Content uploaded by Jani Puttonen
Author content
All content in this area was uploaded by Jani Puttonen on Jan 19, 2015
Content may be subject to copyright.
140
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
1
User Equipment Energy Efficiency versus LTE
Network Performance
Kari Aho∗, Tero Henttonen§, Jani Puttonen∗, Lars Dalsgaard†, Tapani Ristaniemi‡
∗Magister Solutions Ltd., Hannikaisenkatu 41, FIN-40100, Jyv¨
askyl¨
a, Finland.
E-mail: {kari.aho, jani.puttonen}@magister.fi
†Nokia, P.O.BOX 45, FIN-00045 Nokia Group, Finland
E-mail: {lars.dalsgaard}@nokia.com
‡University of Jyv¨
askyl¨
a, P.O. Box 35, FIN-40014, Jyv¨
askyl¨
a, Finland.
E-mail: {tapani.ristaniemi}@jyu.fi
§Renesas Mobile Corporation, Porkkalankatu 24, FIN-00180 Helsinki, Finland
E-mail: {tero.henttonen}@renesasmobile.com
Abstract—The purpose of this article is to analyze the trade-
off conditions between battery saving opportunities at the user
terminal and Long Term Evolution network performance. To
achieve the goal Voice over IP with discontinuous reception
and a vast amount of different settings, including on duration,
inactivity and discontinuous reception cycle timers, have been
studied. An adaptive discontinuous reception with synchronizing
the on duration time with the Voice over IP packet arrival has
been proposed to minimize the delays caused by discontinuous
reception. In addition, a channel quality indicator preamble time
has been introduced to enable channel quality indicator update
prior the on duration period. The quality of service and battery
saving opportunities have been evaluated with a dynamic system
simulator enabling detailed simulation of multiple users and cells
with realistic assumptions. It can be concluded that high battery
saving, i.e. increased talk-time opportunities, can be achieved
without compromising the performance when discontinuous re-
ception is properly adapted. Adaptive discontinuous reception
and channel quality indicator preamble can effectively mitigate
the capacity loss when more stricter DRX settings enabling higher
energy efficiency at the terminal are applied.
Keywords-DRX, VoIP, Battery savings, Energy Efficiency, Ca-
pacity, CQI, Preamble, Adaptivity
I. INTRODUCTION
High peak data rates, low round trip times and high Quality
of Service (QoS) enabled by current and upcoming wireless
cellular technologies such as Long Term Evolution (LTE)
[3][4] have driven the increase of wireless (data) subscribers
to whole new levels. Despite of the fact that the growth is
fueled by various innovative data services, the simple voice
call service and especially Circuit Switched (CS) voice calls
still remain as the main source of revenue for the cellular
operators. However, the situation with CS voice is starting
to change: Future systems, such as LTE, support only Packet
Switched (PS) services meaning that voice calls would also
have to be delivered via PS domain. Thus, this leads to
situation where voice services are offered through Voice over
IP (VoIP) protocols [5]. One of the benefits of sole PS system
from the operator perspective is lower Capital Expenditure
(CAPEX) and Operating Expense (OPEX) due to not having
network elements for both CS (voice) and PS (data).
Generally, the requirement for successful penetration of IP
based voice services, such as VoIP, is that the voice quality
should be comparable to what is available using traditional
CS voice [6]. There are numerous factors affecting VoIP QoS,
which include, e.g., delay, packet loss and packet corruption.
These performance indicators are challenges especially in the
wireless domain due to more unreliable transmission media.
Reliability in LTE networks is addressed through several
radio resource management technologies such as Hybrid ARQ
(HARQ),Link Adaptation (LA),Channel Quality Indication
(CQI),Packet Scheduling (PS) and short Transmission Time
Interval (TTI) of 1 ms. The purpose of this article is to address
LTE performance together with discontinuous reception and
VoIP. Discontinuous reception cycles aim to improve the
energy efficiency at the terminal by allowing possibilities to
turn off the receiver circuitry during certain times. That kind
of solutions are parallel as battery consumption at the terminal
can very well become the limiting factor in providing satisfac-
tory user experience along side of the network performance.
The rest of this article is organized as follows. Section II
covers the motivation and related studies, which are followed
by description of modeling and simulation assumptions related
aspects in Section III. After those, simulation scenario is
presented before simulation results and analysis. Conclusion is
presented in Section V and finally in the Appendix reliability
analysis of the used research tool and results is covered.
II. MOTIVATION AND RELATED STUDIES
Optimizing the VoIP over LTE performance in terms of QoS
and the usage of radio resources has been studied in several
articles [7][8][9]. From those it may be concluded that while
the overall VoIP capacity may be control channel limited, the
situation can be improved effectively by utilizing either packet
bundling or semi-persistent packet scheduling. However, since
the battery life of small hand-held devices might also very
well become a limiting factor in providing satisfactory user
experience, a prominent option to prolong the battery life is
to use downlink Discontinuous Reception (DRX) cycles in
conjunction with VoIP in LTE. DRX cycles, introduced by
141
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
2
Third Generation Partnership Project (3GPP) in [10] and [11],
allow an idle User Equipment (UE) (i.e. UE that is currently
not scheduled, i.e. neither transmitting nor receiving) to save
battery by turning off the radio receiver for a predefined period
(according to predefined, network-signaled parameters). This
produces battery savings at the cost of somewhat reduced
scheduling opportunities at the Evolved Node-B (eNB).
Discontinuous reception has previously been studied both
related to 3rd Generation systems as well as to LTE in several
articles. In [12], the effects of DRX cycles and related timers
to the queue lengths, packet waiting times, and the power
saving factor were studied in Rel’99 wideband code division
multiple access networks. The study showed quantitatively
how to select appropriate DRX cycle values and the related
inactivity timer for various traffic patterns. In [13] the scope
is extended to consider DRX together with delay sensitive
VoIP service over high speed downlink packet access, and
the paper indicated that there are possibilities for high power
savings but VoIP capacity can be compromised if improper
parametrization is applied for DRX.
In [14] the DRX in LTE has been compared to DRX of
3rd Generation networks and it is concluded that LTE DRX is
able to achieve more efficient battery usage through the use of
short and long DRX cycles. In [15] an analysis of DRX with
best-effort type of traffic over LTE networks was conducted.
The paper showed with a single user simulations that a 95 %
reduction of the UE power consumption with a moderate 10-
20 % loss in throughput was achievable. In [16] the analysis
is extended to a short DRX cycle and an inactivity timer.
Both the short DRX cycle and inactivity timer aim to provide
adaptibility to the variable traffic patterns. Both mechanisms
improve the performance over a pure static DRX in terms
of throughput and power consumption. However, short DRX
with inactivity timer shows a gain of 0-3 times over DRX
with just an inactivity timer. In [17] the DRX in LTE has
been analyzed with video streaming and VoIP applications.
