2005 IEEE 16th International Symposium on Personal, Indoor and Mobile Radio Communications

Inter-System Cell Reselection Parameter Optimization

in UMTS

Andrea Garavaglia Christopher Brunner', Dino Flore', Ming Yang', Francesco Pica
,

'QUALCOMM Incorporated 2QUALCOMM GmbH

5775 Morehouse Drive Nordostpark 89
San Diego CA 92121 USA
- - 90411 Nuemberg Germany -

{andrea, chris, dino, myang, fpica}@qualcomm.com

Abstract-With WCDMA networks being deployed in Europe and scenario specific parameter recommendations. Conclusions are
throughout the world, one of the big challenges is to make cell drawn in Section VII.
reselection and handover between GSM and WCDMA work
smoothly. Inter-system cell reselection between WCDMA and
GSM enables the user equipment in idle mode to choose a new cell IT. INTER-SYSTEM CELL RESELECTION
of another system to camp on, thus providing service availability A. WCDMA to GSM (W2G) cell reselection
when WCDMA coverage fades out. In this paper, we study
inter-system cell reselection parameter settings by processing field As specified in [1], during idle mode operations, the UE
measurement data collected in different networks. Performance operates in discontinuous reception (DRX) mode to improve its
metrics (relating to service availability and WCDMA idle mode stand-by time, i.e. reacquiring the serving cell each DRX cycle.
coverage) are computed with our simulation platform to evaluate Depending on the measured serving cell quality Qqualmeas, the
different sets of parameters in WCDMA-network-boundary, UE may trigger intra-frequency, inter-frequency, or
coverage-hole, and entering-a-building scenarios. Based on the inter-system measurements and evaluate the respective cell
simulation results, we discuss optimization trade-offs and reselection criteria. In particular, inter-system measurements are
recommend a set of parameters for each scenario. triggered if
Qqualmeas < Qqualmin + SsearchRA T (1)
I. INTRODUCTION
or if the serving cell does not fulfill the cell selection suitability
In WCDMA, the user equipment (UE) shall regularly search criterion for consecutive Nserv DRX cycles [2]. According to
for a better cell to camp on according to the cell reselection the standard [I], a WCDMA cell is suitable if
criteria [1]. This mechanism, based on parameters provided by
the network, is needed to camp on a good quality cell, and Srxlev > 0 and Squlal > 0 (2)
therefore achieve the desired call setup performance.
TABLE 1. WCDMA TO GSM CELL RESELECTION PARAMETERS AND
In a recent study [6], intra-frequency cell reselection MEASUREMENT QUANTITIES
parameter settings have been optimized for standby time and
Quantity Description Source
camping cell quality. Following the same approach, we focus on
inter-system cell reselection in idle mode, required at the border SsearchRA T System specific threshold [dB] broadcast
of WCDMA coverage to provide seamless service availability Qqualmin Minimum cell quality [dB] broadcast
to the UE by reselecting to GSM. To provide recommendations
for inter-system cell reselection parameter settings, extensive Qrxlevmin Minimum cell RX level [dBm] broadcast
channel measurements from the field have been processed with Pcomp Power compensation [dB] broadcast
our simulation platform. Three scenarios were selected, where
inter-system reselections are expected and which differ in radio Q4hst1, Serving cell RSCP hysteresis [dB] broadcast
frequency (RF) environment and in user mobility: a QofJ&et1,, Neighbor cell RSCP offset [dB] broadcast
network-boundary scenario, a coverage-hole inside the Treselection Cell reselection timer [s] broadcast
WCDMA network and an entering-a-building scenario. Since
the parameters involved in the cell reselection procedures are Qqutalmeas Common pilot channel (CPICH) Ec/No [dB] measured
broadcast by the network and could be set on a per cell basis to Qrxlevmeas Common pilot channel (CPICH) RSCP [dBm] measured
enable a finer optimization of different scenarios, we look at the RLA Received signal Level Averaged [dBm] measured
scenarios separately.
Squtal Qqiualmeas- Qqiualmin [dB] calculated
The paper is organized as follows. Section II provides an
Srxlev Qrxlevmeas - Qrxlevmin - Pcomp [dB] calculated
overview of the cell reselection mechanism and parameters.
Field measurements are described in Section III while the
simulation platform is introduced in Section IV. Performance When GSM cells are measured, the cell reselection criterion
metrics are described in Section V. In Section VI, we present compares the CPICH RSCP of the WCDMA serving cell and
simulation results for different scenarios which motivate monitored cells with the RLA of the GSM cells [2]. If a GSM
cell is ranked higher than any WCDMA cell

