IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 1, NO.

1, JANUARY 2002

169

Successive Interference Cancellation Algorithms for

Downlink W-CDMA Communications
Mohamed F. Madkour, Member, IEEE, Someshwar C. Gupta, Fellow, IEEE, and Yi-Pin Eric Wang

AbstractIn this paper, successive intracell interference

cancellation (IIC) of the wideband-code division multiple access
(W-CDMA) signal at the mobile unit is considered. Three new
interference cancellation techniques suitable for the downlink of
any CDMA system with orthogonal spreading are proposed. No
prior knowledge of users spreading codes or even their spreading
factors are required for interference cancellation. A new term,
effective spreading code, has been introduced in this paper which
is defined as the interfering user physical code as seen by the
desired user within the desired user symbol duration. The mobile
receiver estimates the effective spreading codes of the interfering
users regardless of their spreading factors using fast Walsh
transform (FWT) correlators (instead of the regular correlators)
and uses this information to suppress the intracell multiuser interference. Three different interference-suppressing techniques are
studied: subtraction; combined interfering signal projection; and
separate interfering signal subspace projection. The complexity
of the proposed techniques is low compared to conventional
interference cancellation techniques. For a W-CDMA system and
the IMT-2000 vehicular channel model, a capacity increase of up
to 150% of the original (without IIC) system capacity is shown.
Index TermsDownlink interference cancellation, multirate
CDMA, multiuser detection, spread spectrum techniques, wideband CDMA.

I. INTRODUCTION

HE third generation (3G) cellular mobile communications

systems will support several kinds of communication services, including, e.g., voice, images, and even motion picture
transmission [1]. Therefore, the users will be transmitting their
information signals using different data rates and their performance requirements will vary from application to application.
wideband-code division multiple access (W-CDMA) with variable spreading factor (SF) and multicode modulation as a multirate scheme [2], [4] is emerging as one of the air interfaces
for the 3G mobile communications systems. The high and different user data rates and the large number of users together
with multipath dispersive fading channels cause severe intercell
and intracell multiuser interference. Fundamental investigations
[5][8] have demonstrated huge potential capacity and performance improvements as a result of using multiuser detection
at the expense of increasing complexity of optimum structures.
Manuscript received July 11, 2000; revised October 18, 2000, July 12, 2001;
accepted August 1, 2001. The editor coordinating the review of this paper and
approving it for publication is R. Yates.
M. F. Madkour is with the Mobility Network Consulting, Ericsson Inc, Plano,
TX 75024 USA.
S. C. Gupta is with the Electrical Engineering Department, Southern
Methodist University, Dallas, TX 75275-0338 USA.
Y.-P. E. Wang is with the Ericsson Inc, Research Triangle Park, NC 27709
USA.
Publisher Item Identifier S 1536-1276(02)00188-5.

In general, a major problem with multiuser detectors and interference cancellers is the maintenance of simplicity. Even the
sub-optimal linear detectors require considerable complex processing.
There are several algorithms for interference cancellation for
CDMA systems [9][11]. Most of these algorithms are designed
for the uplink. For uplink interference cancellation, it is assumed
that the receiver knows all the spreading codes. This assumption is however not true for the downlink where the mobile
unit only knows its own spreading code. Furthermore, the complexity of the interference cancellation algorithms proposed up
to date is very high. For the downlink, since interference cancellation has to be performed at a hand-held battery-operated terminal, cost and power consumption are of great concern. In this
paper, we focus on the downlink communications. A successive
intracell interference cancellation mobile receiver is proposed.
The receiver estimates the interfering users effective spreading
codes and uses them to suppress the interference on the received
signal.
Some researchers are trying to estimate the interfering user
actual spreading code and then use it in the cancellation process
using different ways, either by equalization idea or by subtractive cancellation, etc. [13], [14] (very few are dealing with variable spreading factor case). In this paper, we are not trying to
estimate this physical code (because we will deal with variable
spreading factor case), but rather estimating what we call effective spreading code (ESC). This is the effective spreading code
of the interfering user as seen by the desired user within the
desired user symbol duration. This ESC belongs to the Walsh
space seen by the desired user. We combine this estimation
process with the interference cancellation to get a better code
estimate after every iteration and hence better cancellation and
better performance. Three different techniques have been considered in the canceller: subtraction; combined interfering signals projection; and separate interfering signals subspace projection. In this paper, we deal with the intracell multiple-access interference (MAI), therefore, we assume all the own-cell
users use the same scrambling code. Interference from other
cells (intercell interference) is modeled as AWGN. Our work
is different than previous work [10][15] in the sense that it is
done for a multirate system and no prior knowledge of the users
spreading codes or spreading factors are needed at the mobile
unit. Link-level simulation for the W-CDMA system in a frequency-selective fading channel (IMT-2000 vehicular channel
model [1]) is used to test the proposed receivers.
This paper is organized as follows. In Section II, the downlink W-CDMA signals properties are discussed. The signal
model for both fixed and variable processing gain is presented

