverage and Spectral Efficiency7
5. IntelliCell: The Fully Adaptive Smart Antenna
Approach....................................................................... 8
6. System-Level Benefits of IntelliCell......................... 11
7. IntelliCell Architecture ............................................ 16
8. Conclusion............................................................... 17
IntelliCell: A Fully Adaptive Approach to Smart Antennas
INTRODUCTION
Cellular communications has reached mass-market status over the past decade with the
emergence of two very successful standards: CDMA and GSM. Over this same decade, an
important enabling technology, smart antennas, has also matured. Combined with todays
powerful, low-cost processors, advanced smart antenna technology is destined to become an
important part of the cellular landscape over the next decade.

Smart antenna systems utilize multiple antennas at base stations or handsets to better
pinpoint or focus radio energy and thereby improve signal quality. Since cellular
communications systems employ radio signals that interact with the environment and each
other, these improvements in signal quality lead to system-wide benefits with respect to
coverage, service quality and, ultimately, the economics of cellular service. To some extent,
the phrase smart antennas is misleading. There is nothing smart about the antennas
themselves. Whats smart is the sophisticated signal processing applied to simultaneous
signals from an array or collection of multiple antennas.

For nearly a decade, ArrayComm has been at the forefront of developing smart antenna
techniques and intellectual property for commercial cellular systems. IntelliCell is the name
for these techniques and intellectual property. Thru eight years of practical and field
implementation, IntelliCell has been perfected to make smart antennas practical and cost
effective in actual commercial cellular systems. Today, IntelliCell technology is deployed in
more than 90,000 commercial base station deployments worldwide.
1. BASIC CELLULAR ARCHITECTURE
Cellular networks are composed of geographically separated base stations connected to a
backbone network, with each base station serving an area called a cell. (See Figure 1.) In
some systems, cells are further subdivided into sectors, for reasons that will be described later
in this document. The range of each base station may be anywhere from 0.5 km to 15 km,
with 1-3 km as the typical range in digital cellular systems. Handsets communicate with a
nearby base station via radio signals. The information, voice or data, is digitized prior to
transmission in all modern cellular systems. In the United States, most commercial cellular
systems operate in licensed radio frequencies in the region of either 850 MHz or 1.9 GHz.

End-to-end connections with public or private data or telephony networks are made possible
by a backhaul network that connects all of the base stations to a switching/routing function,
which directs users voice or data transmissions to and from their correspondents. Note that
this same network architecture is used for many types of wireless services, including wireless
LANs and point-to-multipoint data services such as LMDS.

ArrayComm 2002 Page 3 of 18
IntelliCell: A Fully Adaptive Approach to Smart Antennas
Figure 1: Basic Cellular Architecture

cell
sector
Data
Networks
Telephony
Networks
Backhaul
Network
base station
Switching/
Routing

In the radio portion of the network, the uplink refers to the communication from the handset
up to the base station: The handset or user terminal suitably digitizes and frames voice or
packet data meant for the network. This digitized data then is modulated using digital and
radio circuitry and transmitted via the antenna in the handset. The antennas and circuitry at
the base station receive the radio signal, demodulate it and send the users information on
into the wired network.

The downlink refers to the reverse direction, where the communication is from the base
station down to the handset or user terminal. The base station suitably digitizes and frames
voice or packet data meant for the subscriber. This digitized data is modulated using digital
and radio circuitry and is transmitted via the antennas at the base station. The antenna and
circuitry at the handset receive the radio signal, demodulate it and send the information on to
the subscriber.

This type of cellular architecture has gained wide acceptance as the most economical and
flexible architecture for delivering mass-market personal wireless services. The decline of
satellite based systems such as Iridium and Globalstar into niche services has proved this
point. Nevertheless, looking forward, cellular systems face a significant challenge as data
services and bandwidth become important. The challenge is to improve the quality of the
communication channel to handle larger traffic loads while maintaining the same cost
structure, despite the scarcity and exorbitant prices of additional spectrum. This challenge is
exacerbated by the expected trend away from todays low-data-rate digital voice services
toward high-data-rate broadband services. Todays cellular systems will require a 10-fold to
40-fold increase in spectral efficiency and capacity to affordably deliver true Internet content.

