Page 1

A Low Interference Time-Slicing Code Assignment

for the 2D-Spread MC-DS-CDMA Systems

Chih-Wen Chang1and Chien-Cheng Kuo

Institute of Computer and Communication Engineering, National Cheng Kung University, Taiwan

Abstract—In addition to some design flexibilities, the low inter-

ference property of code division multiple access (CDMA) makes

multi-carrier direct sequence CDMA (MC-DS-CDMA) suitable for

possible unlicensed transmissions in the licensed bands. Introduc-

ing the concept of time division multiple access (TDMA) into MC-

DS-CDMA may further create a new dimension dealing with the

interference problem. In this paper, we proposed a time slicing

code assignment for the two-dimensional (2D) spread MC-DS-

CDMA. The time slicing technique distributes interference equally

over transmission time intervals (TTIs) by assigning users different

frequency-domain spreading code groups and transmitting in

their corresponding TTIs. Combing with the previous interference

avoidance code assignment and power control mechanism for the

2D-spread MC-DS-CDMA, the proposed time slicing scheme is

proved to be capable of eliminating interference more efficiently. In

terms of the call admission rate, the improvement can be as large

as 10% in one of the considered cases. A good operating point of

time slicing can also maintain a high throughput, while achieving

lower call dropping rate and lower power consumption. The results

of this paper can provide some important and interesting hints for

the system designers.

Index Terms—MC-DS-CDMA, 2D spreading, code assignment,

power control, time slicing, interference mitigation.

I. INTRODUCTION

The unceasing demand for ubiquitous access to network is

a strong driving force behind the revolutions in the wireless

communication technologies. However, the limited spectrum

resources can never quench the thirst. Accordingly, unlicensed

transmission in the licensed band stands out to show its potential

to fit this need, for example, the cognitive radio (CR) [1]. With

no doubt, low interference is the necessary characteristic to

facilitate the unlicensed transmissions. Based on the CDMA

techniques, MC-DS-CDMA possesses some desirable attributes

for the low interference transmissions and additional design

flexibilities [2], [3]. In the MC-DS-CDMA systems, however,

the key point to accommodate users lies in the level of multiple

access interference (MAI) but the amount of available code

resources [4]–[6]. Thus, how to eliminate MAI is a fatal

issue for possibly applying MC-DS-CDMA in the unlicensed

transmissions.

In the literature, several aspects has been explored to elim-

inate interference for MC-DS-CDMA, including the adaptive

subcarrier allocation [7], [8], code and waveform design [9]–

[12], and the novel joint code assignment and power control

techniques [4]–[6]. One systematic way to incorporate the

concept of TDMA into the MC-DS-CDMA may further create a

new dimension dealing with the MAI problem. In this paper, we

propose a time-slicing code assignment for the two-dimensional

(2D)-spread, time and frequency domain, MC-DS-CDMA with

power control. The term ”time slicing” originated from [13]

1The contact author’s e-mail is cwchang@ee.ncku.edu.tw

2This work was supported by the National Science Council, Taiwan, under

the contracts 96-2221-E-006-020 and 97-2221-E-006-085-MY3.

is an example of putting TDMA concept into orthogonal

frequency division multiplex (OFDM) to save power for digital

video broadcasting-handheld (DVB-H) with bursty traffic.

On top of our previous joint interference avoidance code

assignment and power control scheme [4]–[6], the time slicing

here plays an important role in balancing the interferers between

TTIs. With time slicing, TTIs are numbered according to the

number of the frequency spreading code groups (Nf). That is

users associated with different frequency spreading codes may

belong to different groups. Different groups will be arranged

to transmit in their corresponding TTIs. To achieve this, we

define a vacancy quotient (VQ) to quantize the available code

resources for a group of frequency spreading codes. When

receiving a new coming call request, VQ is pre-calculated and

a code belonging to the code group with highest VQ value will

be assigned. Thanks to the aids of VQ and TTI arrangement,

the MAI stemmed from the non-orthogonality of frequency-

domain spreading codes can be largely eliminated and equally

distributed to each TTI.

