4876 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 8, NO.

9, SEPTEMBER 2009

Adaptive Modulation and Coding Techniques for

OFDMA Systems
Romano Fantacci, Fellow, IEEE, Dania Marabissi, Member, IEEE, Daniele Tarchi, Member, IEEE,
and Ibrahim Habib, Senior Member, IEEE

Abstract—The demand for high-speed services and multimedia Differently from the classical OFDM technique, OFDMA al-
applications anywhere and anytime has led to the rise of wireless lows to assign each subcarrier independently to the users, thus,
communications. In particular, WiMAX technology is nowadays enabling a flexible use of resources and making it possible
considered one of the most prominent solutions capable to
provide a Broadband Wireless Access (BWA) in metropolitan to support nomadic or mobile applications. The multicarrier
areas with a simpler installation and lower cost than traditional nature of the OFDMA transmission reaches better results if
wired alternatives. This paper deals with the proposal of efficient integrated with adaptive techniques in order to achieve higher
adaptive modulation and coding techniques to be used in WiMAX efficiency in terms of error rate and throughput.
based wireless networks, that allow to improve network perfor-
mance in the case of Non Line-of-Sight communications, which Our aim is to fully exploit the characteristics of the mul-
are typical in urban environments. Through these techniques it is ticarrier transmission and flexibility offered by the OFDMA
possible to switch the modulation order and coding rate in order scheme for optimizing the use of the network resources by
to better match the channel conditions, and, hence, obtaining assigning adaptively modulation and coding to each data
better performance both in terms of error probability and data transmission, based on actual channel conditions.
throughput.
Among several resource allocation techniques, Adaptive
Index Terms—OFDMA, adaptive modulation and coding,
Modulation and Coding (AMC) allows OFDMA systems to
IEEE 802.16, broadband wireless access, WiMAX.
select the most appropriate Modulation and Coding Scheme
(MCS) depending on the propagation conditions of the com-
I. I NTRODUCTION munication channel: during good propagation conditions a
IGH data rate services, multimedia applications and, in high order modulation scheme with low coding redundancy is
H general, high quality information streams are nowadays
in high demand and will continue for the near future. Wireless
used in order to increase the transmission data rate, while dur-
ing a signal fade, the system selects a more robust modulation
systems are considered to be a viable and attractive solution scheme and a higher coding rate to maintain both connection
to provide high data rates communications, in particular to quality and link stability without increasing the signal power.
mobile users. The use of AMC schemes in wireless communication sys-
Recently, the IEEE 802.16 family of standards [1], [2], tems is a topic widely considered and investigated in the recent
supported by the WiMAX commercial consortium, provides literature [3]–[12]. In [3] the idea of combining the OFDM
the specifications for the Physical and Medium Access Con- technique with adaptive modulation and coding is presented
trol (MAC) layers for a Broadband Wireless Access (BWA) by showing the advantages in terms of overall throughput,
communication protocol. Among several alternatives the IEEE while in [4], [5] the principles of AMC were proposed and
802.16 standards foresee the use of the Orthogonal Frequency investigated. Successively, in [6], [7] the problem of link
Division Multiplexing (OFDM) in order to mitigate the ad- adaptation in OFDM system was introduced by considering
verse effects of the frequency-selective multi-path fading and jointly subcarriers, modulation and power allocation. Adaptive
efficiently contrast the inter-symbol and inter-carrier interfer- resource allocation in OFDMA systems was also considered
ences (ISI and ICI). In particular, we focus our attention on the in [8], where a novel technique for balancing fairness and
Orthogonal Frequency Division Multiple Access (OFDMA) capacity was proposed, and in [9], where an utility based
scheme that is the basis of the IEEE 802.16e standard [2]. approach was discussed. The use of AMC in WiMAX and
OFDMA systems was considered in [10], [11] with special
Manuscript received February 18, 2009; revised May 14, 2009; accepted
June 8, 2009. The associate editor coordinating the review of this paper and focus on WiMAX systems and services. Finally, a joint
approving it for publication was N. Kato. analysis of AMC techniques and admission control schemes
R. Fantacci, D. Marabissi, and D. Tarchi are with the Department in OFDMA systems was proposed in [12].
of Electronic and Telecommunications, University of Florence, Via di S.
Marta, 3, 50139 Firenze, Italy (e-mail: {romano.fantacci, dania.marabissi, Differently from previous papers [3]–[12], the focus here is
daniele.tarchi}@unifi.it). on AMC schemes that are compliant with the IEEE 802.16e
I. Habib is with the Department of Electrical Engineering, City College
of the City University of New York, New York, NY 10031, USA (e-mail: standard [2] due to the use of the slot basis subcarrier
[email protected]). allocation instead of the simple (i.e., individual) subcarrier
This work has been partially supported by Italian National projects IN- allocation [8]–[12]. Moreover, a suitable model of the AMC
SYEME, under grant number RBIP063BPH, and WORLD, under grant
number 2007R989S. process is also outlined here to make possible an extension of
Digital Object Identifier 10.1109/TWC.2009.090253 the obtained results to different application scenarios or user
c 2009 IEEE
1536-1276/09$25.00 ⃝
FANTACCI et al.: ADAPTIVE MODULATION AND CODING TECHNIQUES FOR OFDMA SYSTEMS 4877

