An Adaptive Feedback Method with Channel Change

Indication for Band AMC in a Femto Cell

Changyoun Kim, Jaeki Lee, Changhyun Nam, Jangho Yoon, and Hwang Soo Lee
Division of Electrical Engineering, School of Electrical Engineering and Computer Science, KAIST
Daejeon, Korea
{cykim83, ljk, chang, icespy}@mcl.kaist.ac.kr, hwanglee@ee.kaist.ac.kr

Abstract— The band adaptive modulation and coding (band candidate technologies that can substitute for the wireless local
AMC) scheme is used in orthogonal frequency division multiple area network (WLAN) for indoor environment.
access (OFDMA) systems to increase mobile communication
system throughput, however, an amount of feedback overhead In OFDMA system, the overall system throughput can be
degrades uplink system performance. In this paper, we propose increased by the band adaptive modulation and coding (band
an adaptive feedback method that can significantly reduce the AMC) [1]. However, the channel quality information (CQI)
amount of channel feedback information when a femto base feedback from each mobile station (MS) is required to employ
station (BS) and mobile stations (MS) communicate in the band band AMC. Due to the feedback overhead, the uplink system
AMC mode. Since the channel variation of the MS in the femto performance is degraded. Especially, since the number of BS
cell is relatively small, instead of informing the femto BS of the antennas and MS antennas increase in MIMO-OFDMA
channel quality information (CQI) periodically, MSs only send system, CQI feedback is required for each stream when MIMO
CQI feedback to the femto BS when channel condition changes is used as spatial multiplexing; thus, feedback overhead
above a threshold. Simulation results show that the proposed increases.
scheme can reduce feedback overhead up to 70 % without
resulting in system throughput degradation compared to the To reduce feedback overhead, a lot of research has been
conventional best-M feedback scheme. conducted. Ref. [2] proposed a scheme, which varies feedback
period based on MS’s velocity. According to [3], the MS’s
Keywords-band AMC; channel change indication; feedback; velocity can be measured, but it is difficult to adopt velocity
femto cell; indoor; OFMDA; pico cell; small cell measuring algorithm in real wireless communication systems
due to complexity and accuracy problem. In [4], MSs feedback
I. INTRODUCTION average value of CQI to the BS, and the BS allocates resources
for sub-band CQI to some MSs which have high average value.
Nowadays, users of wireless communication systems want When many MSs operate in the band AMC mode, the method
high data rate services to download and upload high-capacity presented in [4] is well-performed. However, it did not show
data. Also, they want to get service whenever or wherever they tremendous enhancement, when few MSs operate in the band
are. To support these users’ needs, wireless transmission AMC mode like femto cell.
technologies such as orthogonal frequency division multiple
access (OFDMA) and Multiple Input Multiple Output (MIMO) In [5]-[7], opportunistic feedback has been researched to
have been researched widely. reduce feedback overhead. Only MS, which has higher CQI
than threshold value, feedbacks CQI to the BS. Since
Recently, a femto cell has also been researched to extend opportunistic feedback is a contention-based feedback, it is
indoor cell coverage and meet users’ needs. The femto cell is a possible that two or more MSs feedback CQI through the same
small base station (BS) which will be installed at home or in an resource. Since channel of each MS is not predictable, it is
office to increase cell coverage for indoor environment. The difficult to determine how many CQI feedback channels are
femto cell can provide high-capacity data services to users in required.
indoor environment with low price by using the licensed band.
The cell coverage of the femto cell is around 10m. The femto In [8], analog feedback, which just uses one OFDMA
cell was approved as a work item of 3rd Generation Partnership symbol for CQI, was researched. Analog feedback can reduce
Project 2 (3GPP2). Also, 3GPP names the femto cell as Home feedback overhead, but the dynamic range is larger than that of
Node B or Home eNodeB, and it is standardizing the femto quantized feedback such as best-M feedback and opportunistic
cell. There is mention about the femto cell in IEEE 802.16m feedback, so it requires a nice RF transmitter.
system requirement document, and it has been discussed for
Especially, since ongoing standardization activities for
standardization. Also, in ongoing standardization activities
IMT-Advanced system are considering contiguous or non-
such as 3GPP long term evolution advanced (LTE-A) and
contiguous bandwidth with 20 MHz or more or less, a simple
3GPP2 ultra mobile broadband evolution (UMB evolution) for
and cost-efficient channel feedback scheme for indoor
IMT-Advanced system, the femto cell is one of the major
environment is required. In this paper, we propose a simple
This work was supported in part by the IT R&D program of MKE/IITA adaptive feedback method for band AMC in the femto cell
under Contract number 2005-S-609-04 in Korea and by Samsung Electronics.