ArrayComm 2002 Page 4 of 18
IntelliCell: A Fully Adaptive Approach to Smart Antennas
2. COVERAGE
The base station range (cell area) determines the number of base stations required for a
particular coverage area in the early days of deployment, when subscriber density is low. It is,
therefore, one of the key determinants of system economics.

When radio energy propagates in a cellular environment, the received signal level degrades as
the distance between transmitter and receiver increases. This received signal has to exceed
the inherent noise level in the radio receiver by a certain margin in order to be successfully
demodulated. The ratio of the received signal level and this noise is called the signal-to-noise
ratio (SNR). Cellular systems are deployed with a certain average or nominal SNR target.
Everything else being equal, a higher nominal SNR translates into a higher possible data rate
but at the cost of reduced base station range.

For example, GSM systems are typically rolled out so that nominal SNRs are approximately 9
dB; while CDMA systems operate at nominal SNRs of 10 dB. (CDMA systems operate at these
much lower SNRs by introducing large redundancy into transmitted data through a process
known as spreading. In practice, the range of a CDMA base station is limited more by
interference among the users in the system than by receiver noise.)

The challenge for next-generation cellular systems is to maintain comparable coverage
footprints to second-generation cellular systems while delivering higher-data-rate service at
the edge of each cell. With allocated per-user spectral resources being equal, modern
information theory dictates that cell-edge data rates are determined by the received signal
quality, or SNR
1
. However, higher SNR levels in turn require incrementally smaller cells with a
correspondingly large economic implication for the service provider. Estimates show that
conventionally deployed 3G base station systems will have footprints that are two to four
times smaller than those of their 2G counterparts. As discussed below, smart antennas enable
a fundamental improvement to these tradeoffs.
3. SPECTRAL EFFICIENCY
Besides coverage, next-generation cellular systems face another challenge related to spectral
efficiency. Spectral efficiency measures the ability of a wireless system to deliver information
with a given amount of radio spectrum and is directly related to system capacity. It
determines the amount of radio spectrum required to provide a given service (e.g., 10 kbps
voice service, 100 kbps data service) and the number of base stations required to deliver that
service to the end-users. In the latter years of deployment, when subscriber penetration is
high, it is, consequently, one of the primary determinants of system economics.

Spectral efficiency is measured in units of bits/second per Hertz/cell (b/s/Hz/cell). It
determines the total throughput each base station (cell or sector) can support in a given
amount of spectrum. The appearance of a per cell dimension in measuring spectral efficiency
may seem surprising, but the throughput of a particular cells base station in a cellular
network is almost always substantially less than that of a single cell in isolation. This
difference is attributable to self-interference generated in the network.

Radio signals are unruly as compared to electrical signals propagating down wires. In a
cellular system, the radio communication between a user and a base station generates radio
energy thats detectable in places other than the immediate vicinity of the user, the base
station and an imaginary line between the two. For other users in the vicinity, this excess
energy degrades the radio channel, or makes it completely unusable for their conversations.
ArrayComm 2002 Page 5 of 18
IntelliCell: A Fully Adaptive Approach to Smart Antennas
As the user density increases, radio resources are thus eventually exhausted. Systems with
higher spectral efficiency provide more data throughput (services) in a given amount of
spectrum and support more users at a given grade of service before experiencing resource
exhaustion.

The key benefits of higher spectral efficiencies can be enumerated as follows: higher
aggregate capacity (per-cell throughput); higher per-user quality and service levels; higher
subscriber density per base station; small spectrum requirements; and lower capital and
operational costs in deployment.

The spectral efficiency for various systems can be calculated easily via the formula:

Spectral Efficiency = (Channel Throughput/Channel Bandwidth)

This simply sums the throughput over a channel in an operating network and divides by the
channel bandwidth. This calculation is performed for a number of systems in Table 1. As will
be discussed in later sections, the performance of IntelliCell is system dependent; however,
current implementations such as ArrayComms i-BURST system have achieved spectral
efficiencies of 4 bits/sec/Hz/cell. (For other systems, spectral efficiency typically is improved
by a factor of 5 to 20 over those of conventional, non-smart antenna implementations.)