15361276/02$17.00 2002 IEEE

170

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 1, NO. 1, JANUARY 2002

in Section III. The new proposed cancellation techniques are

presented in Section IV. In Section V, we discuss how to reduce
the complexity of the proposed techniques. The simulation
results are presented in Section VI. Finally, Section VII concludes the paper.

where
is a rectangular pulse of duration ; and
is
the th chip of the spreading code of user . Since the spreading
codes are mutually orthogonal, then

II. DOWNLINK W-CDMA SIGNAL PROPERTIES

The base station combines all the signals and scrambles it by a

, which is given by
base station specific scrambling code

The downlink signals for different physical channels within

a cell are transmitted synchronously by the base station [1],
[2]. Typically, orthogonal spreading codes are assigned to distinct physical channels, thereby creating mutually orthogonal
downlink signals. If the channel does not have delay spread,
the orthogonality can be maintained at the despreader output of
the receiver, thereby removing all multiple-access interference
from the same cell. However, for dispersive channels, orthogonality can no longer be maintained at the receiver, giving rise
to intracell multiuser interference, which will result in performance degradation. Such performance degradation could be severe if the near-far problem occurs. For the uplink, power control is used to alleviate the near-far problem. However, power
control cannot alleviate the near-far problem for the downlink.
Thus, in a highly dispersive channel, the system could suffer
significant capacity loss as a result of loss of orthogonality between downlink signals. Also, it is well known that a signal of
wider bandwidth can resolve more multipath [3]. As W-CDMA
has 4 times as much bandwidth as IS-95, the channel experienced by the W-CDMA signal exhibits higher dispersion. As
a result, the degradation due to intracell interference will be
much more than IS-95 [3]. Moreover, W-CDMA utilizes variable spreading factors (SF) to provide various transmission data
, to 512. Typrates. The spreading factor varies from
ically for voice channels, spreading factor 128 is used; and for
very high-speed data services, a spreading factor as low as 4 can
be used. The rate difference between the aforementioned two
cases is a factor of 32. Normally, the same link quality (e.g.,
where
is the energy per information data bit) has to
be maintained for both cases. This implies the signal with a low
spreading factor has to be transmitted with much larger power
than the one with a high spreading factor [6]. This scenario further deteriorates the performance when the orthogonality between the downlink signals is lost.
III. SIGNAL MODEL
A. Fixed Spreading Factor (FSF)
be the signal for the th user after spreading, but
Let
before scrambling.

