Traffic-based Study of Femtocell Access Policy

Impacts on HSPA Service Quality

Shaunak Joshi, Ray C.C. Cheung, Pooya Monajemi, and John D. Villasenor
Electrical Engineering Department, University of California, Los Angeles

sjoshi@ee.ucla.edu, cccheung@ieee.org, pmonajemi@ucla.edu, and villa@ee.ucla.edu

AbstractHome NodeBs (HNB)s can be operated in a fullyclosed manner (in which only user equipment (UEs) within a
closed subscriber group (CSG) are permitted to connect), a
fully-open manner (in which access can potentially be granted
to any nearby mobile station), or somewhere in between. An
additional option is hybrid HNBs, in which a portion of the
HNB resources are reserved for exclusive use of the CSG, and
the remaining resources are allocated in an open manner. The
relationship between access policy and the performance of the
overall network is one of the most interesting open questions
related to HNB deployment. The present paper explores the
relationship between HNB access and quality of service with
specific attention to the differences between HSDPA, HSUPA,
and the relative mix of traffic types.

I. I NTRODUCTION
HNBs are low-power, low-cost cellular base stations designed to serve a very small area, such as a home or office
environment. HNBs can be operated in a closed subscriber
group (CSG) manner (also called closed access), in which
only user equipment (UE)s with proper authorization can
connect, or under an open access policy, in which UEs
outside the CSG are permitted to connect to the HNB. At the
highest level, the open/closed issue is one which balances the
benefits of interference reduction against the costs of HNB
resource sharing. Awareness of these issues has led to the
recent introduction in the standardization process of hybrid
HNBs [1], in which some portion of the HNB resources are
reserved for the exclusive use of the CSG, while the remaining
resources are open.
Factors such as the design and processing limitations of
the HNB baseband hardware typically limit the number of
UEs permitted to connect to a HNB. It is also possible to
operate at intermediate points between fully-closed and fullyopen access. The issue of access policies for co-channel HNBs
(e.g. HNBs sharing frequency bands with macro base stations)
has been examined in the 3GPP industry group. These studies
have explored issues including downlink capacity for networks
with open vs. closed access [2], the effect of the number of
HNBs on the capacity of a macrocell [3], examinations of
interference impacts involving various combinations of HNBs
and macrocells [4], [5], and performance studies on the impact
of coexistence of HNBs and macrocells on both uplink and
downlink [6], [7]. These studies have shown that that a closed
co-channel HNB can lead to interference issues, particularly
if the HNB does not adaptively change its transmit power in

order to mitigate such interference [8]. In [9], [10], [11],

and [12], issues related to HNBs in the context of OFDMbased networks have been examined. In particular, use of selfoptimization and self-configuration of HNBs [9], performance
under limited access [10], and use of dynamic frequency planning to avoid inter-cell interference [12] have been presented.
In [13], we explored the impact of access policy choice
in HSDPA-dominated networks. It was shown that completely
closed access can disrupt HNB service because of the interference generated by neighboring UEs that are near to the HNB,
but not admitted to the CSG. By contrast, completely open
access, while often better than closed access, also involves
some tradeoffs, both due to the loss in data rate arising
from sharing limited HNB bandwidth among potentially large
numbers of UEs, and due to loading impact on nearby macro
base stations. To address this, the work in [13] identified
operating points between fully-closed and fully-open access
that allowed making optimal tradeoffs between HSDPA data
rate maximization and interference minimization.
In the present paper, we direct attention to the interrelationship between traffic type, access policy, and performance.
HSDPA and HSUPA are different in many respects, and it
stands to reason that access policy decisions will impact them
differently. In addition, the way in which the overall network
responds to access policy choices depends in part on the
specific mix of traffic types on the network. To examine
this, we explore the relationship between access policy and
service quality across a broader range of traffic types to
determine optimal network performance (in terms of service
and throughput) for a given scenario.
II. M ODELING E NVIRONMENT
The simulation environment used here is an enhanced
version of the environment described in [13], differing most
significantly in that HSUPA capabilities have been introduced.
Here we provide only a brief review of the information already
presented in [13]. The simulation environment is built on the
static WCDMA circuit-switched simulator developed by Laiho
and Wacker [14], to which we have introduced a time slotted
structure and scheduling methods to introduce dynamism in
the context of packet data services.
The time-slotted structure for HSUPA and HSDPA packet
scheduling in the simulation environment uses a transmission
time interval of 2 msec, used for consistency with the HSPA

978-1-4244-4148-8/09/$25.00 2009
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2009 proceedings.