To understand spectral efficiency calculations consider the PCS-1900 (GSM) system, which can
be parameterized as follows: 200kHz carriers, 8 time slots per carrier, 13.3 kbps of user data
per slot, effective reuse of 7 (i.e. effectively 7 channel groups at 100 percent network load, or
only 1/7
th
of each channels throughput available per cell). The spectral efficiency is therefore:

(8 slots x 13.3 kbps/slot) / 200 kHz / 7 reuse = 0.08 b/s/Hz/cell

Table 1 Spectral Efficiency for various common systems. The spectral efficiency of todays commercial
systems is invariably about 0.1-0.2 bits/sec/Hz/cell, while systems utilizing IntelliCell technology can achieve
spectral efficiencies up to 4 bits/sec/Hz/cell.

Air Interface Carrier BW Peak User
Data Rate
(kbps)
Average Carrier
Throughput
(kbps)
Spectral
Efficiency
Comments
IS95A 1.25MHz 14.4 100 0.08 Source: Viterbi
IS95B 1.25MHz 115 125 0.1 Source: Viterbi
IS95C 1.25MHz 144 200 0.16 Source: Viterbi
cdma2000 5MHz 384 800-1,000 0.16-0.2 Source: Viterbi
GSM 200KHz 13.3
15.2
(13.3*8/7)
0.08 Reuse = 7
GSM (HSCSD) 200KHz 57.6 15.2 0.08
PHS 300KHz 32 12.8 0.04
Effective reuse
= 20
Intellicell
System
Up to 4
Depends on
communication
system

This value of approximately 0.1 b/s/Hz/cell is generally representative of high-mobility 2G and
3G cellular systems, including CDMA systems of all types. It reflects the fact that the classical
ArrayComm 2002 Page 6 of 18
IntelliCell: A Fully Adaptive Approach to Smart Antennas
techniques for increasing spectral efficiency have been exhausted and that new techniques are
necessary. This will be the topic of the next section.

Finally, it should be noted that the value of 0.1 b/s/Hz/cell represents a major stumbling block
for the delivery of next-generation services. Without substantial increases in spectral
efficiency, 3G systems are bound to spectral efficiencies like those of todays 2G systems. In a
typical 3G system with a 5Mhz downlink channel block, this translates into a total cell capacity
of approximately 500 kbps for the entire cell. With services advertised in the range of 144-384
kbps, 1-3 users will completely occupy the cell capacity! This is far from the approximately
250-500 subscribers per cell needed to make the system economically viable, and it
underscores the need for new methods to boost spectral efficiency.
4. THE QUEST FOR BETTER COVERAGE AND SPECTRAL EFFICIENCY
A wide range of techniques and tradeoffs has been developed for enhancing coverage and
spectral efficiency over the past 20 years. The most important and widely used are the
following.

Frequency Planning: A substantial amount of the effort in cellular systems is devoted
to managing interference through the use of a reuse pattern. Traffic channels are
partitioned into groups where, say, each group has one-seventh of the total radio
resources, and channel groups are assigned to base stations in the network in such a
way that any two cells using the same channel group are not adjacent. The resulting
spatial separation ensures that the energy being used for a conversation in one cell has
been sufficiently attenuated by the time it reaches another cell using the same channel
that it does not pose significant interference. Reuse provides interference management,
but at the expense of operational complexity and base station capacity. A given base
stations channel group is a small fraction of the total system resources.
Power control: Power control is a technique whereby the transmit power of a base
station or handset is decreased to near the lowest allowable level that permits
communication. This reduces interference levels in the network, increasing spectral
efficiency. Power control is used in both GSM and CDMA systems on both the uplink and
downlink. CDMA systems require particularly fast and precise power control many
users share the same RF spectrum, and the system capacity is thus highly sensitive to
inadequate interference control.
Modulation and Coding: Modulation and coding techniques can improve the utilization
of spectrum by allowing a faster throughput at a given signal quality. The benefits of
any such techniques are ultimately limited, however, by the Shannon information rate.
Current techniques have brought commercial systems very close to this bound, and any
further spectral efficiency improvements from these techniques will be incremental, at
best.
Sectorization: Sectorized antenna systems take a traditional cell area and subdivide it
into sectors, each covered by its own directional antenna sited at the base station
location. Operationally, each sector is treated as an independent cell. Directional
antennas have higher gain than omni-directional antennas, all other things being equal.
Hence the range of these sectors is generally greater than that obtained with an omni-
directional antenna, roughly 35 percent greater. Sectorized cells can increase spectral
efficiency by reducing the interference presented by the base station and its users to the
rest of the network, and they are widely used for this purpose. Most systems in
commercial service today employ three sectors per site. Although larger numbers of
sectors are possible, the number of antennas and quantities of base station equipment
become prohibitively expensive for most cell sites.