978-3-8007-2909-8/05/$20.00 ©2005 IEEE 1636

 2005 IEEE 16th International Symposium on Personal, Indoor and Mobile Radio Communications
RLA - QoffsetlGsm > Qrxlevmeas(serving cell) + Qhyst1,
RLA - QoffsetjGsM> Qrxlevmeas(monitored cells) - Qoffset1n
for Treselection seconds, the UE reselects to that GSM cell.
B. GSM to WCDMA (G2W) cell reselection
During idle mode operations in GSM, WCDMA cells are
measured when the RLA of the serving cell is lower (or higher,
depending on the setting) than the parameter Qsearch l [5]. If
measurements are made, the reselection criterion is evaluated
and the UE reselects to a suitable WCDMA cell if the following
conditions hold for a period of 5 seconds [3-5]:
* the CPICH RSCP of the candidate WCDMA cell is at least
FDD_Qoffset dB better than the RLA of the serving and of
all suitable non-serving GSM cells
* the CPICH Ec/No of the candidate WCDMA cell is equal
or greater than the parameter FDD_Qmin
TABLE II. GSM TWCDMA CELL RESELECTION PARAMETERS
Quantity J Description Source
FDD_Qmin Minimum CPICH Ec/No threshold [dB] broadcast
Qsearch_I WCDMA measurement threshold [dB] broadcast
FDD Qo]fset RLA offset for WCDMA ranking [dB] broadcast
Although the above conditions may be fulfilled, the
WCDMA cell might not meet the suitability criteria of equation
(2) if the CPICH RSCP is too weak, thus causing the UE to
return to GSM ("G2W-reject"). During a G2W-reject, a loss in
service availability may occur (i.e. pages are not received nor is
access possible during the time the UE attempts the reselection
to WCDMA). If, on the other hand, the above conditions are
fulfilled and a suitable WCDMA cell is found, the phone may
return to GSM after a short time ("ping-pong"), also impacting
service availability. In fact, each time an inter-system cell
reselection takes place, the UE has to perform location area and
routing area updates. These procedures can last several seconds
and during this time the service will not be available.
The outlined cell reselection scheme is based on the 3GPP
Release 99 standard. A fixed RSCP based threshold for G2W
reselection has been added, as optional UE feature, to the
standard [5] to mitigate some of the parameter setting
constraints. Because most UEs in the market do not implement it,
this feature is not considered in the sequel.
III. FIELD MEASUREMENTS
For the purpose of this study, field measurements have been
collected in three different inter-system handover scenarios in
commercial UMTS networks in Europe. The networks were
lightly loaded and characterized by a low base station density.
Both factors lead to higher Ec/No measurements for any given
RSCP and vice versa. The network-boundary scenario is
characterized by the loss of WCDMA coverage when leaving
the WCDMA region. The coverage-hole scenario represents
areas inside the WCDMA region that are subject to bad outdoor
WCDMA coverage. The entering-a-building scenario implies
bad indoor WCDMA coverage within a WCDMA area.
978-3-8007-2909-8/05/$20.00 ©2005 IEEE
(3a) For each scenario, several runs along a metric route selected
to best characterize the specific deployment scenario have been
(3b) made to collect measurements. Table III classifies the scenarios
in terrns of RF environment and mobility.
TABLE III. CLASSIFICATION OF MEASURED DATA
Scenario RF environment and mobility
* WCDMA coverage with high Ec/N.o for a given RSCP
(fewer base stations at boundary to cause interference: less
interference)
* Gradual fading of WCDMA signal strength when moving
Network- into GSM only coverage
Boundary * Sub-urban area, Vehicular channel, average speed of 20
km/h
* Dominant pilot in WCDMA coverage