(3)

(4)
The combined and scrambled signal is given by
(5)
is the transmitted
where is the total number of users; and
power for the th user signal. Assume that the physical channel
(same for all users) is given by the complex low pass equivalent
impulse response
(6)
and are the th complex
where is the number of paths;
path gain and delay, respectively. We assume that the channel
has fixed number of paths and introduces an Additive White
Gaussian Noise (AWGN). Assume that the path gain and the
delay are fixed, and that the channel is stationary over the
longest symbol period. The envelope of each complex path
amplitude has a Rayleigh distribution and each path phase
has a uniform distribution over the interval
.
and are assumed to be known at the mobile unit by channel
estimation (a perfect channel estimation will be assumed
throughout this paper). The received signal can then be given
by
(7)
is the Gaussian noise component which includes
where
both thermal and intercell interference noise. A typical CDMA
receiver for this case consists of fingers for despreading. Despread values are weighted and combined to form a decision
statistic. Without loss of generality, lets assume the demodu(bit number zero of the 1st user signal). The th
lation of
is
where
finger output

(1)
is the th data symbol for the th user;
is the
where
spreading waveform of the th user; is the spreading factor;
and is the chip duration. The spreading waveform is given by
(2)

(8)
represents the th finger correlators output of the 1st
where
the output of its intersymbol interusers desired signal;
the output of the multiuser interference
ference (ISI) terms;
the output of the noise signal (all corresponding
terms; and

MADKOUR et al.: SUCCESSIVE INTERFERENCE CANCELLATION ALGORITHMS

Fig. 1.

171

An example of the various spreading factor case.

to the 0th transmitted bit).

to be

; and

can be shown
(9)

(10)

and

(11)
respectively
(12)

, the ISI term,

, can be neglected. NeAssuming
glecting the ISI is good for large spreading factors (low data rate
voice users) but it might not be so good for low spreading factors (high data rate users). But as the main intracell interference
in 3G will be from the data users with low spreading factor on
the voice users with high spreading factors, we will focus on the
performance of voice users in this paper.

spreading factors is guaranteed. In order to perform interference

cancellation in a multirate CDMA system in a similar manner
as in a single rate system, the concept of effective spreading
code (ESC) is introduced. The ESC for any interfering user
depends on the SF, actual spreading code, and the information
data symbols of that user and has the same duration as the
spreading code of the desired user. Therefore, the ESC used
by the interfering users belong to the Walsh set and vary from
symbol to symbol depending on the information data symbols.
Fig. 1 is an example illustrating the case that three users are
using different spreading factors. The spreading factors for
users 1, 2, and 3 are 16, 8, and 4, respectively. It can be shown
that these signals are mutually orthogonal no matter which
spreading factor (16, 8, or 4) is used for despreading. Assume
user 1s signal is the desired signal. From Fig. 1, to the symbol
, it appears that user 2 uses spreading code (1 1 1 1
1111
) which is a concatenation of
two spreading code actually used by user 2. Similarly, to the
, it appears that user 2 uses spreading code (1 1 1
symbol
1 1 1 1), with a data symbol
1
. Thus, the spreading codes used by the interfering users
of
(users 2 and 3) seem to be varying from symbol to symbol,
depending on the data symbols of the interfering users. The
mathematical illustration of that can be developed as follows.
,
be the spreading factors for users #1, 2, and
Let ,
3, respectively. The spread data signal of these users can be
written as
(13)

B. Variable Spreading Factor (VSF)

In W-CDMA, various spreading factors are used to provide various symbol rates. In this case, Orthogonal Variable
Spreading Factor (OVSF) codes are used [2]. With the
OVSF codes, orthogonality between codes of same or different

and

(14)
(15)

172

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 1, NO. 1, JANUARY 2002

The signal of user #2 can be rewritten as

(16)
and
, the effective spreading wavewhere
form of user #2 for the th symbol of user #1, is
(17)
Similarly, the signal of user #3 can be re-written as
(18)
and
, the effective spreading wavewhere
form of user #3 for the th symbol of user 1, is

(19)
Using this concept, the intracell multiuser interference of VSF
case can be shown to be