III. R ESULTS
Fig. 1 provides information about the system configuration
used in the experiments. The system contains a total of 21
macro base stations located at 7 sectorized sites, 456 UEs,
and 32 HNBs. There are 32 indoor environments with HNBs,
with each HNB associated with two CSG UEs also located
indoors (though potentially separated from the HNB by one
or more indoor walls). In addition there are 8 neighboring UEs
near each HNB, separated from the HNB by an indoor/outdoor

UEs per
environment

Environments

Total # of
UEs

2 CSG UEs

32

2*32 = 64

8 neighboring
UEs

32

32*8 = 256

Indoor environments
without HNBs

2 UEs

32

32*2 = 64

Outdoor

72

72

Total UEs:

456

Environment Type
Indoor environments
with HNBs

Fig. 1.

System configuration.

BS

BS

3000

12

UEs
BS

BS

10

17

Indoor environment with HNB

2000
Indoor
environment without HNB
8 Neighboring UEs

BS

2000

Y-coordinate [m]

standards. Each interval is subdivided into three 667 sec time

slots. For both HSDPA and HSUPA transmissions, respective
uplink and downlink transmissions of control, data, ACK,
etc. are active during appropriately staggered time slots. Each
time slot is independently considered to identify the optimum
power levels. Separate schedulers for HSDPA and HSUPA
independently perform the task of choosing the high speed
UEs to associate with each base station (macro or HNB) for
each packet, and determine the channels that are active for
each time slot. Only one UE per base station is permitted to
be a high speed user during a single time slot. Round robin
scheduling is used in which all UEs connected to a given base
station that desire high speed data take turns receiving (for
HSDPA) or sending (for HSUPA) data. For any given base
station/UE pair, time lags of 3 to 6 time slots are included
to allow for propagation delay and processing. These time
lags are then utilized for communication between other base
station/UE pairs.
Simulation durations range from 500 to 3000 time slots,
which is equivalent at the upper end of this range to 2 seconds
of actual network operation. A high speed downlink packet
transmission is considered to fail when the UE is put to outage
in any of the time slots concerning HSDPA transmission
except CQI. A high speed uplink packet transmission is
considered to fail when the mobile UE is put to outage in any
of the time slots concerning HSUPA transmission. Each of the
HSPA control channels is modeled with specific data rate and
Eb /N0 requirement and subject to closed loop power control.
The HSDPA data channel utilizes adaptive modulation and
coding (AMC) to assign data rates given the allocated power.
In HSUPA transmissions, power is allocated to the high
speed uplink data channel (E-DPDCH) using a loading-aware
approach based on the current load on the base station serving
the UE in question. The amount of additional UE transmit
power that would be required to overload the base station
at the time slot in question is determined. After allowing
3dB of headroom, a certain amount of power is allocated
to E-DPDCH (only if the UE is able to fulfill this power
allocation without surpassing its own overall transmit power
limitation). If the UE is not able to meet the allocated transmit
power, a fraction of the difference between maximum UE
transmit power capability and target power allocation is used.
Scheduling is coordinated so that one UE is sending uplink
data at a time to any given base station. The instantaneous
uplink data rate calculation is based on the received signal-tointerference-noise ratio (SINR) at the serving base station.

