1. GSM FUNCTIONALITY AND
PARAMETER FINE-TUNING:
A CASE STUDY
Issue Date: September 2004
Abstract—This paper documents the effect on the performance of an Ericsson™ global system for mobile communication (GSM) network
realized by evaluating functionality and fine-tuning parameters after completing the pre- and post-launch optimization phases. These
actions were carried out during a base station subsystem (BSS) performance optimization project attempting to further improve the quality
of service (QoS) because network traffic was increasing. Each feature changed is addressed separately, followed by a short technical
description of the philosophy of the changes performed, the exact settings selected, and a statistical evaluation of the results. This paper
demonstrates that network performance gains can be achieved from optimum use of the available functionality and parameter tuning. This
paper can also serve as a reference for optimizing Ericsson CME201 systems.
INTRODUCTION Ongoing changes in functionality and parameter
S ince first deployed in 1992, European global settings are necessary to provide optimum and
system for mobile communication (GSM) constant quality of service (QoS). All system
networks have become a major commercial vendors continuously seek to improve
success. Currently, penetration levels approach functionality by adding improved features with
100 percent in some European countries. The every base station subsystem (BSS) software
rapid increase in subscriber numbers prompted release. If fully exploited, this continuous
network operators to increase investment in evolution of functionality can result in
network infrastructure by embarking on significantly improved QoS and more efficient
aggressive network rollout projects aiming to use of network infrastructure.
expand system coverage and capacity. Increasing This paper describes a review of the functionality
demand for mobile services and competition for and parameter values of an Ericsson™ GSM
market share led many operators to dedicate network containing approximately 150 base
most of their resources to network deployment. transceiver stations (BTSs) with three cells per
Under these circumstances, the industry BTS and main sector configuration of 0-120-240
eventually developed the mindset of “roll out degrees [1]. This review began as an optimization
now and optimize later.” project 6 months after completion of the post-
Until the late 1990s, most network operators launch optimization phase. During this period,
competed purely on coverage, considered the traffic increased substantially and the network
most important differentiator among the offered was expanded to satisfy capacity demand as well
services. Having “signal bars” on phones was all as to extend coverage.
that mattered, even though calls were failing in
A number of features were evaluated and fine-
many cases. Users expected to see signal bars on
tuned. These features are listed below, followed by
their phones everywhere. Focused mainly on an
a short technical description of each and the
agressive buildout strategy, several operators
philosophy of the performed changes, the exact
continued to use default parameter values
settings selected, and a statistical evaluation of the
without fully exploiting the available function-
results. The seven functions discussed below apply
ality of the given system.
to Ericsson BSS R8 and R9 (Releases 8 and 9).
Network pre- and post-launch optimization is a
• Frequency hopping
useful mechanism to ensure good performance
after commercial launch of the service. However, • Mobile station dynamic power control
Michael Pipikakis
as the network expands and traffic increases, the • Cell load sharing
dmpipika@bechtel.com benefits of post-launch optimization may be lost.