ArrayComm 2002 Page 7 of 18
IntelliCell: A Fully Adaptive Approach to Smart Antennas
The reality of all these and other standard radio techniques like them is that their benefits
are, in essence, completely exploited in modern cellular systems. Further improvements in
coverage and spectral efficiency from these techniques will be small and well short of the
orders of magnitude needed for next-generation, broadband, wireless multimedia services. For
these services, new architectures and techniques need to be employed. One such technique is
the use of smart antennas such as those developed by ArrayComm for its patented IntelliCell
technology.
5. INTELLICELL: THE FULLY ADAPTIVE SMART ANTENNA APPROACH
The advent of powerful, low-cost, digital processing components and the development of
software-based techniques have made advanced adaptive antenna systems a practical
reality for cellular communications systems: Arrays of multiple antennas, combined with
digital beam-forming techniques and advanced, low-cost signal processing open a new and
promising area for enhancing wireless communication systems.

Terms commonly used to embrace various aspects of smart antenna system technology
include intelligent antennas, phased arrays, spatial processing, digital beam forming, adaptive
antenna systems, etc. ArrayComm has combined more than 8 years of practical R&D and field
experience into IntelliCell, a battery of techniques and intellectual property that make smart
antenna systems commercially viable.

A base station utilizing IntelliCell employs a small collection (array) of simple, off-the-shelf
antennas (typically 4 to 12) coupled with sophisticated signal processing to manage the
energy radiated and received by the base station. This improves coverage and signal quality
and mitigates interference in the network on both the uplink and the downlink.

The processes on the uplink and downlink are as follows:
5.1 The IntelliCell Uplink (reception at the base station)
Typically, the received signal from each of the spatially distributed antenna elements is
multiplied by a weight, a complex adjustment of amplitude and phase. These signals are
combined to yield the array output. An adaptive algorithm controls the weights according to
predefined objectives such as tuning in to a particular user while tuning out interference
and noise. This processing is performed independently and simultaneously for each of the
users being served by the base station.

These dynamic calculations enable the system to tune itself for optimized signal reception:
The equivalent received signal level is improved by a factor of 10log10 (number of antennas),
which, for example, is 10 dB for a 10-antenna system.

At the same time, interference is rejected by many orders of magnitude, anywhere from 30 to
50 dB if an interfering signal is strong enough to warrant it. This rejection and the analogous
suppression on the downlink are high enough that, in TDD/TDMA implementations of IntelliCell
such as in ArrayComms i-BURST, frequency planning can be done away with completely.

These gains and how they relate to overall gains in signal quality are summarized in Figure 2.

ArrayComm 2002 Page 8 of 18
IntelliCell: A Fully Adaptive Approach to Smart Antennas
Figure 2 Effects of IntelliCell on Signal Quality. The signal level itself is enhanced while the level of interference
is reduced and suppressed.
Reduced
Interference
Noise
Floor
O Or ri ig gi in na al l
S Si ig gn na al l
O Op pt ti im ma al ll ly y E En nh ha an nc ce ed d
Enhanced
Signal Level
( (d dB B) )

5.2 The IntelliCell Downlink (transmit from the base station)
Similar gains occur on the downlink. The signals to be transmitted are multiplied by weighting
factors of different amplitude and phase for each antenna. The weighting factors are chosen
dynamically to ensure that the transmitted signals constructively combine and add at the user
of interest while at the same time presenting no interference to other co-channel users
2
. The
weight factors are again chosen dynamically based on predefined objectives.

These dynamic calculations enable the system to tune itself for optimized signal transmission:
the equivalent transmitted-power signal level is a factor of 20log10 (number of antennas) over
the power emitted by a single antenna at the base station. This is, for example, 20dB for a 10-
antenna system. This is a monumental improvement in equivalent signal level. Because the
signals constructively interfere at the targeted user, for example, 10 1-Watt transmitters at
the base station produce an equivalent incident radiation as a single, 100W transmitter. In
addition, the redundancy introduced through the use of multiple transmitting elements,
combined with the reduction in power amplifier size, increases the base station reliability.
Smaller-power amplifiers are more reliable and less expensive than larger ones, and the loss
of a single transmitting element from the array has only a small effect on base station
downlink performance (as opposed to the case where the base station has only a single
radiating element).

