7/28/2019 Fricke_WWRF14_2005

1/7

Page 1 (7)

AbstractForth generation (4G) wirelesssystems are expected to realize numerousfeatures that are not offered by thirdgeneration (3G) wireless systems. We address

these requirements and introduce an adaptive4G uplink proposal based on a recentlyproposed multiple access technique called

interleave-division multiple access (IDMA),which supports these requirements in an

elegant and simple way.

Index Terms 4G system proposal,

interleave-division multiple access (IDMA), airinterface, uplink

INTRODUCTIONsystems are expected to achieve

the following goals:

Quality of service (QoS) provisioning

Packet-based data services

(probably using mobile IPv6)

Parallel data streams with different

QoS classes (supporting hierarch-

ical source coding and video

streaming, for example)

Low delay

System scalability (w.r.t. data rates,bandwidth, cell sizes, etc.)

High peak rates (about 100 Mbps

for high mobility and 1 Gbps for

low mobility or local wireless

access)

Scalable bandwidth (about 20 MHz

to 100 MHz in 5 MHz slots)

Variable cell sizes (incl. pico,

micro, and macro cells)

High bandwidth efficiency (about 4-10

bits/s/Hz) and simultaneously high

power efficiency

High-speed mobility, high velocities,

fast adaptation

Low computational complexity (iterative

processing).Efficient transmission techniques and

cross-layer design and optimization

techniques are required to implement these

goals.

Due to many factors, direct-sequence

code-division multiple access (DS-CDMA) is

currently an attractive candidate particularly

for the uplink. DS-CDMA offers well-known

features: Dynamic channel sharing, soft

capacity, reuse factor of one, low drop-out

rate and large coverage (due to soft hand-

over), ease of cellular planning, robustness to

channel impairments and immunity against

interference, etc. These advantages are due

to spreading the information over a large

bandwidth (wideband transmission).

IDMA can be considered as a special case

of DS-CDMA. In IDMA, each data stream is

first encoded by a (very) low-rate encoder.

Conceptionally, the same encoder can be

used for each data stream (i.e., even different

users use the same encoder). For

convenience, each data stream is referred to

as a layer. We assume that the encoder is

binary, hence the layers are binary as well.Subsequently the coded bits, usually referred

to as chips in this context, are interleaved.

The main difference between IDMA and DS-

CDMA is that in IDMA each layer is assigned

a layer-specific interleaver, whereas in DS-

CDMA a layer-specific spreader is applied.

As a consequence, in IDMA chip-by-chip

interleaving is done and no spreading code is

used. Finally, the encoded and interleaved

data streams of different layers of the same

user are linearly superimposed (preferably

with different phases and amplitudes) beforetransmission. The data rate can be adapted

by superimposing a variable number of

layers. In contrast to other system designs,

An Uplink Proposal Based On Adaptive

Interleave-Division Multiple Access

Justus Ch. Fricke, Peter A. Hoeher, and Hendrik Schoeneich

4G

7/28/2019 Fricke_WWRF14_2005

2/7

Page 2 (7)

channel coding is an integral part of the

IDMA system design. A receiver-side

separation of the layers can be done

iteratively (turbo processing) by exploiting

the different interleavers, the code

constraints, and a multi-layer detector.

These properties make IDMA an attractive

candidate for the 4G uplink, but also for an

evolution of existing DS-CDMA systems and

other applications such as wireless local area

networks (WLAN), ad-hoc networks, sensornetworks, ultra wideband communications

(UWB), and underwater communications.

In the next section we provide more

background information on IDMA.

Afterwards, we present our system proposal.

Subsequently, we address the goals

mentioned above and point out how they can

be realized by our system proposal. Finally,

the conclusions are drawn.

IDMA Primer

In DS-CDMA, distinct data streamsd

m aredistinguished by different spreading

sequences. Forward error correction (FEC)

coding is typically done before interleaving

and spreading, as shown in the upper part of

Fig. 1. Hence, interleaving is performed on a

symbol-by-symbol basis. Conventionally, the

same FEC encoder and the same interleaver

is used for all data streams dm.

In the lower part of Fig. 1 a DS-CDMA

system is illustrated where the arrangement

of interleaving and spreading is reversed.