requirements with respect to those considered here.

In this paper we propose two AMC approaches for OFDM
systems: in the first scheme we consider the influence of
the wireless channel in terms of channel attenuation due to
multipath fading while in the latter in terms of decoding
errors at the receiving end. For the first approach, our aim
is to develop techniques that improve the system performance
in terms of suitable QoS metrics: keep the error probability
below a specified threshold or maximize the data throughput.
In the second approach the MCS selection is based on the
(a) The bin structure (b) The AMC slot composed by bins
amount of received packets with errors, having as target that
of minimizing that number and reducing the implementation Fig. 1. The AMC slot in IEEE 802.16e.
complexity.
The paper is structured as follows. In Section II, the system
model is presented by focusing the attention on the particular 5 µs separates consecutive frames. We have chosen a balanced
OFDMA structure used in the IEEE 802.16e standard. In division of the frame, dividing it into 40 OFDM symbols for
Section III, the AMC techniques are presented, while in the Downlink subframe and 39 OFDM symbols for the Uplink
Section IV the numerical results obtained by computer simu- subframe, namely equally parceling the 79 symbols frame
lations are shown by comparing the behavior of the proposed defined in the IEEE 802.16e standard as possible size for the
adaptation techniques. Finally, in Section V the conclusions OFDMA frame [13]. The link adaptation process performing
are drawn. the AMC techniques is done in uplink by the BS by measuring
the channel quality according to the procedure sketched in
Fig. 2.
II. S YSTEM M ODEL
In this section an OFDMA system for mobile wireless III. OFDMA-AMC S YSTEM
Metropolitan Area Networks based on a Time Division Du- In this section the proposal of AMC systems exploiting
plexing (TDD) transmission will be described. In OFDM the OFDMA capabilities is described considering that the
communication systems, the available spectrum is accessed modulation order and coding rate of each complex symbol
by a large number of subcarriers. We focus on an OFDM- �� associated to the corresponding �-th subcarrier (0 ≤ � ≤
based wireless communication system where � users can 1023) could be changed according to the physical channel
communicate with the Base Station (BS) and a set of carriers, state, even if, as explained in Section IV, the subcarriers
namely �� , are assigned to the �-th user, grouped in slots. are considered to be grouped among them, and the MCS is
Communications belonging to different users are independent changed accordingly for all the subcarriers of the slot.
among them, because the channel conditions of the set of sub- The selection of the modulation and coding scheme is
carriers assigned to one user are statistically independent from based on the physical channel state information. By assuming
the channel conditions suffered by the subcarriers allocated the channel as reciprocal in frequency, the channel response,
to another user. A slot is composed by a group of adjacent ˆ
�(�), is estimated at the BS side during the uplink communi-
subcarriers subdivided in 48 data subcarriers and 8 pilot cation subframe, and the obtained information are used in the
subcarriers, for a total of 56 subcarriers, according to the IEEE next TDD frame to select the MCS to be used for downlink
802.16e standard (AMC permutation) [2]. This slot structure is communications. The discussion on specific methods to derive
the same for both downlink and uplink communications, and channel estimation is out of the scope of this paper, however,
is based on the concept of bin, which is a set of 9 subcarriers an interested reader can find an extensive surveying of the
(8 data subcarriers and 1 pilot subcarrier), as shown in Fig. 1a: most used approaches in [14]. As a consequence of our MCS
a slot is, therefore, defined as set of bins satisfying the relation selection approach (sketched in Fig. 2), we have assumed here
�� × �� = 6, where �� is the number of consecutive bins that MSs map each subcarrier, in Uplink, with the same con-
and �� the number of consecutive OFDM symbols, as shown stellation and coding rate used during Downlink subframe. It is
in Fig. 1b. Among all the possible permutations, the selected important to stress that even if we assume perfect knowledge
structure is (6 bins, 1 OFDM symbol). The slot is the minimum of the channel impulse response at the receiver, numerical
resource allocation unit in this system. results have been derived by taking into account the delay
The modulation and coding rates are adapted on a frame (i.e., a frame) introduced by the channel estimation algorithm
basis. We consider a frame duration (�� = 8 ms) and a FFT and the channel propagation conditions variation usually due
dimension of 1024 for a 10 MHz channel according to the to MSs mobility. In addition the MCSs are assigned per slot
Scalable OFDMA concept [13]: the subcarrier spacing is a and not per subcarriers, hence, the channel values used to
fixed parameter in all the systems with different bandwidth. select the MCSs are averaged on the slot.
In this paper we assume a TDD structure: this is the most The AMC system is modeled as a Moore’s state machine
suitable solution for data traffic such as IP based multi-media where each state is represented by a couple formed by a
services with multiple QoS requirements. The TDD frame modulation order and a coding rate. For the case considered in
begins with the Downlink subframe, then a guard time of 5 µs this paper the modulations schemes are QPSK, 16QAM and
precedes the Uplink subframe. An additional guard time of 64QAM, and codes are of the block type with coding rates
4878 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 8, NO. 9, SEPTEMBER 2009