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

 Figure 2. A system model of downlink OFDMA systems

Figure 1.The basic concept of a femtocell

in the femto cell move slowly. From this fact, a simple
adaptive feedback method in the femto cell is proposed in this
without above mentioned problems. Channel variation of MSs paper.
in the femto cell is small since users move slowly and some
users do not move in offices or home. By exploiting this III. SYSTEM MODEL
channel characteristic of the femto cell, MSs does not feedback A system model of downlink OFDMA system with I MSs
CQI periodically. Instead, MSs periodically send channel is depicted in Fig 2. The entire channel is divided into L sub-
change indication. When channel variation of the MS is bigger bands. It is assumed that channel is not changed during one
than a threshold value, the BS requests the MS to send CQI frame duration.
feedback. Proposed scheme can reduce feedback overhead up
to 70% in indoor environment. Each MS measures carrier-to-interference-and-noise ratio
(CINR) at every frame, and they feedback best-M sub-bands
This paper is organized as follows: Section II presents the among L sub-bands to the BS when resources are allocated to
basic introduction of femto cell. Section III introduces the MSs for feedback.
system model. An adaptive feedback scheme with channel
change indication is proposed in section IV. The simulation
result is given in section V, followed by the conclusions in Gi (t ) = {g i ,l1 (t ), gi ,l2 (t )," , g i ,lM (t )} (1)
section VI.
where gi ,ls (t ) is CINR of lsth sub-band of MS i at time t, and
II. FEMTO CELL each MS feedbacks Gi(t) to the BS.
The basic concept of a femto cell is depicted in Fig. 1[9].
To provide fairness to MSs, proportional fair (PF)
The femto cell, which is usually connected to the IP backbone,
scheduler is employed to allocate sub-bands to users.
provides wireless communication service to users. As depicted
According to [12], the PF scheduler allocates lth sub-band to a
in Fig. 1, the femto cell can be deployed in home environment
to communicate with a small number of users. However, femto MS which satisfy below criterion at time t.
cell can also be deployed in dense areas such as office and
public places to give wireless communication services to a ri ,l (t )
large number of subscribers. P (l , t ) = arg max i (1 + )
(T − 1) Ri (t − 1)
The femto cell is more competitive than other wireless
communication systems such as Wi-Fi and Bluetooth because ()
where ri,l(t) is the instantaneous transmittable data rate of lth
the small number of femto cell users will use an entire licensed sub-band of user i. Ri(t-1) is the average rate of MS i at the
band in indoor environment [10]. However, to commercialize previous frame, and T is the average window size.
and standardize the femto cell, some problems should be
solved. One main problem is the interference between a macro
cell and the femto cell; also, the interference between femto ∑ ri,l
(T − 1) Ri (t − 1) l∈Si
cells could be problem. In addition, the procedure of handover Ri (t ) = + (3)
from the macro cell to the femto cell should be defined. T T
Moreover, optimized wireless communication technologies where Si is the set of sub-bands allocated to MS i.
for indoor environments can be applied to the femto cell due to Since the PF scheduler uses transmittable data rate and each
the fact that the femto cell has different characteristics MS feedbacks CINR, the BS should calculate the transmittable
compared to the macro cell. Users in the femto cell usually data rate based on feedback CINR information from MSs.
move more slowly than users in the macro cell do. Capacity is used as the transmittable data rate, and it is
Furthermore, some users in the femtocell do not move, so the expressed as follows.
channel variation due to the Doppler shift is negligible [11].
Because of these users’ characteristics, optimized ri ,ls (t ) = log 2 (1 + gi ,ls (t )) (4)
communication method for indoor environments is able to be
used in the femto cell. For example, the femto cell system Also, it is assumed that no error occurs in uplink feedback
performance can be similar to the macro cell system channel.
performance with small amount of feedback because the users
 IV. PROPOSED SCHEME
Users in the femto cell relatively move slowly and they
sometimes do not move; therefore, the channel variation
between the BS and MSs is relatively small. The channel
variation is comparatively small in the femto cell, so the
current channel condition is not largely different from the
previous CQI. Thus, the band AMC scheme can be widely
implemented in the femto cell to increase system throughput
since previous CQI is still available next frame.
For the same reason, if every the MS, which operates in the
band AMC mode, periodically feedbacks CQI, an amount of
feedback overhead will become tremendous. A method that
reduces the feedback overhead by using channel change
indication in the femto cell is proposed in this paper. The
message sequence chart between BS and MS, and the channel
change indication algorithm of MS are described as follows.

