110 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 52, NO.

1, JANUARY 2004

Merging Multicarrier CDMA and Oscillating-Beam

Smart Antenna Arrays: Exploiting Directionality,
Transmit Diversity, and Frequency Diversity
Seyed Alireza Zekavat, Member, IEEE, Carl R. Nassar, Senior Member, IEEE, and Steve Shattil

AbstractIn this paper, a novel merger of multicarrier code-di- of directionality in the BS antenna array and an increased MS
vision multiple access (MC-CDMA) and smart antenna arrays is receiver complexity.
introduced. Here, a group of carriers in the MC-CDMA system Antenna beam-pattern oscillation, a new transmit diversity
is applied to its own -element smart antenna array at the base
station (BS). The smart antennas are located in close proximity
scheme introduced in [6][12], offers both very good proba-
to one another. We generate a transmit diversity gain at the re- bility-of-error performance at the MS receiver via its transmit
ceiver by carefully moving (oscillating) the antenna arrays pat- diversity benefits, and high capacity via SDMA (directionality).
tern. The pattern oscillation is achieved by applying appropriate Both benefits are available while maintaining low MS receiver
time-varying phases to array elements of each smart antenna. The complexity. Specifically, a unique, carefully controlled time-
beam pattern oscillation ensures a mainlobe at the position of the varying phase shift is applied to each antenna array element,
intended user and small oscillations in the beam pattern. This beam
pattern oscillation leads to a time-varying channel with a control-
sweeping the beam pattern over a small angle [less than the an-
lable coherence time; hence, a transmit diversity benefit, in the tenna half-power beam width (HPBW)] in one symbol duration
form of a time diversity benefit, is available at the receiver. Em- . After each , the antenna beam returns to its initial posi-
ploying MC-CDMA with the proposed smart antenna at the BS, we tion, and sweeps the same area of space over the next . This
achieve: 1) directionality which creates high network capacity via beam pattern movement can be carefully controlled to create
space-division multiple access; 2) a transmit diversity gain which a time-varying channel capable of introducing a time diversity
supports high performance at the receiver in the mobile unit; and
3) increased capacity and performance via MC-CDMAs ability to
gain at the receiver side.
support both CDMA and frequency diversity benefits, respectively. Specifically, small beam pattern movement (small enough to
maintain directionality) leads to a time diversity gain as follows.
Index TermsAntenna arrays, code-division multiple access
(CDMA), diversity method.
Assuming a rich-scattering environment (e.g., a city center), and
allowing the beam pattern to move, a notable number of scat-
terers depart the beam patterns mainlobe, while a noticeable
I. INTRODUCTION number of other scatterers enter the beam patterns mainlobe.
As a result, the summation over scatterers in the mainlobe and
A NTENNA ARRAYS located at the base station (BS) en-
hance wireless communication systems via: 1) direction-
ality, which supports space-division multiple access (SDMA);
combining their complex gains to create the observed Rayleigh
or Rician fade, changes significantly from one time to the next.
or 2) more recently, a transmit diversity benefit, i.e., a diversity This translates directly into an induced time diversity. The sim-
scheme that uses the antenna array at the BS to achieve diver- ulations of [8][11] confirm that small beam pattern movement,
sity at the mobile station (MS). Transmit diversity schemes in- less than 5% of the mainlobe width, creates up to seven-fold
troduced recently include delay diversity [1], [2], antenna hop- time diversity in mid-sized city centers.
ping [3], antenna hopping combined with frequency hopping The proposed beam pattern oscillation technique, which sup-
[4], and the phase sweeping technique [5]. The disadvantages ports both directionality and a transmit diversity benefit, repre-
of the above-mentioned transmit diversity techniques are a loss sents a significant innovation to the jitter diversity method in-
troduced in [13], where directionality is maintained and perfor-
mance is improved somewhat by jittering the beam pattern to
Paper approved by R. Kohno, the Editor for Spread Spectrum Theory and reduce the probability of deep fades. In this paper, we merge the
Applications of the IEEE Communications Society. Manuscript received
November 1, 2001; revised June 3, 2002 and April 9, 2003. This work was beam pattern oscillation technique with multicarrier code-divi-
supported by NASA Phase II SRIR: Development of a Wireless Communica- sion multiple-access (MC-CDMA) systems.
tion System to Support Airport Surface Operations. This paper was presented MC-CDMA [14] has emerged as a powerful competitor to
in part at the IEEE Vehicular Technology Conference, Birmingham, AL,
May 2002, in part at the IEEE Third-Generation Wireless Conference, San direct-sequence CDMA (DS-CDMA). In MC-CDMA, a user
Francisco, CA, May 2002, and in part at the IEEE Emerging Technologies sends his data stream over carriers (simultaneously). Mul-
Symposium on Broadband Communications for the Internet Era, Dallas, TX, tiple users are supported by allowing each user to transmit over
September 2001.
S. A. Zekavat is with the Department of Electrical and Computer En-
the same carrier at the same time, and separability (at the re-
gineering, Michigan Technological University, Houghton, MI 49931 USA ceiver) is achieved by assigning each user a unique spreading
(e-mail: [email protected]). sequence (typically values) to apply to the carriers.
C. R. Nassar is with the Department of Electrical and Computer Engineering, Recently, carrier interferometry (CI) codes were introduced
Colorado State University, Ft. Collins, CO 80523-1373 USA.
S. Shattil is with CIAN Systems, Inc., Superior, CO 80027 USA. for MC-CDMA, where complex codes replace the 1 and 1
Digital Object Identifier 10.1109/TCOMM.2003.822180 code values [15][17]. Applying these codes, the MC-CDMA
0090-6778/04$20.00 2004 IEEE
ZEKAVAT et al.: MERGING MULTICARRIER CDMA AND OSCILLATING-BEAM SMART ANTENNA ARRAYS 111