• Locating penalty timers
________________________________
1 Ericsson’s GSM application • Flow control timers
© 2004 Bechtel Corporation. All rights reserved. 17

2. ABBREVIATIONS, ACRONYMS, AND TERMS
BCCH broadcast control channel GSM global system for mobile
communication
BSC base station controller
HSN hopping sequence number
BSIC base station identity code
MAHO mobile assisted handover
BSS base station subsystem
MS mobile station
BTS base transceiver station
QoS quality of service
C/I carrier-to-interference (ratio)
SACCH slow associated control
CLS cell load sharing channel
CP central processor SDCCH standalone dedicated control
CTR cell traffic recording channel
DCR dropped call rate TCH traffic channel
DPC dynamic power control UL uplink
FH frequency hopping
ERICSSON PARAMETERS
ACCMIN minimum signal strength to PSSBQ penalty value for bad quality
access the cell PSSHF penalty value for failed
BSRXSUFF received by the BTS handover
sufficient signal strength PTIMBQ penalty timer for bad signal
level quality
CLSACC CLS traffic accept PTIMHF penalty timer for handover
CLSLEVEL CLS level failure
EVALTYPE evaluation type QCOMPUL uplink signal quality
compensation factor
HIHYST high-signal-strength hysteresis
QDESUL quality desired for uplink
HODWNQA handovers due to downlink
signal quality RHYST region hysteresis
HOTOKCL handovers to K cells RXLEV measured signal strength
level
HOTOLCL handovers to L cells
RXQUAL measured signal quality
HOUPLQA handovers due to uplink
signal quality SSDES signal strength desired
HYSTSEP signal strength level between TALLOC time between TCH
high and low strength cells allocations
KHYST K-criterion hysteresis TCALLS counter for TCH allocation
attempts
LCOMPUL uplink signal strength
compensation factor TCONGAS congestion timer for immediate
TCH assignments
LHYST L-criterion hysteresis
TCONGHO congestion timer for handover
LOHYST low-signal-strength TCH assignments
hysteresis
TURGEN time for urgent handover
MSRXSUFF received by the mobile
sufficient signal strength level
• Cell selection and access
different frequencies. Each cell uses a predefined
• Signal strength measurement criteria in the set of frequencies, among which the connection
locating algorithm hops according to a specified pattern (i.e., cyclic
or random) 217 times per second. The radio
environment between a mobile station (MS) and a
FREQUENCY HOPPING BTS is subject to variations due to multipath
F requency hopping (FH) means that multiple
frequencies are used to transmit speech or
data in a single connection. The basic principle
fading and cumulative interference. FH can
improve the radio environment, providing
frequency diversity against the multipath fading
involves transmitting consecutive bursts at and averaging the overall interference. See [2].
18 Bechtel Telecommunications Technical Journal

3. Parameter Adjustment and Evaluation
Cyclically sequenced baseband FH was 2.3
introduced at launch in traffic channels (TCHs)
2.0
and standalone dedicated control channels
(SDCCHs). With this pattern, all available 1.7
DCR (%)
frequencies of a cell are used with a consecutive
order in a call or signaling connection. For 1.4
instance, a connection in a three-frequency (f1, f2,
f3) cell will show the following burst-to-burst 1.1
pattern:
0.8
…f3, f2, f1, f3, f2, f1, f3, f2, f1, f3, f2, f1,… Cyclic FH Random FH
0.5
With reuse-pattern frequency planning, cyclic 1/12 1/22 2/1 2/11 2/21 3/3 3/13 3/23 4/2 4/12
Date (MM/DD)
hopping may result in connections in cells that
are reusing the same frequencies to get in phase Figure 1. Network DCR After Implementation of Random Hopping Sequence
with one another, hopping “hand in hand” but both TCHs and SDCCHs. The algorithm
losing the benefit of interference averaging. calculates a power order according to BTS
received signal strength and BTS measured
Random FH was proposed, which introduces a
quality. The first term introduces MS power
pseudo random hopping sequence, according to
reduction based on a desired value—signal
parameter hopping sequence number (HSN). Up
strength desired (SSDES)2. The second term
to 63 different FH patterns not correlated with
introduces compensation for bad quality,
one another can be defined. The burst-to-burst
according to a desired value for signal quality—
pattern would look as follows:
quality desired for uplink (QDESUL)2. The MS
…f3, f1, f2, f2, f1, f3, f3, f2, f1, f1, f2, f1,… power capabilities are a limiting factor. The MS
power cannot be reduced beyond the minimum
Carefully choosing HSN values for cells using the
output power of the MS (for phase 2 MSs, the
same frequency groups was expected to increase
dynamic power range is 8 dBm to 33 dBm).
the interference averaging gains of FH.