(20)
and
are the th user th symbol effective
where
spreading code and effective data symbol, respectively.
IV. SUCCESSIVE IIC MOBILE RECEIVERS
From (20), it is obvious that if the ESC; the effective data
symbols and the power levels were known, the intracell multiuser interference would have been calculated and canceled out
on finger basis. Here, instead of using a conventional
from
used correlator at each finger (correlates against the desired
users code), we consider a receiver uses a fast Walsh transform
(FWT) to correlate against all orthogonal codes. This corre-fold increase in complexity. After getsponds to an
ting FWT at each finger, these values are combined according to
the maximum ratio combining (MRC) principle. If a spreading
is used by the base station, at the MRC one would see
code
that the energy of this code is higher than those not used. Thus,
a code detector can compare the MRC outputs with a threshold.
If the energy of a particular code is higher than the threshold,
then that code is detected. Alternatively, the receiver can decide
codes with the highest energy. In either case,
to detect only
it is possible that there are codes present in the composite received signal not detected. This will result in a residual interfer-

ence after cancellation. However, since the interference due to

the codes with the highest energy is removed, the degradation
caused by the residual interference is small. Finally, the modulation values can be also detected after the MRC. Based on
that, we propose three different cancellation techniques. Two
of them are based on combining the estimated interfering signals and then canceling their effects from the received baseband
signal either by direct subtraction or by projecting the combined
interfering signal onto the orthogonal direction to the received
baseband signal. In our third technique, we propose not to combine the interfering users signals. Instead, they are kept separate. Then, the received baseband signal is projected on the
subspace orthogonal to the interfering signals subspace on a
symbol by symbol basis.
A. Combined-Interfering Signals Receivers
The proposed mobile unit receiver structure is shown in
Fig. 2. The receiver estimates the effective spreading codes
of the interfering users regardless of their spreading factors.
The basic idea is decoding the strongest users and then
canceling out their contributions from the received baseband
signal and then using this modified received signal to have a
better estimation of the effective spreading codes. Here, the
strongest user is not known beforehand, but is detected from
the strength of the output correlation values after the FWT
processor. The receiver combines the
correlation values
from the fingers according to the maximum ratio combining
principle. The desired users correlation value is then fed into
the decision circuit to be decoded. Here, we have the choice
to stop or continuing cancellation according to a performance
criterion (e.g., quality of service or CRC check). To continue
correlation values are passed
cancellation, the other
on to a selector which determines the strongest correlation
value (to save processing time, we can also choose to select the
correlation values to cancel out
interfering
maximum
signals at a time) and generates the Walsh code corresponding
to the index of this value. The selected users signal is then
decoded, spread, scrambled and passed on to a regenerated
version of the multipath channel, generating a baseband version
of the interfering signal to be canceled out from the received
baseband signal.
The cancellation process is done using two techniques; subtraction or projection. To illustrate these two techniques, let
;
; and
be the received baseband signal; the regenerated baseband interfering signal; and the modified received
baseband signal, respectively. In the subtraction technique, the
generated baseband interfering signal is directly subtracted from
. In the projection
the received signal
technique, we project the received baseband signal on to the orthogonal direction to the generated baseband interfering signal
where
denotes the
(
inner product of and ). Projection technique is a little bit
more complex than the subtraction technique, however. The interference cancelled in the projection method requires unit energy constraint. It is to be noted that the proposed receiver cancels out the intracell multiuser interference using the combining
correlation values after the MRC and not on a finger by finger

MADKOUR et al.: SUCCESSIVE INTERFERENCE CANCELLATION ALGORITHMS

173

Fig. 2. The proposed downlink W-CDMA combined-interfering signals receiver.