system achieves high capacity by allowing orthogonal users Here, is employed to maintain carrier orthogo-
to occupy the channel simultaneously, and, if further users are to nality, and is a rectangular waveform of unity height over
be supported, an additional users can be added with minimal 0 to . Finally, refers to the th element of user s
interference and high performance via low correlation between spreading sequence, where traditionally,
all users codes.
(i.e., ). However, when CI codes are em-
In this paper, we innovatively apply oscillating-beam an-
ployed
tenna arrays to MC-CDMA systems using CI codes (referred
to as CI/MC-CDMA in this paper). This enables very high
capacity via the merger of SDMA (directionality of antenna (3)
array) and CDMA (inherent in CI/MC-CDMA), and very high
performance via the construction of receivers that exploit both for users (i.e., ). In
transmit diversity and frequency diversity. this way, orthogonal users are supported (similar to case of
At the BS, each set of carriers in the HadamardWalsh codes); moreover, if additional users are to be
CI/MC-CDMA -carrier system is fed into a unique supported, an additional users can be introduced pseudo-or-
-element smart antenna array. The smart antennas are thogonally by adding users with spreading codes characterized
located in close proximity to one another (at the BS). By by
carefully designing the phase shifts applied to each antenna
array element, and by properly setting the distances between (4)
antenna elements, the resulting beam pattern corresponds to
an oscillating beam pattern similar to that in [6] and [7]. This for users , (i.e.,
leads to a time-varying channel with a controllable coherence ).
time. The controllable coherence time is used by the MS to
exploit time diversity and enhance performance. B. Proposed Antenna Array Structure
We locate the -element smart antennas in close proximity
Assuming a CI/MC-CDMA system with subcarriers, we
to each other, facilitating the construction of the entire set of
divide the subcarriers into sets, each with subcarriers
-element antenna arrays via a single structure (e.g., printed
. We apply each set of subcarriers to one -element
circuit), and enabling a single BS mounting. This allows our
antenna array. Fig. 1 shows the th set of subcarriers (with
proposed system to be designed with minimal cost.
frequencies and with codes with phase
Receivers are constructed to exploit the transmit diversity,
which corresponds to an induced time diversity, as well as the offsets ) entering the th antenna array at the BS. The
frequency diversity of the CI/MC-CDMA system. Different th array element of the th smart antenna applies the phase
receiver structures are examined, including equal-gain com- shift , .
bining/maximal-ratio combining/minimum mean-square error The value of , the value of , the distance between the ad-
combining (EGC/MRC/MMSEC ) performed in the time and jacent antenna elements, (see Fig. 1), and the distance be-
frequency domains. tween an element in the th antenna array and the same element
Section II introduces the merger of the beam-sweeping an- in the th antenna array, (which represents the relative
tenna array and CI/MC-CDMA. Section III presents receiver phase difference between these antenna elements), are designed
structures employing EGC across frequency components fol- to generate identical beam patterns for carriers in a single an-
lowed by EGC, MRC, or MMSEC across the time-domain com- tenna array, and identical HPBW across all the smart antennas.
ponents (and vice versa). Section IV presents simulated perfor- We require identical beam patterns for the carriers of an
mance results, while Section V presents a conclusion. antenna array and identical HPBWs across all antenna arrays
to avoid spatial overlaps that would degrade the performance of
some of the mobile users. Specifically, consider Fig. 2(a), which
II. MERGER OF CI/MC-CDMA AND illustrates one of the problems in the case of nonidentical beam
BEAM-SWEEPING ANTENNA ARRAYS patterns (across carriers of the antenna array). In the example of
A. CI/MC-CDMA System Fig. 2(a), we assume that the smart antenna is creating sector-
ization (not adaptive beamforming) for ease in presentation, and
In MC-CDMA, each users bit is transmitted simultaneously that each sector is determined based on the narrowest mainlobe
over narrowband subcarriers. Subcarriers are equally spaced (among the nonidentical beam patterns). Considering mobile ,
in frequency by . To ensure separability of users at the re- it is evident that mobile receives information from both his
ceiver side, a unique spreading code is applied to each users desired sector of space and an adjacent, interfering sector (in-
carriers. Hence, the th users th data bit, , is sent as dicated by the shaded area). As a direct result, mobile s per-
formance is significantly degraded by the signals from the in-
(1) terfering sector.
One solution to the problem of interfering signals from adja-
where is 1 or 1, is the center or carrier frequency, cent sectors (in the case of unequal beam patterns) is shown in
and is the th spreading code, corresponding to Fig. 2(b). Here, by creating sectors based on the widest main-
lobe, it is evident [from Fig. 2(b)] that the lightly shaded sector
(2) (adjacent sector) no longer provides interfering signals to mo-
bile . However, mobile now experiences a different limi-
112 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 52, NO. 1, JANUARY 2004