Parameter Adjustment and Evaluation
Random FH was introduced in all cells, with an
MS DPC was initially introduced with the
HSN per cell based on the base station identity
following settings for desired values and
code (BSIC) plan (HSN = 63 – BSIC), which was
weighting factors:
also planned to differentiate between co-channel
cells. • SSDES = –94 dBm
Old values: HOP = ON, HSN = 0 => Results in • QDESUL = 10
cyclic hopping (HOP is the Ericsson cell level • Uplink signal strength compensation factor
Considerable
parameter to enable hopping) (LCOMPUL)2 = 50 network
New values: HOP = ON, HSN = (63 – BSIC) => • Uplink signal quality compensation factor performance
Results in random hopping (QCOMPUL)2 = 30 gains can be
As can be seen from the results in Figure 1, a These initial values correspond to an aggressive made by fully
considerable improvement in QoS was achieved. power-down regulation aiming to minimize utilizing the
The dropped call rate (DCR) decreased by uplink interference. However, it was observed
available
approximately 20 percent. from analyzing drive test files and cell traffic
recording (CTR) files that the settings could lead functionality
to performance deterioration. For instance, a and fine-tuning
MOBILE STATION DYNAMIC POWER CONTROL connection with received signal strength the network
(RXLEV)2 = –80 dBm and received signal quality
M S dynamic power control (DPC) is a feature parameters
(RXQUAL)2 = 5, given the previous settings,
that controls the output power of an MS so
would be further down-regulated in steps of using statistics
that the BTS receives a desired uplink signal to evaluate
2 dB, despite the obvious quality problem.
strength level. MS DPC helps reduce MS battery
consumption, protects against possible BTS After studying the case, a more reasonable value
the results.
receiver saturation, and reduces overall uplink of SSDES = –88 dBm was introduced, while
interference. QDESUL was set to zero. Also, compensation
factor LCOMPUL, which introduces a slope in the
The MS DPC algorithm is implemented on the
_________________________________
base station controller (BSC) and performed for
2 An Ericsson DPC parameter
September 2004 • Volume 2, Number 2 19

4. power reduction, was set to 100. This setting optimization activities. The minute-Erlang per
corresponds to maximum uplink regulation (no drop index is inversely proportional to the
slope) because the algorithm was expected to DCR index.
work rapidly on “good” signals. Quality
Due to the new settings for SSDES, the average
compensation parameter QCOMPUL was set to
power received on the uplink is greater than
60 to enhance up-regulation in case of inter-
before, so the risk of a connection dropping due to
ference and to give the connection a chance to
weak signal strength on the uplink should
overcome the bad quality by increasing the
decrease. Since the main reason for uplink quality
output power. For a more detailed description of
is also believed to be the strength of the MS
the algorithm, see [3].
transmitted signal power, bad quality drops on the
Figure 2 shows the positive effect of the changes uplink should also decrease with the new settings.
on the MS DPC settings in terms of dropped
In Figure 3, the improvement trend can be
connections due to uplink quality and uplink
verified by examining the handover reasons due
signal strength. The indices “min-ERLANG/
to uplink (UL) quality.
UL_QUAL-DROP” (minutes of traffic carried
before a call drop due to uplink signal quality
occurs) and “min-ERLANG/UL_SS-DROP”
CELL LOAD SHARING
(minutes of traffic carried before a call drop due
1,400
to uplink signal strength occurs) were used.
300
C ell load sharing (CLS) is a feature that
distributes traffic among neighboring cells at
high traffic load to reduce congestion and better
minErl/UL_QA_DROP
1,300
280 use the available resources.
minErl/UL_SS_DROP
1,200 The CLS algorithm works by monitoring traffic
Number of Minutes
minErl/UL_SS_DROP
260
1,100 load for every cell in terms of idle TCHs. When
1,000 240 the number of idle TCHs in a given cell,
expressed as a percentage of the total, falls below
900 220
the CLS level (CLSLEVEL)3, traffic is shifted from
800
200 this cell to prevent it from being congested.
700 Connections close to the cell border, within an
180 area determined by region hysteresis (RHYST)3,
600
Old MS DPC Settings New MS DPC Settings are handed over to any neighboring cell
500 160
5/5 5/7 5/9 5/11 5/13 5/15 5/17 5/19 5/21 5/23 5/25 5/27 considered suitable to accept traffic, i.e., whose
Date (MM/DD) percentage of idle TCHs is greater than the value
Figure 2. Effect of New MS Power Control Settings
CLS traffic accept (CLSACC)3.
