Sie sind auf Seite 1von 10

78

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 11, NO. 1, JANUARY 2012

IDMA vs. CDMA: Analysis and Comparison of


Two Multiple Access Schemes
Katsutoshi Kusume, Member, IEEE, Gerhard Bauch, Senior Member, IEEE,
and Wolfgang Utschick, Senior Member, IEEE

AbstractThis article presents comprehensive comparisons of


interleave division multiple access (IDMA) and direct sequence
code division multiple access (DS-CDMA) in terms of performance
and complexity assuming iterative multiuser detection. IDMA
can be seen as a special case of DS-CDMA with spreading
gain of one using very low rate code and user-specific interleavers for user separation. We focus on three suboptimum
linear detectors: minimum mean square error (MMSE), rake (or
matched filter), and soft-rake detectors from practical concerns.
We analytically prove that the three detectors are equivalent
for asynchronous users of IDMA on frequency flat channels for
complex modulation alphabets. Such equivalence has been shown
only for binary phase shift keying (BPSK) in the literature. The
equivalence guarantees the MMSE solution for IDMA without
computationally expensive matrix inversions or matrix-vector
multiplications. This is generally not the case for DS-CDMA since
DS-CDMA is sensitive to user asynchronism. We also discuss
complexity aspects when the MMSE detector is used where we
focus on essential differences in complexity between IDMA and
DS-CDMA, instead of discussing particular complexity reduction
techniques. Computer simulations are performed in various
scenarios and the performance is analyzed by bit error rate
simulations as well as by extrinsic information transfer (EXIT)
charts. The analysis reveals the advantages of IDMA over DSCDMA in terms of performance and complexity under practical
considerations, particularly in highly user loaded scenarios.
Index TermsInterleave division multiple access (IDMA),
direct sequence code division multiple access (DS-CDMA), multiuser detection, iterative detection and decoding.

I. I NTRODUCTION

NTERLEAVE DIVISION MULTIPLE ACCESS (IDMA)


is a recently proposed spread spectrum multiple access
scheme similar to direct sequence code division multiple
access (DS-CDMA or CDMA for short henceforth). Unlike
CDMA system, however, user-specific spreading sequence is
not used in IDMA system. Instead, bandwidth expansion is
fully exploited for forward error correction code that typically
results in very low rate code as compared to CDMA system. Instead of user-distinct spreading sequences for CDMA
system, user-distinct interleavers are the unique feature to

Manuscript received June 2, 2010; revised December 25, 2010 and April
24, 2011; accepted July 7, 2011. The associate editor coordinating the review
of this paper and approving it for publication was K. B. Lee.
This article was presented in part at the IEEE Global Telecommunications
Conference, November/December 2009 (received the Best Paper Award).
K. Kusume and G. Bauch are with DOCOMO Euro-Labs, Landsbergerstr.
312, 80687 Munich, Germany (e-mail: kusume@docomolab-euro.com). G.
Bauch is also with Universitt der Bundeswehr Mnchen, Werner-HeisenbergWeg 39, 85577 Neubiberg, Germany.
W. Utschick is with Technische Universitt Mnchen, Arcisstr. 21, 80290
Munich, Germany.
Digital Object Identifier 10.1109/TWC.2011.111211.100954

distinguish users for IDMA system. The user separation relies


on iterative multiuser detection.
Interestingly, the principle of IDMA is aligned with many
research results on CDMA that try to find an optimal tradeoff between bandwidth expansion in spreading and in forward
error correction code. For example, in [1], [2], the author concluded that spread-spectrum multiple-access communication
system approaches its ultimate potential only with the use of
very low rate forward error control code with consequently
large bandwidth expansion. Similar conclusions have been also
made in [3] assuming coherent detection and in [4] that very
low rate encoder with short spreading sequences should be
used in terms of capacity and cutoff rates. Thus, IDMA may
be seen as a special case of CDMA with spreading gain of
one using very low rate codes.
The term IDMA appeared in [5], [6] for the first time,
but the main ideas can be found in several references. Chip
interleaving was first introduced in [7] to mitigate burst
impulsive noise disturbance. In [8], the authors compared the
performance of chip-interleaved CDMA system with codespread CDMA system employing very low rate codes. The
code-spread CDMA system has been further studied in [9]
in comparison to CDMA system. These references assume
chip-interleaving, common for all users. In [10], the authors
proposed to apply user-distinct chip-interleavers in uncoded
CDMA system, where the user-specific symbol spreading is
moved before the chip-interleaving. The authors in [11] proposed using distinct interleavers without symbol spreading for
improving the performance of the slotted ALOHA system. The
proposed transmitter structure is quite similar to the IDMA
system in [5], [6], but the assumed receive processing is based
on [12] that is more complex. The authors in [13], [14] proposed a trellis-coded modulation scheme, which is made userspecific by selecting a unique combination of convolutional
code (can be common for all users), modulation constellation,
and symbol interleaver. Several references suggest the use
of user-independent interleavers in CDMA systems for better
user separation [12], [15][17].
Several techniques for IDMA have been studied in a number
of papers in the last years to further enhance the capability of
IDMA. However, only a few papers have been concerned with
comparisons of CDMA and IDMA systems so far. In [18],
[19], only an uncoded system is compared. Although the comparison itself is interesting, uncoded systems are extremely
unfavorable for CDMA systems as we will also discuss in this
article. The authors in [20] compare a coded system, but only
the so-called soft rake detection is used for both CDMA and

c 2012 IEEE
1536-1276/12$31.00

KUSUME et al.: IDMA VS. CDMA: ANALYSIS AND COMPARISON OF TWO MULTIPLE ACCESS SCHEMES

bi,k

Fig. 1.

(Low-rate)
encoder

cj,k

cm,n,k

Symbol
mapper

xn,k

un,k

General transmitter structure for CDMA and IDMA.

IDMA that is, again, unfavorable for CDMA systems. Another


paper [21] compares uncoded and coded systems where the
minimum mean square error (MMSE) detector is utilized
for CDMA systems. That is a more suitable comparison,
but the results are limited to one particular scenario. In
any of the above mentioned references, the main focus is
not the comparison of CDMA and IDMA. The first paper
focusing on the comparisons of CDMA and IDMA appeared
in [22], followed by [23] where further comparisons have
been made. In this article, we present more comprehensive
study that include higher order modulation alphabets, different
choices of detectors, analytical comparisons of the detectors,
complexity analysis, and extrinsic information transfer (EXIT)
chart analysis [24].
This article is organized as follows. Section II introduces
our general system model, which can be seamlessly used for
describing both CDMA and IDMA. In Section III, we review
the principle of iterative multiuser detection and decoding
and then, introduce three suboptimum linear detectors. The
three detectors are analytically compared in order to highlight
some differences between CDMA and IDMA. Complexity
issues are discussed in Section IV where we do not discuss
particular complexity reduction techniques, instead, we focus
on essential difference in complexity between CDMA and
IDMA. Section V presents performance analysis and comparison of CDMA and IDMA using different detectors in
various scenarios. We conclude this article by drawing some
conclusions in Section VI.
II. S YSTEM M ODEL
Fig. 1 illustrates our general transmitter structure, which can
be seamlessly used for both CDMA and IDMA systems. At the
transmitter, information bits , of user , {1, . . . , },
{1, . . . , }, are encoded by a rate encoder. Low rate
code is typically realized by a rate c convolutional code
followed by a rate r repetition code. Then, it follows that
= c r . The resulting coded bits , , {1, . . . , }, are
interleaved by a user-distinct bit interleaver to get ,, ,
which are then mapped onto complex symbols , that are
taken from a 2 -ary symbol alphabet: {1 , . . . , 2 },
where we assume to fulfill the following conditions:
2
2
2
1