Fig. 1. The pth smart antenna array over which the pth group of CI/MC-CDMA signal frequency components are sent.

sector. (Similar problems are noted in the case of antenna ar-

rays generating adaptive beamforming).
To avoid the concerns highlighted by Figs. 2(a) and (b) (and
detailed above), it is important to design MC-CDMA/smart
antenna systems that generate beam patterns that are identical
across carriers.
Criterion 1. Identical Beam Patterns for Carriers in a Single
Antenna Array: For each antenna array, all carriers in the an-
tenna array must create beam patterns that are effectively iden-
tical. That is, for center frequency and carrier separation

(5)

This criteria, which determines a suitable value for , ensures

(a) that all carriers create identical beam patterns in a single an-
tenna array.
Criterion 2: Identical HPBW Across Antenna Arrays: To en-
sure identical HPBW across the antenna arrays, we require

(6)

where is the average wavelength of

the carriers applied to the th antenna array. In addition to (6),
we introduce a condition which ensures small spacing between
adjacent antenna arrays, namely

(7)
(8)

(b) where (7) ensures that the first and second antenna array are sep-
Fig. 2. (a) Interference from the adjacent sector, if the patterns are positioned arated by distance , and, in (8), represents
based on the smallest HPBW. (b) Reduction in some carriers received signal the cumulative distance between the th antenna array and the
power, if the patterns are positioned based on the largest HPBW.
first, and represents the antenna array number. To satisfy (6),
(7), and (8), and are chosen according to
tation that again leads to a degraded performance. Specifically,
consider the mainlobe denoted by dark shading in Fig. 2(b), the (9)
mainlobe of one of the carriers containing information for mo-
bile . Evident from the figure, mobile lies outside of this
carriers mainlobe, meaning that this signal arrives at mobile
at a significantly reduced power. This will be the case for a (10)
growing number of carriers as the user enters the edge of the
ZEKAVAT et al.: MERGING MULTICARRIER CDMA AND OSCILLATING-BEAM SMART ANTENNA ARRAYS 113

where is selected as shown in (7). Using (9) and (10), we

achieve an elegant expression for the normalized antenna array
factor (AF) (considering all antenna arrays)

(11)
where
Fig. 3. Antenna element delay line function.

(12) beam pattern (from the th antenna array) must remain in the
antenna array HPBW. That is
(13)
(16)
Here, refers to the antenna arrays phase shift ,
which is well approximated as a function of time alone (as where denotes the HPBW (equal for all smart antennas), is
shown next in Criterion 3). For ease in presentation, we have the azimuth angle, is the rate of antenna pattern move-
assumed the intended mobile is located at . With ment, and is the amount of antenna pattern move-
, if is sufficiently large (e.g., ment in . The parameter , selected such that ,
), simple mathematical manipulation shows that the guarantees that the received antenna pattern amplitude is within
total antenna array beam pattern in (11) can be approximated the 3-dB beamwidth for the entire symbol duration. Equation
by (16) corresponds to selecting the antenna array control param-
eter according to (based on the calculations of [7])

(14)