basis as most of the interference cancellation schemes [11]. The

process of estimating the ESC and its correlation value and regenerating an estimate of the interfering user signal has to be
done on a symbol-by symbol basis, but when it comes to the
cancellation process (the projection process), we can do it on a
frame-by-frame basis, giving up some performance but reducing
the processing power. What we do is that we save the regenerated interfering user symbols in a buffer until we have a whole
frame and then projecting the received signal vector on it at one
shot (making the system implementable even for very high data
rates).
B. Subspace-Projection Receiver
In this receiver, we propose not to combine the interfering
users signals. Instead, they are kept separate. Then, the received
signal is projected on the subspace orthogonal to the interfering
signals subspace on a symbol by symbol basis. The proposed
mobile unit receiver structure is shown in Fig. 3. The receiver
estimates the effective spreading codes of the interfering users
regardless of their spreading factors using the same procedures
as those in the previous combined-interfering signals receiver.
The difference between the subspace approach and our previous
techniques is that the selected users signals (intracell interfering users) are kept separate from each other and passed individually on to a regenerated version of the multipath channel,
generating an arbitrary number (subspace dimension) of baseband interfering signals. Now a subspace that spans these interfering signals is generated and the received signal is projected
on to a subspace orthogonal to that space. The geometric representation of the subspace projection technique at the th iteration is shown in Fig. 4. The modified received baseband signal
is again passed on to the code detector to have a better estimate of the interfering users effective spreading codes. These

procedures are repeated until a satisfactory estimate has been

reached. It is to be noted that although the code detection procedures have to be done on a symbol-by-symbol basis because
of the VSF, the subspace projection procedures could be done
on both symbol-by-symbol and frame-by-frame basis. However,
as will be seen from the simulation results, symbol-by-symbol
subspace projection gives better results at the expense of the increased computational complexity.
In order to further understand this subspace cancellation approach, a mathematical explanation of cancellation of the
strongest interferers using the Gramm-Schmidt technique [19]
will now be provided. We will drop the symbol index, , for
estimated interferers vectors at the
ease notation. Let the
th iteration be
through
. In order to find the subinterferers vectors at the th iteration,
space that spans the
the Gram-Schmidt procedures [19], operating on the ordered
(descending order according to the correlation values) vector
, is carried out. The Gram-Schmidt
set
procedure is performed as follows.
(21)
(22)

(23)
(24)
denote unit vectors that span a subwhere
defined by interferers
through
. Prospace
is achieved
jecting a vector onto the space orthogonal to

174

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 1, NO. 1, JANUARY 2002

Fig. 3. The proposed downlink W-CDMA separate-interfering signals receiver (subspace approach).

Using (26), (27) can be rewritten as

(28)
For

iterations, the modified received signal vector, , is

(29)

V. REDUCED-COMPLEXITY MOBILE RECEIVER

Fig. 4. Geometric representation of the subspace projection scheme at the ith

iteration.

by left multiplication using the following projection matrix [18]

(the iteration index, , will be dropped for notational simplicity).
(25)
(26)
, is the received baseband signal
where the initial value,
are orthonormal vectors,
vector . Since the columns of
(25) simplifies to
(27)

Here, we are dealing with the fixed data rate IS-95 system.
Our proposed receivers in Sections IIV were dealing with the
multirate W-CDMA system, therefore, we need to do the ESC
estimation per symbol because the user might change its data
rate. In IS-95, we do not have to do that. We just do the ESC
(which is the physical code in this case) estimation one time
(during the first symbol) and then use it in the cancellation
process for the coming symbols. Thus, the receiver structure
shown in Fig. 2 might not be economical as far as the power
consumption is concerned. In order to reduce the computational
complexity, FWT is only used for detecting the codes. Since the
codes change at very slow rate, the code detection block can be
activated at much slower rate than the symbol rate. Once the
codes are detected, the data symbols of interfering users can be
detected by a conventional correlator bank.
VI. SIMULATION RESULTS
We have
users simultaneously transmit their information
data signals with different bit rates. Each user is assigned a
is used
Walsh code with different SF. Note that the same

MADKOUR et al.: SUCCESSIVE INTERFERENCE CANCELLATION ALGORITHMS

Fig. 5. Effect of intracell multiuser interference on the W-CDMA downlink.

Fig. 6.