time

Downlink Uplink Downlink Uplink Downlink

....................... ....................... ....................... ....................... .......................

TTG RTG TTG RTG TTG

Channel estimation Channel estimation

Frame duration

Parameters for Parameters for

the next Frame the next Frame

Fig. 2. BS commands the next frame modulation.

1/2 and 3/4. Despite this we would like to emphasize that the parameter (i.e., the channel attenuation factor) overcomes a
aim of this model is to define a generic approach that can threshold. In this case the state machine enters in a new state
be easily adapted to different AMC techniques as well as to and, as a consequence, a different MCS is used.
different MCSs. The main difference between the two channel state algo-
In particular, we propose here to use two AMC techniques rithms is how thresholds are derived: their values strongly
based on different approaches. The first, named Channel state, influence the behavior of the adaptation algorithm in the per-
is represented in Fig. 3a. This technique aims to adapt the formance measurement. Hence, it is possible to state that the
modulation and coding scheme after deriving an estimation proposed model can be adapted to different user requirements
of the channel behavior in terms of attenuation coefficient. by selecting suitable thresholds for state changes.
Conversely, the second technique, named Error state, has been The Maximum Throughput (MT) algorithm aims to max-
developed on the basis of an approach different from the imize the system throughput without constraints on target
previous one by considering the received frame error rate, as Block Error Probability. The Target BLER (TBLER) algo-
shown in Fig. 3b. rithm, aims to respect a certain target BLER2 (Block Error
For the channel state approach, the actual medium channel Rate) due to specific QoS (i.e., data integrity) requirements.
attenuation factor (�) over a slot is compared with suitable The target value of the BLER can be the same for each SNR
thresholds, in order to determine the MCS to be used in the (i.e., the mean signal-to-noise ratio averaged over the frame
next frame1 . Fig. 3a illustrates the Moore’s state machine for length), or can vary with it.
the case of five thresholds represented, in the figure, by the 1) MT algorithm: The MT algorithm aims to maximize the
letters �, �, �, � and � where � is the lowest threshold overall link throughput by selecting for each SNR value the
value, i.e., � < � < � < � < �. The modulation order MCS that allows the maximum throughput. Hence, the trans-
is chosen independently slot by slot. For simplicity, in the mission efficiency is enhanced but the performance in terms of
following we refer to the generic channel attenuation factor � error rate could be worse, since no constraint on the maximum
omitting the subcarrier index �. acceptable error probability is introduced. Thresholds for the
Conversely, in the error state approach, the MCS switches in parameter � are those values for which the throughput curves
function of the number of detected frame with errors. Fig. 3b related to different MCS are overlapped (Fig. 4). The aim
shows the Moore’s state machine for the error state approach, is to select, for each frame and SNR value, the modulation
under the assumption of an error depth equal to four frames. order and the coding rate that maximize the total throughput.
The QPSK modulation with coding rate 1/2 state is always The final result is that the most efficient scheme in terms of
used at the beginning of a communication session with a par- throughput for each � value is selected. As a consequence, we
ticular user, supposing the BS unable to retrieve information can state that this algorithm is more devoted to such services
from the previous uplink communications phase about the that request the maximum achievable data rate, with a lower
channel conditions of that user. sensibility to the error probability, e.g., video streaming or
VoIP services. This algorithm has no degree of freedom and
A. Channel state technique thresholds can be found by comparing the throughput of the
considered transmission schemes [15]. Towards this end, the
In this section two algorithms named Maximum Throughput classical throughput formulation provided in [16] has been
and Target BLER based on the channel state approach are pro- modified here to better match the slot structure of the proposed
posed . Each adaptation algorithm is basically characterized by system: the throughput (�) can be expressed as a function of
five thresholds. The MCS remains the same until the measured the channel attenuation factor � and used MCS for a fixed
1 In practical applications parameter � can be derived in relation to different
environments by considering the mean of the channel attenuation obtained by 2 We assume in the following to approximate the theoretical block error
standard approaches [14] for all the symbol time for one subcarrier within a probability with the estimated BLER, where the estimation period is consid-
certain frame. ered as long as a frame period.
FANTACCI et al.: ADAPTIVE MODULATION AND CODING TECHNIQUES FOR OFDMA SYSTEMS 4879