From the simulation results, we have several interesting ob-

servations. First, with a proper value of Nf, the call admission

rate can significantly increased and the same system throughput

can be maintained, although time slicing makes the available

code resources less in every TTI. Second, a proper Nf can

also contribute to lower call dropping rate and lower power

consumption. Third, as Nfgets larger, more incoming calls will

be blocked owing to the lack of code resources but the level of

interference. Thus, when we ultimately seek for an interference

free situation via time slicing, it may not always lead to good

results. These observations provide some important hints for the

system designers. Also, one can say that time slicing reachs a

new frontier of dealing with the MAI problem in the MC-DS-

CDMA systems.

The rest of this paper is organized as follows. We address

the motivation and problem formulation in Section II. Some

necessary background and the definition of the VQ value can

also be found in Section II. In Section III, we propose a novel

time slicing code assignment strategy. Simulation results and

conclusion are provided in Section IV and V, respectively.

II. MOTIVATION AND PROBLEM FORMULATION

A. The MAI in the 2D-Spread MC-DS-CDMA

Following the same system model and the assumptions of

[4]–[6], we consider a downlink multi-rate MC-DS-CDMA with

time- and frequency-domain spreading, the so-called 2D-spread

MC-DS-CDMA. In such a system, the orthogonal variable

spreading factor (OVSF) code tree has a two-dimensional struc-

ture, as shown in Fig. 1. In this figure, the time- and frequency-

domain spreading factors are 8 and 4, respectively. As derived in

equations (7)∼(9) of [4], the MAI with the 2D OVSF code tree

results from the non-orthogonality of the frequency spreading

978-1-4244-2517-4/09/$20.00 ©2009 IEEE

Page 2

code tree for frequency-domain

spreading code [1, -1, -1, 1 ]

code tree for frequency-domain

spreading code [1, -1, 1, -1 ]

code tree for frequency-domain

spreading code [1, 1, -1, -1 ]

code tree for frequency-domain

spreading code [1, 1, 1, 1 ]

Fig. 1.

spreading factor is four.

A two dimensional OVSF code tree when the frequency-domain

codes. Moreover, the relationships and the amount of the MAI

can be well described by the grid representation of a 2D

OVSF code tree. In the grid representation, the code resources

are illustrated by a set of rectangles. Code channels with the

same time-domain spreading factor Gt can be represented by

a set of equal-sized rectangles. The area of each rectangle is

proportional to the transmission rate of the corresponding code

channel (or inversely proportional to Gt).

Consideringthesamescenarios

2 of [4]3, the candidate

?

C(4)

8,8

, where C(i)

code in the l-th layer associated with the i-th frequency-

domain code tree, where the time-domain spreading factor

Gt = 1 ∼ 2(l−1)from the top layer (l = 1) to the bottom

layer (l = 4) of the OVSF code tree. The corresponding grid

representation of this scenario is shown in Fig. 2.

To make this example simple, here, we only consider the

candidacy of C(1)

the definition of the related codes in equation (1) of [4]4,

one can know that C(1)

amount of MAI. To be specific, C(1)

by

C(2)

2,2

and

C(1)

2,1

Now, two interesting questions arise: (1) what if the amount

of MAI goes beyond the affordable level of C(1)

(2) Can we further distinguish C(3)

MAI level ? To answer the first question, conventionally, the

coming call request will be blocked. For the second question,

however, the answer is positive when time slicing technique is

applied. Thus, with time slicing, this call request can be served

using C(3)

as

for

8,3,C(4)

shown

a

8,4,C(4)

in Fig.

are

8,7,

codes

8,2,C(4)

request

8,5,C(4)

C(1)

8,8,C(3)

?

