Radio Network

Radio Network Optimization:

Optimization: Contents
Contents

• Reasons for the Need of Optimization

• Performance Data Measurements
• Drive Tests
• Optimization Strategies
• Optimization of Physical Parameters

© TECHCOM Consulting 
• Optimization of Database Parameters
• Example Drive Tests
• Example Drive Tests: Exercises

MN 1790 7-1
 Reasons for
Reasons for the
the Need
Need of
of Optimization
Optimization

Network optimization is an iterative process which should improve the quality and performance of a
network and also run the network more efficiently. As in any optimization problem, also in network
optimization, the network will mostly not run optimal from the very beginning. There can be
mentioned several reasons:

• Systematic inaccuracies

© TECHCOM Consulting 
• Statistical nature of the involved processes like e.g. traffic and RF propagation
• Dynamical nature of the involved processes like e.g. change of the subscriber’s telephone
behavior (e.g. SMS)
• Wrong (or only too rough) planning assumptions, input data and/or planning models
• Increasing number of subscribers
• Installation errors (for example a wrong cabling: transmitting into cell A, but receiving from
cell B)
• Hardware / software trouble
• ...

MN 1790 7-2
 Performance Data
Performance Data Measurements
Measurements

Performance data measurements can help the network operator for example to localise problem
areas as early as possible and also to verify improvements of the network optimisation.
Concerning radio network optimisation there are related performance data measurements foreseen
by GSM (see: GSM 12.04) and in addition also vendor specific ones.
In general performance data measurements can be run continuously, periodically or sporadically, for
a long time or a short time, observing smaller or greater parts of the network.

© TECHCOM Consulting 
The related counters could in principle be actualised continuously during the observation period, but
mostly a scanning method is used. Scanning method means that the system counts the number of
events not continuously but only at particular times. This leads to some uncertainty for the
measurement results. Nevertheless, the error performed can be estimated using statistical methods.
In general, the smaller the scanning interval the higher the precision of the measurement (for
constant observation periods). Typical scanning intervals are 100 ms or 500 ms.

MN 1790 7-3
 Performance Data
Performance Data Measurements
Measurements

Scanner

© TECHCOM Consulting 
Performance data measurement(s)

Counter(s)

Condition(s) for the updating of

the counter value(s)

MN 1790 7-4
 Drive Tests
Drive Tests

Drive tests are performed by the network operator for various reasons:
• To check the coverage in a certain area
• To check the quality of service in a certain area
• To find the answer for customer complaints
• To realise that the network is not properly running

© TECHCOM Consulting 
• To verify that the network is properly running
• To verify that certain optimisation steps have been successful
• ...

Drive tests must be well prepared. Before, during and after the drive test the following steps should
be performed:

MN 1790 7-5
 Drive Tests
Drive Tests

Before drive test • Plan the route where to drive

• Plan the time when to drive
• Determine the MS mode (idle mode/ connected mode) and also the
call strategy (long / short calls)
• Decide which values to focus on (for example: RXQUAL, RXLEV,SQI, ...)

© TECHCOM Consulting 
• Select an appropriate test equipment and check the test equipment
• Think of notes which should be inserted later on in the recording file
During drive test • Monitor the test equipment
• Reconnect dropped calls
• Insert notes in the recording file
• Note interesting events separately (e.g. on a piece of paper)

After drive test • Make back-up files of the captured data

• Replay the captured data and analyse them
• Find out problem areas and problem events
• Use further post-processing tools to display the captured data more
clearly and to graphically display further values
• Perform statistics and summarise the results

MN 1790 7-6
 Optimization Strategies
Optimization Strategies

Before Optimization:
Clean the hardware

© TECHCOM Consulting 
• Performance Data Measurements
• Customer Complaints
• Drive tests

Optimization:
1) Physical parameters
2) Database Parameters

MN 1790 7-7
 Optimization of
Optimization of Physical
Physical Parameters
Parameters

Altering antenna tilt:

• to reduce interference
• to limit coverage area
• to improve coverage (e.g. coverage weakness below main lobe)
• to improve in-building penetration

© TECHCOM Consulting 
Altering the Antenna tilt must be done very carefully to really improve the situation.
Typical down-tilts are between 0° and 10°, however even higher values (up to 25°) have already
been used.
Altering antenna azimuth:
• to overcome coverage weakness between different sectors
• to reduce interference in certain directions
Increasing or decreasing antenna height:
• to reduce or improve coverage
• to reduce interference
Change of antenna type
• to achieve desired ration characteristics

MN 1790 7-8
 Optimization of
Optimization of Physical
Physical Parameters
Parameters

Addition / re-movement of TRXs:

• Depending on the real measured traffic load either TRXs can be removed (switched off or
blocked) or must be added. Not really needed TRXs may interfere other cells.
• The number of needed TRXs and also the configuration of the different channels depend on the
offered traffic, and the subscriber behaviour.