BS

MacroBS

11

1000
BS

BS

BS

BS

15

BS

BS

BS

BS

13

BS

14

-1000

BS

BS

-2000

21

18

BS

BS

19

16

-3000
-3000

-2000 BS20

-1000

1000

BS

17

2000

3000

X-coordinate [m]

Fig. 2. Positions of UEs, macrocells, and HNBs. UEs are indicated by green
dots; the 32 indoor environments with HNBs are indicated by blue circles;
the 32 indoor environments without HNBs are indicated by red circles. Red
Xs indicate the presence of eight neighboring UEs located in the immediate
vicinity of some of the HNBs. These neighbors generally experience a stronger
pilot signal from the HNB than from the macrocell, and thus are candidates
to connect to the HNB if access policy allows. Macro base station directional
antennas are indicated by green lines.

transition wall, and located at a distance from the HNB of less

than 100 meters. There are also 32 indoor environments, each
containing 2 UEs in which there is no HNB present. Finally,
there are 72 outdoor UEs. Signal loss through indoor and
indoor/outdoor transition walls is modeled using a histogrambased wall loss model. The layout of this system is shown
Fig. 2.
One of the challenges in modeling highly complex networks
with hundreds of UEs lies in efficiently presenting performance data. We believe that packet success rates (PSR) are
a useful metric because they directly indicate an important
aspect of performance, and because they can be used to
derive information about the performance of applications such
as media transmissions. Fig. 3 shows a conceptual example
of how PSR information can be conveyed in an intuitive
visual fashion. This figure is a complimentary cumulative
distribution function (CCDF) for the PSR, with the vertical
axis indicating the fraction of UEs and the horizontal axis
providing the minimum PSR rate for the associated fraction of

978-1-4244-4148-8/09/$25.00 2009
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2009 proceedings.

Fraction of UEs

Fraction of UEs

1
0.9

60% of UEs
have PSR > 50%

0.8

Medium

0.7
0.6
0.5
0.4

10% of UEs
have PSR > 5%

0.3
0.2
0.1

Poor

0
0

20

90% of UEs have PSR > 95%

Good

40

50

60

0.8
0.6

csg
csg+4

0.4

csg+8
no HNB

0.2
0

80

95 100

Packet success rate (%)

40

60

80

100

Packet success rate (%)

Fig. 3. An example plot showing how packet success rate can be conveyed
using a complimentary cumulative distribution function (CCDF). A point on
the plot indicates that the associated fraction of UEs (vertical axis) experience
a PSR of at least the value indicated by the horizontal axis. Three illustrative
curves are shown. For example, the good illustrates a system in which 90%
of the UEs experience PSRs of greater than 95%.
1

Fraction of UEs

20

Fig. 5.
HSUPA packet success rates for UEs connected to the macro
base station. Four curves are shown, corresponding to a system with no
HNBs (no HNB), completely closed access (CSG), allowing the four
neighboring UEs to connect (CSG+4), and allowing all eight neighboring
UEs to connect (CSG+8) In contrast with the HNB case in Fig. 5, the PSR
for UEs connected to the macro base station shows a much less significant
improvement with HNB access policy changes, though going from no
HNB to any HNB scenario does give some improvement, due to reduced
interference levels.

0.8

0.6

csg
csg+4

0.4

csg+8

0.2

0
0

20

40

60

80

100

Packet success rate (%)

Fig. 4.
HSUPA packet success rates for UEs connected to the HNBs.
Three curves are shown, corresponding to completely closed access (CSG),
allowing the four neighboring UEs to connect (CSG+4), and allowing
all eight neighboring UEs to connect (CSG+8). Performance improves
significantly once all neighbors are allowed access, as this enables the HNB
to exert scheduling control over HSUPA transmissions.