A. Message Sequence Chart between BS and MS

It is assumed that the BS cannot measure the velocity of
each MS. In the proposed scheme, MSs, which operate in band
AMC mode, do not periodically send CQI. Instead, MSs just
send indication information whether the channel is changed or
not from the previous value. The message sequence chart is
depicted in Fig. 3.
MSs calculate the CINR of each sub-band periodically.
Figure 3. Message sequence chart between BS and MS
And then, it checks if the channel condition is changed or not,
and it sends feedback indication flag to the BS. Based on
feedback indication flag from MSs, the BS allocates uplink
resources for the best-M feedback to the MSs which sent
indication flag with “1”. According to the current feedback
CQI from the MSs which sent indication flag with “1” and the
previous CQI value from the other MSs, the BS sends
downlink resource allocation information to MSs.

B. Channel Change Indication algorithm of MS

The algorithm for MS to feedback CQI is shown in Fig. 4.
Step 1: Each MS periodically measures and find best-M
sub-bands like (1).
Step 2: Each MS checks whether set of best-M sub-bands is
changed or not.

L(t p ) = {l1 , l2 ," , lM }, L(t ) = {l1' , l '2 ," , l ' M }

(5)
Check L(t p ) = L(t ) or L(t p ) ≠ L(t )

where L(tp) is the set of indices of best-M sub-bands when a

MS lastly feedbacked CQI, and L(t) is the set of indices of
best-M sub-bands at current frame.
Step 3: If the set of indices of best-M sub-bands is changed,
the MS concludes that channel condition is changed, so it
notifies the BS that channel is changed. Otherwise, the MS
checks that the variation of CINR of best-M sub-bands. The
variation of CINR is as follows.

Figure 4. Channel change indication algorithm of MS |CINRl,last - CINRl,current| (6)

 Proposed
TABLE I. SIMULATION ENVIRONMENTS 350000 Conventional best-M
Content Description

Feedback Overhead(bits/s)
300000

Wireless channel Indoor hot spot

250000
FFT size 1024
Number of data sub-carriers 768 200000

Number of sub-bands 12 150000

Number of OFDM symbols
47
per one frame 100000

Center frequency 2.5GHz 50000

Bandwidth 10MHz
0
One frame duration 5ms 0 10 20 30 40 50

Number of MSs
Figure 6. Feedback overhead comparison between conventional best-M
and the proposed scheme

Proposed
Figure 5. CQI feedback message 3.0
Conventional best-M

where CINRl,last is CINR of lth sub-band of best-M sub-

Throughput(bit/s/Hz)
2.5
bands when the MS lastly sent best-M feedback, and
CINRl,current is CINR of lth sub-band of best-M sub-bands of
current frame. If at least CINR variation of one sub-band is 2.0

above threshold value, K, the MS concludes that channel

condition is changed, so it informs the BS that channel is
changed. Based on the value of K, the frequency of best-M 1.5