(a) Moore’s state machine for channel state techniques.

(b) Moore’s state machine for error techniques.

Fig. 3. Moore’s state machines: 0=correct frame, err=error frame and x=any frame.

SNR, as: ∙ � is the dimension of the codeword.

4 ⋅ (1 − ������ (�, SNR, �, �� )) ⋅ �� ⋅ ��� ⋅ log2 (� ) ⋅ �� In this paper we have considered 4 bit blocks and 8
�= bit codewords for coding rate 1/2, adjusted with puncturing
4 ⋅ ��
(1) to obtain coding rate 3/4. Both ��MQAM and ����MQAM are
where � = 4, 16, 64 represents the modulation order, functions of SNR, � and modulation order (� ); in particular,
������ stands for the block error probability related to each the symbol error probability ���� in a Rayleigh channel can
modulation–coding rate couple, �� is the number of subcarri- be expressed as follows [16]:
ers per user, ��� is the number of OFDM symbols forming the √ ⎛√ ⎞
4( � − 1) 3SNR ⋅ �2
Downlink OFDMA subframe, �� is the adopted coding rate ���� (SNR, �, � ) = √ ⋅�⎝ ⎠
and �� is the duration of the OFDMA frame. The expression � (� − 1)
of the block error probability ������ can be simply derived as: ⎡ √ ⎛√ ⎞⎤2
2( � − 1) 2
3SNR ⋅ � ⎠⎦
������ = 1 − ������� (2) −⎣ √ ⋅�⎝ (4)
� (� − 1)
where ������� is the probability to have a correct received
block: In Fig. 4 it is shown how thresholds (i.e., � values) for
� ( ) state changes are defined; in particular herein it is supposed to
∑ � )� −�
have a SNR equal to 20 dB. Whenever a different propagation
���MQAM 1 − ��MQAM
(
������� = (3)
�=0
� environment is considered it is straightforward to note that,
due to different performance in terms of throughput, also the
assuming that:
thresholds have different values. The MCSs selected according
∙ the fading process is independent on a frame to frame
to the MT approach are reported in Table I. By solving (1)
basis; for different propagations environments and conditions, it is
∙ a block coder with ���� = 3;
possible to derive off-line the � values to be used in the
∙ a Maximum Likehood decoding algorithm
⌊ �min −1 ⌋ following AMC techniques; in that sense, even if specific to
∙ �= is the number of corrigible errors;
2
����MQAM
a certain environment, the approach presented herein can be
∙ ��MQAM = ∼ is the bit error probability as a extended to almost all environments.
log2 �
function of the symbol error probability in case of Gray 2) TBLER Technique: The main feature of the TBLER
mapping; technique is that of keeping the error rate below a target limit,
4880 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 8, NO. 9, SEPTEMBER 2009

Fig. 4. Thresholds definition (MT technique). Fig. 5. Thresholds definition (TBLER technique).

TABLE I TABLE II
MCS SELECTION FOR THE MT APPROACH . MCS SELECTION FOR THE TBLER APPROACH .

� values MCS � values MCS

� < 0.09 QPSK 1/2 � < 0.46 QPSK 1/2
0.09 ≤ � < 0.2 QPSK 3/4 0.46 ≤ � < 0.65 QPSK 3/4
0.2 ≤ � < 0.33 16QAM 1/2 0.65 ≤ � < 0.95 16QAM 1/2
0.33 ≤ � < 0.51 16QAM 3/4 0.95 ≤ � < 1.38 16QAM 3/4
0.51 ≤ � < 0.57 64QAM 1/2 1.38 ≤ � < 1.95 64QAM 1/2
0.57 ≤ � 64QAM 3/4 1.95 ≤ � 64QAM 3/4