Capacity enhancements using the projective cancellation scheme.

for all physical users (mobile terminals) regardless of their data

rates. That makes the high data rate users have higher average
power than the low data rate users do. For example, the average
is
times that of the single
power of a user with data rate
rate user, therefore, an -rate user can be considered as
single rate users ( effective users). The simulation was performed on the IMT-2000 vehicular channel A model, which has
six multipaths of delays 0, 0.31, 0.71, 1.09, 1.73 and 2.51 s and
,
,
,
, and
dB, respectively. QPSK
power 0,
modulation is used for data with spreading factors ranging from
128 (voice) to 8 (data). The frame length is 10 msec. The RAKE
receiver has six fingers. Finally, a complex gold code of length
1 truncated to 40 960 chips was used as a cell scrambling
2
code.
In Fig. 5, we demonstrate the effect of the intracell multiuser
interference on the performance of the downlink W-CDMA
signal. Different number of users with different spreading
factors are used. We plot the average bit error rate (BER)
as a function of the
of the single rate user
signal-to-Gaussian-noise ratio (we model the intercell multiuser interference as a Gaussian noise). Both AWGN and

175

Fig. 7. Performance evaluation of different combined-interfering signals

receivers.

multipath fading channels are considered. From the plots, it

is clear that as the number of active users increases, the BER
(16
increases. For example, adding a physical user of
effective users) results in a 5-dB SNR loss at 1% BER. The
simulation example in Fig. 6 uses the projection technique to
cancel out the intracell multiuser interference. There are seven
users with spreading factors of 128, 128, 64, 16, 16, 8, and 8
(52 effective users). We cancel out the maximum six correlation
values at a time. We simulate up to 6 iterations. In order to
demonstrate the capacity increase as a result of IIC, the performance of a 36 effective users-system without IIC has been
computed. From the plots, it is clear that using an IIC scheme in
a system of 52 effective users gives a better performance than
a system of 36 effective users. It is about 50% system capacity
increase. In Fig. 7, the BER has been plotted as a function
of the number of iterations for all the combined-interfering
signals schemes using the same system loading as that in the
dB. It is
previous example (52 effective users) at SNR
seen that the projection technique outperforms the subtraction
technique when canceling the same number of interferers at
each iteration. When canceling six interferers at a time using
subtraction technique, the problem of over cancellation starts
to occur after the 2nd iteration (the 3rd iteration is worse than
the second iteration). Therefore, when using subtraction, it is
better to cancel 1 or 2 correlation values at a time so that we
will not have over cancellation. For the projection technique,
we have the choice either to reduce the number of iterations
(feedbacks) and increase the number of the correlation values
canceled at a time (save the symbol processing time) or vice
versa (it is almost a linear relationship). Also it is clear that the
projection technique has a steady state error floor (irreducible
error), however, there is no over cancellation.
Plots in Fig. 8 compare the subspace projection to the combined-interfering signals projection technique. We plot the avas a function of
erage BER of the single rate user
the signal to Gaussian noise ratio for five different cases using
the same system loading as in the previous example (52 effective users). The solid curve is for the noncancellation case (regular RAKE receiver). The bottom curve is the single user per-

176

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 1, NO. 1, JANUARY 2002

Fig. 8. BER versus the SNR for subspace; symbol; and frame projection
interference cancellation schemes.

(feedbacks) to have a good performance. When the subspace

dimension is increased from
to
, a tremendous
improvement in the performance happens, e.g., a BER of 0.2%
is achieved after 7 iterations (compare to 35 iterations when
).
Unlike the uplink, the near-far problem in the downlink can
not be alleviated by using power control. This is simply because
the base station needs to reach users at the cell border using a
stronger power than that required for users close to the base station. It is to be noted that the curve we have for 20 effective users
can be viewed as a representative case at cell border because the
desired user gets one twentieth of total transmitted power. We
showed significant gains achieved by using interference cancellation for this case. Also, the curve for 52 effective users can be
viewed as a representative case at cell center because the desired
user gets only 1/52 of total transmitted power. For this case, the
proposed techniques are also shown to be very effective in improving performance.
VII. CONCLUSION
We proposed new blind intracell interference cancellation
techniques suitable for the downlink of any CDMA system
with fixed or variable spreading factor. No extra information
is required from the base station. A capacity increase of up to
150% of the original (without IIC) system capacity has been
achieved. Even though the problem was addressed for the
W-CDMA, it is also applicable to IS-95 or IS-2000.
ACKNOWLEDGMENT
The authors are thankful to the anonymous reviewers for their
helpful comments and suggestions.
REFERENCES