* Measurements for the G2W direction are obtained by
reversing the time axis (see Section IV)
WCDMA coverage with low Ec/No for a given RSCP (base
*
stations all around: more interference)
Coverage- * Urban environment, Vehicular channel, average speed of 25
Hole km/h with large speed variations between runs
* Measurements are collected while crossing the hole area in
both forward and reverse directions
* WCDMA coverage with medium Ec/No for a given RSCP
(base stations all around but UE is closer to one: medium
level of interference)
* Entering and walking down a large indoor corridor of a large
commercial building
Entering-
a-building * Abrupt decrease in WCDMA signal strength when moving
from WCDMA into GSM only coverage
* Dominant pilot in WCDMA coverage, Pedestrian channel,
average speed of 3 km/h
* Measurements from indoor (GSM) to outdoor (WCDMA)
are obtained by reversing the time axis (see Section IV)
IV. SIMULATION PLATFORM
Our simulation platform imports channel measurement data
from a test phone. A standard compliant UE performing
intra-frequency and inter-system cell reselection is emulated
and the output is aligned with the real UE performance.
Only WCDMA channel measurements are imported, while
we assume a constant good coverage with an RLA of -75 dBm
for the GSM system. Moreover, we do not model different GSM
cells. As a consequence, cell reselections within the GSM
system, which may delay G2W reselection, are not modeled
either. The impact on the results should be very limited due to
the larger size of GSM cells in the 900 MHz band and due to the
fact that only a limited portion of the metric route is recorded
outside of WCDMA coverage.
Intra-system cell reselections and bad RF conditions can
potentially delay the completion of the location area and routing
area update procedures. For simplicity, we assume a constant
delay obtained by averaging the values extracted from the field
data.
1637
 2005 IEEE 16th International Symposium on Personal, Indoor and Mobile Radio Communications
The features listed below significantly reduce the amount of
drive testing required for sufficient statistics (conventionally,
performance metrics are collected once every DRX cycle
directly from the UE):
* Channel measurements are logged at a much higher rate
than one sample per DRX cycle. By shifting the starting
point by a few milliseconds (i.e. longer than the channel
coherence time) and sampling once per DRX cycle, we
gain many realizations of the measured multi-path channel
from one measurement run.
* Different parameter settings can be evaluated based on the
same field data.
* The time axis of the field data can be reversed to emulate
an identical measurement run in opposite direction. This
facilitates a side-by-side comparison between the G2W
and W2G cell reselection schemes.
V. PERFORMANCE METRICS
The two primary metrics we want to consider are the
WCDMA idle mode coverage and the service availability.
The WCDMA idle mode coverage corresponds to the
percentage of the route the UE is camped on WCDMA cells. A
large WCDMA coverage is desirable (consistently aligned in
idle and connected mode) to e.g. make new UMTS services
widely available.
Conceming the WCDMA service availability, we make use
of the following indirect metrics:
* Lower 1%-tile of the WCDMA camping cell Ec/No and
RSCP collected over the metric route. Low Ec/No and
RSCP may lead to missed paging occasions and access
failures, respectively.
* Inter-system cell reselections and G2W-rejects impact
service availability since neither paging nor access work
for the duration of both procedures (several seconds).
Finally, we count the W2G reselection trigger causes, i.e.