It is concluded that DRX can save about 40-45 % of UE
battery power without significantly impacting video quality;
while for VoIP applications the saving can be approximately
60 %. However, the estimations are based on simple analytical
calculations.
A part of the results in this article have earlier been
published in conference articles [1] and [2]. In [1] we have
studied DRX in a LTE network with high number of VoIP
users. The focus was on the impact of DRX cycles and
related timers on the system capacity as well as on battery
saving opportunities. In [2] we proposed a CQI preamble for
improving the VoIP performance with DRX. In this article,
we have extended the work presented in [1] and [2] in
several ways. We introduce an adaptive DRX, where the on
duration time and VoIP packet arrival times are synchronized
for buffering delay minimization. The CQI preamble scheme is
analyzed with higher UE velocity where most of the CQI gains
have been lost proving that most of the loss caused by DRX
arise from the usage of out-dated CQI information. Finally, a
statistical confidence analysis of the used simulation tool is
presented at the end.
Our focus has been to study the combination of VoIP
Fig. 1. Skeleton simulator
and DRX, since the performance degradation due to DRX
is expected to be the highest with real-time delay sensitive
traffic. Secondly, most of the previous work has focused on
radio resource point of view of DRX and not on the battery
saving opportunities. However, the DRX parameterization is
clearly a trade-off between the capacity and battery savings.
Finally, our analysis have been performed with a fully dynamic
system simulator capturing the effects of dynamic nature of
mobility.
III. MODELING AND SIMULATION ASSUMPTIONS
The purpose of this section is to cover briefly the modeling
issues related to dynamic system simulator used in these
studies. Previously general modeling issues are presented
briefly in [18] and VoIP specific issues, e.g., in [7]. In the
following subsections the most critical aspects in terms of this
paper are discussed.
A. Overview of the simulator
The general principle of all network simulators is quite
much the same, where the simulation is configured through
parameters, simulations are run with certain modeling assump-
tions and details and statistics are gathered e.g. by means of
averages, cumulative distribution functions and time traces.
For examples of other network simulators, see [19] and [20].
The simulator used for generating results is a fully dynamic
system simulator, which means that the element of time
passing is considered in details: channel conditions change,
users move, traffic arrives at uplink and/or downlink buffer,
scheduling happens and data is sent in downlink and/or
uplink. The simulator works according to step-wise simulation
principle: at each step, actions are executed in certain order
before proceeding to the next step. The actions that are done
depend on which features are turned on in the simulator (e.g.
DRX can be turned off fully so that no UE uses it) and one
simulation lasts for a certain number of predefined steps.
1) Simulator library: The simulator utilizes a library built
for the purpose of enabling fast simulator development. The
library is coded in C++, and contains many ready-made and
tested implementations (i.e. sub-libraries) for e.g. propagation,
channel, traffic and mobility models, as well as general pur-
pose tools like scenario creation modules, parameter reader,
random number distributions and simulation statistics collec-
tion utilities. At the heart of the library, there is also a so-called
skeleton simulator: a built-in model of a simple simulator that
dynamically loads simulator modules and executes them in the
desired order. This is depicted in Fig. 1.
This kind of modularization of functionalities into stand-
alone libraries enables code reuse in the future: For example,
142
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
3
several simulators may use the exact same modules that do
the same basic functions of the simulators, which enables
easier comparison between such simulators. For example, [21],
which shows a comparison of real network VoIP capacity for
HSPA and LTE, was done with two such simulators, which
enabled an easy comparison of just the system differences
without a massive campaign of verification simulations ensur-
ing the modeling is compatible between the simulators. Also
new simulators can be created with small effort by taking the
skeleton and adding the needed modules on top of that.
2) Simulator structure: The most important part of sim-
ulator is the Signal to Interference plus Noise Ratio (SINR)
calculation engine: It is abstracted so that the SINR is always
calculated between two radio objects (typically eNB and
UE, but UE-to-UE is also possible). Since these objects also
contain the information about the antennas, relative position
and any UE-specific information (such as UE-specific random
number sequences for determining the current channel con-
ditions), the interference calculation can be abstracted quite
easily. Further, there is a class called Physical Resource Blocks
(PRB) manager that acts as an initiator for the calculation
between the radios: Newly scheduled PRBs are inserted to the
PRB manager, which then (at the end of each step) handles
the calculation of C and I for those PRBs that currently exist.
At the end of each TTI, the existing PRBs are destroyed,
keeping the interference calculation machinery generic (i.e.
easily maintainable and modifiable if need be) and safely
isolated from the actual scheduling decisions.
The rest of the simulator consists of inter-working between
modules:
•UEs and eNBs are modeled as entities with a connec-
tion object joining them together, representing the active
Radio Resource Control (RRC) connection.
•The tasks of the UE are to monitor the relevant down-
link channels (i.e. Physical Control Format Indicator
Channel (PCFICH),Physical Downlink Control Channel
(PDCCH),Physical Downlink Shared Channel (PDSCH)
and Physical Broadcast Channel (P-BCH)) and transmit
data in uplink when triggered by protocol stack and when
scheduled to transmit.
•The UE also maintains measurements of serving cell and
neighbor cells, which may trigger measurement report
according to eNB-configured RRC reporting configura-
tion. The UE may also notice a connection failure (called
Radio Link Failure (RLF)) and take appropriate actions
after that (see [10] for further details).
•The tasks of the eNB are to transmit downlink data to
UEs when necessary, handle the scheduling for uplink
and downlink and decide on handovers for UEs.
•The connection object contains information relevant to
both UE and eNB, like scheduling assignments, UL/DL
data generation and protocol stacks handling the data. The
connection also maintains the linking between eNB and
UE, enabling an easy process when handover happens:
The linked eNB is simply exchanged for another and
appropriate actions done separately within the UE and
source/target eNBs, but there is no need to create/destroy
all objects related to traffic models and data buffers since
these common parts are handled within the connection
object that remains.
•In general, many algorithms (such as DRX, CQI or
handover measurements) are further separated to their
own modules as much as possible: Keeping the simulator
as object-oriented makes it relatively easy to maintain and
extend.
B. VoIP Traffic
VoIP traffic is used in the simulations to model an IP based
voice call. The traffic model used is closely based on AMR
codec, and a figure illustrating the anatomy of a VoIP call is
shown in Fig. 2. For more detailed description, see below.
•A VoIP call consists of both downlink and uplink traffic.
The duration of each VoIP call is randomly distributed
according to truncated negative exponential distribution.
The mean, minimum and maximum value of the distri-
bution are given as parameters.
•A VoIP call can be in two states: Active or DTX.
The states have different packet generation patterns and
the duration of each state is distributed according to
parametrized negative exponential distribution. The rel-
ative time a user spends in Active state determines the
Voice Activity Factor (VAF) of the call: For example, a
50 % VAF means that on average, a user spends half of
its time in Active state and half in DTX state.