At the same time, interference is mitigated by 30 to 40 dB if a nearby user (interferer) is close
enough to the base station to warrant it.

An important point here is that the type and performance of the downlink processing used
depends on whether the communication system uses time division duplex (TDD) schemes,
which transmit and receive on the same frequency (e.g., 802.11, PHS and DECT) or frequency
division duplex (FDD) schemes, which use separate frequencies for transmit and receiving
(e.g., GSM, EDGE, W-CDMA, cdma2000). In most FDD systems, fading and other propagation
characteristics are uncorrelated from the uplink radio channel to the downlink one, whereas in
ArrayComm 2002 Page 9 of 18
IntelliCell: A Fully Adaptive Approach to Smart Antennas
TDD systems, the uplink and downlink channels can be considered reciprocal. Hence, in TDD
systems, uplink channel information may be used to achieve spatially selective transmission.
In FDD systems, the uplink channel information cannot be used directly, and other types of
downlink processing must be considered.
5.3 Some Simple Math
Conceptually, all this works as in the simple model shown in Figure 3. Imagine a simple, two-
antenna base station attempting to communicate to two users, User A and User B, on the
same channel. Also imagine that since the signals from these users travel along different
paths to the base station, they arrive in the following combinations at the antenna array:

User A Signal at the base station: (+A, +A)
User B Signal at the base station: (+B, -B)

Note the difference is these signatures: User As signals arrive in phase between the two
antennas and user Bs signals arrive out of phase between the two antennas. These
signatures are commonly referred to as spatial signatures. In a real-world implementation,
these signatures are vectors in an M-dimensional complex space, where M is the number of
antennas.

These signals arrive together at the base station and combine to become:

Base Station Received Signal: (+A + B, +A - B)

Now, very simply, in order for the base station to extract user As signal from the interference
caused by user B, it simply adds the two signals with weight factors (1,1):

Extract User A: (+1, +1) (+A + B, +A - B) = (+A +B) + (+A B) = 2A

and similarly for user B, the weight vector (+1, -1) is used:

Extract User B: (+1, -1) (+A + B, +A - B) = (+A +B) - (+A B) = 2B

ArrayComm 2002 Page 10 of 18
IntelliCell: A Fully Adaptive Approach to Smart Antennas
Figure 3 Simple Model of Adaptive Antennas. Users signals arrive with different relative phases and
amplitudes at the array. Weights are applied in order to extract the signals for particular users.


+1
-1
+1
+1
A - B A + B
User B

User A

2A 2B

In each case, by use of an appropriate weight vector, the base station is able to extract and
separate the signals of user A and user B from one another while simultaneously providing
gain for each!

In a conventional, single-antenna system, this simple process would have been impossible,
and the base station would have been left unable to decipher the two signals. Imagine a base
station with only a single antenna. In this case, the received signals would be modeled as:

Single Antenna Received Signal = (+A +B)

and the base station is left with a confused jumble of signals from User A and User B
intermixed with one another.

Of course, in real-world implementations of IntelliCell, there are multiple complications that
must be handled: There are more co-channel users to decipher; there are multiple other
sources of interference; there are many more antennas; signal levels and phases vary across
the array; and so on not to mention that the base station may not even know or have good
estimates for the spatial signatures!
6. SYSTEM-LEVEL BENEFITS OF INTELLICELL
At the simplest level, IntelliCell systems fundamentally improve the coverage and spectral
efficiency tradeoffs of wireless systems. Wireless system design, nevertheless, still involves a
series of tradeoffs between cost, coverage and capacity. The improvements from IntelliCell
ArrayComm 2002 Page 11 of 18
IntelliCell: A Fully Adaptive Approach to Smart Antennas
fundamentally allow much more flexibility in system-level designs. The benefits are
summarized in Table 2 and further explained in detail in this section.