The indices presented in Figure 2 are TCH drops Drawbacks of the feature are the increased
due to bad uplink quality and low uplink signal number of handovers and a considerable increase
strength related to the traffic carried by the in BSC central processor (CP) load. For a detailed
system. The indices “minErl/UL_QA_DROP” description of the functionality and algorithm,
and “minErl/UL_SS_DROP” express the minutes see [4].
of traffic the system carries before a drop occurs
due to bad uplink quality or low uplink signal Parameter Adjustment and Evaluation
strength. The minute-Erlang method was used The CLS feature was introduced networkwide to
because it is more sensitive to changes and thus cope with unevenly distributed traffic among
more accurately evaluates the effectiveness of cells, to use the available resources efficiently,
24 and to increase the total capacity. The
New MS DPC Settings original parameter set was CLSLEVEL = 23,
23 CLSACC = 55, and RHYST = 75, meaning that
22 CLS evaluations for a cell started when the idle
number of TCHs fell below 23 percent, while a
(%)
21
cell accepted CLS traffic only if 55 percent or
20 more of its resources were idle.
19 Statistical analysis indicated that with these
settings, the success rate of CLS handovers was
18
4/25 5/2 5/9 5/16 5/23 5/30
poor, mainly because of the high values of
Date (MM/DD) _________________________________
Figure 3. Effect on Handovers due to UL Quality of new MS Power Control Settings 3 An Ericsson CLS parameter
20 Bechtel Telecommunications Technical Journal

5. CLSLEVEL and CLSACC. A more reasonable
setting was introduced, where a cell would more 1,000,000
CLS Attempts
easily accept CLS handovers (CLSACC = 25) and CLS Success
would not start CLS evaluations as soon 800,000
Number of Attempts
(CLSLEVEL = 15). Also, RHYST was set to 100, New CLS Settings
maximizing the area around the nominal cell 600,000
border where CLS could take place.
In Figure 4, the impact of the change can be seen. 400,000
Cell load sharing became more effective, since
CLS calculations were limited, practically 200,000
maintaining the same number of successful CLS
handovers. This development had a positive
0
effect on the BSC CP load. 3/6 3/13 3/20 3/27 4/3 4/10 4/17 4/24 5/1 5/8 5/15 5/22 5/29
Date (MM/DD)
Figure 4. Effect of CLS Parameter Changes
LOCATING PENALTY TIMERS originating channel within 10 seconds. Hand-
P enalty timers for bad signal quality over success improved as a direct result of
(PTIMBQ)4 and for handover failure reducing the possibility of attempted connections
(PTIMHF)4 specify the time in seconds for which to a cell suffering from poor quality.
the respective penalty values in decibels, namely The reduction of mobile connections lost during
penalty value for bad quality (PSSBQ)4 and handover can be seen in Figure 6. In addition to
penalty value for failed handover (PSSHF)4, are the improvements in ping-pong effect and
applied to a cell’s neighbors. handover success rate, the timer change also
When an urgent handover is successfully had a positive effect on network dropout
performed that resulted from bad quality due to performance. In a typical GSM network, nearly
downlink, uplink, or both, the originating cell is
penalized with PSSBQ decibels to prevent 96.5 25
immediate hand-back to this cell. The original cell Success
96.0
is penalized because bad radio conditions might Ping-pong 20
Ping-pong Handovers (%)
still be in effect there; also, the original bad 95.5
quality cell is most likely the best cell from
15
a strictly signal strength point of view. Under a 95.0
(%)
similar philosophy, handover to a cell where 94.5 10
a handover failure occurred is inhibited for a time
determined by timer PTIMHF [5]. 94.0
5
Parameter Adjustment and Evaluation 93.5
Penalty Timer Changes
Penalty values PSSBQ and PSSHF were both set 93.0 0
to 50 dB to remove the penalized cells from the 4/17 4/20 4/23 4/26 4/29 5/2 5/5 5/8 5/11 5/14 5/17 5/20 5/23 5/26 5/29 6/1
Date (MM/DD)
locating algorithm evaluations. However, the
Figure 5. Effect of Penalty Timer Changes on Network Handover Performance
lengths of the timers, PTIMBQ = 10 sec and
PTIMHF = 5 sec (original settings), were thought
to be insufficient to give radio conditions in the 0.95
penalized cell a chance to improve. The lengths of 0.90
the two timers must be carefully chosen, on the
0.85
other hand, to predict handover performance of
fast-moving subscribers. A very high value may 0.80
lead to call drops due to handover being inhibited 0.75
(%)
for a time not matching the user’s mobility. The 0.70
new time settings selected were PTIMBQ = 15 sec 0.65
and PTIMHF = 12 sec.