•Only downlink direction is considered in these simula-
tions, but the simulator supports simultaneous traffic in
uplink and downlink (e.g. synchronized so that DTX in
uplink occurs when downlink is in Active and vice versa).
•During Active period, fixed-size packets (i.e. AMR voice
packets) are generated at constant intervals. In this study,
VoIP packet is assumed to be 38 bytes and the interarrival
time between packets is 20 ms [22].
•During DTX period, fixed-size Silence Descriptor (SID)
packets, used for generating comfort noise, are transmit-
ted at constant intervals. In this study, the SID packet
size is assumed to be 14 bytes and the SID packets are
generated at 160 ms intervals.
•Robust Header Compression (ROHC) is not modeled
explicitly but ideal ROHC is assumed by taken it into
account in packet sizes.
•The characteristics of a VoIP call are fully parametrized,
and can be varied between the simulations. It is also
possible to have a mix of different types VoIP calls in
the same simulation.
The traffic model described above is the same as described
in [22], but could also be further enhanced by considering e.g
the effect of explicit ROHC or jitter in packet arrival (i.e. the
Fig. 2. Voice over IP traffic model
143
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
4
Fig. 3. De-coupled TD/FD-scheduler
packet arrival would not always be constant but could vary
a little. However, in this paper, we concentrate on the basic
model, to better account the basic interactions between DRX
and VoIP.
C. Scheduling and Resource Allocation
In this study, we assume dynamic resource allocation for
VoIP over LTE, i.e. we model a scheduler in the eNB that
takes care of physical resource allocations for users. Each TTI,
the eNB sends the scheduling allocations for users connected
to the eNB in Physical Downlink Control Channel (PDCCH),
and UEs read the scheduling assignments and act accordingly
(i.e. do nothing, transmit on uplink or receive in downlink).
Since the scheduler is fully up to eNB implementation, there
is no clear default behavior how the scheduling is done, but
typically the scheduler assigns resources based based on each
user’s channel conditions: each user sends a Channel Quality
Indication (CQI) report to eNB either at periodic intervals or
when requested by the eNB.
Thus, the scheduler could vary the resource allocation of
each user on a TTI basis. But this kind of fully dynamic
scheduling of VoIP packets requires high amount of PDCCH
resources, since each allocation for a UE consumes control
channel resources each time it is signaled. Moreover, the
scheduling assignments can be coded separately, i.e. they
may consume different amount of resources depending on the
selected coding scheme, so the limited availability of PDCCH
resources also results in a varying amount UEs that can be
scheduled per TTI. However, in general in this study we are
assuming static amount of users (i.e., Maximum Schedulable
Users (MSU)) that can scheduled per TTI, i.e., PDCCH is not
explicitly modeled. Exclusion of explicit PDCCH modeling is
done to better account the basic interactions between DRX
and VoIP. The MSU value was chosen by estimating the
average PDCCH capacity, though. Still, the impact of realistic
PDCCH following modeling aspects from [23] and simulation
principles from [24] are briefly addressed in this paper.
The scheduler used in these simulations has been a de-
coupled Time Domain (TD)-Frequency Domain (FD) sched-
uler, presented in [18] and depicted in Fig. 3. This means that
the scheduling is done in two stages: First, a TD-scheduler
selects the candidates for the scheduling (up to MSU users).
Next, a FD-scheduler chooses which resources (i.e. Physical
Resource Blocks (PRBs) to assign to which candidate. After
this, the scheduling allocations are sent to the users. Note, that
if PDCCH is explicitly modeled, then MSU in TD scheduler
is not limiting the amount of scheduling candidates, but the
PDCCH resource check is done in between TD and FD
schedulers.
Scheduling decisions done in TD-scheduler are based on
Head-of-Line (HoL) packet delay, i.e., the user with the oldest
packet in the buffer (i.e. the packet with the largest delay) is
selected first in order to prioritize users who have the worst-
delayed packets. In FD-scheduler, the candidates from TD-
scheduler are treated in the order of delay: The first candidate
is allocated the best PRBs based on CQI information sent by
that user, and then the same is done for the next candidate
and so on. The PRBs are assigned until it is deemed that
the user gets enough PRBs to empty its data buffer 1or
there are no more PRBs available or the maximum limit of
allocated PRBs for one user is reached. Finally, it should be
noted that both HARQ retransmissions and control message
(e.g. handover command in downlink, measurement report
in uplink) transmissions are prioritized by TD- and FD-
schedulers since these types of traffic are considered critical
for the call.
D. Performance Evaluation Criteria
While the simulation enables a wide variety of possible
statistics, here we present the simulation results through a
comparison of Quality of Service (QoS) criterion and Battery
Saving Opportunities (BSO).
1) The QoS Criterion: The QoS criterion, also called
system capacity later in this paper, is defined as a combination
of user and cell outage levels: A user is in outage if too many
of its packets are dropped or have large enough delay, and a
cell is in outage if too many of its users are in outage. More
precisely,
•A cell is in outage if more than 5 % of its users are in
outage.
•A single user is in outage if 2 % or more of the packets
(monitored over the whole call) are not received correctly
within 50 ms (one-way) time.
Note 1: In addition to monitoring delay at the receiving end,
packets can also be discarded at the transmitting side
if the delay of the packets in the buffer reach the
discard delay threshold.
Note 2: The 50 ms delay limit is called the radio network
Delay Budget, and is based on ITU e-Model re-
quirements for good quality voice [25] as well as to
slightly stricter benchmarking requirements of Next
Generation Mobile Networks (NGMN) and 3GPP
evaluations [26] for VoIP.
Since it is very difficult (and not even realistic) to achieve
exactly the same load in each cell even within one simulation,
the overall system capacity is interpolated from a large set of
1With some additional limitations: In this paper one user may be assigned
resources so that it is able to send up to a maximum of two packets during
one TTI. This is called packet bundling
144
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
5
Fig. 4. Power Consumption Reference Model
simulated user amounts per cell (which lead different outage
levels / cell). The target level for this interpolation is level
where at most 5 % of users would be in outage, which is the
absolute upper limit according to the QoS criterion described
above.
2) Battery Saving Opportunity: The BSO is presented as
the time a user spends in DRX state, i.e. the time during which
the UE can be in reduced state of activity. The battery saving
opportunities are calculated according to model presented in
3GPP contribution R2-071285 [27] and illustrated in Figure
4. Power consumption levels are calculated with and without
DRX from which the relative power savings are finally calcu-
lated. In the power calculations one TTI is assumed to be the
time used to receive one transmission. Time spent in active
state, deep sleep and light sleep are collected for each call
and the total ’Active with data Rx’, illustrated in Figure 4, is
calculated for each call in a following manner:
Rxactive = (NV oIP P ackets +NS IDP ack ets)×T xmean,(1)
where T xmean is the average number of transmissions,
NV oI P P ackets and NS IDP ack ets represent the total number
of VoIP/SID packets per call. NV oIP P ack ets and NSI DP ackets
are defined as follows:
NP ackets = (tcall ×µ)/tI A.(2)
In Equation 2 µrepresents the voice activity factor, tcall the
total length of the call and tIA interarrival time for packets
(SID or VoIP).