hence guaranteeing a fixed level of QoS in terms of ������ . Fig. 3b we have that state transitions are allowed only for
Therefore this algorithm is more suitable to services requesting neighbor states (e.g., from QPSK 3/4 to 16QAM 1/2 and
a strict respect of a certain maximum error rate value as best backwards but not from QPSK 3/4 to 16QAM 3/4). In details:
effort traffics. ∙ A transition to a more efficient state occurs only if all
This technique has one degree of freedom, represented by the last four frames are without errors (when we are the
the imposed target BLER; different values of the target BLER most efficient state (i.e., 64QAM 3/4) we still remain in
correspond to different system performance. For each value that state).
of SNR we can derive BLER as a function of �. Conse- ∙ A transition to less efficient state occurs only if the last
quently, we can define five target � values (i.e., thresholds) frame has at least one error (when we are in the less
in correspondence of which state transitions for the AMC efficient state (i.e., 4QAM 1/2) we still remain in that
process occur. The � values assure the target BLER, for a state).
certain SNR and a certain MCS. Thresholds can be derived ∙ The state remains the same if the last frame is without
by selecting the � value according to the specified Target errors independently from the previous three.
BLER (e.g., BLERtarget ). Fig. 5 shows BLER behaviors as a
The error technique has been also implemented for different
function of �. This figure highlights how the thresholds are
memory depth that corresponds to a different consideration on
defined under the assumption of a SNR value equal to 16 dB
the erroneous frames. Due to brevity we have focused herein
and BLERtarget equal to 8 ⋅ 10−3 . As for the MT technique,
our attention only to the memory depth of four frames.
whenever a different propagation environment is considered it
is straightforward to note that, due to different performance in
terms of BLER, also the thresholds have different values. The IV. N UMERICAL R ESULTS
resulting MCSs and related switching thresholds according to
In this Section we describe the numerical results obtained
the BLER approach are shown in the Table II.
through computer simulations [17]. In particular, the perfor-
mance of the proposed adaptive algorithms are compared with
B. Error technique that of the fixed modulation and coding (FMC) algorithm.
The error based technique takes into account the amount of To the best of our knowledge, the FMC scheme is the only
errors in the downlink subframe. Fig. 3b shows a particular benchmark possible because all the AMC techniques available
implementation of the AMC algorithm based on frame errors, in the literature on this subjects are not based on a slot
under the assumption of an error depth3 equal to four. From allocation. In deriving our simulations results we have focused
on a WiMAX environment with:
3 The error depth corresponds to the number of consecutive frames consid-
ered in the error count: a frame is considered erroneous if it has at least one ∙ radio frequency carrier �� = 3.5 GHz;
erroneous symbol ∙ bandwidth of 10 MHz;
FANTACCI et al.: ADAPTIVE MODULATION AND CODING TECHNIQUES FOR OFDMA SYSTEMS 4881

0 1100
10
QPSK 1/2
QPSK 3/4
1000 16QAM 1/2
16QAM 3/4
64QAM 1/2
900 64QAM 3/4
-2
BLERtarget=10-3
−1 BLERtarget=10
10 800 MT

700

Throughput [kb/s]
600
BLER

−2
10 500

400
QPSK 1/2
QPSK 3/4
300
16QAM 1/2
−3 16QAM 3/4
10 200
64QAM 1/2
64QAM 3/4
−2 100
BLERtarget = 10
−3
BLER = 10 0
target
0 5 10 15 20 25 30
MT
10
−4 SNR (dB)
0 5 10 15 20 25 30
SNR(dB)
Fig. 7. Throughput performance comparison between the channel state
approach and FMC alternative.
Fig. 6. BLER performance comparison between the channel state approach
and FMC alternative.