0.60
Figure 5 shows the effect of this change in 0.55
handover performance. The term “ping-pong” Penalty Timer Changes
0.50
indicates the percentage of handovers back to the 4/15 4/25 5/5 5/15 5/25 6/4
_________________________________ Date (MM/DD)
Figure 6. Effect of Penalty Timer Changes on Percent of Mobiles Lost During Handover
4 An Ericsson locating algorithm parameter
September 2004 • Volume 2, Number 2 21

6. 30 percent of the total dropped calls occur during congestion because congestion timers TCONGAS
handover, which is considered a sensitive task in and TCONGHO count every allocation attempt.
the radio environment. By increasing TALLOC, the measured figures for
congestion during handover and assignment
A more reliable way to assess the overall dropout
will be closer to the true, customer-perceived
performance is to determine the MSs lost during
congestion.
handover in relation to the total traffic. This data
is shown in Figure 7, where a clear and steadily Figure 8 shows the measured congestion trend
increasing trend is apparent for the index after the change was performed, indicating that
“minErl/MSLOST” (minute-Erlangs per MS lost the overall measured congestion rate during the
during handover). busy hour is reduced. Reducing the number of
400 channel allocation attempts can also have a
minErl/MSLOST positive effect on the BSC CP load.
350
300 CELL SELECTION AND ACCESS
S
minErl/MSLOST
ome of the parameters controlling MS idle mode
250
behavior during cell selection and system access
are critical for the system’s performance. Minimum
200
signal strength to access the cell (ACCMIN)6 is a
150
cell-level parameter that determines the minimum
Penalty Timer Changes
received signal strength at the MS required to access
100
the system. When an MS first tries to camp to a cell,
4/17 4/24 5/1 5/8 5/15 5/22 5/29 the MS decodes ACCMIN, which is transmitted on
Date (MM/DD) the system information messages of the broadcast
Figure 7. Effect of Penalty Timer Changes on Mobiles Lost During Handover in Relation to Traffic control channel (BCCH), and compares it to the
actual signal strength the MS measures. If ACCMIN
is higher, the MS is not allowed to camp to the cell
FLOW CONTROL TIMERS
because the MS is considered to be at poor radio
F low control timer time between TCH
allocations (TALLOC)5 gives the time in slow
associated control channel (SACCH) periods
conditions.
Parameter Adjustment and Evaluation
(480 msec) between consecutive TCH allocation Depending on the setting of ACCMIN, the cell
attempts, from the channel allocation algorithm, radius (in idle mode) can be modified. ACCMIN
if the first TCH allocation attempt fails. The timer was originally set to –107 dBm to improve the
is used during assignment when the BSC customer perception of the available coverage.
attempts to find an idle TCH for data or speech However, such perceived improvement was
and also during handover. No candidate list is achieved at the risk of an increased number of call
prepared from the locating algorithm before the set-up failures, since MSs at poor radio conditions
timer expires unless an urgency is detected, in were allowed to access the system. Additionally,
which case the new list for handover is sent the mobile equipment static sensitivity is limited
within the time specified by the timer for urgent to approximately –104 dBm for most of the
handover (TURGEN)5. handsets available, so lower signals are not
practically measurable.
Parameter Adjustment and Evaluation
Parameter TALLOC specifies the pace at which A lower ACCMIN value also meant that fewer
allocation attempts counted by the Ericsson BSC subscribers were able to respond to paging
counter for TCH allocation attempts (TCALLS)5 messages and that poor paging performance
are repeated when congestion is counted by the could result [6].