=1 = 0 and 2
=1 = 1. Note that ,, ,
{1, . . . , }, {1, . . . , }, denotes the -th bit
of the -th transmit symbol , of user , i.e., code bits
1,, , . . . , ,, from the encoder after the interleaver are
mapped onto symbol , by the symbol mapper.
In case of CDMA, the complex symbols , are further
spread by a user-distinct spreading code , , which
may depend on symbol index , if a scrambling code is
used. The resulting spread signals are often called chips by
convention of CDMA systems. In case of IDMA, chips are
equivalent to symbols as IDMA applies no spreading codes.

79

The channel is modeled as an order Finite Impulse


Response (FIR) filter with
channel
} taps , having the nor {
E , 2 = 1, where E {} denotes
malized energy: =0
expectation. The chip time/duration is denoted by c . Besides
, we also consider a user delay , which accounts for
user-asynchronous transmissions due to imperfect network
synchronization. Then, the maximum delay of user becomes ( + )c . We define the maximum total delay over
all users as: c max ( + ) and s c / in the
number of chips and symbols, respectively, where rounds
the argument to the nearest larger integer.
It is convenient to consider an effective channel that is a
discrete convolution of the channel taps with the spreading
code:
, , c +2
(1)
scenarios,
where [ is an amplitude reflecting
]T near-far
T
T
c +1
0 , 0, , , , , 0c
is a channel impulse response vector including a user delay , and
]T
[
, T, , 0T 2 is the spreading code extended
with some zeros; this is simply to make sure that the convolved
sequence in (1) becomes sufficiently long for the later use
in (2), (3), and (5). We denote by 0 an all-zero column
vector of dimension and by ()T matrix transpose.
With the effective channel, the system model can be described at the symbol level, instead of at the chip level. This allows us to work on a unified model that is valid independent of
the use of spreading codes. Using
the effective channel, the
]Tre[
ceive signal vector + + , , +(+1) 1
s
may be expressed as: = =0
(, ) + ,
[
]T
where we define a symbol vector: ,1 , , ,
[
]T
, a noise vector: + + , , +(+1) 1
, and an effective channel matrix for symbol with
delay :
[
]
(, ) 1 (, ), , (, )
(2)
and (cf. Eq. 1)
[
]T
(, ) , [ + 1], , , [( + 1) ] .
(3)
We now develop a time-variant sliding window model
for the detection of symbol , . Since there are s
prethe] input symbols of
[ T and post-cursors,
T
s , , T , , T+s (2s +1) , are taken into
account in the sliding window that includes sufficient statistics
for the detection of , . Then, we finally get the sliding
window model for the detection of , :
= + ,

(4)
]T
where we define a receive vector: T , , T+s ,
[
]T
a noise vector: T , , T+s , and an effective
channel matrix of dimension (s + 1) (2s + 1) can
be defined as:

( s , s ) (, 0)

..
..
..

.
.
.
.
[

(, s ) ( + s , 0)

(5)

80

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 11, NO. 1, JANUARY 2012

xn,1
hn,1

Ts

Ts

hn+1,1

hn+Ls ,1

n
yn

xn,K
hn,K

Ts

Ts

hn+1,K

hn+Ls ,K

Fig. 2. General multiple access channel model in the symbol level. s


denotes symbol duration. The effective (time-variant) vector channels ,
include channel memory , user delay , spreading/scrambling code , ,
and near-far factor .
0
Lm (cm,n,1 )

multiuser

yn

1
1
1


Lm
a (cm,n,1 )

Lm (cm,n,K )


Lm
a (cm,n,K )

Fig. 3.

1
K

Ld (bi,1 )

Ld (cj,1 )

Ldc (cj,1 )

0
detector

APP
decoder

APP
decoder

Ldc (cj,K )

Ld (bi,K )

Ld (cj,K )

Structure of iterative multiuser detection and decoding scheme.

We also define a partition of the effective channel matrix:


[[s ,1 , , s , ] , . . . , [,1 , , , ] ,
. . . , [+s ,1 , , +s , ]] .
Then, (4) can be rewritten in the column-wise form:
=

, , + .

(6)

=1 =s

This column-wise channel model is illustrated in Fig. 2


and will be found useful later for describing the detection
algorithms.
III. I TERATIVE M ULTIUSER D ETECTION AND D ECODING
The overall receiver structure is depicted in Fig. 3. The
receiver for a system of simultaneously transmitting users
is comprised of a multiuser detector and a posteriori probability (APP) decoders that are connected via interleavers.
Both multiuser detector and APP decoder are soft-in-softout component, which iteratively exchange soft information
by accepting and computing log-likelihood ratios (LLRs) as
the soft information. According to the turbo principle [25],
the output LLRs must be independent of the respective input
LLRs. That is ensured by computing the so-called extrinsic
LLRs.
The multiuser detector, receives a priori LLRs
( about )code
,,
bits sent from the decoder, defined as: m
a
P{
=+1}
ln P ,, =1 , which is initialized to 0 before the first
{ ,,
}
iteration (i.e., there is no a priori information from the decoder
at the beginning: ,, are equally likely to be +1 and 1).

(
)
,, for
With the a priori LLRs about code bits m
a
users and the observation , the goal of the multiuser detector
is to compute the a posteriori LLRs:
}
{

=
+1

(
)

,,
},
m ,, ln {
(7)

P ,, = 1
(
)
,,

and then sending the extrinsic LLRs: m


e
(
)
(
)
m

,, a ,, , to the decoder after the dein-

terleaving operation by 1
. By using the total probability
theorem [26], we get:
(
)
m
e ,, =

{
}
p { , = }
P ,, =
ln

,, =+1

p { , = }

,, =1

P ,, =

},

(8)

where {+1, 1} denotes the -th bit of the symbol ,


code bits that are mapped on , are assumed
{ independent} of
each other due to the bit interleaver, and P ,, = =
(
(
(
) ))
1
m

tanh

1
+

a
,, /2
2
Each of decoders (receives) the deinterleaved version of
,, from the multiuser detector
the extrinsic LLRs m
e
d
as the channel LLRs c (, ) as illustrated in Fig. 3. From the
received channel LLRs and according to the code constraints,
the decoder computes a posteriori LLRs d (, ), which are
improved LLRs about the code bits. We do not further discuss
the decoding algorithm, since it is the standard function [27].
The channel LLRs dc (, ) are subtracted from d (, )
to get the extrinsic LLRs de (, ) (cf. Fig. 3): de (, )
d (, ) dc (, ), which are
( sent to
) the multiuser detector
m

as the new a priori LLRs a ,, after being interleaved


by . In turn, the( multiuser
) detector computes and delivm

ers new LLRs ,, . The error performance can be


improved by some iterations. Finally, the decoder computes
a posteriori LLRs( about information
bits d (, ). Taking its
)
d

sign: , sign (, ) gives an estimate of information


bit.
Now, we consider three suboptimum linear detectors:
MMSE, rake, and soft rake detectors, which are derived for
given input a priori information from decoders. These detectors are chosen mainly because of concerns about complexity.
A. Minimum Mean Square Error Detector
In [28], it is shown that the MMSE solution with
unbias constraint adheres the turbo principle without
any heuristic adjustment of statistics that is commonly
done in the literature. By writing the filter output as:
, H + where ()H denotes Hermitian conjugate
transpose and the affine term is introduced to be jointly
optimized with , the optimization problem may be stated
as: {, , , } arg min, s.t.:{H , = }1
where the MSE reads as: E
, , 2 .