E. Discontinuous Reception (DRX)
The discontinuous reception (DRX) feature was introduced
to LTE as a network-configured feature (meaning it can be also
turned off) that provides improved battery saving opportunities
for the UE. DRX is specified at MAC level (see [11]), but is
Fig. 5. Diagram of DRX operation
turned on by eNB by RRC (see [28]) signaling when eNB
transmits the DRX parameters to the UE.
DRX consists of a cycle of alternating active period (called
On Duration in MAC specification), during which UE func-
tions just as without DRX, and an inactive period (also referred
as DRX period), during which UE is not mandated to receive
PDCCH (which means any scheduling assignment during that
time would be lost so in practice eNB will not schedule the
UE during that time) and can save power by turning off its
receiver hardware. A set of timers, illustrated in Figure 5 and
covered below, further control the operation of DRX cycle:
1) DRX Cycle Timer: specifies the periodic repetition of
the on duration (active) time, which is followed by a possible
period of inactivity time.
2) On Duration Timer (ODT): represents the minimum
time in Downlink (DL) subframes that the UE is required to
monitor the PDCCH at each DRX Cycle.
3) Inactivity Timer (IAT): specifies the number of con-
secutive subframes that UE shall stay active and monitor
PDCCHs. Timer is (re-)started every time when the UE is
scheduled and successfully decodes a PDCCH indicating new
data transmission.
4) HARQ Round Trip Time (RTT): specifies the minimum
amount of subframes before a DL HARQ re-transmission is
expected by the UE. UE can enter to inactivity during RTT if
not otherwise required to monitor the PDCCH.
5) Retransmission Timer (RTxT): specifies the maximum
number of consecutive subframe(s) that UE waits for incoming
retransmission after HARQ RTT. UE is not allowed to enter
inactivity while retransmission timer is running in order to
be able to receive the HARQ transmissions. However, when
HARQ transmissions are prioritized, as the case is in this
study, the retransmission timer does not have substantial
impact. RTxT is stopped when retransmission is received.
F. CQI preamble concept
According to 3GPP specifications ([10], [11], [28]) UE only
has to do measurements once during a DRX cycle, which
means that the UE will typically do measurements (CQI or
other) only during the active period. However, performing
the measurements, processing and transmitting them and eNB
receiving the measurements consumes time, so the active
time may have passed before up-to-date CQI information is
available in the e-Node B scheduler. Moreover, the lack of
up-to-date information can lead to lowered performance. This
procedure is illustrated in Fig. 6.
Fig. 6. Example of CQI measurement procedure with normal DRX.
Assuming total CQI delay 3, on duration 2 and DRX cycle 7 TTIs
145
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
6
Fig. 7. Example of CQI measurement procedure with CQI preamble.
Assuming total CQI delay 3, on duration 2 and DRX cycle 7 TTIs
To avoid possible performance loss, this paper considers a
so-called CQI preamble scheme as a potential enhancement
for the CQI measurement operation described above. With the
CQI-preamble scheme, a CQI preamble time is applied before
the actual on duration time takes place. During preamble time
the UE turns its receiver circuitry on to perform and report
the CQI measurements in addition to other possible Radio
Resource Management (RRM) measurements. Naturally, when
performing the measurements, the normal requirements, such
as CQI measurement granularity and the CQI measurement
period (i.e. minimum time since previous measurement), still
apply. With the preamble scheme UEs could be scheduled
with up-to-date CQI information for any potential DL data
transmission as Fig. 7 illustrates. The downside of this oper-
ation is increased activity time, which leads to higher power
consumption and shorter talk time from the UE/user point of
view.
The current 3GPP specifications already allow this kind of
improved operation. UE is allowed to do CQI measurements
(and fall back to inactivity) before actual on duration. Though,
according to [11] UE is not allowed to send the CQI report
on PUCCH before the active time / on duration begins. Thus,
from the UE side, CQI Tx part of the preamble scheme would
be achieved through slightly prolonged on duration to allow
the transmission. Changes could, however, be needed for the
e-Node B scheduling: eNB should consider the time when the
last CQI report has been received from a UE that has its on
duration started / ongoing. Once CQI report is received during
preamble time or within CQI requirements (e.g., period) UE
becomes schedulable, not before. Also, changes could also
possibly be needed in eNB implementation for evaluating
and indicating the length of CQI preamble time for DRX
purposes. CQI preamble time consists of the time before CQI
is measured, transmitted, received and processed. In this paper
preamble length is estimated to be as long as the total sum of
CQI delay (defined through parameters) and the time that it
takes to measure the CQI (1 TTI). Moreover, UEs are assumed
to be active throughout the preamble time to better understand
the ’worst-case’ scenario.
G. Adaptive DRX
As described above, DRX operates in predefined cycles
of active and inactive states. Moreover, different UEs can
(and should) be allocated with slightly different offset from
which their cycle timers start so that even amount of candidate
set for scheduling remains balanced in the downlink. That
is handled by eNB, which signals the DRX configurations
Fig. 8. Example of possible problems without adaptive DRX
Fig. 9. Simulation scenario
to each UE. However, if the offset setting and timers are
configured ’blindly’, i.e. without considering the expected
packet arrival times, possible problems could occur especially
for delay critical services such as VoIP. Fig. 8 illustrates
this through simple example: Blind offset setting can lead
to situation where a single VoIP packet can have additional
delay of 18 ms, which, in the context of the typical 50 ms
delay budget, means that almost 40 % of the delay budget is
already consumed before the first transmission, thus reducing
the time for possible retransmissions.
To mitigate the negative effects of offset setting this paper
shows that DRX offset could be adapted to the data flow tim-
ing. In terms of the example presented in Fig. 8 it would mean
that instead of having the DRX cycle starting 2 TTIs before the
packet generation (blind setting) the DRX would be adapted to
start during the time when packets are generated. This could
be achieved in real networks, e.g., by monitoring the data flow
some time before configuring the DRX. Even though there can
be some jitter for the packets in real networks, on average the
adapted offset would not cause as high additional delays as
the blind offset would in the worst case.
IV. SIMULATION SCENARIO AND RESULTS ANALYSIS
Simulations were run in a hexagonal macro cellular scenario
with three tiers, see Fig. 9. Scenario consists of 7 active base
stations with Inter Site Distance (ISD) of 500 m. Each base
station has 3 sectors resulting into layout containing 21 active
cells. Statistics are collected from 6 middle cells. Third, i.e.,
the outer tier is simulated as interfering tier, which adapts to
the load of the statistic cells. Users are not allowed to move
to the third tier but only within two inner tiers. The main
simulation parameters are shortly listed in Table I.