users. Accordingly, good performance is exemplified by the

upper curve in Fig. 3 which shows the curve passing through
(fraction of UEs=0.9, PSR=95%), meaning that 90% of the
UEs have a PSR of at least 95%. The middle curve illustrates
a system with medium performance; in this case the passage
of the curve through (fraction of UEs=0.6, PSR=50%) means
that 60% of the UEs have a PSR of at least 50%, and
thus conversely, that 40% of the UEs have a PSR of less
than 50%. Finally, the lower curve illustrates a system with
poor performance. In this example, the curve passes through
fraction of UEs=0.1, PSR=5%), which means that 90% of the
UEs have a PSR of below 5%.
Fig. 4 illustrates the PSR complimentary cumulative distribution function for UEs connected to the HNBs. Three curves
are shown, corresponding respectively to a fully closed access
policy (curve labeled CSG), an access policy in which the
CSG plus four neighbors are granted access (CSG+4), and
an access policy in which the CSG plus eight neighbors are

granted access (CSG+8). This figure, and in particular the

negative slope of the CSG and CSG+4 curves indicates a wide
range of HSUPA packet success rates for UEs served by HNBs
when there are other nearby UEs performing HSUPA transmissions to a macro base station and not to the HNB. Under
such a scenario, the uplink interference generated by other
cells (in other words by the neighboring UEs not connected
to the HNB) changes significantly from time slot to time
slot, often resulting in interference with HSUPA transmissions
involving the HNB-associated UEs. When the access is opened
to allow all neighbors to utilize the HNB (CSG+8 curve)
there is a marked improvement in PSR. This is due to the
reduced interference levels from other (non-HNB) cells. Most
fundamentally, when an HNB allows a formerly excluded UE
to connect to it, the HNB can then exert scheduling control
over the UE HSUPA transmissions. This benefit of allowing
the HNB to directly manage transmissions in its vicinity in
many cases will be well worth the price paid in resource
sharing.
Fig. 5 shows the PSR for UEs connected to the macro base
station. Curves for the three access policy scenarios CSG,
CSG+4, and CSG+8 are again given, as well as a curve
no HNB corresponding to the case where there are no
HNBs present. In contrast with the UEs connected to the
HNB in Fig. 4, in Fig. 5 there is much less improvement
in the PSR as a function of HNB access policy change. The
most significant change occurs in PSR improvement in the
transition from a system with no HNBs to one with HNBs,
regardless of access policy. This is due to the overall reduction
in interference experienced by the UEs connected to the macro
base station when other UEs are offloaded onto the HNB,
and thus perform transmissions using lower power. It is also
worth commenting on why going from CSG only to CSG+8
has a lower impact in Fig. 5 than in Fig. 4, because both macro

978-1-4244-4148-8/09/$25.00 2009
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2009 proceedings.

0.8
csg
csg+4
csg+8

0.6

Fraction of UEs

Fraction of UEs

0.8

0.4

0.2

0.6
csg
csg+4
csg+8
no HNB

0.4
0.2

0
0

20

40

60

80

100

Packet success rate (%)

Fig. 6.

HSDPA packet success rates for UEs connected to the HNBs.

base stations and HNBs impose scheduling control over their

respective UEs. The impact is not symmetric because UEs
associated with a HNB transmit with much lower power than
UEs associated with a macro base station. Thus, the HSUPA
transmissions of HNB UEs have a much lower impact on
macro base station UEs than the other way around.
It is also important to recognize that (in both Fig. 5 and
Fig. 4) the set of UEs represented by the curves expands or
contracts as access policies change. Thus, for example, there
are more total UEs represented by the CSG curve in Fig. 5
than by the CSG+8 curve, since the act of opening access
has moved some UEs off of the macro base station and onto
the HNB. A UE that remains connected to the macro base
station across all access policies will have a proportionally
larger impact on the overall packet success rates in Fig. 5 as
the access is progressively opened. A converse effect occurs
with Fig. 4. While these effects do not impact the specific
observations above, there are scenarios and circumstances
under which this is a factor, as discussed below.
Fig. 6 shows the PSR performance for HSDPA for UEs
connected to the HNB. Unsurprisingly given the proximity of
the HNBs to the UEs, the PSR rates are very high regardless of
access policy. In contrast with HSUPA, in which interference
from macro-connected UEs impeded transmissions of HNBconnected UEs, with HSDPA UEs connected to the macro base
station do not generally cause large enough downlink out-ofcell interference to disrupt downlink transmissions from the
HNB to the UEs connected to it. This is in part due to the
fact that HSDPA downlink power in a transmission from a
macro base station to a UE is independent of the served UE
location with respect to HNBs in the system. By contrast, the
loading on the HNB is impacted by HSUPA transmissions
from nearby, macro-connected UEs.
Fig. 7 shows the performance for HSDPA for UEs connected
to the macro base station. One of the most notable features of
this figure is the reduction in PSRs in going from a system