8,1,C(3)

8,2,C(3)

8,3,C(3)

8,4,C(4)

8,1,C(4)

8,6,C(4)

2l−1,n(n = 1,...,2(l−1)) denotes the n-th

8,8and C(3)

8,1for the request. According to

8,8and C(3)

8,1are affected by the same

8,8and C(3)

2,1,C(2)

, respectively.

8,1are interfered

?

2,2,C(3)

???

8,8and C(3)

8,8in terms of the

8,1?

8,1from C(1)

8,1.

B. The Concept of Time Slicing

In [13], time slicing turns off radio frequency (RF) front-end

when the desired service is not in transmission. In this paper,

3Owning to the page-limit, the Fig. 2 of [4] is not reiterated in this paper.

4Related Codes ≡ {The used codes positioned in the same column of the

grip representation}. Codes belonging to the same related codes group interfere

each other.

: used code

?????????????

?????????????

?????????????

?????????????

Layer 2

Layer 3

Layer 4

Aggregate

Layer 2~4

) 3 (

1 , 8

C

(

1 , 8

C

) 3 (

2 , 8

C

(

2 , 8

C

) 3 (

3 , 8

C

(

3 , 8

C

) 3 (

4 , 8

C

) 4

C

) 4

) 4(

4 , 8

) 4(

5 , 8

C

) 4(

7 , 8

C

) 4(

6 , 8

C

) 4 (

8 , 8

C

????????????????

?????????????

?????????????

?????????????

?????????????

?????????????

?????????????

?????????????

?????????????

?????????????

?????????????

?????????????

?????????????

) 4

) 1 (

8 , 8

C

) 4 (

8 , 8

C

) 4 (

7 , 8

C

) 4 (

6 , 8

C

) 4(

5 , 8

C

) 4 (

4 , 8

C

) 4 (

3 , 8

C

) 4 (

2 , 8

C

) 4 (

1 , 8

C

) 3 (

1 , 8

C

) 3 (

2 , 8

C

) 3 (

3 , 8

C

) 3 (

4 , 8

C

) 1 (

1 , 2

) 2(

1 , 2

C

C

) 1 (

8 , 8

C

) 2(

2 , 2

) 3 (

2 , 2

C

C

Fig. 2.

MC-DS-CDMA system with time and frequency domain spreading.

The grid representation of Fig. 2 of [4] for the code resources in the

based on the interference level during each TTI, time slicing

allocates TTIs to users utilizing different frequency spreading

codes. Figs. 3(a) and 3(b) respectively shows the traditional

continuous transmission and time slicing principle during two

consecutive TTIs for the 2D-spread MC-DS-CDMA. As shown

in the figure, all the K users with 2D spreading codes transmit

simultaneously using continuous transmission scheme, while

these codes are divided into two groups to transmit during two

different TTIs when time slicing is applied. One should note

that the length of the TTI is unchanged when time slicing is

applied. Recall the example shown in Fig. 2, what will it be

when we apply time slicing to it ? Fig. 4 shows the answer. As

shown in the figure, the 1stand 2ndcode trees are arranged to

transmit during the m-th TTI, while the 3rdand 4thcode trees

are allocated the (m+1)-th TTI. Now, C(3)

free. Fortunately, the aforementioned coming call request can

now be served.

8,1can be interference

C. Vacancy Quotient of a Frequency Spreading Code Group

Assume that the number of frequency spreading codes and

groups of which are Gf and Nf, respectively. It gives M =

Gf/Nf frequency spreading codes in one group. A user is

requesting a code with time-domain spreading factor equal

to Gt. Then, the possible candidate codes are

1 ≤ i ≤ Gt and 1 ≤ j ≤ Gf. The vacancy quotient (VQ)

of the k-th frequency spreading code group can now be defined

as following.

?