(17)
(15) where

(18)
Equation (15) demonstrates that careful parameter selection
in the bank of antenna arrays leads to an AF that is very close Here, is the BS-MS distance, and is a term that al-
to the AF generated by a typical single element antenna array. lows the time-varying phase, , applied to the th antenna,
Specifically, the AF in (15), a tight approximation to the antenna to compensate for the difference in antenna array/mobile dis-
arrays exact AF in (11), is identical to the AF of an -element tance (between antenna array and antenna array 1). Equation
linear array with one exception. There is a time-varying param- (18) indicates that, because antenna arrays are very close to-
eter in the AF equation, namely, the time-varying phase offset gether relative to array/mobile distance, such a compensation is
. In traditional antenna arrays, the value of in (15) is a not necessary.
fixed value (e.g., denoted in [18]). By applying a time-varying After each , returns to its value (returning
, we create a movement of the beam pattern, i.e., the posi- the beam pattern to its original position) and then
tion of the peak value, and the direction of the entire mainlobe, recreates an identical spatial movement over the next dura-
is a function of time. By careful control of the parameter , tion. Assuming a small HPBW, (17) can be simplified [using
we will move the beam pattern over each symbol duration . (6)] to
We will repeat the same beam pattern movement over each ,
and, as a direct consequence, we denote the beam pattern move-
(19)
ment generated by as a beam pattern oscillation.
The beam pattern oscillation is carefully controlled to ensure
that, over each period of the oscillation (i.e., over each symbol When coupling the of (19) with the AF equation of (15),
duration ), two criteria are achieved. The first criterion is that we observe the following. The phase offset, , which con-
the mean and variance of the Rayleigh fading must be constant trols the directionality of the antenna array [as seen in the AF
over the symbol duration , i.e., a constant large-scale fading equation of (15)], is a linearly increasing function of over
must be observed. The second criterion, which leads directly to each symbol duration . That is, graphically, the phase offset
a time diversity benefit in the proposed system, is that the beam that controls directionality is the sawtooth waveform shown in
pattern movement ensures independent fades within the rep- Fig. 3. When applying this to the AF of (15), we observe that
etition period . Details regarding how to select to meet over each period of , the beam pattern sweeps a small area
these two criteria are provided in the next two subsections. of space, then returns to its starting position, where it again
Criterion 3: Constant Large-Scale Fading: To ensure con- sweeps the same area of space. This is highlighted conceptu-
stant large-scale fading over each symbol time duration , the ally by Fig. 4.
114 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 52, NO. 1, JANUARY 2004

Fig. 4. Antenna beam pattern movement over each symbol duration T .

The periodic nature of the sweeping facilitates excellent pa- i.e., a time diversity gain is observed. A detailed statistical char-
rameter estimation. Specifically, as we will show in the next sub- acterization of the parameters in (20) and (21) can found in [8]
section, the beam-pattern oscillation is capable of generating and [9], and combined with the channel impulse response model
independent fades over each period of duration . When the of (20), we can determine the time-varying nature of the channel
beam pattern returns to its initial position and sweeps the same via the well-known statistical measure of coherence time .
area over the next symbol duration, it creates the same set of The evaluation as a fraction of symbol duration is de-
independent fades over the next symbol duration . With tailed in [8], where simulation results with a beam pattern move-
an identical set of parameter fades from one symbol duration ment less than 5% of the antenna arrays mainlobe width, in the
to the next, tracking and estimation of the channel gains and rich-scattering environment of a mid-sized urban area, lead to
phase offsets becomes as simple as using a bank of locked an available diversity gain on the order of . Using this
loops or feedforward tracking loops, or as simple as using phase result in our work, we assume beam-pattern oscillations lead to
differentials between adjacent symbol times. an -fold transmit diversity.
Criterion 4: Independent Fades: Movement of the antenna
array beam pattern based on the time-varying antenna array III. RECEIVER DESIGN
phases in (17) or (19) results in a time-varying channel. Specif-
The th users transmitted signal, without an antenna array
ically, based on our earlier work in [8] and [9], the time-varying
[using (1) and (2)], corresponds to
channel is well represented by the linear time-varying impulse
response model