particular, this allows throughput advantages at the expense

maximum Doppler deviation �� ≃ 408 Hz (mobility
∙ of a lower data integrity at the receiving ends. The opposite is
terminals up to 125 km/h); valid under low BLERtarget conditions. In this case, in order to
∙ coherence time �� ≃ 1.03 ms; meet the data integrity requirements at the receiving end, the
∙ maximum delay spread equal to 20 µs (worst case); possibility of using bandwidth efficient modulation schemes
∙ ITU-R vehicular channel model A, with 6 paths [18]. is reduced and hence data throughput degradations arise.
Using an oversampling factor 8/7 we can structure an Fig. 7 shows the performance in terms of data throughput
OFDMA symbol over a 11.429 MHz bandwidth. The useful of the MT and TBLER methods in comparison with differ-
symbol time is �� = 89.6 µs and the guard time �� is 1/8 of ent FMC schemes. We can note from this figure that the
an OFDMA symbol duration, hence, a maximum of 11.2 µs throughput behavior has an opposite trend with respect to the
delay spread can be accepted [13]. BLER curves (Fig. 6): reducing the target block error rate,
The considered number of active users within the system the throughput value gets lower and vice versa. Moreover,
is 25; however it is also important to stress that under the Fig. 7 highlights that the best throughput value is allowed
assumption of an ideal knowledge at the BS side of the by the MT method. The better behavior of the MT scheme
channel parameters and frequency synchronization for all users is highlighted in the figure being the MT curve the envelop
our numerical results are independent from the number of of all the other curves given in the figure. In particular, this
simultaneous users. In actual applications the presence of other allows a greater spectral efficiency, at low SNR conditions, and
users can introduce some performance degradations mainly lower error probability values (high SNR). As a consequence,
due to the presence of different carrier frequency offsets (CFO) this technique represents a good solution for applications as
among the users. This drawback may be reduced by adopting realtime or UDP services where a low integrity information
suitable countermeasures; among them a suitable and efficient loss is accepted in favor of an increased bit rate.
method seems to be that proposed in [19]. The main advantage compared to the FMC scheme is the
Fig. 6 shows the TBLER and the MT techniques perfor- increase in terms of data rate under the same constraints.
mance in terms of BLER. For the TBLER the system keeps On the basis of the obtained results we can state that the
working with the lowest MCS, namely QPSK modulation with selection of the best adaptive algorithm depends on the specific
coding rate 1/2, until the SNR allows to meet the error rate system requirements. While the MT technique represents a
constraint, then the system switches to higher transmission well suited method for applications requesting a strict respect
schemes since a better spectral efficiency is achieved without of throughput, the TBLER technique is well suited for generic
loosing the target performance tie. data applications where a target error rate needs to be fulfilled
Two different target values of BLER have been reported in based on the requested QoS level. As a matter of fact, the
the figure, i.e., BLERtarget = 10−2 and BLERtarget = 10−3 . TBLER technique has been designed to offer the possibility
The best BLER performance is obtained for the lowest target of switching the target value whenever the QoS requirement
value. For the MT scheme the BLER performance are worse of a specific application needs it.
compared to the TBLER technique due to the lack of control As for the error state AMC algorithm, we have derived its
on the error rate for the MT method. performance in the case of different values of the memory
In order to introduce the numerical results shown in Fig. 7 depths and compared the obtained results with those of the
we point out from Fig. 6, that it is straightforward to note FMC scheme. Figs. 8 and 9 show performance comparisons in
in that if the BLERtarget increase, the BLER performance is terms of BLER and throughput, respectively, of different AMC
worse. This is mainly due to the fact that high bandwidth techniques based on errors counting with the FMC alternative.
efficient modulation schemes are used more frequently. In In particular, we have considered here to have a memory depth
4882 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 8, NO. 9, SEPTEMBER 2009

0
10
guarantees the highest efficiency while respecting the error
target, and MT that, conversely, aims to maximize the system
throughput without controlling the error rate performance. A
−1
10 frame error rate approach has been then considered by taking
into account the effect of the errors in previous frames. It
has been highlighted in the paper by means of computer
simulations that all the proposed techniques allow to satisfy
BLER

−2
10
different QoS constraints in terms of BLER or throughput and
allow significant advantages with respect to the FMC where
QPSK 1/2
QPSK 3/4
the modulation and coding scheme are fixed.
−3 16QAM1/2
10
16QAM 3/4
64QAM 1/2 R EFERENCES
64QAM 3/4
Error state memory 3/4
Error state memory 2
[1] IEEE Standard for local and metropolitan area networks - Part 16: Air
Error state memory 1 Interface for Fixed Broadband Wireless Access Systems, IEEE Std.
−4
10
0 5 10 15 20 25 30
802.16-2004, Oct. 2004.
SNR(dB) [2] Amendment to IEEE Standard for Local and Metropolitan Area Net-
works - Part 16: Air Interface for Fixed Broadband Wireless Access
Fig. 8. BLER performance comparison between the error state approach and Systems - Physical and Medium Access Control Layers for Combined
FMC alternative. Fixed and Mobile Operation in Licensed Bands, IEEE Std. 802.16e-
2005, Dec. 2005.
1100 [3] T. Keller and L. Hanzo, “Adaptive modulation techniques for duplex
QPSK 1/2
QPSK 3/4 OFDM transmission,” IEEE Trans. Veh. Technol., vol. 49, no. 5, pp.
1000 16QAM 1/2
16QAM 3/4 1893–1906, Sept. 2000.
64QAM 1/2 [4] A. J. Goldsmith and S.-G. Chua, “Adaptive coded modulation for fading
900 64QAM 3/4
Error state memory 3/4
Error state memory 2
channels,” IEEE Trans. Commun., vol. 46, no. 5, pp. 595–602, May
800 Error state memory 1 1998.
700
[5] W. T. Webb and R. Steele, “Variable rate QAM for mobile radio,” IEEE
Trans. Commun., vol. 43, pp. 2223–2230, July 1995.
Throughput [kb/s]