CQI feedback varies. When K is small, many best-M CQI

feedbacks are required. However, with large K, relatively
small number of best-M CQI feedback is required, but 1.0
0 10 20 30 40 50
system performance degrades due to incorrect channel Number of MSs
information. For this reason, setting the optimal K value is
important. Figure 7. Downlink channel throughput comparison between conventional
best-M feedback and the proposed method
Step 4: The BS allocates uplink resources to MSs, which
inform the BS that channel is changed, to get best-M CQI V. SIMULATION RESULTS
feedback.
Step 5: The MSs, which get the resources to feedback best- A. Simluation Environments
M CQI, inform the BS of best-M sub-bands CQI. Simulation environments are given in Table 1. Channel
model is indoor hot spot model of IEEE 802.16m evaluation
Similarly, [13] uses the band AMC differential CINR
methodology document (EMD) [14]. Indoor hot spot model is
feedback, which is that the MS sends the differential CINR of
used to simulate for the case that a large number of MSs are
M selected sub-bands through its fast-feedback channel. Since
operating in a femtocell. Thus, we perform the simulation
the band AMC differential CINR feedback sends feedback
according to the number of MSs from 5 up to 50. The
information through fast feedback channel, it uses 8 sub-
distribution of velocity of MSs also follows IEEE 802.16m
carriers per each sub-band. Thus, because ongoing
EMD. According to the EMD, 20 percent of MSs move at
standardization activities for IMT-Advanced system consider
3km/h, and 80 percent of MSs do not move.
wider bandwidth than current system, CQI of more sub-bands
may be required for the band AMC, which increases more fast- In conventional best-M feedback, since the BS does not
feedback channels. However, if fast-feedback channel is used know velocity of each MS, it should require periodic CQI to
in the proposed scheme, the proposed scheme only uses one every MS. The period of best-M feedback should be set based
fast-feedback channel for channel change indication, so on the worst case – the possible biggest channel variation. The
feedback overhead does not increase in indoor environments. worst channel variation case of the femto cell is the channel
variation of MSs moving at 3km/h. However, when MSs move
In this paper, we assume that MSs use channel change
3km/h or slower, channel is not changed significantly every
indication flag, which consists of 1 bit, to inform about whether
frame. For this reason, it is not necessary to feedback every
channel is changed or not. If channel condition is changed, the
frame. Therefore, in this simulation, MSs feedback every two
channel change indication flag is set as “1”. Otherwise, it is set
frames to reduce feedback overhead. For fair comparison with
as “0”.
conventional scheme, MSs also informs the BS whether
channel is changed or not at every two frames. The message VI. CONCLUSION
sequence chart and the channel change indication algorithm of In this paper we propose a simple adaptive feedback
MS follow Fig. 3 and Fig. 4, respectively. The threshold value, method with channel change indication flag for the band AMC
K, is set as 1dB. It is assumed that 32 bits message is used for when the BS and MSs operate in the femto cell. Since users in
best-4 CQI feedback. 32 bit message format is depicted in Fig. indoor move slowly or they do not move, the proposed method
5. As shown in Fig. 5, CQI feedback message consists of reduces the feedback overhead up to 70% without resulting in
BITMAP and CQI value. The number of bits of BITMAP and channel throughput degradation compared to conventional
CQI value follows [13]. best-M feedback. The proposed scheme does not require
To measure the channel throughput, the feedbacked CINR velocity measuring algorithm. Also, in the proposed scheme,
value is used. In addition, PF scheduler is used to allocate since each MS uses dedicated feedback channel, the same