Ericsson congestion timer for immediate TCH
To improve call set-up performance and
assignments (TCONGAS)5 or by the Ericsson
minimize the risk of SDCCH dropped
congestion timer for handover TCH assignments
connections, ACCMIN was set to the quoted
(TCONGHO)5. A decision was made to change
mobile static sensitivity of –104 dBm and the
the original setting from two SAACH periods to
SDCCH drop rate was monitored. The expected
four to limit the number of allocation attempts
improvements were verified by a 22 percent
per event (assignment or handover). Multiple
reduction in SDCCH drops, as shown in Figure 9.
allocation attempts increase the overall measured
_________________________________ _________________________________
5 An Ericsson flow control parameter 6 An Ericsson access parameter
22 Bechtel Telecommunications Technical Journal

7. 50
SIGNAL STRENGTH MEASUREMENT CRITERIA IN
45
THE LOCATING ALGORITHM Bid Congestion Handover
40
T he locating algorithm implemented in the
BSC controls cell selection in dedicated (i.e.,
call) mode and determines handover decisions.
35
30
Bid Congestion Assignment
The main objectives of handover are to maintain 25
(%)
call continuity and quality and to control cell size 20
and handover borders to minimize total network 15
interference. 10
The inputs to the locating algorithm are signal 5
TALLOC = 2 TALLOC = 4
strength and quality measurements from the MS 0
(the so-called mobile assisted handover [MAHO]) 5/11 5/13 5/15 5/17 5/19 5/21 5/23 5/25 5/27 5/29 5/31 6/2 6/4 6/6 6/8
Date (MM/DD)
and from the BTS. The output is a list of candidate Figure 8. Difference of Measured Congestion After Flow Control Timer Change
cells for handover, ranked in descending order
according to preferences and constraints intro-
strength level (MSRXSUFF)7 and received by
duced by other features and by the settings of the
the BTS sufficient signal strength level
algorithm itself. The locating algorithm works
(BSRXSUFF)7. High-signal-strength cells are
continuously for all active MSs and completes a
ranked according to the L criterion and the
cycle every SAACH period (480 msec).
rest according to the K criterion.
The signal strength measurements reported by
• Ericsson-3, where ranking is performed only
the MS and the BTS are evaluated according to
according to the K criterion, but two separate
comparison criteria that can be selected with
hysteresis values are used.
different settings in the locating algorithm. The
first is the signal strength or K criterion and the Parameter Adjustment and Evaluation
second is the path loss or L criterion. They are used Before this exercise, only the K criterion was used
to compare reported values for serving and for handover calculations. The hysteresis was set
neighboring cells to determine the optimum cell to K-criterion hysteresis (KHYST)7 = 4 dB.
ranking and the handover borders. Hysteresis is a signal strength offset that is added
In the K-criterion mode, the comparisons are to the actual reported value for the serving cell to
performed purely according to the received prevent unnecessary ping-pong handovers at the
signal strength (i.e., cells measured with higher border between two cells. The L criterion was
signal strength are ranked higher). Hence, an introduced in an attempt to further improve
increase in the output power of a cell signifies network handover performance. Sufficient
expansion of its service area. This criterion seeks condition parameters MSRXSUFF = –86 dBm and
to maximize the carrier-to-interference (C/I) ratio BSRXSUFF = –92 dBm determine the breaking
by maximizing “C.” point between L and K ranking. Cells reporting
with signal strength values greater than both
In the L-criterion mode, path loss is taken into levels are considered suitable for L ranking,
account. Cells with lower path loss are ranked where an increased hysteresis value, L-criterion
higher, and the output power of each cell does hysteresis (LHYST)7 = 7 dB, is used. The
not affect the calculations. The criterion actually remaining cells are K ranked with a hysteresis
favors cells with low output power; thus, KHYST = 4 dB.