In the following subsections the performance of VoIP over
LTE downlink is analyzed with respect to system capacity and
power saving opportunity criteria covered in Section III-D.
The analysis for the trade-off between increased talk-time
opportunities and LTE performance is based on vast amount
146
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
7
TABLE I
SIMULATION PARAMETERS.
Simulation time 1 million steps
Time resolution 1 step, i.e., 0.0714 ms
e-Node B Max. Tx power 46 dBm
Transmission Time
Interval (TTI) 1 ms
Pathloss model Modified
Okumura-Hata [29]
Channel profile Typical Urban (TU)
UE velocity [3, 30] kmph
Handover Hard, 3 dB threshold
MSU [8, PDCCH]
CQI resolution 2 PRBs per CQI (fullband)
CQI delay 2 ms
CQI measurement period 5 ms
VoIP packet size 38 bytes [22]
VoIP packet 20 ms
interarrival time
SID packet size 14 bytes
SID packet 160 ms
interarrival time
VoIP call length Negative exponential
distribution,
truncated, mean 20 s
min 5 s
max 60 s
Activity / Silence Negative exponential
period length distribution,
mean 2.0 s
Layer 3 packet 50 ms
discard threshold
Max. VoIP packet delay 50 ms
HARQ transmissions (max.) 7
HARQ RTT 8 TTIs
Retransmission Timer 10 TTIs
On duration timer [1, 2, 3, 4] TTIs
Inactivity timer [2, 10, 20] TTIs
DRX cycle timer [10, 20, 40] TTIs
of different DRX adjustments and enhancements, including
on duration, inactivity and DRX cycle timers as well as CQI
preamble scheme and adaptive DRX.
A. On Duration Timer Impact
Figure 10 shows the impact of on duration timers (fixed
inactivity timer of 2 TTIs) to the LTE downlink performance
in terms of maximum number of VoIP users that can be served
per cell with acceptable QoS. The power saving opportunities
for those cases are illustrated in Figure 11, respectively. As
those Figures show, with DRX cycle timer 10 TTIs the on
duration results in significant power saving opportunities with
only minor impacts on DL capacity, providing that the timer
value is higher than one TTI. Similar trend can be seen with
20 TTI cycle with the exception that the LTE performance
numbers are lower than with 10 TTI but at the same time
the power saving opportunities are higher (˜
75-90 % versus
˜
55-80 %). The reason behind the poor performance when the
on duration is one TTI is that the scheduling opportunities
are very scarce and with high probability there will be many
users competing of that scheduling slot. Competition results
into missed scheduling opportunities, which again leads to
highly increased queuing delays for VoIP packets due to DRX
cycles. Increased queuing delays lead to increased amount of
discarded packets already at the transmitting end and thus
Cycle 10 TTIs Cycle 20 TTIs Cycle 40 TTIs
150
200
250
300
350
Capacity per cell
VoIP DL Capacity, On duration timers
No DRX
On duration 1 TTIs
On duration 2 TTIs
On duration 3 TTIs
On duration 4 TTIs
Fig. 10. VoIP DL capacity, ODTs, IAT 2 TTIs
Cycle 10 TTIs Cycle 20 TTIs Cycle 40 TTIs
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Power consumption ratio
DL Power saving opportunities, on duration timers
No DRX
On duration 1 TTIs
On duration 2 TTIs
On duration 3 TTIs
On duration 4 TTIs
Fig. 11. DL power saving opportunities, ODTs, IAT 2 TTIs
to poor performance also from the system level perspective.
Similar phenomena is seen and emphasized when the cycle
length is very long. However, long cycles could possibly be
taken into use when the cell is not fully loaded, assuming that
DRX adapts to the different system loads.
B. Inactivity Timer Impact
The trade-off between VoIP DL capacity and DL power
saving opportunities when inactivity timer is also varied on top
of on duration and DRX cycle timers is shown in Figure 12
and 13. When compared to on duration timer results presented
above the longer inactivity timer does not provide much higher
capacity improvements, especially with cycles 10 and 20 TTIs.
Power saving opportunities are also much lower when longer
inactivity timer is used. With cycle 40 TTIs the inactivity
timer can provide benefits as the cycle is so long that users
will have multiple packets to be transmitted once they become
active from DRX. Thus, longer inactivity timer allows users to
deplete their packet buffers and possibly transmit new packets
arriving their buffers while the inactivity timer is running.
However, this combined effect of prolonged inactivity and on
147
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
8
Cycle 10 TTIs Cycle 20 TTIs Cycle 40 TTIs
150
200
250
300
350
Capacity per cell
VoIP DL Capacity
No DRX
IAT 2 TTIs, ODT 2 TTIs
IAT 2 TTIs, ODT 3 TTIs
IAT 10 TTIs, ODT 2 TTIs
IAT 10 TTIs, ODT 3 TTIs
IAT 20 TTIs, ODT 2 TTIs
IAT 20 TTIs, ODT 3 TTIs
Fig. 12. VoIP DL capacity with different IATs and ODTs
Cycle 10 TTIs Cycle 20 TTIs Cycle 40 TTIs
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Power consumption ratio
DL Power saving opportunities
No DRX
IAT 2 TTIs, ODT 2 TTIs
IAT 2 TTIs, ODT 3 TTIs
IAT 10 TTIs, ODT 2 TTIs
IAT 10 TTIs, ODT 3 TTIs
IAT 20 TTIs, ODT 2 TTIs
IAT 20 TTIs, ODT 3 TTIs
Fig. 13. DL power saving opportunities with different IATs and ODTs
duration timer together with long cycle cannot reach signifi-
cantly difference in power savings to justify the compromise
in the system capacity. More optimal trade-off point can be
achieved with timer settings described above.
C. DRX Performance with Realistic PDCCH
To further verify the trends presented above the performance
of DRX is studied with realistic PDCCH modeling. This
means that PDCCH resources can be exhausted even with a
few users and PDCCH transmission can be erroneous leading
to varying amount of users that can be scheduled per TTI.
Previously presented results assumed fixed number of users
that can be scheduled per TTI (MSU).
The performance of DRX with different activity timers
and realistic PDCCH is illustrated in Figure 14 and 15. As
the capacity Figure shows the absolute capacity numbers are
on lower level, which indicates that the averaged number of
schedulable users per TTI is actually a bit lower than the one
that was assumed in this study as a basis (MSU 8). Apart from
the absolute numbers the performance follows the same trends
presented without detailed PDCCH modeling.