resources. feedback resource is not used by two or more MSs.
However, in the proposed scheme, the BS requires
B. Simluation Results feedback about whether the channel is changed or not to the
In this sub-section, we will show the uplink feedback MS which has the same channel condition. Since this feedback
overhead and downlink throughput of the proposed scheme is unnecessary, the research to reduce this feedback is required.
through simulation results. Also, it is required to develop the proposed scheme to be
proper for MIMO system.
In Fig. 6, the feedback overhead per second is depicted
according to the number of MSs in one femto cell. The amount
of feedback overhead increases as the number of MSs REFERENCES
operating in the band AMC mode increases. However, the [1] S. Kim and C. Kang, “Throughput analysis of band AMC scheme in
proposed scheme reduces feedback overhead even though the broadband wireless OFMDA system”, IEEE Wireless communications
number of MSs is small, since the BS allocates uplink and Networking Conf, Vol.3, pp. 1305-1310, 2006.
resources for CQI feedback to the MS which informs the BS [2] H. and H. Kim, “An adaptive determination of channel information
feedback period in OFDMA systems,” IEEE ISCE pp. 1-5, 2006.
that channel is changed. The proposed scheme can reduce
[3] C. Tepedelenliglu and G. B. Giannakis, “On velocity estimation and
feedback overhead up to 70% compared to best-M feedback correlation properties of narrow-band mobile communication channels,”
scheme. Even though we simulate the feedback overhead for IEEE Trans. Veh. Technol., Vol. 50, pp.1039-1051, July. 2001.
the case that the number of MSs is up to 50 because we also [4] T. Kand and H. Kim, “Opportunistic feedback assisted scheduling and
consider indoor hot spot area, the proposed scheme can reduce resource allocation in OFDMA systems,” IEEE ISCC pp. 1-5, Oct. 2006.
the feedback overhead regardless of the number of MSs in a [5] R. Agarwal, V. Abhishek, R. Vannithamby, and J. Cioffi, “Opportunistic
femto cell. feedback in OFDMA downlink,” IEEE GLLOBECOM, pp.3320-3324,
Nov. 2007.
In uplink of the proposed scheme, fewer resources are [6] T. Tanq, and R. Heath, “Opportunistic feedback for downlink multiuser
required to send feedback information to the BS, so more diversity,” IEEE Comm. Letters, vol.9, pp 948-950, Oct. 2005.
resources can be used for uplink data transmission. Thus, the [7] X. Wu, J. Zhou, G. Li, and M. Wu, “Low overhead CQI feedback in
uplink throughput degradation caused by feedback information multi-carrier systems,” IEEE GLOBECOM, pp. 371-375, Nov. 2007
is reduced due to less feedback overhead. [8] D.L Ruyet and B. Ozbek, “Partial and analog feedback for MISO
precoding systems,” IEEE ICC, pp2767-2772, Jun. 2007.
Fig. 7 shows the downlink throughput based on feedbacked [9] 3GPP2 S.P0126, Femto Cell Requirements.
CQI according to the number of MSs in one femto cell. When [10] ABI Research, Femtocell Access Points: Fixed-Mobile Convergence for
the number of MSs in one femto cell increases, the downlink Residential, SMB, and Enterprise Markets, 2006.
throughput increases in both the proposed method and the best- [11] H.Hashemi, “The indoor radio propagation channel,” Proceedings of the
M feedback method because of multiuser diversity. Also, even IEEE, vol. 81, pp 943-968, Jul. 1993.
though MSs send less feedback information to the BS, the [12] H. Kim and Y. Han, “A Proportional Fair Scheduling for Multicarrier
downlink throughput is not degraded. Thus, we believe that the Transmission Systems,” IEEE Commun. Letters vol. 9, pp. 210-212,
proposed scheme is the method that reduces the feedback Mar. 2005.
overhead without degrading the downlink performance. [13] IEEE P802.16Rev2/D4–2008, “IEEE standard for Local and
Metropolitan Area Networks – Part 16: Air Interface for Broadband
Wireless Access Systems,” April 2008.
[14] IEEE 802.16m-08/004r1, “Project 802.16m Evaluation Methodology
Document (EMD)”, Mar. 2008.