KUSUME et al.: IDMA VS. CDMA: ANALYSIS AND COMPARISON OF TWO MULTIPLE ACCESS SCHEMES

The constrained optimization can be solved by using


Lagrangian multipliers. The solution can be written as
,

1
, ,

1
H
, , ,

and

81

TABLE I
P ROPORTIONAL
EXPRESSION
OF CONDITIONAL PROBABILITY DENSITIES
{
}
p , = () OBTAINED BY THE DETECTORS .

,
, , H
,

E { } =
E { },
,
where
and the soft symbol replicas
, are{ computed as:
}

, E {, } =

.
,,

=1
, , H
, ,


,
We
{ also define2 } ,
2
=
,

P
{
=
}

, 2 ,
E ,

} ,
{
H
H
)(
) = + 2 .
and E (
Here, we assume the noise vector to be a zero
mean proper complex Gaussian with the covariance
matrix 2 and we assume {that transmit symbols }are
)(
)H =
mutually
E }(
{ T uncorrelated:
diag s , . . . , T , . . . , T+s ,
with

[
]T
,1 , . . . , , .
, where the
The MMSE estimate reads as:
, H
,
soft interference cancellation:
+ , , ,
,

(9)

is performed. Assuming that the MMSE estimate is Gaussian distributed, we compute the conditional mean and variance. Then, we end up with the conditional probability
density MMSE () in Table I, which is proportional to
p { , = }, i.e., p { , = } MMSE ().
Therefore, MMSE () can be used for computing the extrinsic
LLR in (8).
Note that the MMSE detector has been shown to be equivalent to the so-called jointly Gaussian detector for the BPSK
alphabet in [29] and for general complex alphabets in [28].
B. Rake Detector
Rake detector, also known as maximum ratio combining or
matched filter, is a conventional single user detection scheme
for DS-CDMA. The rake detector collects the frequency
diversity available from multipath channels, but unlike the
MMSE detection, it does not explicitly attempt to suppress
interference. Thus, the rake detector may suffer from performance degradation in the presence of strong interference, but
it is much simpler than the MMSE detector.
If the rake detector is integrated in the iterative detection and decoding technique, interference suppression
can still be performed as the soft interference cancellation by exploiting a priori information from the decoders.
Iterative rake detection has been studied, e.g., in [30]
[32]. After the soft interference cancellation as in (9), the
rake detector tries to maximize the desired signal portion
(rake)

by being matched to the channel response:


,
(
})
{
H
H
2
,
,
where
= , , + , , E ,
, = ,{s,...,s } , , + . Assuming that
==0
}
{
the residual interference plus noise term H
, ( , E , )
is Gaussian, its probability density is characterized by its mean
and variance. Then, we end up with the conditional probability
density Rake () in Table I, which can be plugged into (8)
for computing the extrinsic LLR.

(
MMSE
Rake
Soft rake

MMSE () exp
Rake () exp

SR () exp

1
, , )2
H
, , (

1
H
, , ,
)
(

, )2
H
,
,

(s +1)

=1

H
, , ,
T
2
(
, , )

T ,

C. Soft Rake Detector


In the IDMA literature, the so-called soft rake detector has
been most frequently considered, e.g., [18][20], [22], [33],
[34]. This is analogous to the rake detector for DS-CDMA
as the standard detection scheme. The soft rake detector
is also frequently called elementary signal estimator (ESE).
Unlike the MMSE approaches, the soft rake approach does
not attempt to equalize the remaining interference after the soft
interference cancellation. Instead, the soft rake detector applies
the following approximation in (8): p { , = } =
(s +1)
p {+1 , = } , where the elements of the
=1
observation vector (s +1) : +1 T =
T , , + T , , are assumed to be independent of each
other where denotes a (s + 1) dimensional unit vector
whose -th element is 1. Assuming that T , is Gaussian
distributed, we compute its mean and variance to get the
conditional probability density SR () in Table I, which can
be used for computing the extrinsic LLR in (8).
D. Analytical Comparisons of Detectors
This section provides analytical comparisons of the three
detectors in the previous sections. Using Table I, the three
detectors can be conveniently compared. The MMSE detector is most complex due to matrix inversions 1
, , matrix
vector multiplications 1

,
and
vector
inner
products
, ,
1
H

).
The
rake
detector
is
simpler
because
,
, ,
,
no matrix inversion has to be computed, but there are still
the vector inner products. The simplest among these three
detectors is the soft rake, which does not require any of
the above mentioned complex operations. Several numerical
comparisons of the detectors by means of computer simulations will be provided in Section V. However, under certain
conditions, some of the solutions can analytically be shown to
be equivalent. In what follows, we discuss such equivalence
for CDMA and IDMA systems.
All the detectors generally give different solutions for
CDMA systems. However, the MMSE and rake detectors are
equivalent:
MMSE () = Rake () = SR ().
(for CDMA under Assumption III.1.)
Assumption III.1. users are perfectly synchronous ( =
0, ) and are transmitting over flat channels ( = 0, )
using orthogonal spreading codes ( ).
The equivalence under this assumption is intuitive and
straightforward and thus, the proof is omitted. Note, however,

82

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 11, NO. 1, JANUARY 2012

that perfect user synchronous transmission is generally difficult to achieve and maintain in practical systems such as the
uplink of cellular systems or in ad hoc networks [35] due to,
e.g., different propagation delays for users.
The equivalence for CDMA can be further extended for
IDMA systems under a more relaxed condition. Users are not
restricted to be synchronous. That might be more relevant in
practical systems. The equivalence holds not only between the
MMSE and rake detectors, but also for the soft rake detector.
Now, we make the following assumption:
Assumption III.2. asynchronous users are transmitting
over flat channels ( = 0, ).
Under Assumption III.2, the effective channel matrix
in (5) has dimension (s + 1) (2s + 1) where
s = c = max because IDMA does not apply spreading
code, i.e., = 1 and symbol delay is equivalent to
chip delay. [The effective channel
]T vector can be written
= 0, +1 s +1 .
as: , = 0T 0, 0Ts
Other effective channel vectors , of are simply
time shifted versions of , and thus, they have at most
one non-zero element 0, . Therefore, the covariance
s
H
2
matrix =
=s , , , +
=1
has non-zero elements only in its main diagonal. Then,
2
T
, = , , H
, = , 0, +1 +1
is diagonal as well. With the diagonal structure,
1
=
we get the following equalities: H
, ,