(22)
For carrier , this carrier is the th carrier sent on
(20)
antenna array , and can be written in isolation using
Here, is the observation time and is the response time of
the impulse at each . In addition, represents the number
of multipath components in the beam patterns mainlobe, which
changes with time as the beam pattern moves; represents (23)
the angular position of the th scatterer in the coverage area; where is as shown in (3) and (4). Now, considering the
represent the gain due to the th scatterers posi- antenna array, the signal in (23) enters the th smart antenna
tion in the antenna array mainlobe [see (15)] and array. The output of the th element of the th antenna array,
is the corresponding phase offset; is after application of phase offset in (19), is simply
the amplitude component due to channel effects; charac-
terizes the phase offset due to the channel effects; is the time
of arrival of the propagation path from the th scatterer; and
represents the impulse function. (24)
In the case of a flat-fading channel, i.e., the delay spread is The presence of creates a frequency offset; however, with
much less than the symbol duration, the channel model of (20) selected according to (19) (and considering
simplifies to ), and assuming typical parameter values for mobile com-
munication transmissions, it is easily shown that the frequency
offset induced by is less than 0.5% of a 1-MHz bandwidth.
Hence, we ignore this frequency offset in our presentation. The
total downlink transmitted signal, considering all antenna ele-
(21)
ments (all ), all carriers (all and ) and all users (all ) is
The arrival and departure of scatterers, which leads directly [from (24)]
to the time-varying nature of in (21), causes the channel
impulse response to vary as a function of time. In the case of a
flat-fading channel, illustrated in (21), the summation of com-
plex gains (from 1 to ) changes, and, if this change is rapid
enough, multiple fades are observed over one symbol duration, (25)
ZEKAVAT et al.: MERGING MULTICARRIER CDMA AND OSCILLATING-BEAM SMART ANTENNA ARRAYS 115

Fig. 5. Mobile receiver.

At the receiver side, considering the transmit diversity, which ments), and combining the two summations over and for
leads to an -fold time diversity, the received signal in simplicity in presentation, (26) can be rewritten as
can be divided into time slots , where
, and each time slot has independent fade. For
, this signal corresponds to

(28)

Here, is the normalized AF of the smart antenna

arrays introduced in (14). Assuming a beam pattern directed
toward the intended user and movement of the beam pattern
by less than 5% of its HPBW, we can approximate
, and we can allow at the mobile.
The CI/MC-CDMA receiver for user , , is
shown in Fig. 5. In this figure, a bank of bandpass filters sepa-
rates the MC-CDMA signal into its multiple carriers. Next,
(26)
each carrier is multiplied by a local oscillator generating

Here, is a white Gaussian noise, independent for

differing carriers and different time slots ; (29)
is the fade on the th carrier in the th time slot (due Once each of the MC-CDMA carriers (located in parallel
to fading), and is the phase offset in the th branches) has been returned to baseband, the baseband signal is
carrier and th time slot (due to fading) (hereafter, phase is as- integrated over each interval over which the fade is constant,
sumed to be tracked and removed), and is as shown i.e., over , .
in (12). From [19], the fades over the subcarriers After applying the spreading code of user to separate users,
(applied to the th smart antenna) are the received signal for each subcarrier , ,
correlated Rayleigh random variables with correlation coeffi- and time interval , , corresponds to
cient between the subcarrier fade and the subcarrier fade,
characterized by

(27)
(30)
where is the coherence bandwidth of the channel. Ap- where and is a zero-mean Gaussian
plying the summation over (the set of antenna array ele- random variable with variance . The first term
116 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 52, NO. 1, JANUARY 2004

in (30) represents the th component of the desired signal, the where

second term is multiaccess interference (MAI), and the third
term is noise. (35)
With diversity components, over frequency and
over time, the combiner can be designed using EGC, MRC,
or MMSEC, applied in frequency, then in time (or vice versa).
Examples of the combiner follow. D. EGC (Frequency)-MRC (Time) Scheme
In this subsection, we first apply EGC across the frequency
A. EGC-EGC Scheme diversity components available via the MC-CDMA transmit
EGC represents the optimal method for combining informa- signal, leading to a decision variable per time diversity com-
tion sent at multiple times or over multiple frequencies in the ponent (index )
cases of slow, flat-fading channels or additive white Gaussian
noise (AWGN) channels. In the proposed scenario, where
beam pattern movement creates different fades at different
times, and the channels frequency diversity leads to different
fades at different frequencies, EGC combining represents a
suboptimal scheme. However, because, as the name suggests,
EGC combining does not require tracking the unique fades for
combining purposes, it is very simple to apply. For this reason,
we consider the use of EGC applied across the time 13 compo-
nents, followed by EGC across the frequency components, or
vice versa, leading to the combiner output (36)

(31) Next, we apply MRC to combine across the time diversity

components created via beam pattern movement, leading to the
final decision statistic at the output of the combiner
B. MRC (Frequency)-EGC (Time) Scheme
MRC represents the optimal combining strategy in a (37)
single-user system that experiences unique fades on its di-
versity components. Here, we apply MRC across the frequency where
diversity components (a suboptimal scheme due to the multiuser
nature of MC-CDMA). We follow this with an EGC across the (38)
time diversity components, leading to a combiner output

