QUALITY EVALUATION OF VOIP SERVICE OVER IEEE 802.

11 WIRELESS LAN


Andrea Barbaresi, Massimo Colonna, Andrea Mantovani and Giovanna Zarba


Telecom Italia, via G. Reiss Romoli 274, I-10148 Torino (TO), Italy




Abstract: Wireless broadband access, in particular that based on IEEE 802.11 WLAN
technology, is reaching significant penetration rates in most countries and can be
considered a valid and cheap alternative to xDSL or optical fiber technology, particularly
in low density areas and small villages, to offer a bundle of services including VoIP. In
this paper we evaluate the performances of the VoIP service in a WLAN network by
means of an event-driven simulator developed by Telecom Italia. Copyright ECRR
2007

Keywords: VoIP, WLAN, IEEE 802.11, MOS, quality evaluation.





1. INTRODUCTION

The growing interest in Voice over IP (VoIP) service
is essentially due to the fact that nowadays Internet-
Protocol (IP)-based networks are widely deployed
and the broadband access is offered at very low cost.
Traffic voice, generally offered through the circuit-
switched telephone system (PSTN networks), is
moving over the packet switched Internet with the
support of new protocols, SIP and H.323 defined by
ITU and IETF, respectively.

Broadband access, in particular that based on IEEE
802.11 Wireless LAN (WLAN) technology, is
reaching significant penetration rates in most
developed countries due to heightened market
competition: at the beginning WLANs have been
essentially used as cable replacement for residential
or office environments; in recent years, public
hotspots have been deployed in airports, hotels,
trading centres essentially because people, by means
of very low cost wireless cards available for PCs and
PDAs, begin to expect and demand anytime and
anyway Internet access. Moreover, WLAN networks
have been installed by some municipalities in order
to offer broadband services, including VoIP (named
VoWLAN), at low cost to all citizens; this kind of
networks is a cheaper alternative to xDSL or optical
fiber technology particularly in low density areas and
small villages.

In the last years, many works detailing the
throughput performance of 802.11b WLANs have
been presented. Most of them, focus either on
analytical modeling or on simulations for data
transfer (Bianchi, 2000; Tay and Chua, 2001).
Experimental analysis of the VoIP service employing
a G.711 codec are shown in (Anjoum, et al., 2003)
and in (Garg and Kappes, 2003), while simulation
analysis are shown, for example, in (Hole and
Tobagi, 2004) and in (Coupechoux, et al. 2004).

This paper analyses the performances of the VoIP
service in a WLAN network with low bit rate codecs
(mainly the 1.2. kbit/s AMR codec); this analysis is
carried out by means of an event-driven simulator
developed by Telecom Italia that simulates all the
elements involved in a VoIP system including the
signalling protocols, the conversation dynamics and
the ITU-T quality evaluation metrics (MOS and E-
model). The completeness of our simulator makes our
work an enhancement of previous simulative works.

The paper is organized as follows: Section 2 presents
some general information about the protocol stack in
the VoIP system with an emphasis on the WLAN
access technique. Section 3 shows the bi-directional
traffic model for VoIP sources used in the
simulations, while Section 4 presents the QoS
evaluation. In Section 5 a general description of the
simulated scenario is provided, as well as information
about simulation hypothesis. Section 6 shows the
main simulation results obtained and, finally, Section
7 summarizes the main results coming from this
study.


2. VOIP OVER WLAN

2.1 VoIP protocol stack

Taken into account that packet-switched technology
can deliver services more cost efficiently than todays
circuit switched technology, an efficient voice
encoding and decoding mechanism is vital. The
purpose of a voice coder (vocoder) - also referred to