C(j)

Gt,i

?

for

V Qk=

kM

?

j=(k−1)M+1

Gt

?

i=1

e(j)

Gt,i, 1 ≤ k ≤ Nf ,

(1)

Page 3

Code K

Code 2

…

Code 1

Frequency spreading

code domain

Time

m th TTI

m+1 th TTI

(a) Continuous transmission

Code n

Code 2

…

Code 1

Frequency spreading

code domain

Time

m th TTI

with allocating

1st code group

m+1 th TTI

with allocating

2ndcode group

Code K

Code n+2

…

Code n+1

(b) Time slicing principle

Fig. 3.

principle during two consecutive TTIs for the 2D-spread MC-DS-CDMA.

(a) The traditional continuous transmission and (b) time slicing

) 3 (

1 , 8

C

(

1 , 8

C

) 3 (

2 , 8

C

(

2 , 8

C

) 3 (

3 , 8

C

(

3 , 8

C

) 3 (

4 , 8

C

) 4

C

) 4

) 4(

4 , 8

) 4(

5 , 8

C

) 4(

7 , 8

C

) 4(

6 , 8

C

) 4 (

8 , 8

C

) 4

) 1 (

1 , 2

) 2(

1 , 2

C

C

) 1 (

8 , 8

C

) 2(

2 , 2

C

) 3 (

2 , 2

C

Fig. 4.Applying time slicing to the example of Fig. 2.

where e(j)

otherwise e(j)

candidate code is a code which can preserve orthogonality either

in time or frequency domain between any other codes in use. A

group with a higher value of VQ will be assigned to the coming

user.

Gt,i= 1 if C(j)

Gt,i= 0. One should note that the available

Gt,iis an available candidate code;

III. THE TIME SLICING CODE ASSIGNMENT

As shown in Fig. 5, the proposed time slicing code assign-

ment mainly consists of four steps. In the first step, according

to the VQ value, the new coming user will be assigned a

frequency-domain spreading code group, in other words, being

allocated to a particular TTI as mentioned before. In the second

step, a code group will be randomly assigned if there are more

than one group having the same VQ value in the first step. The

third step is to apply the so-called interference avoidance (IA)

code assignment to pick a code within its corresponding code

group [4], [5]. At last, the user will transmit signals using the

power control mechanism in [6].

IV. SIMULATION RESULTS

In this section, we illustrate the effectiveness of the proposed

time slicing code assignment for the power-controlled MC-DS-

CDMA with 2D spreading, in terms of call admission rate,

1st step?Pick a code group with highest

VQ value

2nd step?Pick a code group randomly

4th step?Power control mechanism

New call arrival

Satisfying all power constraints

This is a new call

Call blocking

Call dropping

NO

YES

Exceed the call dropping threshold

NO

YES

NO

YES

If more than two groups tie

3rd step?Pick a code which produces less

MAI by the IA code assignment

strategy

NO

YES

Fig. 5.

MC-DS-CDMA with 2D spreading.

The flowchart of the proposed time slicing code assignment for the

call blocking rate, call dropping rate, received SINR, power

consumption and total throughput.