600 [6] C. Y. Wong, R. S. Cheng, K. B. Letaief, and R. D. Murch, “Multiuser

OFDM with adaptive subcarrier, bit and power allocation,” IEEE J.
500
Select. Areas Commun., vol. 17, no. 10, pp. 1747–1758, Oct. 1999.
400 [7] M.-K. Chang and C.-C. J. Kuo, “Power control, adaptive modulation and
subchannel allocation for multiuser downlink OFDM,” in Proc. IEEE
300 VTC2004-Fall, Los Angeles, CA, USA, Sept. 2004.
200
[8] Z. Shen, J. G. Andrews, and B. L. Evans, “Adaptive resource allocation
in multiuser OFDM systems with proportional rate constraints,” IEEE
100 Trans. Wireless Commun., vol. 4, no. 6, pp. 2726–2737, Nov. 2005.
[9] G. Song, Y. Li, and M. Inc, “Utility-based resource allocation and
0
0 5 10 15 20 25 30 scheduling in OFDM-based wireless broadband networks,” IEEE Com-
SNR (dB) mun. Mag., vol. 43, no. 12, pp. 127–134, Dec. 2005.
[10] C. Tarhini and T. Chahed, “On capacity of OFDMA-based IEEE802.16
Fig. 9. Throughput performance comparisons between the error state WiMAX including adaptive modulation and coding (AMC) and inter-
approach and FMC alternative. cell interference,” in Proc. IEEE LANMAN 2007, June 2007, pp. 139–
144.
[11] H. Wang, H. P. Schwefel, and T. S. Toftegaard, “History-based adaptive
modulation for a downlink multicast channel in OFDMA systems,” in
of 1, 2, 3 and 4. Due to similar performance results, memory Proc. IEEE WCNC 2008, Las Vegas, NV, USA, Apr. 2008, pp. 1588–
depths 3 and 4 are represented in the figures by a curve. 1592.
[12] S. E. Elayoubi and B. Fourestie, “Performance evaluation of admis-
This allows us to state that while increasing complexity a sion control and adaptive modulation in OFDMA WiMax systems,”
high memory depth does not provide significant performance IEEE/ACM Trans. Networking, vol. 16, no. 5, pp. 1200–1211, Oct. 2008.
improvement. To justify this behavior, it is possible to note [13] H. Yaghoobi, “Scalable OFDMA physical layer in IEEE 802.16 Wire-
lessMAN,” Intel Technology J., vol. 8, no. 3, Aug. 2004.
that higher is the memory depth lower is the block error rate, [14] A. Molisch, Wireless Communications. Chichester, UK: John Wiley &
but also lower is the throughput. This can be explained by Sons, 2005, ch. 19.
considering that a higher memory depth corresponds to aver- [15] D. Takeda, Y. C. Chow, P. Strauch, and H. Tsurumi, “Threshold
controlling scheme for adaptive modulation and coding system,” in Proc.
age the error rate on a longer time interval thus corresponding IEEE PIMRC 2004, Barcelona, Spain, Sept. 2004.
to a more proactive approach to track the channel variations. [16] J. G. Proakis, Digital Communications, 3rd ed. McGraw-Hill Interna-
tional, 1995, ch. 12.
[17] MUDiSP3 - MUltirate Digital Signal Processor v3. [Online]. Available:
V. C ONCLUSION http://mudisp3.sourceforge.net/
This paper has discussed the use of adaptive modulation and [18] Guidelines for evaluation of radio transmission technologies for IMT-
2000, ITU-R Recommendation M.1225, Feb. 1997.
coding schemes in an OFDMA based system. In particular, [19] D. Marabissi, R. Fantacci, and S. Papini, “Robust multiuser interference
two approaches have been proposed by taking into account cancellation for OFDM systems with frequency offset,” IEEE Trans.
the channel behavior both in terms of attenuation and frame Wireless Commun., vol. 5, no. 11, pp. 3068–3076, Nov. 2006.
error rate. The first approach selects the most suitable MCS
on the base of the estimation of the channel attenuation on
the uplink. This technique has been implemented according
to two different algorithms: TBLER, aiming to keep the error
rate under a specified value and to employ the MCS which
FANTACCI et al.: ADAPTIVE MODULATION AND CODING TECHNIQUES FOR OFDMA SYSTEMS 4883