as coder/decoder, or simply codec - is to use the
analog signal coming from human speech and
transform and compress it into digital data. Various
voice compression schemes have been developed: we
will focus on two most ITU-T low bit rate codecs,
namely G.723.1 and G.729, and on the 3GPP codec
for the UMTS system, namely AMR. G.723.1 is a
codec that has two bit rates associated with it, 5.3 and
6.3 kbit/s, whose mode of operation can change
dynamically at each frame. It encodes speech in
frames of 30 ms using the MP-MLQ technique for
the high rate codec and the ACELP technique for the
low rate one. The G.729 codec uses the CS-ACELP
coding technique and operates at 8 kbit/s with an
input frame of 10 ms. The AMR codec has been
chosen by 3GPP for the compression of voice signals
in 3G mobile communications and it is an evolution
of the GSM-EFR codec. It consists in a single
integrated speech codec with eight source rates, i.e:
12.2 (GSM-EFR), 10.2, 7.95, 7.40 (IS-641), 6.70
(PDC-EFR), 5.90, 5.15 and 4.75 kbit/s and it works
with speech frames of 20 ms, using MR-ACELP as a
coding scheme.

SIP is the main signalling protocol today for the
VoIP service. It has been developed by the IETF to
provide a simple, scalable and easy-to-implement
protocol from an IP perspective. SIP defines packet
exchange procedures for setting up, modifying and
tearing down multimedia sessions. Although SIP
works with most transport protocols, its optimal
transport protocol is RTP. This protocol, defined in
RFC 1889, operates on the layer above UDP/IP and
provides delivery monitoring of its payload types
through sequencing and timestamping. RTP is
optionally augmented by a control protocol, RTCP,
also defined in RFC 1889, that enables exchanges of
control information between session participants with
the goal of providing quality-related feedback. IP
packets with voice frames or signalling messages are
then sent to the lower layers of the protocol stack
(LLC, MAC and PHY) for the transmission on the
wireless medium. In our study these layers are
carried out according to the IEEE 802.11 standard.


2.2 IEEE 802.11

In the IEEE 802.11 MAC Layer, the fundamental
mechanism to access the medium is called
Distributed Coordination Function (DCF) and it is
based on the Carrier Sense Multiple Access with
Collision Avoidance (CSMA/CA) protocol. DCF
describes two techniques that can be employed for
packet transmission: the default scheme is a two-way
handshaking technique called Basic Access
mechanism.

According to the Basic Access mechanism (Fig. 1) a
station with a packet to transmit, monitors the
channel activity until an idle period equal to a DIFS
(Distributed InterFrame Space) has been observed. In
case the medium is sensed busy, a random backoff
interval is selected. The backoff time counter is


Fig. 1. Basic Access mechanism.

decremented as long as the channel is sensed idle,
stopped when a transmission is detected on the
channel, and reactivated when the channel is sensed
idle again for more than a DIFS period. The station
transmits when the backoff time reaches 0. If two or
more stations start transmission simultaneously, a
collision occurs. Unlike wired networks (e.g. with
CSMA/CD), in a wireless environment collision
detection is not possible. Hence, a positive
acknowledgement (ACK) is used to notify to the
station that the transmitted frame has been
successfully received. The ACK transmission is
initiated at a time interval equal to the SIFS (Short
InterFrame Space) after the end of the reception of
the previous frame. If this ACK is not received, the
station retransmits the packet. The DCF adopts a
binary exponential backoff technique. The backoff
time is uniformly chosen in the interval [0; CW]
defined as the Backoff Window (Contention
Window). At the first transmission attempt, the
CWmin value is considered, and it is doubled at each
retransmission up to CWmax.

Concerning the Physical Layer, the most widely used
standard, IEEE 802.11b, specifies a particular form of
spread spectrum technology, named CCK
(Complementary Code Keying), which allows a
maximum data rate of 11 Mbit/s. Other data rates,
used when the channel deteriorates, are 5.5, 2 and 1
Mbit/s. An alternative solution is the IEEE 802.11a
standard that allows optionally a maximum data rate
of 54 Mbit/s while data rates of 6, 12, and 24 Mbit/s
are mandatory. The 802.11a standard uses OFDM.