Now, distinct data streams are distinguished

by different interleavers. This special case of

DS-CDMA is called code-spread CDMA [1],

chip-interleaved CDMA (cI-CDMA ) [2] or

interleave-division multiple access (IDMA) [3]

in the literature. We will use IDMA

throughout this paper. In IDMA, FEC

encoding and spreading may be done jointly

by a single low-rate encoder, subsequently

denoted by ENC. The spreader has no

special task. Furthermore, it is important to

note that interleaving is done on a chip-by-

chip basis. The implications of this fact will be

addressed in the following. Fig. 2 shows themain principle of IDMA. For illustrative

purposes, a repetition code is assumed in

this figure.

As an alternative to the described re-

arrangement between interleaving and

spreading, IDMA may be interpreted as DS-

CDMA without spreading or as DS-CDMA

with a spreading length of one. Multiple code

words can be linearly superimposed in order

to enhance the data rate per user, a conceptwhich is called multi-code technique in

UTRA. In conjunction with IDMA, this concept

has been proposed in [4]. As mentioned

earlier, each encoded and interleaved

sequence xm is referred to as a layer. Due to

interleaving, the layers may be interpreted as

typical sequences. In case of multi-code

transmission, a set of typical sequences is

superimposed. Accordingly, with IDMA a high

bandwidth/power efficiency can be achieved.

In order to enable real-valued processing, we

assume a superposition of binary (pseudo-)

random sequences with uniformly distributedphase offsets in the following. Each layer is

assigned a specific phase and amplitude.

4G Uplink ProposalFor the uplink, an efficient yet simple

multiple access scheme supporting a high

number of active users is needed. Using

IDMA, we can handle a large number of

users, each of them using a different

interleaver. As the interleavers operate on

the spreaded bit sequence, the interleaver is

more efficient given the same number ofuncoded information bits. Hence, low delays

can easily be achieved. Furthermore,

Fig. 1: Comparison of conventional DS-

CDMA and IDMA. The data sequence ofthe mth layer is denoted as dm and the

corresponding encoded and interleavedsequence is denoted as xm.

7/28/2019 Fricke_WWRF14_2005

3/7

Page 3 (7)

resource allocation simplifies. The system

design is suitable for synchronous as well as

asynchronous networks.

We propose the transmitter and receiver

structure shown in Fig. 3 [4]. After low-rate

encoding, the interleaving is done with a

layer specific interleaver m. Assuming a

linear channel, the impact of the transmission

medium can completely be represented by

the vector h = [h 0,...,hL], where hl denotes the

lth coefficient of the equivalent discrete-time

intersymbol-interference channel model with

memory length L. At the receiver side the

individual layers, which are experiencing

multilayer interference (MLI), are detected by

a low complexity multilayer detector (MLD). It

has been widely recognized that iterativereceiver structures can achieve better

performance than separated detection and

decoding [5], [6]. For an IDMA-based system,

an iterative (turbo-like) receiver that

exchanges extrinsic information between the

MLD and the decoder (DEC) can be applied.

This receiver structure is of low complexity

and provides reliable soft outputs. The MLD

uses channel estimates delivered by a

channel estimator. We propose a pilot-layer

aided channel estimator (PLACE) [7]. Given

the log-likelihood values from the mth layerdelivered by the decoder after the last

iteration, an evaluation module is used to

calculate an estimate of the bit error rate

(BER) of the mth layer. The BER estimate is

used for soft link adaptation.

For the radio air interface we are

considering the parameters shown in Table I.

These parameters are suitable for the

proposed system and are comparable to the

uplink proposal in [8], which uses the same

bandwidth, chip duration, and is also based

on code-spreading [1]. Therefore, the

proposal in [8] is combinable with our

proposal. However, it is also possible to use

other parameters than given in Table I. A

spreading factor (SF) between 8 and 32B/5MHz should be chosen to achieve good

results.

TABLE IRADIO AIR INTERFACE PARAMETERS FOR AN

IDMA-BASED UPLINK.

Wireless Access (MC-)IDMA or

(MC-)OCDM/IDMADuplex scheme FDD or TDD

Frequency bin width B Multiples of 5 MHzTotal bandwidth e.g. 40 MHz

Roll-off factor 0.22Chip duration Tc 244 ns5 MHz/BBlock duration TB 1156.25 s

Chips per block TB/TCData modulation BPSK with layer specific

phase offsetsSpreading factor SF 8 32B/5 MHz

Code rate(with spreading)

1/SF

Fig. 2 Illustration of IDMA. For each layer, a different (pseudo-random) interleaver is used.