(32)
E. MMSEC (Frequency)-EGC (Time) Scheme
It is worth noting that applying MRC across the time diversity MMSEC is an excellent combining strategy in a multicarrier
components, followed by EGC across the frequency diversity system, as it attempts to jointly minimize the MAI and the noise,
components, leads to the same decision statistic of (32) at the while simultaneously exploiting the frequency diversity benefit.
output of the combiner. Applying this combining first across the frequency components
in a CI/MC-CDMA system leads to the combiner output (per
C. EGC (Time)-MRC (Frequency) Scheme time index , and assuming users on an carrier
In this subsection, we present the combiner that first performs system) [20]
an EGC across the time components (per carrier), leading to, for
carrier , the decision statistic (39)

where is the noise variance of , i.e.,

and corresponds to
for or
(40)
else.
(33) Next, applying a simple EGC across the time diversity com-
ponents leads to
Next, MRC is performed across the carrier components,
leading to the final decision statistic output

(34)
(41)
ZEKAVAT et al.: MERGING MULTICARRIER CDMA AND OSCILLATING-BEAM SMART ANTENNA ARRAYS 117

(a) (b)

(c) (d)
L
Fig. 6. Simulation results for = 7 independent fades, K N
= 32 users, and = 32 subcarriers. (a) MRC, MMSEC, and EGC first in freq., then EGC in time.
(b) MRC, MMSEC, and EGC first in time, then EGC in freq. (c) EGC first in time, then EGC, MRC, and MMSEC in freq. (d) EGC first in freq., then EGC, MRC,
and MMSEC in time.

F. EGC (Time)-MMSEC (Frequency) Scheme G. EGC (Frequency)-MMSEC (Time) Scheme

We also note, for the sake of completeness in presentation,
In this subsection, we first perform EGC across the time that if EGC is combined first across frequency components, and
diversity components, leading to decision statistic per carrier MMSEC is then applied across the resulting time diversity
(index ) components, the output decision statistic of the combiner corre-
sponds to

(42)
(44)
Next, we apply a MMSEC across carrier components, leading
to the final decision statistic output by the combiner
where is defined in (36), is defined in (38),
, and is the
noise variance of , i.e., .
(43) IV. SIMULATED PERFORMANCE
where , , and are as defined in (42), (35), and (40), For simulation purposes, we consider the following.
respectively, and is the noise variance of , 1) subcarriers in the CI/MC-CDMA system. (We
i.e., . assume subcarriers for the following reason.
118 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 52, NO. 1, JANUARY 2004