Cycle 10 TTIs Cycle 20 TTIs Cycle 40 TTIs
150
200
250
300
350
Capacity per cell
VoIP DL Capacity
No DRX
On duration 1 TTIs
On duration 2 TTIs
On duration 3 TTIs
On duration 4 TTIs
Fig. 14. VoIP DL capacity, realistic PDCCH
Cycle 10 TTIs Cycle 20 TTIs Cycle 40 TTIs
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Power consumption ratio
VoIP capacity vs. Power saving, Max bundled PDUs 2, Inactivity timer 2 TTI
No DRX
On duration 1 TTIs
On duration 2 TTIs
On duration 3 TTIs
On duration 4 TTIs
Fig. 15. DL power saving opportunities, realistic PDCCH
D. Performance with CQI Preamble
DRX and CQI preamble scheme (see Section III-F) perfor-
mance in terms of VoIP capacity and power saving opportu-
nities are illustrated in Figs. 16 and 17, respectively. In those
figures three types of DRX scenarios are presented in addition
to ’no DRX’ case:
•Ideal CQI, which equals to the case where CQI is updated
regardless of the DRX and power savings are unaffected.
In reality this would not be possible but it is considered
as a reference to benchmark the performance.
•Normal DRX, which equals to the case where normal
DRX settings are used and thus CQI is updated only when
UE is active.
•Preamble 3, which equals to the proposed scheme where
UE wakes up before the actual on duration time to
perform the measurements and to send them to e-Node
B. Preamble length of 3 TTIs is assumed for this study.
When normal DRX operation is compared to the case with
ideal CQI feedback (or no DRX case) it can be seen that
the VoIP capacity is rather sensitive to the up-to-date CQI
information availability. With cycle length of 20 TTIs the
difference between those is around 15 % and with cycle of
148
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
9
Cycle 10 TTIs Cycle 20 TTIs Cycle 40 TTIs
0
50
100
150
200
250
300
350 VoIP DL capacity
Capacity per cell
no DRX
Ideal CQI, On duration 2 TTIs
Ideal CQI, On duration 3 TTIs
CQI Preamble 3 TTIs, On duration 2 TTIs
normal DRX, On duration 2 TTIs
normal DRX, On duration 3 TTIs
normal DRX, On duration 4 TTIs
Fig. 16. VoIP DL capacity with and without DRX preamble, 3 kmph
Cycle 10 TTIs Cycle 20 TTIs Cycle 40 TTIs
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1DL Power saving opportunities
Power consumption ratio
no DRX
Ideal CQI, On duration 2 TTIs
Ideal CQI, On duration 3 TTIs
CQI Preamble 3 TTIs, On duration 2 TTIs
normal DRX, On duration 2 TTIs
normal DRX, On duration 3 TTIs
normal DRX, On duration 4 TTIs
Fig. 17. DL power saving opportunities with and without DRX preamble,
3 kmph
40 TTIs even more. Moreover, as it can be seen from Fig. 16,
mere increase of on duration time might not be adequate as
UE can be scheduled at any point during that time, likely with
outdated CQI information. This implies that some mechanism
to guarantee newer CQI information for scheduling should be
considered, CQI preamble scheme for instance.
When CQI preamble scheme is benchmarked against normal
DRX cycle of 10 TTIs with different activity timer values
it can be seen from Fig. 16 that preambles do not provide
extra value as the capacity is not compromised with so short
cycle. However, even though (normal) DRX cycle of 10 TTIs
implies roughly 30-45 % power consumption versus ’no DRX’
the higher cycles imply possibilities for even higher power
savings, as Fig. 17 shows. Thus, more focus should be paid
on longer cycles, which could guarantee longer UE talk times.
With CQI preamble scheme and longer cycles of 20 or
even 40 TTIs the deterioration in terms of VoIP capacity with
DRX can be mitigated quite well. The gain from preamble
scheme over normal DRX becomes higher when the DRX
cycle length is longer, which is quite intuitive as then the
CQI information available in the scheduler is older (without
Cycle 10 TTIs Cycle 20 TTIs
0
20
40
60
80
100
120
140
160
Capacity per Cell
VoIP DL Capacity
No DRX
Normal DRX, on duration 2 TTIs
Ideal CQI, on duration 2 TTIs
Fig. 18. VoIP DL capacity with and without ideal CQI, 30 kmph
preambles). Preamble scheme can even outperform ’ideal CQI’
case in some situations due to preamble scheme forcing the
periodic CQI measurements being synchronized with data
transmissions. With ’ideal CQI’ the periodic reporting happens
every 5 ms regardless of the data flow / DRX.
In terms of downlink power saving opportunities the per-
formance of CQI preamble scheme is expectedly lower than
with normal DRX operation, as Fig. 17 implies. However, as
the performance in terms of system capacity is much more
robust against potential losses, the loss in battery savings with
preambles can be considered as acceptable to guarantee more
satisfactory service provision. Moreover, this paper evaluated
only the scheme where UE stays active for the whole duration
of preamble time but in principle after the UE has performed
the measurements the UE could fall back to inactivity before
actual on duration. By turning receiver circuitry off during
those periods higher battery savings could be expected.
Finally, the findings and conclusions from CQI impact with
DRX are confirmed in Fig. 18 where VoIP performance with
all users moving with higher velocity is illustrated. As [30]
indicates and this study confirms with higher UE velocity the
frequency selectivity gain of CQI is lost to most extent and
thus no additional gain could be achieved even with ’ideal
CQI’.
E. Performance with Adaptive DRX
Finally, the purpose of this section is to cover how adaptive
DRX presented in Section III-G affects to the performance.
Figs. 19 and 20 illustrate the performance in capacity and
power consumption ratio, respectively. As the capacity figure
shows, adaptive DRX can provide noticeable gains, especially
with longer cycles than 10 TTIs where practically no losses
are seen if on duration is long enough. In terms of battery
savings the adaptive DRX leads to slight increase in battery
consumption. This due to the fact that blind offset setting could
further induce the packet bundling and reduce the amount of
retransmissions due to increased delays and thus reduce the
time needed to keep receiver circuitry active.
149
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
10
Fig. 19. VoIP DL capacity, adaptive DRX
Fig. 20. DL power saving opportunities, adaptive DRX
V. CONCLUSION
The purpose of this article was to evaluate the trade-off
conditions between energy efficiency at the user terminal and
LTE performance. That goal is achieved by studying VoIP
over LTE with various DRX related timers, which limit the
scheduling freedom of users while increasing battery saving
opportunities at the terminal. Increased power saving oppor-
tunities lead to increased talk-times for the terminals and thus
more satisfactory user experience. The study was conducted
with dynamic system simulator modeling LTE network with
high level of detail.