© TECHCOM Consulting 

MN 1790 7-9
 Optimization of
Optimization of Physical
Physical Parameters
Parameters

Cell sectorization / cell splitting:

Can be used for:
• Coverage enhancements (since the antenna gain of sectorised antennas is higher than that of
omni directional antennas)
• Interference reduction

© TECHCOM Consulting 
• Capacity enhancements, but only if together with the sectorisation also the number of TRXs is
increased (compare Erlang-B loss formula)

Depending on how the splitting is performed:

• it may be a more or less expensive and difficult (time consuming) solution
• coverage weakness between the main lobes may appear
• the capacity will be reduced if the total number of TRXs remains constant

MN 1790 7 - 10
 Optimization of
Optimization of Physical
Physical Parameters
Parameters

Implementation of Antenna near Pre-Amplifiers:

Link imbalances are one reason for poor quality, increased call drop rate and increased
handover failure rate. In case of an unbalanced link, the uplink and downlink coverage ranges
differ. Often the downlink range is bigger than the uplink range. This problem can be overcome
by using antenna near preamplifiers which improve the sensitivity and the noise figure of a base
station system. Looking to the link budget: The better the sensitivity of the base station, the
more fare the possible uplink range. In any case, a proper running network requires a balanced

© TECHCOM Consulting 
link.

Implementation of Repeaters:
A repeater (see GSM 11.26 (ETS 300 609-4) and GSM 05.05) is a bi-directional (full duplex) RF
amplifier and is used to overcome coverage holes in a base station area. Typical applications of
repeaters are the coverage of problem zones like tunnels, valleys, in buildings, ...
A repeater receives, amplifies and retransmits the downlink signal from a donor base station
into an area with weak or no coverage, and the uplink signal from mobile stations which are
located in such an area. Repeaters extend but do not replace base stations.

MN 1790 7 - 11
 Optimization of
Optimization of Database
Database Parameters
Parameters

Frequency Changes:
• To overcome e.g. sever cases of downlink interference (therefore it is advisable to have some
spare frequencies).
• May influence other areas.
• Re-planning may become necessary.

© TECHCOM Consulting 
• In high-density areas often difficult.

Strategies:
• Using spare frequencies in severely interfered regions.
• TCH – BCCH change as temporary solutions in low TCH traffic load areas.
• Re-planning of TCH and BCCH frequencies.

MN 1790 7 - 12
 Optimization of
Optimization of Database
Database Parameters
Parameters

Frequency Hopping: Cyclic or pseudo random hopping?

f0
Example of cyclic hopping:
f1 HSN = 0
f2

© TECHCOM Consulting 
f3
f4
f5

Time in TDMA frames

Optimum frequency diversity

due to averaging of Rayleigh fading
⇒ rural, coverage limited areas

MN 1790 7 - 13
 Optimization of
Optimization of Database
Database Parameters
Parameters

Frequency Hopping: Example of pseudo random hopping: HSN = 1-63

Cell1 (e.g. HSN =14) Re-use cell 15 (e.g. HSN =23)

0 1 2 3 4 5 6 7 8 9 10 TDMA Nr. 0 1 2 3 4 5 6 7 8 9 10
f0 f0
f1 f1

© TECHCOM Consulting 
f2 f2
f3 f3
f4 f4
f5 f5
time

Co-channel interference averaging

Collision probability: 1/number of hopping frequencies
⇒ interference limited areas (hot spots)

MN 1790 7 - 14
 Optimization of
Optimization of Database
Database Parameters
Parameters

Radio Link Failure (RLF) / Radio Link Timeout (RLT):

(see also GSM 04.08 and GSM 05.08)
Calls which fail due to radio coverage problems or which suffer under unacceptable voice or data
quality (due to e.g. interference) which cannot be improved by power control or handover are either
released or re-established in a defined way.

© TECHCOM Consulting 
The criterion for the detection of a radio link failure by the MS is the success rate of decoding DL-
SACCH messages.
The criterion for the determination of a radio link failure by the BS is either the success rate of
decoding UL-SACCH messages or it is based on RXLEV / RXQUAL measurements.

The MS checks the DL with the help of a radio link (failure) counter running in the MS.
The BS checks the UL with the help of a radio link (failure) counter running in the BS.

MN 1790 7 - 15
 Optimization of
Optimization of Database
Database Parameters
Parameters

Radio Link Failure (RLF) / Radio Link Timeout (RLT):

The algorithm for the modification of the radio link failure counter S is the following:

Starting value for the Radio Link Failure Counter: Radio_Link_Timeout

© TECHCOM Consulting 
In case of successful decoding of SACCH messages: Snew=Sold+2
In case of non-successful decoding of SACCH messages: Snew=Sold-1
value range for S: 0≤ S≤ Radio_Link_Timeout
Radio link failure is detected if: S=0

This algorithm is only running after assignment of a dedicated channel (i.e. in connected mode).

The starting value Radio_Link_Timeout for the MS counter is sent on the BCCH system information
type 3 or on the SACCH system information type 6 in the information element ‘Cell Options’.