T
T
1
T +1 1
=

)
,
0,
,

+1
+1
+1

0,
,

2
1
T
1
T
H

)
,

0,
, , ,
+1 , +1
,
becomes 0, , if = + 1, and 0 otherwise. Using these
equalities, it can be shown that:
MMSE () = Rake () = SR ()
(
)
2
2
H
0, + 2 Re{0,
, +1 }
= exp
.
T +1 , +1
(for IDMA under Assumption III.2.)
The equivalence is also applicable for broad band transmissions using orthogonal frequency division multiplexing
(OFDM) that transforms broad band signal into a set of narrow
band signals. Several references can be found that combine
IDMA with OFDM modulation [36], [37]. It should be pointed
out that the equivalence guarantees the MMSE solution without matrix inversions or matrix vector multiplications. This is
usually not the case for CDMA, if users are asynchronous.
We note that some part of the equivalence shown above
for IDMA systems have been discussed in the literature for
BPSK in [29], [38], [39] whereas our proof is more general for
complex modulation alphabets = {1 , . . . , 2 } fulfilling


the two conditions: 2=1 = 0 and 21 2=1 2 = 1
that are described in Section II.
IV. C OMPLEXITY A NALYSIS OF MMSE D ETECTOR FOR
CDMA AND IDMA
The complexity of the iterative MMSE detector arises from
the fact that the MMSE filter has to be computed for symbols of users that is repeated over iterations. There
have been a number of complexity reduction techniques in the

literature, e.g., [40][42]. In this section, we do not discuss


particular complexity reduction techniques. Instead, we focus
on essential difference in complexity between CDMA and
IDMA.
The complexity is compared under the assumption that
both systems operate with the common parameters: number
of users , number of information bits per transmission
block , modulation alphabet , bandwidth, and chip delay c (computed from common channel memories and
user delays ).
Let us start with a comparison of the number of rows of
channel matrix (s + 1) . Since s c / , it
holds that:
(CDMA) (s + 1) (c / + 1)
c + 1 (IDMA) ,
with equality for = 1. Thus, the number of rows of channel
matrix for CDMA is larger than for IDMA. For the number
of columns of channel matrix (2s + 1), one can write
(IDMA)
(2c + 1)

=
(CDMA)
(2s + 1)

for large c .

Thus, IDMA generally results in a much larger than


CDMA. We note that the difference of grows much faster
than that of with respect to the spreading factor .
Now, we compare the number of symbols . Since bandwidth expansion for IDMA is fully exploited by low rate code
before the bit-interleaver, we can write:
(IDMA) = (CDMA) .
Since we need MMSE filters for symbols, IDMA has to
compute times larger number of filters than CDMA.
From the analysis above, we see that the MMSE detection
for IDMA may be much more complex than for CDMA. As
we will observe later, IDMA generally performs well with the
simpler soft rake detector in many scenarios. But, in scenarios
where the MMSE detector should be used rather than the soft
rake to get better performance, IDMA can be more complex
than CDMA. Nevertheless, when we recall the symbol-wise
complexity reduction technique according to the time-average
of covariance matrix [40], [41], the matrix inversion have to
be computed only once over symbols, and then the number of
symbols is not the main complexity factor anymore. This
technique can be exploited over time-invariant channels.
In the presence of user asynchronism, however, CDMA
either does not perform well without using user-distinct scrambling codes, or cannot use the time-average technique due
to the use of scrambling sequences that effectively make the
time-invariant channel vary over time. Therefore, IDMA may
be more likely to gain tremendous complexity reductions by
the time-invariant approach without compromising the performance as compared to CDMA, when channel is quasi static.
If channel varies quickly within each transmission block, then
the time-invariant approach may not be applied. Then, the
MMSE detector for IDMA is much more complex than for
CDMA.

KUSUME et al.: IDMA VS. CDMA: ANALYSIS AND COMPARISON OF TWO MULTIPLE ACCESS SCHEMES

TABLE II
M OST COMMONLY USED SIMULATION PARAMETERS .

10

IDMA

information bits
= 1024
convolutional code
c = 1/2, mem 4, [31, 27]8 , trellis is terminated
repetition code

r = 1/4
bit interleavers
user-independent, uniform random
interleaver size
= 2056
= 8224
modulation
QPSK (Gray labeling)
symbols
= 1028
= 4112
spreading codes
= 4, OVSF codes

scrambling codes UMTS uplink long codes

user delay
= 1

K=8

10

10
BER

CDMA

10

10

This section presents comparisons of CDMA and IDMA by


means of computer simulations. The simulation parameters are
summarized in Table II. Information bits , of block size
for user are encoded by the rate c = 1/2 memory 4 standard convolutional code with generator polynomial [31, 27]8.
The block size of = 1024 is used for obtaining most
simulation results presented later. The convolutional encoder
starts in the initial state 0 and ends in the terminal state 0. The
trellis is terminated by additional 4 termination bits, resulting
in ( + 4)/c = 2056 code bits.
For IDMA, the convolutionally coded bits are further encoded by the rate r = 1/4 repetition code to get ( +
4)/c /r = 8224 code bits. The repetition coding alternates
the sign of input bit to balance the number of ones and minus
ones. The code bits , are, then, interleaved by user-distinct
interleaver to get ,, . Therefore, the interleaver size
is accordingly determined as ( + 4)/c = 2056 and
( + 4)/c /r = 8224 for CDMA and IDMA, respectively.
Interleavers are randomly generated and are updated for
each transmission frame and for all users. Then, the performance presented later is averaged over a large number
of interleaver realizations. We note that, although a common
interleaver is typically used for all users in case of CDMA,
user-independent interleavers are used also for CDMA,
since some literature suggest the use of user-independent
interleavers in CDMA systems for better user separation capability [12], [15][17]. A much simpler interleaver generation
scheme may be desired in practical systems [43], [44].
After bit-interleaving, code bits ,, are mapped onto
complex symbols , , which are elements of a quaternary
phase shift keying (QPSK)
to
Gray
constellation

according

{1/
2
+
j/
2,
1/
2

j/
2,
1/
2+
labeling:

QPSK

j/ 2, 1/ 2 j/ 2}. We note that Gray labeling does


not represent the best choice in conjunction with iterative
detectors [45], but we do not attempt to optimize the symbol
mapping in this article. Two code bits [1,, 2,, ] are
mapped onto , .
For CDMA, QPSK symbols , are spread by a user
distinct spreading code, which is constructed from a short code
followed by a long code. User distinct short codes of spreading
factor = 4 are taken from the orthogonal variable spreading factor (OVSF) codes [46], [47]. Long codes are generated
according to the uplink long scrambling sequences [47] and

K=6

10

10

V. P ERFORMANCE A NALYSIS OF CDMA AND IDMA

83

single user
CDMA, MMSE
IDMA, soft rake (=MMSE, rake)
1

3
4
Eb/N0 in dB

Fig. 4. BER performance of highly loaded CDMA (MMSE) and IDMA (soft
rake) systems with = 6, 8 users after 10 iterations on an AWGN channel.