20

40

60

80

100

Packet success rate (%)

Fig. 7. HSDPA packet success rates for UEs connected to the macro base
station.

with no HNBs to the CSG case. This is due to two factors.

First, as discussed earlier, the system includes 32 HNBs having
8 neighbor UEs each. Thus, there are a total of 256 UEs that
are very near to, but in the case of CSG operation, excluded
from access to HNBs. When the HNBs are turned on under
a CSG access policy, the potential gain in PSR enabled by
lower transmission power of the two CSG UEs now connected
to the HNB is more than outweighed by the interference experienced by the neighbor UEs due to the HNB transmissions. As
access is opened up, more of these neighbors are connected to
the HNB, thus the fraction of overall macro base station UEs
that are also non-HNB-connected HNB neighbors is reduced,
and the overall PSRs rise. Even in the CSG=8 case, however,
when all neighbors have been admitted to the HNBs, the PSR
remains lower than it was with no HNBs. This is because
there are some macro-connected, non-neighbor UEs that, due
to distance to the macro base station and/or location in poor
propagation environments, have poor PSR regardless of HNB
access policy. These UEs constitute a higher proportion of the
macro-connected UEs as access policy is opened, and thus
have a larger impact on the PSR computation.
Figures 8 through 11 present throughput results. These figures show HSDPA and HSUPA throughputs for two different
scenarios: HSPDA dominant, and HSUPA dominant. In the
HSDPA dominant scenario, 90% of the UEs are engaged in
HSDPA only, while the remaining 10% are simultaneously
using all of HSDPA, HSUPA, and voice. In the HSUPA
dominant scenario, 90% of the UEs are engaged in HSUPA
only, while the remaining 10% are simultaneously using all of
HSDPA, HSUPA, and voice. In the two scenarios studied, the
UEs engaging in each type of service are chosen randomly in
accordance with the probabilities given above. While typical
relative usage patterns of different traffic types will obviously
vary, exploring these heavily HSUPA- or HSDPA- dominant
extremes helps to focus on the specific interaction of traffic

978-1-4244-4148-8/09/$25.00 2009
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2009 proceedings.

90

macroBS UE throughput
HNB UE throughput

3.5
3
2.5
2
1.5
1
0.5
0
No HNB

2CSG

S e rv e d U E T h ro u g h p u t [k b p s ]

Served UE Throughput [M bps]

2CSG + 1 2CSG + 2 2CSG + 3 2CSG + 4 2CSG + 5 2CSG + 6 2CSG + 7 2CSG + 8

80
70
60
50
40
30
20
10
0
No HNB

Fig. 8.
HSDPA throughput for UEs in the HSDPA-dominant scenario.
The dark/red bars correspond to macro-connected UEs; the light/blue bars
correspond to HNB-connected UEs.

2CSG

2CSG + 1 2CSG + 2 2CSG + 3 2CSG + 4 2CSG + 5 2CSG + 6 2CSG + 7 2CSG + 8

Fig. 10. HSUPA throughput for UEs in the HSDPA-dominant scenario.