7/28/2019 Fricke_WWRF14_2005

4/7

Page 4 (7)

Extensions of Pure IDMAIn [9] it is shown that the error performance

of IDMA can be improved by means of

orthogonal code-division multiplexing/

interleave-division multiple access

(OCDM/IDMA). This optional extension of

IDMA is depicted in Fig. 3. The data stream

of one user is transmitted via Gm parallellayers. Orthogonal codes are used for the

parallel layers. All layers of one user share

the same interleaver.

It is also possible to perform the spreading

over multiple sub-carriers to obtain an MC-

IDMA system. This is an interesting

alternative to IDMA for systems with a huge

number of sub-carriers, or an extension of

existing or planned MC-CDMA systems [10],

[8].

Soft Link AdaptationOne enormous advantage of the low-cost

IDMA-receiver considered in [11] is the

inherent output of reliability information. In

[12], a rate/power adaptation strategy based

on the soft outputs is described.

Besides of delivering hard decisions after

the final iteration, the receiver in Fig. 3

calculates estimates of the BER in an

evaluation module. The evaluation module

compares the estimated BER with a

predefined target BER, Pt. Since the BER

degrades with increasing number of layers

because of increased multi-layerinterference, the number of layers transmitted

in the next block is decreased at the

transmitter side if the estimated BER is

higher than the target BER. If the estimated

BER is below the target BER, the

transmission power can be reduced. It was

shown in [9] that the proposed iterative

receiver is appropriate for the suggested

adaptation strategy.

A large variety of data rates can besupported by the multi-code technique. As

opposed to conventional adaptive

modulation/channel coding techniques, the

modulation scheme is fixed (and even binary)

and the same channel code is used for all

layers. Power adaptation/savings are

particularly useful for the uplink.

Multiple Antennas (MIMO)Multiple antenna systems are a key

technology for 4G. An extension of IDMA to

multi-antenna (MIMO) systems is

straightforward since each layer is allocated

a specific interleaver. In case of spatial

multiplexing, different interleavers are

allocated to the different antennas. In case

of transmit diversity, the transmit signal is

simply repeated. Therefore, neither a special

code design is needed (e.g. orthogonal or

arithmetic space-time code design), nor

further increase in complexity is caused. In

[13] excellent results have been reported for

multiple transmit antennas.

Despite co-located antennas, IDMA also

supports distributed antennas since thesignal design does not rely on orthogonality.

Fig. 3 IDMA-based system proposal. For conceptional purposes, only the mth layer is shown.

7/28/2019 Fricke_WWRF14_2005

5/7

Page 5 (7)

Cross-layer DesignTo achieve high data rates in 4G, the

overhead between the different layers (e.g.

physical layer and data link control (DLC)

layer) should be minimized by an optimized

cross-layer design. Medium access can be

managed quite easily for IDMA systems as it

can be done mainly by allocating layers. The

use of different layers for the transmission of

different information from the same user can

be combined with automatic repeat request(ARQ) protocols at the link or transport layer

for intelligent retransmission of data in an

extra layer. That could be the retransmission

of a whole block, retransmission of unreliable

bits, or additional code bits for adaptive

hybrid Type-II or Type-III ARQ schemes

without interrupting the data flow on other

layers, which again helps to fulfill QoS

parameters. Furthermore, a decision on a

repeat request can be based on the bit error

probability instead of an error detecting code.

The soft outputs provided on the physical

layer may be used up to the transport layer,

e.g. to avoid unreasonable decrease of the

TCP congestion window because of fading

conditions.

Discussion of the Goals andhow they can be Achieved byMeans of IDMA

Quality of ServiceThe quality of service (QoS) is mainly

defined by a maximum bit error rate, aminimum data rate, and a maximum delay

(especially for packet based services). These

parameters are highly dependent on the

application, e.g. text message, voice

transmission, or video transmission. IDMA-

based systems can be made highly adaptive

in order to guarantee a certain QoS level. We

do not seek quasi error -free transmission, but

apply the mentioned soft link adaptation

strategy to guarantee a certain bit error rate

for a layer or group of layers allocated to a

user or application (unequal errorprotection). On the other hand, we keep the

transmission power as low as possible for

longer battery life and less emitted radiation.

The bit error rate that can be toleratedis application-dependent, e.g. voice

transmission allows higher bit errorrates than data transmission. Instead ofusing adaptive modulation and/or

channel coding, in IDMA the number oflayers and the transmission power aremodified to meet this requirement [4].

The data rate is an essential QoSparameter, for example text messagingservices need much lower data rates

than video transmission. The data rateis adapted in a similar way as the targetBER is. With a higher number of layers

assigned to a user, its data rate ishigher. To ensure a certain BER thepower can be adapted as well.