are user-distinct. In this article, long codes are always used


since the performance degradation is severe otherwise [22].
If the number of users, , exceeds the spreading factor ,
then the common short code [+1, 1, +1, 1] is used for all
users and long codes distinguish the users.
After the spreading (no spreading for IDMA), the signal
is transmitted over channel. We consider a small amount of
user delays = 1 to account for user asynchronous
transmission. Therefore, even on an AWGN channel, signals
for different users at the receiver are generally non-orthogonal
for CDMA systems. Since we assume no guard interval
between transmission blocks, interblock interference is taken
into account, but it is not cancelled in the iterative process.
All channels are assumed perfectly known at the receiver.
A. AWGN Channel
We start with the performance comparisons of CDMA and
IDMA for = 6, 8 users on an AWGN channel. Fig. 4
illustrates the BER performance after 10 iterations. The results
are obtained by using the MMSE detector for CDMA to avoid
the performance degradation by other two detectors, whereas
for IDMA we use the soft rake detector, which is equivalent to
other detectors on an AWGN channel (cf. Section III-D for the
analytical proof). For both CDMA and IDMA, we see that the
performance approaches the single user bound above certain
b /0 values, but the performance of IDMA is superior to
that of CDMA and the performance gap grows as the number
of users increases. Although not presented here, we note that
the difference becomes even greater if simpler detector is used
for CDMA, instead of the MMSE.
B. Multipath Channels
Next, we evaluate the performance over multipath channels.
Fig. 5 shows the BER performance over randomly generated
16 paths ( = 15, ) channels following a uniform delay
profile. The performance of single user transmission over an
AWGN channel is also included as a comparison. There is
performance degradation for CDMA against IDMA for = 8

84

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 11, NO. 1, JANUARY 2012

10

K=8

10

10
BER

K=4
3

10

10

10

10

single user (AWGN)


CDMA, MMSE
IDMA, soft rake
IDMA, MMSE
1

4
5
Eb/N0 in dB

Fig. 5. BER performance of highly user loaded CDMA and IDMA systems
with = 8 users after 10 iterations on randomly generated 16 paths channels.
The performance of the systems with = 4 users is plotted as a comparison.

users that is not present for = 4 users. We also observe that


the soft rake detector does not perform as well as the MMSE
detector anymore for IDMA.
C. Uncoded Transmissions
So far, the convolutional code is always employed for the
performance evaluation. While it is reasonable to assume a
certain forward error correction (FEC) code to be employed
in modern communications systems, it is still of interest to
consider uncoded systems to avoid a rate loss, e.g., in high
SNR environments. We study such an uncoded scenario in this
section.
We use the term uncoded when no convolutional code or
any other code except repetition code is employed. Precisely
speaking, repetition code is still an FEC code of simplest
kind. However, to put an emphasis on comparisons to CDMA
systems (only spreading and no FEC codes), the performance
with only repetition code for IDMA systems is called uncoded.
Fig. 6 shows the uncoded BER performance of CDMA and
IDMA systems with = 4, 6 users on an AWGN channel
after 20 iterations. The MMSE and soft rake detectors are
used for CDMA and IDMA, respectively. The single user
performance is also plotted as the lowest achievable bound.
It can be observed that the performance of IDMA approaches
the single user bound above certain b /0 values, whereas the
performance of CDMA is strongly suffered from the multiple
access interference. Although the performance over iterations
is not shown in Fig. 6, we observed that the performance
does not improve over any iterations for CDMA. That is due
to the lack of decoder for CDMA in case of no FEC code and
therefore, no improvement can be obtained from decoder. Note
that introducing another user-distinct chip-interleaver after
spreading could bring performance gain over iterations [10] for
CDMA, which is outside the scope of this article. IDMA, on
the other hand, employs the repetition code of rate r = 1/4.
The repetition code is a weak code, but it still brings some

Fig. 6. Uncoded BER performance of CDMA (MMSE) and IDMA (soft


rake) systems with = 4, 6 users on an AWGN channel after 20 iterations.

improvements over iterations and thus, the iteration works for


IDMA, even though no convolutional code is used.
D. Analysis Using Extrinsic Information Transfer Charts
We analyze the system using EXIT chart [21], [24], [38],
[48], which is a convenient tool for studying iterative processing techniques. The EXIT chart shows the average mutual
information between the transmitted code bits and the input
LLRs, which are obtained from the other soft-in soft-out block,
versus the average mutual information between the transmitted
code bits and the extrinsic output LLRs, which are sent to the
other block.
In our system, the horizontal axis is the mutual information

( , m
a ( )) at the input of the multiuser detector, which
is also the mutual information at the output of the decoder
(, de ()) (cf. Fig. 3 and (8)). The vertical axis is the mu
tual information ( , m
e ( )) at the output of the multiuser
detector, which is also the mutual information at the input of
the decoder (, da ()).
Curves for the relation between input and output mutual
information can be plotted for both soft-in soft-out blocks:
the multiuser detector and the decoder in our system. The
trajectory between the two curves starting from the multiuser

detector without a priori information, i.e., ( , m


a ( )) =
d
(, e ()) = 0, illustrates the performance improvements by
iterations. The average mutual information between code bits
and LLRs () can be computed by numerical evaluation [24].
For the EXIT chart computation, it is assumed that the input
LLRs are statistically independent and Gaussian distributed.
1) Coded transmissions: Fig. 7 illustrates the EXIT curves
of the decoders and the detectors for CDMA and IDMA
systems with = 4 or 8 users over an AWGN channel at
b /0 = 5 dB with block size of = 100000. These EXIT
curves correspond to the BER performance in Fig. 4.
Two EXIT curves are plotted for the decoder: only with the
rate c = 1/2 convolutional code for CDMA and additionally
with the rate r = 1/4 repetition code for IDMA. Since
vertical and horizontal axes are, respectively, input and output
mutual information for the decoder, it can be observed that

KUSUME et al.: IDMA VS. CDMA: ANALYSIS AND COMPARISON OF TWO MULTIPLE ACCESS SCHEMES

I (c, Lm
(c)) = I (c, Lda(c))
e

0.8

K=4, CDMA
AWGN channel
K=8, CDMA
Eb/N0=5dB MMSE
rake

AWGN channel
0.9 Eb/N0=10dB
0.8
I (c, Lm
(c)) = I (c, Lda(c))
e

85

IDMA, soft rake


(=MMSE, rake)

0.7

0.6

0.6

decoder:Rc=1/2

0.4

0.2

0
0

CDMA, MMSE

K=8, IDMA
K=4, IDMA
soft rake
decoder:Rc=1/2 &Rr=1/4
0.2
0.4
0.6
0.8
1
I (c, Lm
(c)) = I (c, Lde(c))
a

0.5
0.4
0.3
0.2
0.1

K=4
K=6

decoder:Rr=1/4

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
I (c, Lm
(c)) = I (c, Lde(c))
a

Fig. 7. EXIT curves of CDMA and IDMA systems with = 4, 8 users


over an AWGN channel at b /0 = 5 dB and EXIT curves of decoders
(cf. Fig. 4 for the corresponding BER performance).