The dark/red bars correspond to macro-connected UEs; the light/blue bars
correspond to HNB-connected UEs.
70

Served UE Throughput [kbps]

S e rv e d U E T h ro u g h p u t [M b p s ]

60

50

40

30

20

10

0
No HNB

2CSG

2CSG + 1 2CSG + 2 2CSG + 3 2CSG + 4 2CSG + 5 2CSG + 6 2CSG + 7 2CSG + 8

0
No HNB

Fig. 9.
HSDPA throughput for UEs in the HSUPA-dominant scenario.
The dark/red bars correspond to macro-connected UEs; the light/blue bars
correspond to HNB-connected UEs.

type, access policy, and network performance.

Fig. 8 considers the HSDPA throughput in the HSDPA
dominant scenario. The vertical axis gives the average (per
UE) throughput for those UEs connecting to the macro base
stations (dark bars), and to the HNB (light bars). Different
positions on the horizontal axis represent different policies for
HNB access. At the left end of the axis, the case with no
HNBs is considered. In this case all of the UEs, including the
two UEs that will later be part of the CSG, connect (or attempt
to connect) to the macro base stations. The next position
on the horizontal axis is labeled 2CSG, and considers that
performance when the HNBs are present and the access is
closed so that only the two CSG mobiles can utilize them. As
one moves further to the right on the axis, successively larger
numbers of neighbors are granted access to the HNBs.
As the access policy is changed toward a more open policy,
the average HNB-connected UE throughput decreases. This
is because the HNB needs to share its bandwidth among
increasing numbers of neighbors. At the same time, as more
users are allowed onto the HNB, the number of UEs served
per each macrocell decreases, so the throughput enjoyed by
the macrocell-connected UEs increases.

2CSG

2CSG +
1

2CSG +
2

2CSG +
3

2CSG +
4

2CSG +
5

2CSG +
6

2CSG +
7

2CSG +
8

Fig. 11. HSUPA throughput for UEs in the HSUPA-dominant scenario.

The dark/red bars correspond to macro-connected UEs; the light/blue bars
correspond to HNB-connected UEs.

Fig. 9 considers the HSDPA throughput in the HSUPAdominant scenario. Unsurprisingly, the overall throughput rates
are significantly higher than for the HSDPA-dominant case
in Fig. 8, because in the HSUPA-dominant scenario there are
fewer UEs competing for downlink data bandwidth, and downlink interference levels are lower. In addition, the decrease in
HNB-connected UE throughput with increasingly open access
is significantly less pronounced than in Fig. 8 because, again,
there are fewer downlink demands on the HNB, so the resource
sharing that accompanies open access imposes a smaller perUE penalty.
Fig. 10 considers the HSUPA throughput in an HSDPA
dominant scenario. Fig. 11 considers the HSUPA throughput in
an HSUPA dominant scenario. Both of these figures show that
the HSUPA throughput increases with more open access, as
contrasted with the decrease observed in HSPDA throughput
in Figs. 8 and 9. This indicates that, for HSUPA throughput,
while more open access means that each HNB shares its
resources among a larger group of UEs, with respect to
HSUPA, this potential disadvantage is more than outweighed
by the increased ability of the HNB to control the scheduling

978-1-4244-4148-8/09/$25.00 2009
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2009 proceedings.

of UEs in its vicinity. Interestingly, of the four throughput

plots in Figs. 8 through 11, Fig. 10 is the only plot in which
macro-connected UEs experience a small decrease in average
throughput as the HNB access becomes more open. In fact,
each UE that remains connected to the macro base station
experiences a relatively small increase in HSUPA throughput
with increasingly open access. This small increase is due to
a modest decrease in uplink interference levels because with
more open access there is a smaller load on the macro base
stations (the decrease is modest because the fraction of overall
UEs using HSUPA is only 10%). Despite this, the plot shows
an overall decrease in macro UE HSUPA throughput because
more open access moves increasing numbers of neighboring,
high-rate HSUPA UEs onto the HNB. The loss of these UEs
from the pool of macro-connected UEs used in computing
the averages shown in the plot outweighs the improved interference effects noted above. Nevertheless, while the plot
is therefore accurate, it does not explicitly convey the fact
that from the standpoint of any specific UE using HSUPA in
this scenario, the throughput will generally improve with more
open access.