Fig. 9. Performance comparison between the subspace cancellation and other

proposed cancellation schemes.

formance curve (for comparison issue). The dashed curve and

the curve with a star are for the combined-interfering signals
projection technique (6 iterations, cancellation of 6 interferers
at a time) using frame and symbol projection, respectively. The
dash-dotted curve is for the subspace projection. It can be seen
that the subspace projection technique outperforms other cancellation techniques, however it is more complex. With SNR
of 18 dB, the subspace cancellation for 52 effective users gives
0.1% BER which is the same as that for 20 effective users with
no cancellation which means about 150% capacity increase.
Also, in this figure, the performance curve of the centralized
decorrelating detector has been shown for a number of 52 effective users. To get this performance curve, we followed the
methodology in [8]. It is clear that the decorrelating detector
outperforms our proposed receivers which is expected because
it has the knowledge of users SF as well as the spreading codes.
The BER as a function of the number of iterations for the different proposed cancellation techniques is plotted in Fig. 9. It is
seen from the plots that the subspace approach outperforms the
other cancellation techniques, however, it needs more iterations

[1] T. Ojanpera and R. Prasad, Wideband CDMA for Third Generation Mobile Communications. Boston, MA: Artech House, 1998.
[2] Specifications of Air-Interface for 3G Mobile System Version 1.0, Association of Radio Industries and Business (ARIB), 1999.
[3] J. Proakis, Digital Communications, 3rd ed. New York: McGraw-Hill,
Inc, 1995.
[4] T. Ottoson and A. Svensson, Multi-rate schemes in DS/CDMA
systems, in Proc. IEEE VTC95, Chicago, IL, July 2628, 1995, pp.
10061010.
[5] A. Klein, Data detection algorithms specially designed for the downlink
of CDMA mobile radio systems, in Proc. IEEE VTC97, Phoenix, AZ,
May 1997, pp. 203207.
[6] A. Hottinen, H. Holma, and A. Toskala, Multi-user detection for
multi-rate CDMA communications, in Proc. IEEE Int. Conf. Commun.,
Dallas, TX, June 2428, 1996.
[7] M. J. Juntti, System concept comparisons for multi-rate CDMA with
multi-user detection, in Proc. IEEE VTC98, Ottawa, Canada, May
1998, pp. 3640.
[8] M. F. Madkour and S. C. Gupta, Performance analysis of a wireless
multirate direct-sequence CDMA using fast Walsh transform and decorrelating detection, IEEE Trans. Commun., vol. 48, pp. 14051412, Aug.
2000.
[9] A. Johansson and A. Svensson, Successive interference cancellation
in multiple data rate DS/CDMA systems, in Proc. IEEE VTC95,
Chicago, IL, July 2628, 1995.
[10] P. Teder, G. Larsson, B. Gudmundson, and M. Ewerbring, CDMA with
interference cancellation: A technique for high capacity wireless systems, in Proc. IEEE Int. Conf. Commun., Geneva, Switzerland, 1993.
[11] P. Patel and J. Holtzman, Analysis of a simple successive interference
cancellation scheme in DS/CDMA system, IEEE J. Select. Areas
Commun., vol. 12, pp. 796807, June 1994.