Fig. 8. EXIT curves of uncoded CDMA and IDMA systems with = 4, 6


users over AWGN channel at b /0 = 10 dB and EXIT curves of decoder
(cf. Fig. 6 for the corresponding BER performance).

higher output mutual information (, de ()) for a given


input (, da ()) is obtained by employing the additional
repetition code. We note that EXIT curves of the decoder are
independent of channel and SNR value in our system and
solely depend on the quality of the input mutual information
since there is no direct input from the channel (cf. Fig. 3).
The EXIT curves are computed for the three detectors in
case of CDMA, whereas only the soft rake detector is evaluated in case of IDMA because all the detectors are equivalent
on an AWGN channel (cf. Section III-D). In contrast to the
decoder, EXIT curves of the multiuser detector depend on
the channel and SNR value since the detector receives the
observations from the channel. To be more precise, EXIT
curves of the multiuser detector are influenced not only by the
channel and SNR value, but also by all the parameters after
the bit interleaving such as modulation alphabet, spreading
and scrambling codes, user delays and near-far factors, and
the number of users simultaneously accessing the channel.
Detection algorithm naturally influences the quality of the
output LLRs of the detector as well.
In order for the iterative process to reach the single user
bound, there must be an open tunnel between the EXIT curves
of detector and decoder. The convergence behavior is then
determined by the width of the tunnel; narrower the tunnel is,
larger the number of iterations must be.
For CDMA, it can be observed in Fig. 7 that, for a given

input mutual information ( , m


a ( )), the MMSE detector

results in much higher mutual information ( , m


e ( )) than
the other two detectors. With = 4 users, there is the widest
tunnel opened between the EXIT curves of the decoder and the
detector using the MMSE detector. Thus, the MMSE detector
requires the least number of iterations to end up with the single
user bound among the three detectors (at the price of larger
complexity per iteration). With = 8 users, there is no open
tunnel by using any detector for CDMA and convergence to
the single user bound is not possible. That agrees with the
BER performance results in Fig. 4 at b /0 = 5 dB.
In case of the IDMA system, there are open tunnels for both

cases of = 4 and 8 users. In the latter case, although the


tunnel is narrow, the single user bound should be reached after
a sufficient number of iterations. That can be also evidenced
from the BER performance at b /0 = 5 dB in Fig. 4.
2) Uncoded Transmissions (without Convolutional Code):
Fig. 8 illustrates the EXIT curves of the uncoded CDMA
system using the MMSE detector and the IDMA system
using the soft rake detector with = 4 or 6 users at
b /0 = 10 dB. The EXIT curve of the decoder is plotted
for the rate r = 1/4 repetition code as well. There is no
decoder for the uncoded CDMA system. The respective BER
performance curves can be found in Fig. 6 at b /0 = 10 dB.
For the CDMA system, there is no improvement in terms of
mutual information over iterations due to the lack of decoder.
The mutual information is, for instance, about 0.81 for the
CDMA system with = 4 users. This value is obtained
without iteration and it remains so, no matter how many
iterations are performed. For the IDMA system, on the other
hand, starts with a much lower mutual information of about
0.21, then it improves over iterations and eventually ends up
with the single user bound of about 0.95, which is higher
than 0.81 for the CDMA system. Note that, as mentioned
earlier, introducing another user-distinct chip-interleaver after
spreading could bring performance gain over iterations [10]
for CDMA, which is not considered in this article.
VI. C ONCLUSIONS
This article presented comprehensive comparisons of IDMA
and CDMA systems. Since our focus was on the performance
and complexity under practical considerations, we concentrated on the three suboptimum iterative linear detectors:
MMSE, rake, and soft rake. The three detectors were analytically shown to be equivalent for IDMA over flat channels
for asynchronous users. It was then pointed out that the
equivalence guarantees the MMSE solution for IDMA over flat
channels without computationally expensive matrix inversions
and matrix vector multiplications. This is not the case for
CDMA system in general since CDMA is sensitive to user

86

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 11, NO. 1, JANUARY 2012

asynchronism. The equivalence can be extended to frequency


selective channels using OFDM modulation.
We performed computer simulations in various scenarios.
In case of CDMA, it was observed that the performance
degradation is severe by the rake and soft rake detectors
as compared to the MMSE detector in most scenarios. For
IDMA, the simplest soft rake detector performs nearly as
well as the most complex MMSE detector in many scenarios
except highly user-loaded scenarios or for a channel with poor
frequency characteristic.
In highly user-loaded scenarios by increasing the number
of users exceeding the spreading factor of CDMA systems
, i.e., > , IDMA clearly outperforms CDMA and
the performance gap becomes larger. IDMA also outperforms CDMA, if convolutional code is not employed, because
CDMA then becomes an uncoded system and it does not
improve the performance over iterations whereas IDMA still
improves due to a repetition code.
We analyzed the system using EXIT charts, which revealed
predictions of iterative processing behavior assuming a sufficiently large block size. In highly user-loaded scenarios, the
soft rake detector for IDMA does not perform well anymore
and/or requires a large number of iterations. Then, a large
block size would be necessary to avoid increasing correlations
in extrinsic information over many iterations. That may limit
its applicability to delay-sensitive applications where short
block size is used. In some cases, there are clear advantages
of employing the MMSE detector, which is however more
complex and one has to consider complexity reduction techniques, e.g., [40][42].
Instead of discussing particular complexity reduction techniques, we focused on essential difference in complexity
between CDMA and IDMA, when using the MMSE detector.
By comparing the number of symbols and the dimension
of covariance matrix to be inverted, we observed that the
MMSE detection for IDMA may be considerably more complex than for CDMA. However, if we exploit the symbolwise complexity reduction technique according to the timeaverage of covariance matrix [40] over time-invariant channels, matrix inversion has to be computed only once over
symbols, and then the number of symbols is not the main
complexity factor anymore. In the presence of user asynchronism, however, CDMA either does not perform well without
using user-distinct scrambling codes, or cannot use the timeaverage technique due to the use of scrambling sequences,
which effectively make time-invariant channel vary over time.
Therefore, IDMA may be more likely to gain tremendous
complexity reductions by the time-invariant approach without
compromising the performance as compared to CDMA, when
channel is quasi static.
There are other aspects discussed in the literature that
are essential for comparing IDMA and CDMA, but are not
covered in this article. In [33], [34], the authors explain
the better information exchange achievable by IDMA than
CDMA by utilizing a generalized Tanner graph. A good
information exchange plays an important role to guarantee
the independence assumption for iterative multiuser detection.
The authors in [21], [49] develop evolution techniques for
optimizing power allocation among users in IDMA to max-