Table 2 System-Level Benefits of IntelliCell

Gain System-Level Significance
Selective Uplink Gain
Receive processing at
base station
Increased Range, Coverage, Link budget
10*log10(M) gain
13dB 17dB diversity gain
Lower terminal transmit power
Uplink multipath immunity
Lower-complexity equalization
Uplink Interference Mitigation
Receive processing at
base station
Improved Signal Quality
Robust to interference from multiple uplink
interferers
30dB 40dB interference immunity
Higher spectral efficiency
Selective Downlink Gain
Transmit strategy based
on uplink information
and feedback from
terminal
Increased Range, Coverage, Link budget
20*log10(M) gain
13dB 17dB diversity gain
Reduced base station PA sizing
Reduced downlink multipath
Lower-complexity equalization at terminal
Downlink Interference Mitigation
Transmit strategy based
on uplink information
and feedback from
terminal
Improved Signal Quality
Automatically reduces signal transmission to co-
channel interference
Increases system-wide downlink signal quality
30dB 40dB interference immunity
Higher spectral efficiency
6.1 Selective Uplink Gain: 10log10(M)
As mentioned before, IntelliCell significantly improves uplink link budgets by a factor of the
number of antennas. More formally, this can be seen as follows. As in the simple example
above, M copies of the same signal, s, are received, one per antenna, where M is the number
of antennas. Assuming that the signals arrive with the same power, an appropriate application
of weights will lead to the signals adding together coherently:

Uplink Received Signal After Processing = s + s + + s

(There is a subtlety that has been avoided here. In reality, the signals do not arrive at the
array with the same power due to fading processes that differ among the antennas. This will
be discussed later in this white paper.)

Similarly, after the application of the weights, the noise processes, N
i
, in each of the antenna
receivers add:

Uplink Noise After Processing = N
1

+ N
2

+ + N
M

ArrayComm 2002 Page 12 of 18
IntelliCell: A Fully Adaptive Approach to Smart Antennas

Since the received signals add coherently and the noise powers add independently (since the
noise processes are independent and identically distributed), this leads to:

Multiple Antenna Uplink SNR = (Ms)
2
/ M
2
= Ms
2
/
2
= M (Single Antenna SNR)

where s
2
is the signal power,
2
is the noise power per antenna, and as before, M is the
number of antennas.

Taking log10 of both sides, we obtain a 10log10(M) gain in signal-to-noise ratio. With 10
antennas and under typical propagation conditions, this leads to approximately a doubling of
range and quadrupling in coverage. Of course, this increase in range can be traded off against
other system parameters such as the required user terminal transmit output power.
6.2 Selective Downlink Gain: 20log10(M)
A similar calculation to the above can be performed on the downlink. In the downlink case, M
copies of the same signal, s, are transmitted, one per antenna. Assuming that the signals
arrive with the same power, an appropriate use of the transmit weighting factors will lead to
the signals adding together coherently at the handset:

Downlink Received Signal at Handset = s + s + + s

In this case, the receiver noise floor at the handset is independent of the antenna weightings
used for base station transmit, so the multiple antenna and single antenna SNR can be
compared as follows:

Multiple Antenna Downlink SNR
= (Ms)
2
/
2
= M
2
s
2
/
2
= M
2
(Single Antenna Downlink SNR)

In contrast to the uplink gain of 10log10(M), the downlink gain is, therefore, 10log10(M
2
) =
20log10(M). (The extra factor of M compared to the uplink should not be surprising, since we
have assumed M times more total radiated power in the multi-antenna case). For example, 10
1-Watt base station amplifiers thus perform like a single antenna with a 100W power output.
This level of effective output power is often overkill in a balanced link budget. In fact, the base
station radios are often wideband in an IntelliCell system in order to support the large system
capacity. Such wideband radios are highly price sensitive to output power, and the 20log10(M)
gain mitigates the output power requirements for these radios.

As mentioned before, a significant issue in forming the downlink weights is obtaining
knowledge of the downlink signature. In the uplink, the base station can employ any number
of passive techniques such as using training sequences to obtain the uplink signature. In the
downlink, however, the methods are highly dependent on the type of communication system
employed. TDD systems have a reciprocal property in that the downlink signature is more or
less proportional to the uplink signature. In FDD systems, the relation between uplink and
downlink is not so simple, and a complex nonlinear mapping between the two along with other
techniques based on direction of arrival or feedback from the handset is often necessary.