3. VOIP TRAFFIC MODEL

In order to carry out a realistic analysis of VoIP, the
conversation between two users has been modeled
using the conversational speech model specified in
the ITU Rec. P.59. According to this model, the
conversation between two users A and B can be
modeled as a four state Markov chain (Fig. 2):
State A: A talking B silent;
State B: A silent B talking;
State D: Double talking;
State M: Mutual silence.

This model is characterized by the transition
probabilities between the states p1, p2, p3,
considering negligible the transitions between the
states A-B and D-M, due to their rare occurrence.
Furthermore, the sojourn time in A, B, D, and M
states is modeled by a random variable with



Fig. 2. State transition model for a conversation.

exponential distribution, respectively defined by
parameters
A
,
B
,
D
and
M
. The typical values for
all these parameters are listed in ITU-T Rec. P.59.


4. QUALITY EVALUATION: THE E-MODEL

The telephone industry employs a subjective rating
system known as the MOS (Mean Opinion Score) to
measure the quality of telephone connections. MOS
is defined in ITU-T Rec. P.800 and is based on the
opinions of many volunteers who listen to a sample
of voice traffic and rate the quality of the
transmission. They rate the voice samples from 1 to 5
with 5 being excellent and 1 being bad: the voice
samples are then awarded a MOS.

An alternative method for the voice quality
evaluation is represented by the E-model defined in
ITU-T Rec. G.107 as well as other associated ITU-T
Recommendations: it is an analytic model of voice
quality evaluation used for network planning
purposes that reflects the effects of different types of
impairments on the end-to-end speech transmission
performance. E-model bases its own behaviour on the
following consideration: psychological factors on
the psychological scale are additive, which means
that each impairment factor which affects a voice call
can be computed separately, even if this does not
imply that such factors are uncorrelated, but only that
their contribution to the estimated impairments are
separable.

A basic result of the E-Model is the calculation of the
R-factor, which is a simple measure of voice quality
ranging from a best case of 100 to a worst case of 0.
User satisfaction and the corresponding R and MOS
ranges are shown in Fig. 3. The R-factor uniquely
determines the MOS through the following relation:

MOS=1, R<0
MOS=1+0.035R+710
-6
R(R-60) (100-R), 0<R<100
MOS=4.5, R>100 (1)

The operational range for PSTN voice quality
corresponds to MOS >= 3.6. The desirable range of
operation for toll quality is MOS >= 4.

The R-factor is composed by several additive terms
each one representing a specific source of voice


Fig. 3. Voice quality levels.

quality degradation:

R= R
0
- I
s
- I
d
-(I
e
-eff) + A (2)

In this formula,
R
0
represents in principle the basic signal-to-
noise ratio, including noise sources such as
circuit noise and room noise;
I
s
, the simultaneous impairment factor, is the
sum of all impairments which may occur more or
less simultaneously with the voice signal;
I
d
, the delay impairment factor, represents all
impairments due to delay of voice signals;
I
e
-eff, represents all the impairments caused by
low bit rate codecs and by packet-losses;
the advantage factor A allows for compensation
of impairment factors when there are other
advantages of access to the user.

R
0
and I
s
are not a function of the underlying packet
network and consequently they can be assigned their
default values listed in ITU-T Rec. G.107; I
d
can be
evaluated with the following expression (Atzori, et
al., 1994):

I
d
=0.024d+0.11(d-177.3)u(d-177.3) (3)

where d is the end-to-end delay (measured in ms) and
u() is the Heaviside step function; I
e
-eff is finally
calculated using the expression (G.107, 2005):

I
e
-eff = I
e
+(95- I
e
)p/(p+Bpl) (4)

where p is the packet loss rate (expressed in % of
received frames) and I
e
and Bpl are codec-dependent
parameters whose provisional values for the codecs
previously described are listed in ITU-T Rec. G.113.