In some applications, e.g. real-timespeech transmission, a large delay isinconvenient, in other applications evencritical, e.g. packet loss in TCP based

networks. To achieve small delays, theblock length for IDMA transmission canbe chosen to be quite small. This is

possible because the chip-by-chipinterleaving is done. Note that the blocklength given in Table I is four times

shorter than in [8].

System ScalabilityUsually, asymmetrical data links (higher

data rates in the downlink than in the uplink)

are anticipated for cellular systems. This

suggestion is based on state-of-the-art

services (e.g. internet applications). As new

applications (e.g. video telephony) are

demanding high uplink data rates as well, thisassumption is questionable for future

systems [6]. An adaptive and bandwidth

efficient IDMA-based transmission system is

future proof as it can cope possible future

need for higher uplink data rates.

For reasons of scalability and ease of

implementation we propose to divide the

available bandwidth into bins. Since the 4G

frequency allocation not known yet, the bin

width is assumed to be a multiple of 5 MHz.

That could be 220 MHz as considered in [8],

but also an inhomogeneous allocation asshown in Fig. 4. Following this proposal, the

available frequency bins need not to be

contiguous. Even dynamic bandwidth

7/28/2019 Fricke_WWRF14_2005

6/7

Page 6 (7)

allocation as in cognitive radio technologies

may be considered.

High Bandwidth and Power

EfficiencyWithout adaptive power allocation, a

spectral efficiency of about 3.75 bits/s/Hz can

be reached for one transmit antenna without

asymptotic loss in power efficiency compared

to binary antipodal modulation (2-PSK). With

adaptive power allocation, a spectral

efficiency of up to 8 bits/s/Hz (corresponding

to 256-QAM) for one transmitter antenna has

been reported [14]. With the parameters

given in Table I, the maximum theoreticthroughput for a single antenna system is

122.95 Mbps for a spectral efficiency of 3.75

bits/s/Hz, which is higher than the uplink

results presented in [8].

In conjunction with the proposed

adaptation strategy, a continuous operation

near the theoretical limit can be achieved.

Due to soft link power adaptation the mobile

device always uses minimum power for

transmission what results in long battery life.

Efficient Resource AllocationResource allocation in IDMA is done

mainly by allocating layers (multi-code

technique). If the data rate provided is not

enough, additional layers may be allocated to

a user or application. The number of layers

used for transmission can be reduced if the

data rate is higher than needed or, if the data

rate cannot be reduced for QoS reasons, the

transmit power can be increased until the

target BER is achieved. The data rate

corresponding to one layer can be tuned by

the spreading factor to avoid overhead.However, since the total load of the

system is limited and at some point it may not

be possible to allocate further layers. In this

case scheduling techniques using time-

division multiplex as in other CDMA systems

may be used. Doing so, it is advantageous

that IDMA is able to use frame lengths that

are very short in time, which again avoids

wasting bandwidth. Criterions for the

allocation should be the QoS parameters

(BER and delay).

High-speed Mobility

High-speed mobility implies that thereceiver must be able to track fast fading

channels. Towards this goal, we propose

pilot-layer aided channel estimation (PLACE)

[7]. In PLACE, one layer for every user

carries training symbols for channel

estimation. The data-carrying layers are

linearly superimposed onto the training layer

as described above. PLACE is especially

useful for the uplink as for every mobile

terminal the channel is different. In

conjunction with semi-blind channel

estimation, PLACE has been shown to

achieve excellent performance even for

rapidly changing frequency-selective fading

channels. Vehicular speeds larger than 1000

km/h (!) at a carrier frequency of 2 GHz are

feasible. In contrast to conventional channel

estimation schemes the performance even

improves with increasing vehicular speed due

to time diversity.

Low Computational ComplexityThe transmitter (e.g. the mobile terminal in

case of the uplink) has an extremely low

complexity, c.f. Fig. 3.The receiver (e.g. the base station in case

of the uplink) is more complex, since the

multilayer interference must be suppressed.

Towards this goal, knowledge of the

interferers is advantageous or even

necessary. This knowledge is available at the

base station. A possible low-complexity

receiver is the simplified version of the Wang

& Poor receiver [15] derived in [11]. This

receiver is able to cancel any type of

interference (multilayer interference, multi-

user interference, multiantenna interference,intersymbol interference, etc.) jointly. The

receiver is based on the Gaussian

Fig. 4 Possible spectrum allocation for IDMA-

based 4G services.