This article indicates that for dynamic and PDCCH con-
trolled scheduling, short DRX cycle timers together with
appropriate on-duration timer is an attractive choice for LTE
energy efficiency: Substantial power saving opportunities are
achievable with minor trade-off in terms of maximum VoIP
over LTE DL capacity. Regardless, this article also points out
that in lower load situations different (more stricter) DRX
adjustments could be used as then the capacity might not be
compromised and the power savings would be higher. This
article also shows that prolonging inactivity timer together
with on duration and DRX cycle timers does not justify the
TABLE II
CONFIDENCE INTERVALS FOR INTERPOLATED VOIP CAPACITY
Capacity [UEs/cell]
MEAN 260.305
STANDARD DEVIATION 3.85448
CONFIDENCE INTERVAL, 90 % ±0.44 %
CONFIDENCE INTERVAL, 95 % ±0.52 %
CONFIDENCE INTERVAL, 99 % ±0.69 %
slightly increased performance (capacity) as the trade-off in
battery saving opportunities is too high.
VoIP performance in terms of capacity may, however, be
reduced by some extent if sub-optimal DRX settings are
chosen. One main reason for reduced performance, especially
with long cycles, is linked to outdated CQI information.
CQI preamble scheme, introduced in this paper, mitigates the
reduced performance quite well in terms of VoIP capacity
when dynamic user scheduling is assumed. Moreover, the
improvements can be achieved with acceptable trade-off in
terms of battery saving opportunities.
Finally, this study introduces concept where DRX (offset)
would adapt the data flow. The result show significant im-
provement in terms of capacity in situations where blind DRX
configuration show losses. Moreover, adaptive DRX brings
only minor impact to the power consumption ratio.
VI. ACKNOWLEDGMENTS
This study is a collaborative work between Magister So-
lutions Ltd., University of Jyv¨
askyl¨
a, Nokia, Nokia Wireless
Modem Research (nowadays a part of Renesas Mobile Corpo-
ration) and Nokia Siemens Networks. The authors would like
to thank all of their co-workers and colleagues for their com-
ments and support. Finally, special thanks go to Hannu-Heikki
Puupponen from University of Jyv¨
askyl¨
a for his contributions
on providing confidence analysis results.
APPENDIX
Research tool used in this study was presented briefly
above. This appendix is aimed to deepen the presentation by
providing statistical confidence analysis for the used system
level tool.
Statistical analysis is based on evaluating the performance
with a few selected test cases, namely VoIP simulations
with different random number generator seeds. Based on the
seed all random generators are initialized, these include e.g.
starting position and direction of the movement for the UEs.
Even though, all of the simulation results depend on random
processes, the results are reproducible with certain level of
accuracy, which is defined in this appendix.
An interval estimation can be used to define a confidence
interval, which means that the sample φ, is within a defined
interval with a certain probability. This probability can be
expressed as follows
P(a≤φ≤b) = 1 −α(3)
where the interval [a, b]is a (1−α)×100% confidence interval
of φ. A probability that the φis not within the interval is
150
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
11
α. When the number of samples nis ≤30 the standardized
normal distribution, N(0,1), can be used to define confidential
interval, which is
(x−zα/2×s/√n, x +zα/2×s/√n).(4)
In Eq. 4 the xis the average value, zα/2is the critical value
taken from the standardized normal distribution N(0,1),sis
the standard deviation and nis the number of samples i.e. in
this case the number of simulation runs.
The simulation environment used for confidence analysis is
the same macro cellular layout used for DRX studies. Main
parameters are as well similar to the simulations presented
above with the exception that dynamic scheduling MSU is
assumed to be 6. This is done so that simulations take PDCCH
restrictions better into account.
Confidence intervals are calculated for radio system capacity
i.e. at the point where 5 % VoIP users are in outage. Capacity
is interpolated from different cell loads i.e. from another one
where the system outage level is below 5 % and another where
it is above that. Thus, the required simulation amounts for
confidence analysis equal two times n= 31.
The results for dynamic LTE system level tool are shown
in Table II. As that table shows the difference even with the
highest confidence interval are very minor. Thus, this gives
the confidence that the results produced with the tool for this
study are also well within the required level of reliability and
reproducibility.
REFERENCES
[1] K. Aho, T. Henttonen, J. Puttonen, and T. Ristaniemi, “Trade-Off
between Increased Talk-Time and LTE Performance,” in Proceedings
of International Conference on Networks (ICN), April 2010.
[2] K. Aho, T. Henttonen, J. Puttonen, and L. Dalsgaard, “Channel Quality
Indicator Preamble for Discontinuous Reception,” in Proceedings of
IEEE Vehicular Technology Conference (VTC’S10), May 2010.
[3] H. Ekstr¨
om and A. Furusk¨
ar and J. Karlsson and M. Meyer and S.
Parkvall and J. Torsner and M. Wahlqvist, “Technical Solutions for the
3G Long Term Evolution,” IEEE Communications Magazine, vol. 44,
pp. 38–45, March 2006.
[4] A. Toskala, H. Holma, K. Pajukoski, and E. Tiirola, “UTRAN Long
Term Evolution in 3GPP,” in Proceedings of IEEE International Sympo-
sium on Personal, Indoor and Mobile Radio Communications (PIMRC),
September 2006.
[5] F. Persson, “Voice over IP Realized for the 3GPP Long Term Evolution,”
in Proceedings of IEEE Vehicular Technology Conference (VTC’F07),
September 2007.
[6] Requirements for Evolved UTRA (E-UTRA) and Evolved UTRAN (E-
UTRAN), 3GPP TR 35.913, Rev. v9.0.0, December 2009.
[7] J. Puttonen, T. Henttonen, N. Kolehmainen, K. Aschan, M. Moisio, and
P. Kela, “Voice-over-IP Performance in UTRA Long Term Evolution
Downlink,” in Proceedings of IEEE Vehicular Technology Conference
(VTC’S08), May 2008.
[8] Y. Fan, P. Lunden, M. Kuusela, and M. Valkama, “Performance of VoIP
on EUTRA Downlink with Limited Channel Feedback,” in Proceed-
ings of International Symposium on Wireless Communication Systems
(ISWCS’08), October 2008.
[9] D. Jiang, H. Wang, E. Malkamki, and E. Tuomaala, “Principle and
Performance of Semi-Persistent Scheduling for VoIP in LTE System,” in
Proceedings of International Conference on Wireless Communications,
Networking and Mobile Computing (WiCom’07), September 2007, pp.
2861–2864.
[10] Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved
Universal Terrestrial Radio Access Network (E-UTRAN); Overall de-
scription; Stage 2, 3GPP TS 36.300, Rev. v9.3.0, March 2010.
[11] Evolved Universal Terrestrial Radio Access (E-UTRA) Medium Access
Control (MAC) protocol specification, 3GPP TS 36.321, Rev. v9.2.0,
March 2010.
[12] S.-R. Yang and Y.-B. Lin, “Modeling UMTS Discontinuous Reception
Mechanism,” in IEEE Transactions on Wireless Communications, vol. 4,
January 2005, pp. 312–319.
[13] K. Aho and I. Repo and T. Nihtil¨
a and T. Ristaniemi, “Analysis of
VoIP over HSDPA Performance with Discontinuous Reception Cycles,”
in Proceedings of International Conference on Information Technology
(ITNG), April 2009.