[7] , Simulation results for home NodeB to macro UE downlink coexistence within the block of flats scenario, 3GPP, Sorrento, Italy, 3GPP
document reference R4-080151, Feb. 2008, 3GPP TSG RAN WG4
Meeting 46.
[8] 3GPP TR25.8200 v1.0.0 release 8 (11-2007).
[9] D. Lopez-Perez, A. Valcarce, G. de la Roche, and J. Zhang, Ofdma
femtocells: A roadmap on interference avoidance, Communications
Magazine, IEEE, vol. 1, no. 1, pp. 1112, June 2009.
[10] A. Valcarce, D. Lopez-Perez, G. de la Roche, and J. Zhang, Limited
access to OFDMA femtocells, in Proc. of PIMRC 2009, vol. 3, Sep.
2009, pp. 22512255.
[11] D. Lopez-Perez, A. Valcarce, G. de la Roche, E. Liu, and J. Zhang,
Access methods to WiMAX femtocells: A downlink system-level case
study, in Proc. of ICCS 2008, vol. 1, Nov. 2008, pp. 16571662.
[12] D. Lopez-Perez, G. de la Roche, A. Valcarce, A. Juttner, and J. Zhang,
Interference avoidance and dynamic frequency planning for WiMAX
femtocells networks, in Proc. of ICCS 2008, vol. 1, Nov. 2008, pp.
15791584.
[13] D. Choi, P. Monajemi, S. Kang, and J. Villasenor, Dealing with loud
neighbors: The benefits and tradeoffs of adaptive femtocell access, in
Proc. of Globecom, 2008.
[14] J. Laiho and A. Wacker, Radio network planning process and methods
for WCDMA, Annals of Telecommunications, vol. 56, no. 5-6, pp. 317
331, 2001.

IV. C ONCLUSIONS
We have examined the impact of HNB access policy as it
relates to network traffic type as well as to the type of base
station (HNB or macro) to which the UEs are connected. In
particular, it was shown that for HSUPA, more open access
provides the HNB with increased opportunities to exert control
over the scheduling of HSUPA transmissions for UEs in its
vicinity, thus reducing the amount of interference from nearby
UEs not connected to the HNB and leading to higher overall
packet success rates. HNB access policy has a much less
significant impact on the PSR of UEs connected to the macro
base station, as the UEs connected to the HNB transmit with
much lower power and thus provide less interference.
There are many potential future avenues of investigation,
including a more comprehensive examination of the effect
of different relative mixes of HSDPA, HSUPA, and voice
traffic, different UE and base station locations, and different
scheduling mechanisms.
R EFERENCES
[1] Overview of 3GPP release 9 v0.0.6 (2009-06).
[2] Nortel, Vodafone, Open and closed access for home NodeBs, 3GPP,
Athens, Greece, 3GPP document reference R4-071231, Aug. 2007,
3GPP TSG-RAN WG4 Meeting 44.
[3] Nokia Siemens Networks, Initial home NodeB coexistence simulation
results, 3GPP, Orlando, USA, 3GPP document reference R4-070902,
Jun. 2007, 3GPP TSG-RAN WG4 Meeting 43bis.
[4] Orange, Home BTS consideration and deployment scenarios for
UMTS, 3GPP, Kobe, Japan, 3GPP document reference R4-070825, May
2007, 3GPP TSG RAN WG4 Meeting 43.
[5] Q. Europe, 3GPP home NodeB interference analysis, 3GPP, Sorrento,
Italy, 3GPP document reference R4-080329, Feb. 2008, 3GPP TSG RAN
WG4 Meeting 46.
[6] Ericsson, Simulation results for home NodeB uplink performance in
case of co-channel deployment within the block of flats scenario, 3GPP,
Sorrento, Italy, 3GPP document reference R4-080153, Feb. 2008, 3GPP
TSG RAN WG4 Meeting 46.

978-1-4244-4148-8/09/$25.00 2009
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2009 proceedings.