Finally, another important complication arises because the signals received by the base station
and transmitted by the base station flow through different electronic circuitry. In the simplest
case, this induces a simple multiplicative transformation on spatial signatures, and in the
worst case, this induces a linear or even nonlinear transformation on the signatures.
ArrayComm 2002 Page 13 of 18
IntelliCell: A Fully Adaptive Approach to Smart Antennas
Correcting for this effect is known as calibration, and techniques for calibration are important
elements of the techniques in IntelliCell.
6.3 Selective Uplink and Downlink Gain: Diversity
All wireless systems suffer some degree of fading, which is the unavoidable consequence of
reflections with short time lags constructively and destructively interfering at the receiving
antenna. Since the environment is dynamic, the fades themselves are time varying. The
consequence for wireless system designers is that the air interface must be robust to sudden
outages (for example, using interleaving of symbols), and margins against fading must be
introduced into link budgets and cell planning, which reduces coverage.

In the above calculation, one simplifying step was to restrict the discussion to the non-fading
case. For example, all the received signal power levels between the antennas on the uplink
were assumed to be equal. In reality, the signals across the array will fade just as in a
conventional single-antenna system.

This fading is substantially mitigated, however, in the multiple-antenna case. When one
antenna fades in the array, chances are that others do not. The output of the array is,
therefore, much smoother over time. Thus, there is a reduction in the needed margin against
fading, which is often referred to as a diversity gain.

The calculation of this gain depends on the targeted outage probability, the detailed
assumptions regarding the fading process and the number of antennas. Under a large class of
assumptions, the averaging provided by the array yields reductions in the equivalent margin
of 13-17 dB.

It is important to note that this diversity gain is in addition to the standard 10log10(m) gain:
Both the conventional, single-antenna system and the multiple-antenna system require a
fading margin, but the multi-antenna system requires a much lower margin.

ArrayComm 2002 Page 14 of 18
IntelliCell: A Fully Adaptive Approach to Smart Antennas
Figure 5 Signals Across a 4-Antenna Array. One antenna undergoes a large fade, while the other antennas
show slight variation. Though individual antennas fade in and out, the total composite signal is far more stable.


Figure 5 shows real data collected from a 4-antenna array in a suburban environment. The
horizontal axis denotes a 2.5-second interval, while the vertical axis indicates the power level.
Each of the lower curves shows the power levels for each antenna, respectively, while the
upper curve denotes the composite power. The composite power is simply the power level
obtained after applying appropriate weights via IntelliCell processing. The individual antennas
in the array fade in and out independently. As illustrated, the total composite power is far
more stable. A conventional single-antenna system would need provisions and margins
against the deeper fades exhibited by a single antenna, while an IntelliCell system requires a
far smaller margin.

It is interesting that the common industry practice is to ignore such fading margins in
published link budgets such as those for CDMA and GSM. However, fading is a very real effect
that must be accounted for in the planning of real cellular systems. A testament to this is that
most cellular systems employ two receive diversity antennas on the uplink at the base station.
(Together with a single downlink antenna, this yields the characteristic 3-antenna-per-sector
configuration most common today.)

6.4 Uplink and Downlink Interference Mitigation
The uplink and downlink interference mitigation provided by IntelliCell processing is perhaps
the most remarkable of the system benefits. In practical implementations, uplink interference
can be suppressed between 30 to 50 dB. (The simple mathematical example above illustrated
infinite suppression because interference from user Bs signal was completely eliminated by
the weights applied to derive user As signal, and vise versa.)
ArrayComm 2002 Page 15 of 18
IntelliCell: A Fully Adaptive Approach to Smart Antennas

In the downlink direction, practical issues often mean that the ability to null or mitigate
transmitted interference is less than can be achieved in the uplink. As before, the performance
in the downlink direction is dependent on the particular communication system. In a TDD
system, practical issues limit the nulling performance to between 30 and 40 dB.

It should be noted that within the fully adaptive approach of IntelliCell, the weighting and the
receive and transmit interference suppression and mitigation are performed continuously and
dynamically in very short time frames. This is a necessity: The radio environment is time
varying on the scale of tens of milliseconds; and the network and service environment itself is
time varying with in-cell and out-of-cell interferers appearing and disappearing with
fluctuating power levels. The null placement is so precise that any slight variation in the
signatures of the interferers can result in a 30 to 40dB null to be cut to a 10dB null,
substantially reducing the network performance.
6.5 Higher Spectral Efficiency
The reduction in interference due to IntelliCell allows an increase either in the number of
subscribers utilizing the spectrum or in the overall signal quality, which enables higher data
throughput. The upshot of the reduction in interference network-wide is, in either case, an
increase in spectral efficiency. The order-of-magnitude increases possible are in the range of
20-40X versus non-IntelliCell implementations. As shown before, spectral efficiency of existing
systems is no mystery and can be calculated from their well-documented performance
characteristics. These values typically yield spectral efficiencies of 0.1 bits/sec/Hz/cell. Fully
adaptive implementations utilizing IntelliCell have been shown to achieve factors of 20X
improvement in actual field performance. Furthermore, an approach such as ArrayComms i-
BURST, which designs and optimizes an air interface from the bottom up to utilize IntelliCell,
can achieve spectral efficiency of 4 bits/sec/Hz/cell, or 40X conventional performance.