imize the spectral efficiency. A power optimization scheme


that can provide good performance prediction with reasonable
complexity has practical interest and importance.
R EFERENCES
[1] A. Viterbi, Spread spectrum communicationsmyths and realities,
IEEE Commun. Mag., vol. 17, no. 3, pp. 1118, May 1979.
[2] A. J. Viterbi, Very low rate convolutional codes for maximum theoretical performance of spread-spectrum multiple-access channels, IEEE J.
Sel. Areas Commun., vol. 8, no. 4, pp. 641649, May 1990.
[3] M. Bickel, W. Granzow, and P. Schramm, Optimization of code rate
and spreading factor for direct-sequence CDMA systems, in Proc. 1996
IEEE Int. Symp. Spread Spectrum Tech. App., vol. 2, pp. 585589.
[4] J. Hui, Throughput analysis for code division multiple accessing of the
spread spectrum channel, IEEE J. Sel. Areas Commun., vol. 2, no. 4,
pp. 482486, July 1984.
[5] L. Ping, L. Liu, K. Y. Wu, and W. K. Leung, A unified approach
to multiuser detection and space-time coding with low complexity and
nearly optimal performance, in Proc. 2002 Annual Allerton Conf.
Commun., Control, Comput., pp. 170179.
[6] L. Ping, K. Y. Wu, L. Liu, and W. K. Leung, A simple, unified approach
to nearly optimal multiuser detection and space-time coding, in Proc.
2002 IEEE Inf. Theory Workshop, pp. 5356.
[7] S. Tachikawa and G. Marubayashi, Spread time spread spectrum
communication systems, in Proc. 1987 IEEE Global Telecommun.
Conf., vol. 16.5, pp. 615619.
[8] P. Frenger, P. Orten, and T. Ottosson, Code-spread CDMA using
maximum free distance low-rate convolutional codes, IEEE Trans.
Commun., vol. 48, no. 1, pp. 135144, Jan. 2000.
[9] S. Abeta and H. Kobayashi, Performance of very low rate channel
coding in W-CDMA reverse link, in Proc. 2002 IEEE Int. Symp.
Personal, Indoor Mobile Radio Commun., vol. 2, pp. 961965.
[10] R. H. Mahadevappa and J. G. Proakis, Mitigating multiple access
interference and intersymbol interference in uncoded CDMA systems
with chip-level interleaving, IEEE Trans. Wireless Commun., vol. 1,
no. 4, pp. 781792, Oct. 2002.
[11] Y. Li, S. Zhou, and J. Wang, Improving the performance of ALOHA
system utilizing coded multi-user detection, in Proc. 2003 IEEE Veh.
Technol. Conf., vol. 3, pp. 21442147.
[12] M. Moher, An iterative multiuser decoder for near-capacity communications, IEEE Trans. Commun., vol. 46, no. 7, pp. 870880, July
1998.
[13] F. Brnnstrm, T. M. Aulin, and L. K. Rasmussen, Iterative multi-user
detection of trellis code multiple access using a posteriori probabilities,
in Proc. 2001 IEEE Int. Conf. Commun., vol. 1, pp. 1115.
[14] , Iterative detectors for trellis-code multiple-access, IEEE Trans.
Commun., vol. 50, no. 9, pp. 14781485, Sept. 2002.
[15] S. Brck, U. Sorger, S. Gligorevic, and N. Stolte, Interleaving for outer
convolutional codes in DS-CDMA systems, IEEE Trans. Commun.,
vol. 48, no. 7, pp. 11001107, July 2000.
[16] A. Tarable, G. Montorsi, and S. Benedetto, Analysis and design of
interleavers for CDMA systems, IEEE Commun. Lett., vol. 5, no. 10,
pp. 420422, Oct. 2001.
[17] , Analysis and design of interleavers for iterative multiuser receivers in coded CDMA systems, IEEE Trans. Inf. Theory, vol. 51,
no. 5, pp. 16501666, May 2005.
[18] W. K. Leung, L. Liu, and L. Ping, Interleaving-based multiple access
and iterative chip-by-chip multiuser detection, IEICE Trans. Commun.,
vol. E86-B, no. 12, pp. 36343637, Dec. 2003.
[19] L. Ping, L. Liu, and W. K. Leung, A simple approach to near-optimal
multiuser detection: interleave-division multiple-access, in Proc. 2003
IEEE Wireless Commun. Netw. Conf., vol. 1, pp. 391396.
[20] L. Ping, L. Liu, K. Wu, and W. K. Leung, Interleave-division multipleaccess, IEEE Trans. Wireless Commun., vol. 5, no. 4, pp. 938947, Apr.
2006.
[21] K. Li, X. Wang, and L. Ping, Analysis and optimization of interleavedivision multiple-access communication systems, IEEE Trans. Wireless
Commun., vol. 6, no. 5, pp. 19731983, May 2007.
[22] K. Kusume and G. Bauch, CDMA and IDMA: iterative multiuser
detections for near-far asynchronous communications, in Proc. 2005
IEEE Int. Symp. Personal, Indoor Mobile Radio Commun., vol. 1, pp.
426431.
[23] K. Kusume, G. Bauch, and W. Utschick, IDMA vs. CDMA: detectors,
performance and complexity, in Proc. 2009 IEEE Global Telecommun.
Conf..