A. Simulation Environment

In the simulation, we follow the same assumptions as [4]–

[6]. In the downlink MC-DS-CDMA system, the subcarriers

carrying the same data bits experience independent flat Rayleigh

fading channel. The background noise is modelled by the

white Gaussian noise with power spectrum density of N0 =

−103.2 (dBm). A new call is generated by the Poisson arrival

process with arrival rate (λ) of 1/2 per time unit and the

paired departure rate (μ) can be one of {1/32, 1/48, 1/64,

1/80, 1/96, 1/112, 1/144, 1/176}. Thus, there are on average

λ/μ = 16 ∼ 88 active calls in the system. We simulate

5,000 incoming calls for each combination of λ and μ. With

128 parallel substreams, the frequency-domain spreading factor

(Gf) is 8 and the time-domain spreading factors (Gt) are 4,

8, 16, or 32. Each call requests a code of 8R (Gt = 4), 4R

(Gt= 8), 2R (Gt= 16), or R (Gt= 32) with a probability

according to the code traffic pattern [ 1 1 2 8 ], where R is

the unit data rate. A code traffic pattern of [ a b c d ] means

that the times of requesting data rates 8R, 4R, 2R, and R are

proportional to a : b : c : d, respectively. The data rate of each

user is fixed during his call holding time. To clearly indicate

the traffic load brought by the active calls with different data

rates, we define an effective traffic load in the following. With

the time-domain spreading factor selecting from Gt=[ 4 8 16

32 ] and the code traffic pattern of [ a b c d ], the effective

traffic load (ρ) is defined as

μ×8R × a + 4R × b + 2R × c + R × d

a + b + c + d

For λ = 1/2 and μ = 1/80 and the code traffic pattern [ 1 1 2

8 ], the effective traffic load ρ is 250% of the utilization of the

time-domain resources. The maximum transmission power of a

substream is Po = 1 mW and that of BS is Ptotal = 10W.

The target signal to interference and noise (SINR) ratio is 5dB.

A new call will be blocked if it can not reach the target SINR

or without sufficient code resources. Furthermore, an active call

ρ =λ

×

1

32R

. (2)

Page 4

1 1.52 2.53 3.54 4.55 5.5

40

50

60

70

80

90

100

Effective traffic load (ρ)

Average admission rate (%)

Nf=1

Nf=2

Nf=4

Nf=8

Fig. 6.

Nf= 1,2,4,8 code groups.

Average call admission rate for various effective traffic loads with

will be dropped if it call can not reach the SINR target for a

consecutive period of time (Tdrop), where Tdrop= 0.01 seconds

is assumed to be a frame time interval [14]. Note that the power

control mechanism will be executed 15 times per frame. The

transmission time interval (TTI) is 2 milliseconds. The number

of code groups (Nf) can be one of [ 1 2 4 8 ]. Note that

Nf = 1 is actually equivalent to the conventional continuous

transmission scheme.

B. Performance Comparison

Fig. 6 shows the average call admission rate for various

effective traffic loads with Nf= 1,2,4,8 code groups. One can

find that with Nf= 2 and 4, the proposed time slicing code as-

signment outperforms the conventional continuous transmission

with Nf = 1. For example, at ρ = 4.5 with Nf = 4, the call

admission rate can be raised from 67% to 77%. However, with

Nf= 8, the time slicing technique can not be effective, owing

to the lack of code resources. That is because, with Nf = 8,

only one eighth of the code resources can be supplied in each

TTI. This phenomenon can also be explained by Fig. 7.

Fig. 7 shows the average percentage of call blocking (a)

because of the surplus MAI and (b) because of the lack of code

resources for various effective traffic loads with Nf= 1,2,4,8

code groups. As shown in the figure, with Nf= 1 and 2, 100%

of the call blocking results from the surplus of MAI, which

leads to the unsatisfactory received SINR. On the contrary, with

Nf = 8, 100% of the call blocking is caused by the short

supply of the code resources. However, it is interesting to find

that, with Nf = 4, the call blocking event can happen due to

the both reasons. Somehow, we can say that in our considered

cases, Nf= 4 can be a better and flexible time slicing operating

point.

Fig. 8 shows average call dropping rate for various effective

traffic loads with Nf= 1,2,4,8 code groups. From this figure,

one can see the effectiveness of the time slicing technique

to distribute interference into every TTI. Comparing with the

performance curve of Nf = 1, Nf = 2 and Nf = 4 can

preserve lower call dropping rate while allowing higher call

admission rate simultaneously (Fig. 6).

Fig. 9 shows the average received SINR for various effective

traffic loads with Nf = 1,2,4,8 code groups. Recalling the

simulation setup, the power control scheme in this paper is to

target the received SINR at five dB. Thus, with Nf = 8, the

12345

0

10

20

30

40

50

60

70

80

90

100

Effective traffic load (ρ)

(a)

Percentage of blocking (%)

Nf=1

Nf=2

Nf=4

Nf=8

12345

0

10

20

30

40

50

60

70

80

90

100

Effective traffic load (ρ)

(b)

Percentage of blocking (%)

Nf=1

Nf=2

Nf=4

Nf=8

Fig. 7.

and (b) because of the lack of code resources for various effective traffic loads

with Nf= 1,2,4,8 code groups.