Todays CDMA system, IS-95, operates with a spreading performance gains when added to MC-CDMA/antenna array
gain of 32; todays IEEE802.11a WLAN system operates systems.
with 52 carriers; and tomorrows wideband (W)CDMA
system will operate with a processing gain of 128. Hence, V. CONCLUSIONS
we believe that selections of 32, 52, or 128 are all
excellent choices, and we select because it leads At the BS, antenna arrays are used to send CI/MC-CDMA
to consistency with earlier work, e.g., [14], [21].) signals, while at the MS, a single antenna is used for reception.
2) antenna arrays with carriers per array. The phase shifts introduced to the BS antenna array elements
3) A CI/MC-CDMA system with orthogonal users control the antenna pattern such that it moves (oscillates), en-
on the carriers. ables the system to benefit from both directionality and transmit
4) A frequency selectivity resulting in four-fold frequency diversity.
diversity over the entire bandwidth, i.e., High network capacity is achieved as the resulting mobile
. system benefits from both CDMA and SDMA, and high perfor-
5) Beam pattern movement resulting in independent mances are demonstrated in this paper by receivers that exploit
fades in the duration . diversity in both the frequency and time domains. Performance
Simulation results are provided in Fig. 6(a) for EGC, MRC, curves for these receivers show the significant performance im-
and MMSEC combining across frequency terms, followed by provement available via the small beam-pattern oscillation.
EGC in the time domain. It is seen that the performance results
of EGC-EGC exceed those of MMSEC-EGC and MRC-EGC REFERENCES
for signal-to-noise ratios (SNRs) greater than 7 dB. A similar [1] N. Seshadri and J. H. Winters, Two signaling schemes for improving
conclusion can be made when considering the performance the error performance of frequency-division duplex (FDD) transmission
curves of Fig. 6(b), where EGC, MRC, or MMSEC combining system using transmitter antenna diversity, Int. J. Wireless Inform. Net-
works, vol. 1, pp. 4960, Jan. 1994.
schemes are applied in the time domain, followed by EGC [2] J. H. Winters, The diversity gain fading of transmit diversity in wire-
across the frequency terms. less systems with Rayleigh, in Proc. ICC/SUPERCOMM, vol. 2, New
Orleans, LA, May 1994, pp. 11211125.
In Fig. 6(c), simulation results are provided for EGC across [3] H. Olofsson, M. Almgren, and M. Hookin, Transmitter diversity
the time domain terms, followed by EGC, MRC, or MMSEC with antenna hopping for wireless communication systems, in Proc.
in the frequency domain. EGC-EGC curves demonstrate the IEEE Vehicular Technology Conf., vol. 3, Phoenix, AZ, May 1997, pp.
best performance among these schemes. Only minor differences 17431747.
[4] R. W. Heath, Jr. and A. Paulraj, Transmit diversity using decision-di-
are observed among the many performance curves of Fig. 6(d), rected antenna hopping, in Proc. IEEE Communications Mini-Conf.,
where EGC is applied in the frequency domain, followed by an June 610, 1999, pp. 141145.
EGC, MRC, or MMSEC combining in the time domain. [5] A. Hiroike, F. Adachi, and N. Nakajima, Combined effects of phase
sweeping transmitter diversity and channel coding, IEEE Trans. Veh.
The simulation results of Fig. 6(a)(d), presenting perfor- Technol., vol. 41, pp. 170176, May 1992.
mance curves for the proposed CI/MC-CDMA/antenna array [6] S. A. Zekavat, C. R. Nassar, and S. Shattil, Combined directionality
and transmit diversity via smart antenna spatial sweeping, in Proc.
system with a moving beam pattern, are compared with a care- 38th Allerton Annu. Conf. Communication, Control and Computing,
fully selected benchmark system. As a benchmark for compar- Oct. 2000, pp. 203211.
ison, we use the CI/MC-CDMA system coupled with a slightly [7] , Smart antenna spatial sweeping for combined directionality and
different antenna array structure. A bank of antenna arrays are transmit diversity, J. Commun. Networks, Special Issue on Adaptive
Antennas for Wireless Commun., vol. 2, no. 4, pp. 325330, Dec. 2000.
used, but this time the antenna array system is used to achieve [8] S. A. Zekavat and C. R. Nassar, Smart antenna arrays with oscillating
a directionality benefit and no time diversity benefit. That is, beam patterns: characterization of transmit diversity using semi-elliptic
in the benchmark system, the beam pattern generated by the an- coverage geometric-based stochastic channel modeling, IEEE Trans.
Commun., vol. 50, pp. 15441556, Oct. 2002.
tenna array structure achieves an improved network capacity via [9] , Semi-elliptic coverage geometric-based stochastic channel mod-
directionality, but the beam pattern does not move, meaning that eling for smart antenna arrays with oscillating beam patterns, in Proc.
no time diversity benefit is available to this system. This allows IEEE Globecom, San Antonio, TX, Nov. 2529, 2001, pp. 32373241.
[10] , Fading channel characterization for oscillating beam pattern
us to determine the benefit of small beam-pattern oscillations in smart antennas using geometric-based stochastic channel modeling
a CI/MC-CDMA/antenna array system. with circular coverage scenario, in Proc. IEEE Vehicular Technology
The performance of the benchmark system, a system without Conf., Atlantic City, NJ, Oct. 711, 2001, pp. 14521456.
beam pattern movement, is illustrated by the top curves of [11] , Geometric-based stochastic channel modeling for adaptive an-
tennas with oscillating beam patterns, in Proc. IEEE Int. Symp. Per-
Fig. 6(a)(d). This benchmark system is able to exploit the sonal, Indoor, Mobile Radio Communications, PIMRC01, San Diego,
four-fold frequency diversity that can be achieved by use of CA, Sept. 2001, pp. 130134.
CI/MC-CDMA, but does not benefit from any additional time [12] , Diversity gains in wireless systems with oscillating beam smart
antennas: An evaluation via geometric-based stochastic channel models
diversity gains. We observe that this benchmark system suffers with point scatterers, in Proc. SPIE Int. Symp. Convergence of Infor-
from approximately a 7-dB performance degradation when mation Technologies and Communication, ITCOM, Denver, CO, Aug.
compared with the proposed MC-CDMA/antenna array system 2024, 2001, pp. 118124.
[13] O. Norklit, P. C. F. Eggers, and J. B. Andersen, Jitter diversity in mul-
with a moving beam pattern. That is, the time diversity benefit tipath environments, in Proc. IEEE Vehicular Technology Conf., vol. 2,
available by moving the beam pattern, when coupled with the 1995, pp. 853857.
frequency diversity benefit available via MC-CDMA, leads to [14] S. Hara and R. Prasad, Overview of multicarrier CDMA, IEEE
a 7-dB performance gain relative to a system capable only of Commun. Mag., vol. 35, pp. 126133, Dec. 1997.
[15] B. Natarajan, C. R. Nassar, S. Shattil, Z. Wu, and M. Michelini, High
exploiting the frequency diversity benefit. In other words, small performance MC-CDMA via carrier interferometry codes, IEEE Trans.
beam-pattern oscillations are capable of offering significant Veh. Technol., vol. 50, pp. 13441353, Nov. 2001.
ZEKAVAT et al.: MERGING MULTICARRIER CDMA AND OSCILLATING-BEAM SMART ANTENNA ARRAYS 119