improvement in C/I ratio is attempted by
3.0
decreasing the total interference. However,
L ranking can sometimes lead to a locally lower
2.5
C/I ratio than K ranking. Two cell ranking
SDCCH Drop Rate (%)
algorithms are available, set by BSC parameter
evaluation type (EVALTYPE)7 [5]: 2.0
• Ericsson-1-2, which uses both L and 1.5
K ranking. The candidate cells are separated
into high- and low-signal cells by comparing
1.0
received signals to the following parameters
for downlink and uplink, respectively:
0.5
received by the mobile sufficient signal 4/30 5/5 5/10 5/15 5/20 5/25 5/30 6/4 6/9
_________________________________ Date (MM/DD)
Figure 9. SDCCH Drop Rate Before and After ACCMIN Change
7 An Ericsson locating parameter
September 2004 • Volume 2, Number 2 23

8. 3.0
The Ericsson-3 algorithm was also tested. The
main difference from the previous K-ranked-only
2.5
algorithm is that, depending on the received
downlink signal strength, one of two hysteresis
2.0
values is used. The signal strength level
(%)
between high and low strength cells
1.5
(HYSTSEP)7 = –86 dBm parameter specifies
whether the serving cell is a high or low strength
1.0
K Only K+L Ericsson-3 cell, allowing a larger high-signal-strength
Criterion
hysteresis (HIHYST)7 = 7 dB or a smaller low-
0.5
signal-strength hysteresis (LOHYST)7 = 4 dB to
5/14 5/16 5/18 5/20 5/22 5/24
5/26 5/28 5/30 6/1 6/3 6/5 6/7 6/9
Date (MM/DD)
be applied. The purpose of the high hysteresis
Figure 10. Handovers per Call per Evaluation Criterion values for both tested algorithms is to prevent
unnecessary handovers in the cell borders when
3.0 x 106
radio conditions permit.
2.5 x 106 In Figure 10 the reduction in the total number of
handovers in the system due to the increased
Number of Handover Causes
2.0 x 106 hysteresis in both testing cases can be verified. It
is noteworthy that the L-criterion algorithm
1.5 x 106 seems to introduce the highest (25 percent)
reduction in the handovers, as expressed by the
1.0 x 106 handovers per call index.
0.5 x 106
Figure 11 shows the following handover areas:
HOTOKCL LOWHYST HIHYST
handovers to K cells (HOTOKCL)7, handovers to
HOTOLCL HOUPLQA HODWNQA
0
K Only K+L Eric-3 L cells (HOTOLCL)7, low hysteresis (LOWHYST),
5/1 5/5 5/9 5/13 5/17 5/21 5/25 5/29 6/2 6/6 (HIHYST), handovers due to uplink signal quality
Date (MM/DD) (HOUPLQA)7, and handovers due to downlink
Figure 11. Handover Causes per Evaluation Criterion signal quality (HODWNQA)7.
0.90
280
The portion of handovers performed with the
0.85
L criterion in the first case and with the HIHYST
0.80 250 value in the second can well justify the previous
deviation. Up to 30 percent of total handovers in
minErl/MSLOST
0.75 220
both cases take place with the use of the increased
(%)
0.70 hysteresis values, which means that the
190
0.65 handovers are actually delayed. The result is a
160 total handover reduction, if averaged over the
0.60
Over Total Handovers whole network.
0.55 minErl/MSLOST 130
K Only K+L Eric-3 As already mentioned, handover is considered a
0.50 100
task with a high risk of call drop. Figure 12 shows
5/1 5/4 5/7 5/11 5/14 5/17
5/20 5/23 5/26 5/29 6/1 6/4 6/7
Date (MM/DD)
the effect of the tested settings in call drop
Figure 12. Mobiles Lost During Handover per Evaluation Criterion performance of the handover algorithm.
96.5 25
Handover dropouts, expressed as a percentage of
HO Success
96.0
total handovers, may initially convey that the
Ping-pong 20 situation worsened with the new settings.
Ping-pong Handovers (%)
95.5 Nevertheless, what matters is the absolute
15 number of failures actually experienced by the
95.0
subscriber; since the total number of handovers
94.5 10 decreased, this difference is not substantial.