KUSUME et al.: IDMA VS. CDMA: ANALYSIS AND COMPARISON OF TWO MULTIPLE ACCESS SCHEMES

[24] S. ten Brink, Convergence behavior of iteratively decoded parallel


concatenated codes, IEEE Trans. Commun., vol. 49, no. 10, pp. 1727
1737, Oct. 2001.
[25] J. Hagenauer, The Turbo principle: tutorial introduction and state of
the art, in Proc. 1997 International Symp. Turbo Codes Related Topics,
pp. 111.
[26] A. Papoulis, Probability, Random Variables, and Stochastic Processes,
3rd edition. McGraw-Hill, 1991.
[27] L. R. Bahl, J. Cocke, F. Jelinek, and J. Raviv, Optimal decoding of
linear codes for minimizing symbol error rate, IEEE Trans. Inf. Theory,
vol. 20, no. 2, pp. 284287, Mar. 1974.
[28] K. Kusume, G. Dietl, and W. Utschick, On the unbias constraint
for iterative minimum mean square error detection, in Proc. 2009
International Symp. Wireless Personal Multimedia Commun..
[29] X. Yuan, K. Wu, and L. Ping, The jointly Gaussian approach to iterative
detection in MIMO systems, in Proc. 2006 IEEE International Conf.
Commun., vol. 7, pp. 29352940.
[30] A. Glavieux, C. Laot, and J. Labat, Turbo equalization over a frequency
selective channel, in Proc. 1997 International Symp. Turbo Codes
Related Topics, pp. 96102.
[31] P. D. Alexander, A. J. Grant, and M. C. Reed, Iterative detection
in code-division multiple-access with error control coding, European
Trans. Telecommun., vol. 9, pp. 419425, Sep./Oct. 1998.
[32] L. Liu, W. K. Leung, and L. Ping, Simple iterative chip-by-chip
multiuser detection for CDMA systems, in Proc. 2003 IEEE Veh.
Technol. Conf., vol. 3, pp. 21572161.
[33] H. Schoeneich and P. A. Hoeher, A hybrid multiple access scheme
delivering reliability information, in Proc. 2004 International ITG Conf.
Source Channel Coding, pp. 437442.
[34] , Adaptive interleave-division multiple accessa potential air
interface for 4G bearer services and wireless LANs, in Proc. 2004
IEEE IFIP Int. Conf. Wireless Optical Commun. Netw., pp. 179182.
[35] K. Kusume, R. Vilzmann, A. Mller, C. Hartmann, and G. Bauch,
Medium access in spread-spectrum ad hoc networks with multiuser
detection, EURASIP J. Advances Signal Process., vol. 2009.
[36] I. M. Mahafeno, C. Langlais, and C. Jego, OFDM-IDMA versus
IDMA with ISI cancellation for quasi-static Rayleigh fading multipath
channels, in Proc. 2006 International Symp. Turbo Codes Related
Topics Connection International ITG Conf. Source Channel Coding.
[37] L. Ping, Q. Guo, and J. Tong, The OFDM-IDMA approach to wireless
communication systems, IEEE Wireless Commun., vol. 14, no. 3, pp.
1824, June 2007.
[38] B. Cristea, D. Roviras, and B. Escrig, Turbo receivers for interleavedivision multiple-access systems, IEEE Trans. Commun., vol. 57, no. 7,
pp. 20902097, July 2009.
[39] L. Ping, Approximate MMSE-APP estimation for linear systems with
binary inputs," IEEE Commun. Lett., vol. 9, no. 2, pp. 172174, 2005.
[40] M. Tchler, A. C. Singer, and R. Koetter, Minimum mean squared error
equalization using a priori information, IEEE Trans. Signal Process.,
vol. 50, no. 3, pp. 673683, Mar. 2002.
[41] K. Kusume, G. Dietl, W. Utschick, and G. Bauch, Performance of
interleave division multiple access based on minimum mean square error
detection, in Proc. 2007 IEEE Int. Conf. Commun., pp. 29612966.
[42] K. Kusume, Complexity reduction of iterative minimum mean square
error detection and decoding techniques, in Proc. 2008 IEEE International Symp. Personal, Indoor Mobile Radio Commun.
[43] K. Kusume and G. Bauch, Some aspects of interleave division multiple
access in ad hoc networks, in Proc. 2006 Int. Symp. Turbo Codes
Related Topics Connection Int. ITG Conf. Source Channel Coding.
[44] , Cyclically shifted multiple interleavers, in Proc. 2006 IEEE
Global Telecommun. Conf.
[45] J. Tan and G. L. Stber, Analysis and design of symbol mappers for
iteratively decoded BICM, IEEE Trans. Wireless Commun., vol. 4,
no. 2, pp. 662672, Mar. 2005.
[46] F. Adachi, M. Sawahashi, and K. Okawa, Tree-structured generation
of orthogonal spreading codes with different lengths for forward link
of DS-CDMA mobile radio, Electron. Lett., vol. 33, no. 1, pp. 2728,
Jan. 1997.
[47] 3G TS 25.213, 3rd generation partnership project; technical specification group radio access network; spreading and modulation (FDD)
(release 4), June 2002.
[48] K. Kusume and G. Bauch, A simple complexity reduction strategy for
interleave division multiple access, in Proc. 2006 IEEE Veh. Technol.
Conf. (Fall).
[49] P. Wang, L. Ping, and L. Liu, Optimized power allocation for multiple
access systems with practical coding and iterative multi-user detection,
in Proc. 2006 Int. Symp. Turbo Codes Related Topics Connection Int.
ITG-Conf. Source Channel Coding.

87

Katsutoshi Kusume (M05) received the B.Sc.


in Information Science from Tokyo Institute of
Technology, Tokyo, Japan, in 1997, the M.Sc. in
Communications Engineering and the Dr.-Ing. with
honor in Electrical Engineering, both from Munich
University of Technology (TUM), Germany, in 2001
and 2010, respectively.
In 2002, he joined DOCOMO Euro-Labs, Munich, Germany, where he has been active in the
areas of multiple antenna systems, multicarrier transmission, spread spectrum communications, iterative
(turbo) processing, and ad hoc networking. In 2009-2010 he has contributed to
the 3GPP RAN1 meetings with a focus on downlink MIMO transmission for
LTE-Advanced. Since June 2011 he has been with Wireless Application Group
of Radio Access Network Development Department of NTT DOCOMO Inc.,
Yokosuka, Japan, where he has been involved in several development issues
for DOCOMOs LTE commercial service Xi.
He received the best paper award of IEEE Global Telecommunications
Conference in 2009 and the Werner von Siemens Excellence Award for his
master thesis.
Gerhard Bauch received the Dipl.-Ing. and Dr.Ing. degree in Electrical Engineering from Munich
University of Technology (TUM) in 1995 and 2001,
respectively, and the Diplom-Volkswirt (master in
economics) degree from FernUniversitaet Hagen in
2001.
In 1996, he was with the German Aerospace
Center (DLR), Oberpfaffenhofen, Germany. From
1996-2001 he was member of scientific staff at
Munich University of Technology (TUM). In 1998
and 1999 he was visiting researcher at AT&T Labs
Research, Florham Park, NJ, USA. In 2002 he joined DOCOMO EuroLabs, Munich, Germany, where he has been managing the Advanced Radio
Transmission Group. In 2007 he was additionally appointed Research Fellow
of DOCOMO Euro-Labs. From 2003-2008 he was an adjunct professor at
Munich University of Technology. In 2007 he was a visiting professor teaching
courses at the University of Udine in Italy and at the Alpen-Adria-University
Klagenfurt in Austria. Since February 2009 he has been a full professor at
the Universitat der Bundeswehr Munich.
He received best paper awards of the European Personal Mobile Communications Conference (EPMCC) 1997, Globecom 2008, Globecom 2009
and the IEEE International Conference on Communications (ICC) 2009, the
Texas Instruments Award of TUM 2001, the award of the German Information
Technology Society (ITG in VDE) 2002 (ITG Foerderpreis) and the literature
award of the German Information Technology Society (ITG in VDE) 2007
(ITG-Preis). He was appointed as distinguished lecturer of the IEEE Vehicular
Technology Society 2011-2013.
Wolfgang Utschick (SM06) was born on May 6,
1964. He completed several industrial education programs before he received the diploma and doctoral
degrees, both with honors, in electrical engineering
from the Technische Universitat Munchen, Germany
(TUM), in 1993 and 1998, respectively. In this
period he held a scholarship of the Bavarian Ministry
of Education for exceptional students. From 1998
to 2002, he codirected the Signal Processing Group
of the Institute of Circuit Theory and Signal Processing, TUM. Since 2000, he has been consulting
in 3 GPP standardization in the field of multielement antenna systems. In
2002, he has been appointed Professor at the TUM where he is Head of
the Fachgebiet Methoden der Signalverarbeitung. He teaches courses on
signal processing, stochastic processes and optimization theory in the field of
digital communications. Dr. Utschick was awarded in 2006 for his excellent
teaching records at TUM, and in 2007 received the ITG Award of the German
Society for Information Technology (ITG). He is a senior member of the
German VDE/ITG where he has been appointed in the Expert Committee
for Information and System Theory in 2009. He is currently also serving
as a Chairman of the national DFG Focus Program Communications in
Interference Limited Networks (COIN). Wolfgang Utschick is an Associate
Editor for the IEEE T RANSACTIONS ON S IGNAL P ROCESSING and an editor
and founder of the Springer Book Series Foundations in Signal Processing,
Communications and Networking. Since 2010 he serves for the IEEE Signal
Processing Society as an elected member of the Technical Committee of
Signal Processing for Communications and Networking.

Das könnte Ihnen auch gefallen