[16] C. R. Nassar, M. Michelini, B. Natarajan, and S. Shattil, Introduction Carl R. Nassar (S90M90SM02) received the
of carrier interference to spread spectrum multiple access, in Proc. Bachelors, Masters, and Ph.D. degrees, all from
IEEE Symp. Emerging Technologies in Wireless Communications Sys- McGill University, Montreal, QC, Canada.
tems, Richardson, TX, Apr. 1213, 1999. He is an Associate Professor of Telecommunica-
[17] Z. Wu, B. Natarajan, C. R. Nassar, and S. Shattil, High performance, tions in the Department of Electrical and Computer
high capacity MC-CDMA via carrier interferometry, in Proc. IEEE Engineering, Colorado State University, Fort Collins.
Int. Symp. Personal, Indoor, Mobile Radio Communications, vol. 2, San There, as Founder of the RAWCom (Research in Ad-
Diego, CA, Sept./Oct. 2001, pp. 1116. vanced Wireless Communications) Laboratory, he di-
[18] C. A. Balanis, Antenna Theory. New York: Wiley, 1997. rects research in the areas of multiple access, mul-
[19] J. W. C. Jakes, Ed., Microwave Mobile Communications. New York: tiantenna, and multicarrier systems. In addition, he
Wiley, 1974. serves as the Vice President of Wireless Research for
[20] S. Haykin, Adaptive Filter Theory, 2nd ed. Englewood Cliffs, NJ: Idris Communications, an innovate telecommunications firm dedicated to the
Prentice-Hall, 1991. advancement of multicarrier technologies in wireless. He is the author of two
[21] S. A. Zekavat, C. R. Nassar, and S. Shattil, The merger of a single os- books and over 90 journal articles and conference publications, has served re-
cillating-beam smart antenna and MC-CDMA: transmit diversity, fre- peatedly as conference chair, and was the recipient of awards for his research
quency diversity and directionality, in Proc. IEEE Emerging Technolo- efforts from both the state of Colorado and his university.
gies Symp. Broadband Communications for Internet Era, Dallas, TX,
Sept. 2001, pp. 107112.

Steve Shattil received the B.S. degree in physics

from Rensselaer Polytechnic Institute, Troy, NY, the
M.E. degree in electrical engineering from Univer-
Seyed Alireza Zekavat (S87M02) received the sity of Colorado, Boulder, and the M.S. degree in
M.Sc. degree in electrical engineering from Sharif physics from the Colorado School of Mines, Golden.
University of Technology, Tehran, Iran, in 1993, He is currently Chief Scientist at CIAN Systems,
and the Ph.D. degree in electrical and computer Inc., Superior, CO, where he is leading development
engineering from Colorado State University, Fort in carrier interferometry and other coding technolo-
Collins, in 2002. gies. He has over 15 years experience in the wireless
From 1993 to 1998 he was with Civil Aviation Col- industry. Prior to founding CIAN, he was Founder
lege of Technology, Tehran, Iran. Since August 2002, and Chief Technical Officer of Genesis Telecom.
he has been with Michigan Technological University Genesis invented highly bandwidth-efficient antenna array technologies for
(MTU), Houghton. He is the Director of the Labora- broadband wireless communications. During his time at the Colorado School of
tory for Wireless Communication Research at MTU. Mines, he built the first laser that generates carrier interferometry signals. Over
His research interests include wireless communications, radar theory, statistical the last decade, he has led the development of baseband and RF processors for
modeling, adaptive beam forming, and neural networks. He has published over broadband products, and is world renowned in both academia and industry for
35 journal and conference papers, and has coauthored the book Multicarrier his innovation, leadership, and the advancement of wireless communications.
Technologies for Wireless Communications (Boston, MA: Kluwer, 2001). He has authored dozens of domestic and international patents.