Average percentage of call blocking (a) because of the surplus MAI

1 1.52 2.53 3.54 4.555.5

0

1

2

3

4

5

6

Effective traffic load (ρ)

Average droping rate (%)

Nf=1

Nf=2

Nf=4

Nf=8

Fig. 8.

Nf= 1,2,4,8 code groups.

Average call dropping rate for various effective traffic loads with

received SINR can be perfectly maintained at five dB because

only one frequency-domain spreading code can be used during

each TTI in this case. Note that the MAI in the 2D-spread

MC-DS-CDMA system results from the non-orthogonality of

the frequency-domain spreading codes. However, the dynamics

of the MAI level leads to the loosed controlled received SINR

as in [6].

Fig. 10 shows the average power consumption for various

effective traffic loads with Nf = 1,2,4,8 code groups. From

this figure, one can see that with the help of the time slicing,

the average power consumption can be largely reduced. For

example, at ρ = 3.5, the power consumption can reduce from

0.0163(W) to 0.0135(W) and 0.0004(W) with four (Nf = 4)

and eight code groups (Nf= 8), respectively.

Fig. 11 shows the average total throughput for various

effective traffic loads with Nf = 1,2,4,8 code groups. To

facilitate the comparison, the throughput statistics are collected

every eight TTIs because the largest number of the frequency

domain spreading code groups is Nf= 8. Also, the throughput

is evaluated based on the unit data rate R. Obviously, the

throughput performance of Nf = 2 and 4 can be maintained

at the same level as that of Nf = 1, although the available

code resources in every TTIs can only be 1/Nf. However, the

low call admission rate of Nf = 8 (Fig. 6) results in the low

throughput performance.

Page 5

1 1.522.533.54 4.555.5

5

5.1

5.2

5.3

5.4

5.5

5.6

5.7

5.8

5.9

6

Effective traffic load (ρ)

Average received SINR (dB)

Nf=1

Nf=2

Nf=4

Nf=8

Fig. 9.

1,2,4,8 code groups.

Average received SINR for various effective traffic loads with Nf=

1 1.52 2.53 3.54 4.55 5.5

0

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0.016

0.018

Effective traffic load (ρ)

Average power consumption (W)

Nf=1

Nf=2

Nf=4

Nf=8

Fig. 10.

Nf= 1,2,4,8 code groups.

Average power consumption for various effective traffic loads with

1 1.52 2.53 3.54 4.55 5.5

200

250

300

350

400

450

500

550

600

650

Effective traffic load (ρ)

Average total throughput (R/8TTI)

Nf=1

Nf=2

Nf=4

Nf=8

Fig. 11. Average total throughput for various effective traffic loads with Nf=

1,2,4,8 code groups.

V. CONCLUSION

In this paper, we introduced the concept of TDMA into

MC-DS-CDMA and proposed a low interference time slicing

code assignment for the 2D-spread MC-DS-CDMA with power

control. The time slicing here was to equally distributed interfer-

ence over every TTI. To facilitate the time slicing technique, we

defined a VQ value to indicate the available code resources for

a group of frequency spreading codes. After this VQ-evaluation

process, a user can be assigned a frequency-domain spreading

code group and arranged to transmit in a proper TTI. Cooper-

ating with IA code assignment and power control mechanism

for the 2D-spread MC-DS-CDMA, the proposed scheme has

shown its ability to perform much better than the traditional

continuous transmission scheme. In terms of call admission rate,

the improvement can be 10%. Unfortunately, time slicing can

not be operated ultimately to reach an interference free situation.