While the details of these performance numbers are complicated, a simple calculation justifies
this order-of-magnitude performance increase. Current, commercially available smart antenna
systems based on TDD/TDMA air interfaces are able to achieve an intra-cell reuse of 3. In
these SDMA (spatial division multiple access) systems, the same conventional resource (in this
case, a timeslot of the same carrier) is used multiple times in the same sector or cell. Todays
conventional TDMA systems such as GSM operate at a reuse of 7 i.e., the same timeslot and
frequency combination is used in every 7
th
cell. The order-of-magnitude performance gain is,
therefore, 3 divided by 1/7, or a factor of 21.
7. INTELLICELL ARCHITECTURE
IntelliCell systems employ a highly integrated approach in base station design, with the smart
antenna architecture incorporated from the outset. Figure 6 shows the block diagram of such
an architecture.

IntelliCell systems make use of ordinary, off-the-shelf antennas. Received radio signals are
digitized and accumulated by the receiver bank. This received data is then packaged and
processed in the spatial temporal processing block. This block is the heart of the IntelliCell
system and typically involves the use of high-performance digital signal processors and ASICs.
This block extracts and demodulates the various signals of interest and appropriately packages
the results for transport through the network interface and on into the network.

ArrayComm 2002 Page 16 of 18
IntelliCell: A Fully Adaptive Approach to Smart Antennas
At the same time, data is being received from the network bound for subscribers. The spatial
temporal processing block communicates with the transmitter bank to indicate how the data is
to be weighted across the different antennas. The operations in the transmitter bank consist of
many, albeit simple, multiplication operations. Finally, the modulated data is routed through
power amplifiers (PAs), one for each antenna, and transmitted across the array.

Figure 6 Typical IntelliCell Base Station Architecture



8. CONCLUSION
IntelliCell technology uses sophisticated signal processing techniques in combination with
small arrays of standard, off-the-shelf antennas to manipulate signals at the base station and
dynamically control transmission and reception. Conventional radio systems indiscriminately
broadcast energy, creating interference for other users. Using IntelliCell processing, base
stations optimize radio transmission and reception by selectively amplifying signals to / from
users of interest and rejecting unwanted signals. This substantially increases the signal quality
and suppresses and mitigates interference on both the uplink and downlink radio channels,
resulting in increased coverage and spectral efficiency.

The conventional techniques used to increase coverage and spectral efficiency have been
exploited over the past 20 years to the point that gains from these techniques are
incrementally small. Despite these efforts, the spectral efficiency of commercial air interfaces
is typically only about 0.1-0.2 bits/sec/Hz/cell, independent of the technologies used. This
performance is well short of what is needed to deliver broadband wireless data services
ArrayComm 2002 Page 17 of 18
IntelliCell: A Fully Adaptive Approach to Smart Antennas
ArrayComm 2002 Page 18 of 18
economically. IntelliCell offers a solution with spectral efficiencies of up to 4 bits/sec/Hz, or 40
times those of conventional systems.

Finally, though this white paper has focused on IntelliCell implementations at the base station,
the technology is equally applicable to handsets and subscriber units. Todays trends in
handset component costs and processing capabilities point to this being the next major
frontier for IntelliCell and smart antenna technologies.

1
It may be argued that higher cell edge data rates may be easily obtained by allocating a large bandwidth
to a particular subscriber at one time. However, this solution proportionally increases handset output
power requirements on the uplink and per-user power levels on the downlink, proportionally increases
handset battery power requirements, and degrades system capacity.
2
The somewhat confusing technical names for these two effects are respectively constructive
interference of signals at the user and destructive interference of signals at other co-channel users.