5. SIMULATION METHODOLOGY

Several dynamic simulations have been carried out in
order to evaluate the QoS for the VoIP service. The
network architecture used in all the simulations is
depicted in Fig. 4. It is an infrastructure network
formed by a single Access Point (AP) and a



Fig. 4. Simulated architecture

multiplicity of Stations (STA) connected to it. Each
VoIP call is established between a STA and the host
inside the IP fixed network. The IP network does not
introduce delays and losses on the VoIP packets.
Hence, the performance evaluation is limited to the
WLAN component of the network. Moreover, since
we are interested on the specific behaviour of the
MAC layer, we also assume that the wireless channel
is ideal: path loss and fading phenomena are both
absent and the only source of errors on received
packets is the collisions caused by the contemporary
transmission of two or more stations (including the
AP).

In order to perform the simulations, some specific
modules for VoIP service have been used:
VoIP traffic generator: it produces the VoIP
traffic for the peer entities involved in the voice
call;
VoIP client: it uses RTP protocol to send VoIP
frames over the network and it includes a buffer
(dejittering buffer) to compensate jitter effects
when receiving VoIP frames from the peer
entity;
VoIP sessions control module: it uses the SIP
protocol to establish and release VoIP calls.

The simulation work has been focused on the
evaluation of the QoS experienced by VoIP users
with respect to various parameters; the final objective
is to find out the maximum capacity that can be
achieved by the MAC layer, in terms of the
maximum number of active STAs (i.e. STAs that
have a call in progress at the same time), with an
acceptable quality (MOS 3.6).

Simulations have been conducted with the following
additional hypothesis:
Mean VoIP call duration = 120 s;
Mean VoIP traffic per user = 250 mErlang;
802.11 Multple Access technique: Basic Access.


6. SIMULATION RESULTS

This section shows the main results of the
simulations, with a focus on the following
application-layer parameters:
MOS: average value of the Mean Opinion Score;
Transmission delay: end-to-end delay of voice
frames between the STA and the host;
Packet loss: percentage of packets that are
discarded at the receiving side if they arrive to
early and the dejittering buffer is full or to late
after their nominal playout time;
Number of playout blocks: number of times per
session in which the dejittering buffer becomes
empty.

The previous parameters are obtained averaging the
different values calculated in each call.


6.1 Performance with an AMR codec

This first simulation aims at characterizing the
performances of the WLAN network for the four
802.11b transmission rates. This simulation has been
carried out considering the AMR codec with the fixed
rate of 12.2 kbit/s, a frame length of 20 ms and a
dejittering buffer of 60 ms at the receiver (that means
that the buffer can contain up to three AMR frames).

Fig. 5 shows the downlink average MOS values for
the four transmission rates of the standard and the
uplink values for the 11 Mbit/s rate, while Table 1
shows the average transmission delay and the average
discarded packets for the 11 Mbit/s transmission rate
in both downlink and uplink.

To understand the behaviour of the network, let us
focus on the MOS values for the 11 Mbit/s
transmission rate (Fig. 5). When the number of active
STAs is low, less than 6, all the offered voice traffic
is immediately transmitted by the equipment without
waiting in the MAC layer buffers; this is confirmed
by the value of the transmission delay that is about 61
ms (Table 1) and it is constituted by the sum of the
dejittering buffer delay (60 ms), the MAC protocol
delay (about 0.84 ms) and the transmission delay on
the air interface (about 0.06 ms). The collisions and
the subsequent retransmissions, that are however
always present, give a minimal contribute. At the
same time, the percentage of discarded packets is
null. In this case MOS is mainly determined by I
e

(that is equal to 5 for the AMR codec) and for this
reason it remains constant to the maximum value of
4.47. When the offered traffic increases, i.e. when the
number of active STAs reaches the 7 units, the
network approaches its throughput limit and
contemporarily the number of collisions becomes
more relevant: packets have to wait more in the MAC
buffers causing an increase of the transmission delay
above all and minimally of the discarded packets; the
final result is a MOS decrease. When the number of
active STAs equals or exceeds the 8 units, the offered
traffic is higher then the maximum throughput and
the voice packets have to wait much longer in the
MAC buffer before their transmission: this is
evidenced by the rapid increase of the delay curve. At
the same time, also the discarded packets increase but
they are still quite low (less than 1%) and they have a

Table 1 Average transmission delay and discarded
packets for the 11 Mbit/s transmission rate