(%)
To emphasize this point, the index
94.0
“minErl/MSLOST,” giving Erlang minutes of
5
93.5
K Only K+L Eric-3
traffic carried out per handover dropout, is also
depicted. Inspecting this index, it is clear that the
93.0 0
5/1 5/4 5/7 5/11 5/14 5/17
5/20 5/23 5/26 5/29 6/1 6/4 6/7
L-criterion algorithm appears much improved,
Date (MM/DD) while the performance of the Ericsson-3
Figure 13. Handover Success Rate and Ping-pong Rate per Evaluation Criterion algorithm is rather ambiguous.
24 Bechtel Telecommunications Technical Journal

9. The superiority of the L-criterion algorithm over TRADEMARK
the Ericsson-3 algorithm is also apparent in Ericsson is a trademark or registered trademark
Figure 13; the ping-pong handovers (i.e., of Telefonaktiebolaget LM Ericsson.
handovers back to the originating cell within
10 seconds) are reduced in both cases. This
reduction is a direct consequence of the hysteresis REFERENCES
values of 7 dB introduced in both algorithms.
[1] “Radio Network Parameters and Cell Design
However, the L-criterion algorithm shows the
Data” – Ericsson CME20 Documentation.
best performance in this field, meaning that more
[2] “User description, Frequency Hopping” –
accurate and reliable handover decisions Ericsson CME20 Documentation.
accompany this algorithm, exactly as predicted [3] “User description, MS Dynamic Power Control” –
by theory. Ericsson CME20 Documentation.
[4] “User description, Cell Load Sharing” –
The only disadvantage of the L-criterion algorithm Ericsson CME20 Documentation.
appears to be the handover success percentage, [5] “User description, Locating” – Ericsson CME20
half a decimal unit below the previous figures. The Documentation.
same applies for the Ericsson-3 algorithm, which [6] “User description, Idle Mode Behaviour” –
can be attributed to the lower number of handover Ericsson CME20 Documentation.
commands. It can be assumed that, due to various
radio problems, a significant number of handover
failures always exist in the network. This assumed BIOGRAPHY
value can be highlighted or hidden, in a statistical Michael Pipikakis is a network
sense, depending on the volume of the total planning and wireless tech-
nology manager for Bechtel's
sample. It is believed that careful optimization and
Europe, Africa, Middle East,
individual neighbor cell inspection of the and Southwest Asia Region. He
network’s handover performance can further supports ongoing and new
improve this figure. projects and new business
development; writes guidelines
As a result, the K-L combination algorithm was and procedures for mobile
eventually introduced. Further improvement can network design, planning, and
optimization; and participates in technology forums.
be achieved by fine-tuning sufficient level
parameters BSRXSUFF and MSRXSUFF to Michael is a mobile networks specialist with 17 years
identify a balanced breakpoint for cell ranking. of experience in the telecommunications industry,
including more than 11 years in RF planning, design,
Also, different LHYST values can be tried. optimization, and management of the end-to-end
performance of cellular networks.
Before joining Bechtel, Michael held various
CONCLUSIONS management positions in the Vodafone Group's radio
A ll of the optimization-related changes were system design and optimization department and
development department over a 10-year period;
made in a controlled manner so that their worked for Cellnet UK and GEC Marconi UK; and was
effectiveness could be measured and evaluated. a telecommunications operator in the Greek Navy.
At the end of the project, the average daily DCR From 1999 to 2003, he was a member of the Vodafone
was reduced by 30 percent, and the average Global Forum for UMTS design harmonization.
minute-Erlang per drop was increased by almost Michael has a BEng Honors in Electronics Engineering
45 percent. At that point, a foundation was with Computing and Business from Kingston
created for further fine-tuning as the network University in Surrey, England, and an HND in Radio
Communications Systems Design from the Polytechnic
expands in response to increases in traffic and School of Athens, Greece. He is a member of the
subscriber base. Institution of Electrical Engineers.
As has been shown, considerable network
performance gains can be made by fully utilizing
the available functionality and fine-tuning the
network parameters using statistics to evaluate
the results.
September 2004 • Volume 2, Number 2 25