7/28/2019 Fricke_WWRF14_2005

7/7

Page 7 (7)

assumption and turbo processing in

conjunction with the low-rate encoder. Its

complexity is linear with respect to the

number of layers, number of chips/layer,

number of users, number of receive

antennas, number of channel taps, and the

number of iterations.

ConclusionsIn this paper, we addressed requirements

expected from 4G systems and illustratedwhy we believe that these requirements can

be solved in the uplink by means of IDMA in

an elegant and simple way. Towards this

goal, an IDMA-based adaptive air interface

has been proposed (KIEL system). We

believe that this novel concept can generate

some fruitful discussion. The common

advantages of conventional DS-CDMA are

maintained, since IDMA is just a special form

of DS-CDMA. As a consequence, existing

DS-CDMA systems may be enhanced by

IDMA as well.Specific contributions of our research

group include (i) the proposal to use IDMA

with binary data layers and layer-specific

phase offsets (in order to enable real-valued

processing), (ii) adaptation based on log-

likelihood ratios, (iii) orthogonal code-division

multiple access/interleave-division multiple

access (OCDM/IDMA), (iv) pilot-layer aided

channel estimation (PLACE), (v) unequal

error protection (UEP) in conjunction with

IDMA, and (vi) cross-layer optimization for

IDMA.

Future work should include code and

interleaver design, automatic repeat request,

synchronization issues, and peak-to-average

power reduction.

REFERENCES

[1] 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.

[2] R. Mahadevappa and J. Proakis, Mitigating multipleaccess interference and intersymbol interference in

uncoded CDMA systems with chip-levelinterleaving, Trans. Wireless Commun., vol. 1, no.

4, pp. 781792, Oct. 2002.[3] L. Ping, L. Liu, K. Wu, and W. Leung, A unified

approach to multiuser detection and space-timecoding with low complexity and nearly optimal

performance, in Proc. 40th Allerton Conference onCommunication, Control, and Computing,

Monticelli, Illlinois, Oct. 2002.[4] H. Schoeneich and P. Hoeher, Adaptive interleave-

division multiple access A potential air interfacefor 4G bearer services and wireless LANs, in Proc.

WOCN 2004, Muscat, Oman, June 2004, pp. 179182.

[5] S. Kaiser and A. Svensson (eds.), Broadband

multi-carrier based air interface, Wireless WorldResearch Forum (WWRF), Tech. Rep., April 2004,version 1.5.

[6] J. G. Andrews, Interference cancellation for cellular

systems: A contemporary overview, IEEE Wireless

Commun. Mag., vol. 12, pp. 1929, April 2005.[7] H. Schoeneich and P. Hoeher, Semi-blind pilot-

layer aided channel estimation with emphasis on

interleave-division multiple access systems, 2005,in preperation.

[8] H. Atarashi, N. Maeda, S. Abeta, and M.Sawahashi, Broadband packet wireless accessbased on VSF-OFCDM and MC/DS-CDMA, inProc. IEEE Int. Symp. Personal, Indoor and MobileRadio Commun. (PIMRC), September 2002, pp.992997.

[9] H. Schoeneich and P. Hoeher, A hybrid multipleaccess scheme delivering reliability information, inProc. Int. ITG Conf. on Source and ChannelCoding, Erlangen, Germany, Feb. 2004, pp. 437442.

[10] S. Kaiser, U.-C. Fiebig, N. Matoba, A. Jeffries, M.de Courville, and A. Svensson, Broadband multi-

carrier based air interface, Wireless WorldResearch Forum, Tech. Rep., May 2002.

[11] L. Liu, W. Leung, and L. Ping, Simple iterativechip-by-chip multiuser detection for CDMA

systems, in Proc. IEEE VTC 2003, 2003, pp.21572161.

[12] P. Hoeher, Adaptive modulation and channelcoding using reliability information, in Proc. 5th Int.

OFDM-Workshop, Hamburg, Germany, Sept. 2000,pp. 14.114.4.

[13] K. Y. Wu, Li Ping, and W. K. Leung, Multi-layer

turbo space-time codes for high-rate applications,in Proc. GLOBECOM 2004, December 2004, pp.37583762.

[14] Li Ping and L. Liu, Analysis and design of IDMAsystems based on SNR evolution and powerallocation, in Proc. IEEE VTC 2004-Fall,September 2004, pp. 10681072.

[15] X. Wang and H. Poor, Iterative (turbo) soft

interference cancellation and decoding for codedCDMA, IEEE Trans. Commun., vol. 47, no. 7, pp.10461061, July 1999.