Downlink Uplink
Number
of active
STAs
Delay
[ms]
Discard.
packets
[%]
Delay
[ms]
Discard.
packets
[%]
1 61.0 0 61.0 0
6 61.6 0 61.3 0
7 73.5 0.03 61.5 0
8 190 0.35 61.8 0
9 360 0.75 61.9 0

3
3.2
3.4
3.6
3.8
4
4.2
4.4
4.6
4.8
5
1 2 3 4 5 6 7 8 9 10
Number of active STAs
M
O
S
1 Mbit/s 2 Mbit/s 5.5 Mbit/s 11 Mbit/s 11 Mbit/s; uplink


Fig. 5. Downlink average MOS for the IEEE 802.11b
transmission rates and uplink MOS for the 11
Mbit/s rate.

limited effect on the MOS degradation. The
maximum number of active STAs in order to have an
acceptable voice quality (MOS 3.6) is 9.

The uplink MOS, on the contrary, remains constant
to the value of 4.47 due to the low delay (with 9
active STAs it is equal to 62 ms) and to the null value
of the discarded packets. This difference with the
downlink is due to the IEEE 802.11b multiple access
technique that has been designed to be fair between
all the equipments. In our scenario, instead, the AP
has to transmit the half of the voice traffic and so it
follows to be damaged.

When the transmission rate reduces, the network
performance get worse and the maximum number of
active STAs supported by the network decreases to 7
(5.5 Mbit/s), 5 (2 Mbit/s) and 3 (1Mbit/s) units. This
performance reduction is not directly proportional to
the rate reduction due to the increasing efficiency of
the MAC protocol due to the fixed duration of some
MAC overhead elements (DIFS, SIFS, PHY layer
preamble, ACK).

Another effect able to affect the overall QoS level of
VoIP service is the increase of the time needed to
setup and to release the call because of the increasing
downlink transmission delay. The mean value of the
session setup delay, for the 11 Mbit/s transmission
rate, increases from 0.3 s (with 1 active STA) to 1.3 s
(with 9 active STAs), while similarly the mean value
of the session release delay increases from 0.06 s
(with 1 active STA) to 1.0 s (with 9 active STAs).
Fig. 6 shows the call setup and call release CDFs for
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Delay [seconds]
C
D
F
1 STA 7 STA
8 STA 9 STA
Call Release
Call Setup


Fig. 6. CDF for the call setup and call release delays
with the 11 Mbit/s transmission rate.

the 11 Mbit/s transmission rate. When the number of
active STAs is high, i.e. 8 or more, the call release
procedure suffer higher delays more frequently than
the call setup procedure even if the latter involves a
greater number of SIP messages. Nevertheless, even
with these high delays, all the calls are correctly setup
and released and there are no calls that are dropped
during their progress due to the excessive delay of
voice packets.


6.2 Effect of buffer length and VAD

In the following, we analyse the impact on
performance due to dejittering buffer length
variations and VAD (Voice Activity Detection).

Fig. 7 shows the downlink MOS values versus the
maximum number of active STAs for different
lengths of the dejittering buffer (20 ms, 40 ms, 60 ms
and 120 ms). Each variation of this length causes an
identical variation of the transmission delay: the
MOS changes accordingly but the maximum number
of active STAs supported by the network remains
unchanged (at least for the buffer lengths used in our
simulations). The optimum length for this buffer can
be determined looking at the number of playout
blocks per session: they have to be minimized
because even if they do not influence the voice
quality expressed through the E-model, they impact
on the overall quality perceived by the user because
they produce silence spikes during the
conversation. Fig. 8 shows the number of playout
blocks versus the buffer length in three different
cases; a low number of playout blocks can only be
obtained with a buffer length at least equal to 40 ms.