Romano Fantacci (F’05) born in Pistoia, Italy, grad- Daniele Tarchi (S’98, M’06) was born in Flo-
uated from the Engineering School of the Universit rence, Italy, in 1975. He received the MSc degree
di Firenze, Italy, with a degree in electronics in 1982. in Telecommunications Engineering and the PhD
He received his Ph.D. degree in telecommunications degree in Informatics and Telecommunications En-
in 1987. After joining the Dipartimento di Elettron- gineering from the University of Florence, Italy, in
ica e Telecomunicazioni as an assistant professor, he 2000 and 2004, respectively. He is now a research
was appointed associate professor in 1991 and full assistant at the University of Florence, Italy. His
professor in 1999. His current research interests are research interests are in both Data Link and Phys-
digital communications, computer communications, ical Layers, with particular interests to Resource
queuing theory, satellite communication systems, Allocation algorithms in wireless networks, Link
wireless broadband communication networks, ad- Adaptation and Adaptive Modulation and Coding
hoc and sensor networks. He has been involved in several European Space Techniques, MAC Protocols for Broadband wireless access. He has been
Agency (ESA) and INTELSAT advanced research projects. He is the author of involved in several national projects (Insyeme, Rescue, Pattern and Women)
numerous articles published in prestigious communication science journals. as well as European projects (Nexway, Newcom, Satnex, COST289). He
He guest edited special issues in IEEE journals and magazines and served is currently serving as Associate Editor for IEEE T RANSACTIONS ON
as symposium chair of several IEEE conferences, including VTC, ICC and W IRELESS C OMMUNICATIONS, and has been reviewer of several technical
Globecom. Professor Fantacci received the IEE IERE Benefactor premium papers submitted to journals and magazines and TPC for several symposium
in 1990 and IEEE COMSOC Award Distinguished Contributions to Satellite of ICC and Globecom of the last years. He has been publicity chair for a
Communications in 2002. He is currently serving as Associate Editor for Workshop within IWCMC’09.
T ELECOMMUNICATION S YSTEMS , I NTERNATIONAL J OURNAL OF C OMMU -
Ibrahim Habib received the PH.D. degree from the
NICATIONS S YSTEMS , IEEE T RANSACTIONS ON C OMMUNICATIONS and
City University of New York, the M.Sc. degree from
Area Editor for IEEE T RANSACTIONS ON W IRELESS C OMMUNICATIONS.
Polytechnic University of New York, and the B.Sc.
degree from Ain Shams University, Cairo, Egypt
Dania Marabissi (M’00) born in Chianciano, Italy.
all in Electrical Engineering. In 1991 he joined the
She received the degree in Telecommunications En-
Faculty of the City University of New York where
gineering and the PhD degree in Informatics and
is now a Professor and Director of the Advanced
Telecommunications Engineering from the Univer-
Networking Research Laboratory. From 1997 till
sity of Florence in 2000 and 2004, respectively.
2001 he was with AT&T Labs and then Telcordia
She joined the Electronic and Telecommunications
Technologies Research Labs where he worked on the
Department at University of Florence in 2000 where
architecture design and planning of next generation
now works as assistant professor. She currently con-
IP networks and optical metro networks. From 2001 till 2008 he has been
ducts research on physical layer design for broad-
consulting with a number of industrial and research labs both in the USA and
band wireless systems. In particular her interests
Europe. His research interests spans different areas of networking including
include WiMAX and OFDM systems, resource al-
architecture design and traffic engineering in next generation networks, quality
location strategies, channel estimation and synchronization. She has been
of service enabling and optimization of resources in wireless networks,
involved in several national and European research projects and is author
and designing service enabling platforms for healthcare applications. He
of technical papers published in international journals and conferences.
has published extensively in those areas. He was a Guest Editor for a
She is currently serving as Associate Editor for IEEE T RANSACTION ON
number of special issues in the J OURNAL ON S ELECTED A REAS IN C OM -
V EHICULAR T ECHNOLOGY.
MUNICATIONS (JSAC), IEEE C OMMUNICATIONS M AGAZINE, W IRELESS
C OMMUNICATIONS M AGAZINE, and John Wiley J OURNAL ON W IRELESS
C OMMUNICATIONS AND M OBILE C OMPUTING. He is currently an editor of
the IEEE T RANSACTIONS ON W IRELESS C OMMUNICATIONS, as well as a
number of Elsevier networking journals. He has co-chaired and chaired many
symposia and workshops in various IEEE conferences. He is frequent invited
guest speaker at several universities and research labs both in the US and
internationally.