[14] L. Zhou, H. Xu, H. Tian, Y. Gao, L. Du, and L. Chen, “Performance
Analysis of Power Saving Mechanism with Adjustable DRX Cycles in
3GPP LTE,” in Proceedings of IEEE Vehicular Technology Conference
(VTC’F08), September 2008.
[15] T. Kolding, J. Wigard, and L. Dalsgaard, “Balancing Power Saving and
Single User Experience with Discontinuous Reception in LTE,” in Pro-
ceedings of IEEE International Symposium on Wireless Communication
Systems 2008 (ISWCS’08), October 2008.
[16] J. Wigard, T. Kolding, L. Dalsgaard, and C. Coletti, “On the User
Performance of LTE UE Power Savings Schemes with Discontinuous
Reception in LTE,” in Proceedings of IEEE International Conference
on Communications (ICC’09), June 2009.
[17] C. Bontu and E. Illidge, “DRX Mechanism for Power Saving in LTE,”
IEEE Communications Magazine, vol. 47, pp. 48–55, June 2009.
[18] P. Kela, J. Puttonen, N. Kolehmainen, T. Ristaniemi, T. Henttonen, and
M. Moisio, “Dynamic Packet Scheduling Performance in UTRA Long
Term Evolution Downlink,” in Proceedings of International Symposium
on Wireless Pervasive Computing (ISWPC), May 2008.
[19] D. B¨
ultmann and M. M¨
uhleisen and K. Klagges, “openWNS - open
Wireless Network Simulator,” in Proceedings of the European Wireless,
May 2009.
[20] J. C. Ikuno, M. Wrulich, and M. Rupp, “System Level Simulation of LTE
Networks,” in Proceedings of the IEEE Vehicular Technology Conference
(VTC’S10), May 2010.
[21] K. Aho, I. Repo, J. Puttonen, T. Henttonen, M. Moisio, J. Kurjenniemi,
and K. Chang, “Benchmarking of VoIP over HSDPA and LTE Per-
formance with Realistic Network Data,” in Proceedings of the IEEE
International Symposium on Wireless Pervasive Computing (ISWPC),
May 2010.
[22] R. Cuny and A. Lakaniemi, “VoIP in 3G Networks: An End-to-
End Quality of Service Analysis,” in Proceedings of IEEE Vehicular
Technology Conference (VTC’S03), Jeju Island, Korea, April 2003.
[23] D. Lopez, C. U. Castellanos, I. Z. Kovacs, F. Frederiksen, and K. I.
Pedersen, “Performance of Downlink UTRAN LTE Under Control
Channel Constraints,” in Proceedings of IEEE Vehicular Technology
Conference (VTC’S08), May 2008.
[24] J. Puttonen, H.-H. Puupponen, K. Aho, T. Henttonen, and M. Moisio,
“Impact of Control Channel Limitations on the LTE VoIP Capacity,”
in Proceedings of the International Conference on Networks (ICN’10),
April 2010.
[25] One-way transmission time, ITU recommendation G.114, 2003.
[26] “LTE Physical Layer Framework for Performance Verification,” 3GPP
TSG RAN WG1 LTE contribution R1-070674, February 2007.
[27] “DRX Parameters in LTE,” 3GPP TSG RAN WG2 LTE contribution
R2-071285, March 2007.
[28] Evolved Universal Terrestrial Radio Access (E-UTRA) Radio Resource
Control (RRC) protocol specification, 3GPP TS 36.331, Rev. v9.1.0,
December 2009.
[29] Selection procedures for the choice of radio technologies of the
UMTS, Universal Mobile Telecommunications System (UMTS 30.03)
TR 101.112, Rev. v3.1.0, November 1997.
[30] N. Kolehmainen, J. Puttonen, P. Kela, T. Ristaniemi, T. Henttonen, and
M. Moisio, “Channel Quality Indication Reporting Schemes for UTRAN
Long Term Evolution Downlink,” in Proceedings of IEEE Vehicular
Technology Conference (VTC’S08), May 2008.
151
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
12
Kari Aho received his M.Sc., L.Sc. and Ph.D.
degrees in information technology in the field of
mobile telecommunications from the University of
Jyv¨
askyl¨
a in 2006, 2009 and 2010, respectively.
His Master’s, Licentiate and Ph.D. thesis focused
on various third generation cellular network perfor-
mance enhancements, which were addressed mainly
through VoIP and MBMS services. Currently he is
working as project manager at Magister Solutions in
the area of future HSPA evolution. Current research
interests are VoIP performance, uplink transmit di-
versity and downlink multipoint transmission and reception issues.
Tero Henttonen received his M.Sc. degree in ap-
plied mathematics from the University of Helsinki
in 2001. His master’s thesis focused on compressed
mode effects in WCDMA. He is currently working
as Principal Researcher at Renesas Mobile Corpora-
tion, with research areas related to 3GPP Long Term
Evolution; especially on mobility performance and
heterogeneous networks. Current research interests
include mobility in LTE-A and in heterogeneous
networks.
Jani Puttonen received his M.Sc. and Ph.D. degree
in information technology in the field of telecommu-
nications from the University of Jyv¨
askyl¨
a in 2003
and 2006, respectively. His Ph.D. thesis focused on
IP level mobility management in heterogeneous net-
works. He is currently working as a Senior Research
Scientist at Magister Solutions with research areas
related to 3GPP Long Term Evolution; especially
RAN features and performance. Current research
interests are Minimization of Drive Tests and Self
Organizing/Optimizing Networks.
Lars Dalsgaard received his M.Sc. degree in
Telecommunications from Aalborg University in
1995. His master’s thesis focused on applying
OFDM multicarrier transmission within GSM. He
is currently working as a senior specialist at Nokia
Devices within work areas related to 3GPP Long
Term Evolution with special focus on UE perfor-
mance, power consumption and mobility. Current
interests include mobility and Carrier aggregation
within LTE-A, mobility in general including 3GPP
inter-system and heterogeneous network.
Tapani Ristaniemi was born in Kauhava, Finland,
in 1971. He received the M.Sc. in mathematics
in 1995, Ph.Lic. in applied mathematics in 1997,
and Ph.D. in telecommunications in 2000 from the
University of Jyv¨
askyl¨
a, Jyv¨
askyl¨
a, Finland. During
2001-2003 he was a professor of telecommunica-
tions at the Department of Mathematical Informa-
tion Technology, University of Jyv¨
askyl¨
a, and 2003-
2006 a professor of wireless data communications,
Institute of Communications Engineering, Tampere
University of Technology, Finland. In 2006 he joined
the University of Jyv¨
askyl¨
a, where he holds a professorship of computer
science. His research interests include signal processing for communications,
biosignal processing and system engineering for wireless communications.