unsuitability defined in (2) and measurement rules defined in (1).
These indicate which parameters play a significant role in
different loading conditions and RF environments.
VI. SIMULATION RESULTS
Since the measurement data was collected in low site density
networks under light load (cf. Section III), W2G reselections are
more frequently triggered by unsuitability due to low RSCP
rather than by the measurement rules. Accordingly, the
parameter Qrxlevmin is the main driver for the W2G reselection
while Qqualmin only plays a secondary role. Favoring
WCDMA, the parameters Qsearch_I and FDD Qoffset have
been set such that WCDMA cells are always measured and
reselected as soon as they meet the quality criteria. As a
consequence, the G2W reselection is mainly driven by the
parameter FDD Qmin. As indicated in Table IV, Qrxlevmin
settings of -1 15, -1 3, and -1 1 1 dBm are tested together with
FDD_Qmin settings of -12, -10, and -8 dB.
978-3-8007-2909-8/05/$20.00 ©2005 IEEE
TABLE IV. PARAMETER SET FOR INTER-SYSTEM CELL RESELECTfON
Parameter Setting Comments
-115. -113,--111 In this RF, main driver for W2G
Qrx/evmin LdBm] reselection
To avoid ping pong:
SsearchRAT+
-14 [dB] FDD_Qmin > SsearchRA T+
Qqzialmin Qqualmin
~~~~~~Qquialmin
In this RF, secondary driver for
Qqztalmin Qqualmin -18 [dB] ~W2G reselection
Qoffset1,+Qhvst1s 3 [dB] Equation (3) is generally
fulfilled
Treselection I [s] Common to intra-frequency cell
reselection
FDD_Qmin -12, -10, -8 [dB] Drives G2W cell reselection
Qsearch I Always measure W To prioritize WCDMA
FDD_Qoffset -infinite To prioritize WCDMA
A. Network-Boundary Scenario
In the network-boundary scenario, the route starts in good
and ends in poor WCDMA coverage, while the GSM coverage
remains strong all the way. Here, "WG -G" denotes the
direction in which WCDMA coverage deteriorates and
"G ->WG" the reverse direction. As indicated in Figure 1, the
WCDMA quality and strength are low in WG - G direction. In
G --WG direction, the WCDMA quality is better because of the
high FDD Qmin settings, which have to be exceeded for 5s
before reselecting to WCDMA. The downside is that the UE
stays in GSM longer (cf. Figure 3). Figure 2 shows that a high
FDD_Qmin is the most effective way to reduce inter-system
reselections and G2W-rejects, thus improving service
availability at the expense of a slight reduction in WCDMA idle
mode coverage. A decrease in Qrxlevmin increases the number
of inter-system cell reselections in G - WG direction because of
fewer G2W-rejects, and decreases the serving cell strength.
Therefore, we recommend Qrxlevmin = -111 dBm and
FDD Qmin = -8 dB. Figure 3 indicates that in WG ->G
direction most reselections are triggered by unsuitability due to
low RSCP. In G - WG direction, W2G reselections (statistically
insignificant) are neither caused by unsuitability nor by
measurement rules but by missing channel measurements.
B. Coverage-Hole Scenario
In the coverage-hole scenario, the route starts and ends in
points with good WCDMA coverage, denoted by A and B,
while crossing an area with poor WCDMA coverage. Here, we
use different channel measurements in A-B and B-4A
directions instead of reversing the time-axis, and results are
provided for both directions. As indicated in Table III, due to the
high interference, Ec/No is low for a given RSCP. Accordingly,
W2G reselections triggered by the measurement rules are more
significant (cf. Figure 6). Because of the relatively high RSCP
levels corresponding to a certain Ec/No (cf. Figure 4), it is
possible to use a low Qrxlevmin setting to effectively reduce
inter-system reselections and G2W-rejects and to increase the