The quality/bandwidth ratio can be enhanced using
the technique called Voice Activity Detection
(VAD). With VAD, during inactivity periods, the
coding scheme does not process speech fragments,
but it generates a Silence Descriptor (SID) that
contains a set of characteristics that describe
background ambient noise: this SID is sent to the
receiver, which decodes it and plays a comfort
noise. Moreover, SID is coded with a lower bit rate
than speech frames. The generation of SID frames is
not continuous and the algorithm taken into account

3
3.2
3.4
3.6
3.8
4
4.2
4.4
4.6
4.8
5
1 2 3 4 5 6 7 8 9 10
Number of active STAs
M
O
S
20 ms 40 ms 60 ms 120 ms
11 Mbit/s 1 Mbit/s


Fig. 7. Downlink average MOS for different
dejittering buffer lengths with the transmission
rates of 1 Mbit/s and 11 Mbit/s.

0
100
200
300
400
500
600
20 40 60 120
Buffer length [ms]
P
l
a
y
o
u
t

B
l
o
c
k
s

[
f
r
a
m
e
]
3 STA @ 1 Mbit/s
9 STA @ 11 Mbit/s
7 STA @ 11 Mbit/s


Fig. 8. Downlink average playout blocks versus the
dejittering buffer length.

is the following: when inactivity is detected, the
codec generates 7 frames similar to the speech ones,
1 SID frame, 2 NOTX frames (NOTX frames are
frames in which nothing is transmitted), 1 SID frame
and then a periodic sequence of 7 NOTX frames and
1 SID frame up to the end of the silence period. Table
2 compares the maximum number of active STAs
with and without VAD for the four 802.11b
transmission rates. Improvements are minimal
because there are however SID frames that consume
bandwidth and they are limited to the 2 Mbit/s and
the 5.5 Mbit/s transmission rates.


6.3 IEEE 802.11a PHY layer

Fig. 9 shows the downlink average MOS for several
IEEE 802.11a transmission rates in the same
conditions of the previous analysis. The maximum
number of active STAs ranges from 24 (6 Mbit/s) to
53 (54 Mbit/s). Comparing these curves with those
related to the 802.11b PHY layer (Fig. 5), we notice
that the 6 Mbit/s transmission rate has better
performance than the 11 Mbit/s one. This
improvement is due to the higher efficiency of the
802.11a PHY layer. Moreover, also with this PHY
layer, the usage of VAD has a marginal effect (Table
2).

6.4 Comparison with the G.723 and G.729 codecs

In this section, we will compare the AMR codec with
Table 2 Maximum number of active STAs with and
without VAD

Transmission rate No VAD VAD
1 Mbit/s 3 3
2 Mbit/s 5 6
5.5 Mbit/s 7 8
11 Mbit/s 9 9
6 Mbit/s (802.11a) 24 27
54 Mbit/s (802.11a) 53 55

3
3.2
3.4
3.6
3.8
4
4.2
4.4
4.6
4.8
5
15 20 25 30 35 40 45 50 55
Number of active STAs
M
O
S
6 Mbit/s 12 Mbit/s 36 Mbit/s 54 Mbit/s


Fig. 9. Downlink average MOS for several IEEE
802.11a transmission rates.

3
3.2
3.4
3.6
3.8
4
4.2
4.4
4.6
4.8
5
1 2 3 4 5 6 7 8 9 10
Number of active STAs
M
O
S
AMR (12.2 kbit/s) G.729 (8 kbit/s) G.723 (5.6 kbit/s)
11 Mbit/s 1 Mbit/s


Fig. 10. Downlink average MOS for three low bit rate
codecs at 1 Mbit/s and 11 Mbit/s.

two other ITU-T low bit rate codecs; in particular we
will analyze the G.723.1 one with the fixed rate of
5.6 kbit/s and the G.729 one with the rate of 8 kbit/s.
Also for these codecs the frame length and the
dejittering buffer length have been supposed
respectively of 20 ms and 60 ms. With a low offered
traffic, the downlink MOS, shown in Fig. 10, is
exclusively determined by the intrinsic quality of the
coded, i.e. by its I
e
value. Instead, when the offered
traffic increases, the transmission delay prevails on I
e

and this reduces the MOS differences.