A good operating point of time slicing can not only maintain a

high throughput but also achieve lower call dropping rate and

lower power consumption. In our considered cases, we found

that four groups of the frequency-domain code (Nf= 4) can be

an good operating point. An interesting research direction for

the future work can be the adaptive adjustment of the number

of code groups based on the MAI level.

REFERENCES

[1] S. Haykin, “Cognitive radio brain-empowered wireless communications,”

IEEE Journal on Selected Areas in Communications, vol. 23, no. 2, pp.

201–220, Feb. 2005.

[2] L.-L. Yang and L. Hanzo, “Multicarrier DS-CDMA: a multiple access

scheme for ubiquitous broadband wireless communications,” IEEE Com-

munications Magazine, vol. 41, no. 10, pp. 116–124, Oct. 2003.

[3] A. Attar, M. R. Nakhai, and A. H. Aghvami, “Cognitive radio trans-

mission based on direct sequence MC-CDMA,” IEEE Trans. on wireless

communications, vol. 7, no. 4, pp. 1157–1162, April. 2008.

[4] L.-C. Wang, C.-W. Chang, and H. Huang, “An interference avoidance

code assignment strategy for downlink multi-rate MC-DS-CDMA with

TF-domain spreading,” IEEE Trans. Wireless Communications, vol. 6,

no. 7, pp. 2508–2518, July 2007.

[5] C.-W. Chang and L.-C. Wang, “Effects of subcarrier power allocation on

an interference avoidance code assignment strategy for multi-rate MC-

DS-CDMA systems,” IEEE Trans. on Vehicular Technology, vol. 57, pp.

1513–1526, May 2008.

[6] C.-W. Chang and C.-C. Kuo, “A joint power control and interference

avoidance code assignment strategy for downlink MC-DS-CDMA with

2D spreading,” in IEEE Vehicular Technology Conference, May 2008, pp.

2316–2320.

[7] S. P. Chang and K. B. Lee, “Transmit power allocation for BER perfor-

mance improvement in multicarrier systems,” vol. 52, no. 52, pp. 1658–

1663, Oct. 2004.

[8] H. Long and Y. H. Chew, “An adaptive subcarrier allocation scheme for

MC-DS-CDMA systems in the presence of multiple access interference,”

in IEEE International Conference on Communications, vol. 5, Paris,

France, June 2004, pp. 2894–2898.

[9] C.-M. Yang, P.-H. Lin, G.-C. Yang, and W. C. Kwong, “2D orthogonal

spreading codes for multicarrier DS-CDMA systems,” in IEEE Interna-

tional Conference on Communications, vol. 5, Anchorage, Alaska, May.

2003, pp. 3277–3281.

[10] H. Wei, L.-L. Yang, and L. Hanzo, “Time- and frequency-domain spread-

ing assisted MC-DS-CDMA using interference rejection spreading codes

for quasi-synchronous communications,” in IEEE Vehicular Technology

Conference, vol. 1, Milan, Italy, Sept. 2004, pp. 389–393.

[11] ——, “Interference-free broadband single and multicarrier DS-CDMA,”

IEEE Communications Magazine, vol. 43, no. 2, pp. 68–73, Feb. 2005.

[12] S. Sureshkumar, E. Shwedyk, and H. H. Nguyen, “MAI-minimized

signature waveforms for MC-DS-CDMA systems,” in IEEE Vehicular

Technology Conference, vol. 1, Stockholm, Sweden, June 2005, pp. 305–

309.

[13] X. Wang, Y. Wu, H.-C. Wu, and J.-Y. Chouinard, “A time slicing technique

for mobile multimedia communications using MSE-OFDM system,” in

IEEE Vehicular Technology Conference, May 2008, pp. 1231–1235.

[14] H. Holma and A. Toskala, WCDMA for UMTS, radio access for third

generation mobile communications, 3rd ed.

2004.

John Wiley & Sons Ltd.,