WCDMA idle mode coverage. Therefore, a Qrxlevmin setting
of -115 dBm is recommended. We also recommend a low
setting of FDD_Qmin (-12 dB) to increase the WCDMA idle
1 638
 2005 IEEE 16th International Symposium on Personal, Indoor and Mobile Radio Communications

mode coverage at the expense of a tolerable reduction in serving network-boundary scenario

cell quality. ............ ....... ....... ....... ....... ..........................

~~G.tWG .1 .111)
.8 .... .... .

G.'W G-(>WG,
--1 l)+
C. Entering-A-Building Scenario .9
In the entering-a-building scenario, the results are given for -10, ...............
G->WG (-12, -1 13).
......................
. .........................
1 -11
................. ....

WG -G (i.e. entering the building) and G -WG directions (i.e. C->WO (- - 2.

o -11 ...... .... .....
exiting the building) along the same route. The scenario is .
....... ........................ ........................

LU .12
similar to the network-boundary case. In both cases, 0

inter-system reselection is not avoidable. Due to RF, the = -133

WCDMA signal deteriorates more abruptly here but pedestrian *I1 -14 :1,0,. ING>.G..
1....:
1)
............ ...- ...

:..... .......
user velocity compensates for the quick drop. As indicated in
Figure 7, a medium interference level and Ec/No values can be -15 ... ...

observed, due to one dominant pilot and several nearby cells. -16 .. *.W13>G(1.
6 113i)
The tail of the RSCP distribution is relatively low on the -17 1...........
WG -G direction especially for low Qrxlevmin settings. *WG- >G (- 12, l1l5......
5 -
-i
-lb
Therefore, as in the network-boundary scenario, a Qrxlevmin of 22 -120 -118 -116 -114 -112 -110 -108 -106 -104 -102 -100
-11 ldBm is recommended. Figure 8 indicates that a high 1 %-tile of RSCP
FDD Qmin is most effective in reducing inter-system Figure 1: WCDMA serving cell quality in dB versus strength in dBm
reselections and G2W-rejects while maximizing the service network-boundary scenario
availability. In addition, a high FDD Qmin = -8 dB improves 07r ............................. .... .......... .......

the tail of the RSCP distribution at the expense of a slight 4G->WG (-12, -1:1 1 )
reduction in WCDMA idle mode coverage, especially in indoor
areas. As in the network-boundary scenario, most reselections in
0.5-
WG -G direction are triggered by unsuitability (cf. Figure 9).
Note that a stationary indoor UE close to a site may experience - (-10, 1
.7->WG (-l 1, -1:1 3)
high Ec/Io values for low RSCP and may require an even higher G->W(1S (-I 0, -11 )
FDD_Qmin to avoid frequent G2W-rejects. 0 3 .................... ...

:V9. i- >WG t2, 115)

VII.CONCLUSIONS ...............................
02 WG->G (-12, -1 t1) ...........
......................................................................

We have introduced a concept to generate cell reselection

parameter recommendations that significantly reduces the 0.1 ;>lG it G:-
-.i1i)-
..-Il) WG->G (-0 W111>G
amount of driving test required otherwise. Then we applied this WG->G:(-8, -111) WG->G (-12 -113)
approach to different scenarios. Simulations show that the InIL u
'iG->G (-2. -115i)
126 1.28 13 132 134 1.36 138 1.4
combination of Qrxlevmin = -11 5dBm and FDD_Qmin = -12dB # of inter-system reselections
works well in the coverage-hole scenario, while
Qrdevmin-- 11IdBm and FDD_Qmin = -8dB performs well in Figure 2: Number of G2W-rejects versus number of inter-system cell
both network boundary and entering-a-building scenarios. reselections
If parameters can not be set on a cell basis, the best
compromise would be to apply the most convenient settings network-boundary scenario