7. CONCLUSIONS

In this paper we have analysed the performance of
the VoIP service when it is provided to WLAN users.
By means of an event-driven simulator that simulates
all the elements involved in a VoIP system, we have
derived for the AMR codec the maximum number of
active STAs supported by an IEEE 802.11b and
802.11a network: it ranges from 3 to 9 in an IEEE

802.11b network and from 24 to 53 in a 802.11a one.
We have also derived the minimum value for the
dejittering buffer length (it is 40 ms) and we have
evaluated the effect of the VAD option. Finally, the
AMR performances have been compared with those
of two popular ITU-T low bit rate codecs (G.723.1
and G.729).


ACKNOWLEDGEMENTS

This work has been partially performed within the
framework of the IST-AROMA project (Casadevall,
et al., 2006) (www.aroma-ist.upc.edu), which is
partially founded by the European Community.


REFERENCES

G. Bianchi, Performace analysis of the IEEE 802.11
Distributed Coordination Function, IEEE
Journal on Selected Areas in Communications,
Vol. 18, No. 3, March 2000.
Y. C. Tay, K. C. Chua, A capacity analysis for the
IEEE 802.11 MAC protocol, Wireless
Networks, Vol. 7, Issue 2, March 2001.
F. Anjum et al., Voice Performance in WLAN
Networks - An Experimental Study,
Proceedings of IEEE GLOBECOM03, Vol. 6,
December 2003.
S. Garg, M. Kappes, An experimental study of
throughput for UDP and VoIP traffic in IEEE
802.11b networks, Proceedings of IEEE
WCNC03, March 2003.
D. P. Hole, F. A. Tobagi, Capacity of an IEEE
802.11 wireless LAN supporting VoIP,
Proceedings of IEEE ICC 2004, 2004.
M Coupechoux et al., Voice over IEEE 802.11b
capacity, Proc. of the 16th ITC Specialist
Seminar on Performance Evaluation of Wireless
and Mobile Networks, September 2004.
L. Atzori, M. L. Lobina, Speech playout buffering
based on a simplified version of the ITU-T E-
model, IEEE Signal Processing Letters, Vol.
11, No. 3, March 2004.
P. F. Casadevall et al., Overview of the AROMA
Project, IST Mobile & Wireless
Communications Summit, 2006.


BIOGRAPHY

Andrea Mantovani
received the Laurea degree
in Electronic Engineering
in 2003 from Politecnico di
Torino, Italy. After a first
job experience as
electronic designer, in
2005 he started his
activities in Telecom Italia,
Turin, Italy. His main
research interests concern wireless communications
and, in particular, on the analysis of real time packet
switched services over 3G mobile networks and
WLANs.

Andrea Barbaresi received
a Laura degree cum laude
in Electronic Engineering
from the University of
Ancona, Italy, in 1999. He
has been working in the
R&D center of Telecom
Italia since 2000, in radio-
access and cellular
planning group. His
activities are in the field of mobile communication
systems, especially radio resource and QoS
management, system-level analyses, optimization of
3G and B3G systems performances. Since 2002, he
has been participating in several IST European
research projects as Telecom Italia representative.

Massimo Colonna received
the Laurea degree in
Electronic Engineering
from Politecnico di Torino,
Italy, in 1997. In the same
year he joined CSELT, the
R&D Center of Telecom
Italia, Turin, Italy, in the
wireless access networks
group. His activity was
initially related to the fixed wireless access systems
(BFWA, LMDS) and then, from 2000, to the wireless
LANs. Actually he is involved in research projects
dealing with radio resource management in
3G/WLAN integrated networks and with statistical
localization within mobile networks.

Giovanna Zarba received
the Laurea degree in
Electronic Engineering
from Politecnico di Torino,
Italy, in 1992 and she
joined CSELT, the R&D
Center of Telecom Italia in
the same year. Firstly, her
activity was related to the
design and development of
microwave components for satellite applications.
From 2004 she is active in the radio-access and
cellular planning department and she has been
involved in research projects related to 3G and
WLAN networks.