MADKOUR et al.: SUCCESSIVE INTERFERENCE CANCELLATION ALGORITHMS

[12] X. Wang and H. V. Poor, Blind multi-user detection: A subspace approach, IEEE Trans. Inform. Theory, vol. 44, pp. 677690, Mar. 1998.
[13] M. Golanbari and G. E. Ford, Signal monitoring on the downlink of cellular CDMA communications with interference cancellation, in Proc.
IEEE Wireless Communications and Networking Conf., WCNC99, New
Orleans, LA, Sept. 2125, 1999.
[14] K. Hooli, M. Latva-aho, and M. Juntti, Multiple access interference suppression with linear chip equalizers in WCDMA downlink
receivers, in Proc. IEEE Global Telecommunications Conf. 1999,
GLOBCOM99, vol. 1A, pp. 467471.
[15] R. Kohno, H. Imai, M. Hatori, and S. Pasupathy, Combination of an
adaptive array antenna and a canceller of interference for direct-sequence spread-spectrum multiple access system, IEEE J. Select. Areas
Commun., vol. 8, pp. 675681, May 1990.
[16] M. F. Madkour, S. C. Gupta, and Y.-P. E. Wang, A novel downlink
blind interference cancellation in a W-CDMA mobile communications
system, in Proc. IEEE Wireless Communications and Networking
Conf., WCNC99, New Orleans, LA, Sept. 2125, 1999.
, A novel downlink W-CDMA blind interference cancellation
[17]
using the subspace approach, in Proc. 33rd Asilomar Conf., Pacific
Grove, CA, Oct. 2427, 1999.
[18] G. E. Bottomley, Improved successive cancellation of DS-CDMA
signals using signal orthogonalization, in Proc. IEEE Int. Conf.
Universal Personal Communications, ICUPC96, Cambridge, MA,
Sept.Oct. 292, 1996, pp. 141144.
[19] W. L. Brogan, Modern Control Theory. Englewood Cliffs, NJ: Prentice-Hall, 1982.

Mohamed F. Madkour (S98-M00) was born in

Alexandria, Egypt, on May 25, 1968. He received
the B.Sc. (with honors) and M.Sc. degrees in
electrical engineering from Alexandria University,
Alexandria, Egypt, in 1991 and 1994, respectively,
and the Ph.D. degree in electrical engineering from
Southern Methodist University (SMU), Dallas, TX,
in 2001.
He joined the Department of Electrical Engineering, Alexandria University, in 1991, as a
Teaching Assistant and Lab Instructor. From 1996 to
1999, he was a Research Assistant at the Department of Electrical Engineering,
SMU. He has been a member of the Mobility Network Consulting Goup,
Ericsson Inc., Plano, TX, since 1999. His current research interests include
mobile, spread-spectrum communications and third generation wireless.
Dr. Madkour received the SMU Outstanding Graduate Student Award
in 1999. He also received the Best Graduate Paper Award from the IEEE
Communications and Vehicular Technology Societies, Dallas Chapter in 1999.

177

Someshwar C. Gupta (S61-M63SM65-F99)

received the B.S. (with honors) and M.S. degrees
in mathematics from Punjab University, India, the
B.S. (with honors) degree in electrical engineering
from Glasgow University, U.K., and the M.S.E.E.
and Ph.D. degrees from the University of California,
Berkeley.
He taught at Carnegie Mellon University and
Arizona State University prior to joining Southern
Methodist University, Dallas, TX, in 1967. Presently,
he is the Cecil H. Green Professor of Engineering.
He has considerable teaching, research supervision, academic administration,
and industrial experience. He has been very active in the areas of cellular and
personal communications for the last 18 years. He is the author or co-author
of three widely used textbooks and a over 100 technical papers. Almost all his
research has been supported by federal agencies such as NSF, NASA, USAF,
DCA, AFOSR, and industrial companies such as GE, Ericsson, Nortel, LCCI,
and Raytheon. He has supervised over 60 Ph.D. dissertations.

Yi-Pin Eric Wang received the B.S. degree in electrical engineering from National Taiwan University
in 1988, and the M.S. and Ph.D. degrees, both in electrical engineering, from the University of Michigan,
Ann Arbor, in 1991 and 1995, respectively.
He has been a member of the Advanced Development and Research group of Ericsson Inc., Research
Triangle Park, NC, since 1995. His work focuses on
wireless communications, including mobile satellite
communication systems and terrestrial cellular systems. His research interests include coding, modulation, synchronization, and interference cancellation in CDMA systems.