considering the entire RNC area (e.g. figuring out which 1 r --... . - -- .................... ........
scenarios and environments would be the most representative
and strategic for that area). 0.98 [

When load increases, a further tuning of the above settings is

necessary to optimize the performance with higher interference 085 a) WG->G (-12,- 15)+
0785
levels. Finally, connected mode parameters shall be considered D

for consistency with idle mode parameters. ao

E 0.7 ..
WG-6 > -12/-010-8, -1. 1)
....... ...

06
REFERENCES E:' 0.75
.G->WG (-12,-1 15)
[1] 3GPP TS 25.304, "User Equipment (UE) procedures in idle mode and .G>1 WG (-12.-113)
0.7
procedures for cells reselection in connected mode"
G->WG(-8 ,C1 -111)
[2] 3GPP TS 25.133, "Requirements for support of radio resource 0.66 ... ...
management"
[3] 3GPP TS 04.18, "Mobile radio interface layer 3 specification, Radio 0.1
Resource Control Protocol" 02 03 0.4 0.5 0.6 07 08 09
0 I
WN2)G tnggered by unsuitability versus total W2G tniggers
[4] 3GPP TS 03.22, "Functions related to Mobile Station (MS) in idle mode
and group receive mode" Figure 3: WCDMA idle mode coverage in percent versus W2G triggered
[5] 3GPP TS 05.08," Radio subsystem link control" by unsuitability relative to all W2G triggers
[6] D. Flore, C. Brunner, F. Grilli, V. Vanghi, "Cell Reselection Parameter
Optimization in UMTS", submitted to ISWCS 2005

978-3-8007-2909-8/05/$20.00 ©2005 IEEE 1639

 2005 IEEE 16th International Symposium on Personal, Indoor and Mobile Radio Communications

covrage-hole sceriario entering-a-boilding scenario

-lb -ii
K I .~~~~~~~~~~~~~B8>A (-8. 11

-16.5 - -.. .. B-.-A)-.10-,--1.111. -12 : tWG1S-1 WG

H -8-113 )

-17
-.B--4j-12 -115~~~~~~~~~~~~B>A:(12 -13)- G S->V(10 111

LU
B->A12, 113)
c'-17.5
-15

-18 A4->B (10,1 2 1)

-18
- 4->B{-212,-113)
-18,5j -17 +%NG ->G(G1 -2 0 -i 2,-ill
-A->8(1 2, -1 15)

19I)-114.5 -114 -113.5 -113 -112.5 I1.12) -111.5 -111

Id
-120
- - 1-
-118
--.,
-118 -114 -112 -110 -108 lO06 -104 -102 -100
1 %-tile of 03CP 1 %-tile sftRSCP

Figure 4: WCDMA serving cell quality in dB versus strength in dBm Figure 7: WCDMA serving cell quality in dB versus strength in dBm

coverage-hole scenario
0.5
entering-a-building scenano
*A->5 (-12, 1li) o>WG3(-12,-111)
0 18 0.45-
0.4-

0.14- 0.35-
o- 03-
co
3~0.25-
c-ri
'2 0-08 t0 092
WG->WG (-12,-113):
*.*A.R(J2.113) ~~~~~~~~B->A
(- 12-1 1 1 0.15
0.1 ...................

WG->G(8,-11) , W >WG
(-12; -115
A*>->8(-10--111)1. 0 05 G 'NG 1 1) WG.->Gi21f -jilt
(.8,
002
r L . -V v k-'- , -, ir u i
2 2.5 3 3.5 1.1S 1-2 1 .25 1.3 1.35 1.4
# of inter-system resetectrons #of inter-system resietections

Figure 5: Number of G2W-rejects versus number of inter-system cell Figure 8: Number of G2W-rejects versus number of inter-system cell
reselections reselections

coverage-hole scenario entering-a-buildirig scesaris

098 .
~~~~~~~~A->
8(12. -1:15)*
A->B (-12, -113Yi 0 98
0.98 ............12...1..
0.92
56 0.94 ........ .. ........ .... A->8-(-8,47 11),4 WG->G (-12,-1I13).
< 0884
25CD
E5 0 92 .. ..
...

0 8
4-B5>A(1 2,. 115) 8.2 G-WG (12. -:115) -
2o 0~9 ... -.. A...-.. .....113). 0
-*G.-->WGf(-12, 113)
5.' *~~~~~~~~~B->A(-12,~-111) *G4>WG (1.l2, -1 11)
*8> A(110,- 1 11)

*B->A (-8,-.11 1)
,G->WGS (-8, -11 1)
n R
0.6
0.4 0.5 06 0.7 0.8 0.9 0.2 0 0.3 0.4 0-1 0.5 086 0.7 0.8 0.9 1
WV2i triggered by unsuitatility per total W2Gl triggers iAV2G triggered by orisoitability versos total WN2G triggers
Figure 6: WCDMA idle mode coverage in percent versus W2G triggered Figure 9: WCDMA idle mode coverage in percent versus W2G triggered
by unsuitability relative to all W20 triggers by unsuitability relative to all W2G triggers

978-3-8007-2909-8/05/$20.00 ©2005 IEEE 14

1640