MN 1790 7 - 16
 Optimization of
Optimization of Database
Database Parameters
Parameters

Handover and Power Control:

Reasons for the optimisation of handover parameters:

• To reduce the number of call drops
• To reduce the number of handovers

© TECHCOM Consulting 
• To maximise the time duration the MS spends in the best cell
• To improve the speech quality

Handover types: intra- / inter- cell, BTS, BSC, MSC handovers

Handover causes:
• (Bad) RXQUAL
• (Low) RXLEV
• (far) DISTANCE
• (Power Budget) PBGT

MN 1790 7 - 17
 Optimization of
Optimization of Database
Database Parameters
Parameters

Thresholds needed for handover evaluation:

Abbreviation Remarks
L_RXLEV_UL_H RXLEV threshold on the uplink for handover process to commence (outgoing HO)
L_RXQUAL_UL_H RXQUAL threshold on the uplink for handover process to commence (outgoing HO)
L_RXLEV_DL_H RXLEV threshold on the downlink for handover process to commence (outgoing HO)

© TECHCOM Consulting 
L_RXQUAL_DL_H RXQUAL threshold on the downlink for handover process to commence (outgoing HO)
MS_RANGE_MAX Threshold for the maximum allowed distance between MS and current BTS (outgoing
HO)
RXLEV_UL_IH RXLEV threshold on the uplink for intracell (interference) handover
RXLEV_DL_IH RXLEV threshold on the downlink for intracell (interference) handover
RXLEV_MIN(n) Minimum RXLEV required for a MS to be allowed to be handovered to neighbour cell
(n) (incoming HO)
HO_MARGIN(n) Threshold for power budget process
MS_TXPWR_MAX Maximum transmission power a MS may use in the serving cell
MS_TXPWR_MAX(n) Maximum transmission power a MS may use in the adjacent cell (n)
BS_TXPWR_MAX Maximum transmission power the BTS may use

MN 1790 7 - 18
 Optimization of
Optimization of Database
Database Parameters
Parameters

Level triggered handover:

Depending on the measured and averaged RXQUAL_XX and RXLEV_XX values the system (MS
and / or BS) may increase or decrease the output power or may handover the call.
Remarks to the corresponding handover threshold settings:
L_RXLEV_UL_H and L_RXLEV_DL_H should be set some dB (e.g. 5 dB) above the effective (+
diversity gain, + preamplifier) receiver sensitivity limit:

© TECHCOM Consulting 
Receiver sensitivity levels due to GSM 05.05:
For DCS 1800 class 1 or class 2MS: -100 dBm
For DCS 1800 class 3 MS: -102 dBm
For GSM 900 small MS: -102dBm
For other GSM 900 MS: -104 dBm
For normal BTS: -104 dBm

Example settings: L_RXLEV_DL_H = –95 dBm

L_RXLEV_UL_H = –102 dBm

MN 1790 7 - 19
 Optimization of
Optimization of Database
Database Parameters
Parameters

Level triggered handover:

There should be a level hysteresis between the threshold RXLEV_MIN(n) for incoming handover
and the threshold L_RXLEV_XX_H for outgoing handover:
RXLEV_MIN > L_RXLEV_XX_H + 4....10 dB

© TECHCOM Consulting 
The size of this hysteresis should be related to the standard deviation of the long term fading
(typically 4...10 dB) and should be large enough to avoid ping-pong handovers and small enough to
allow fast handovers.

Example setting: RXLEV_MIN = -90 dB

MN 1790 7 - 20
 Optimization of
Optimization of Database
Database Parameters
Parameters

Level triggered Power Control:

Depending on the measured and averaged RX_QUAL and RX_LEV values the system (MS and /
or BS) may increase or decrease the output power or may handover the call.

For the power control and handover threshold settings the following considerations should be taken
into account:

© TECHCOM Consulting 
L_RXLEV_UL_P (Lower) RXLEV threshold on the uplink for power increase

U_RXLEV_UL_P (Upper) RXLEV threshold on the uplink for power reduction

L_RXLEV_DL_P (Lower) RXLEV threshold on the downlink for power increase

U_RXLEV_DL_P (Upper) RXLEV threshold on the downlink for power reduction

MN 1790 7 - 21
 Optimization of
Optimization of Database
Database Parameters
Parameters

Level triggered Power Control:

To avoid consecutive power increase or decreases directly after each other the difference between
upper and lower power control thresholds should be large enough (e.g. 10 dB).
To allow the system to perform power control before handover is executed, the lower power control
level thresholds should be about 10 dB above the lower handover level thresholds.

© TECHCOM Consulting 
Example settings:
L_RXLEV_DL_H = –95 dBm, L_RXLEV_DL_P = -85 dBm, U_RXLEV_DL_P = -75 dBm
L_RXLEV_UL_H = –102 dBm, L_RXLEV_UL_P = -92 dBm, U_RXLEV_UL_P = -82 dBm

MN 1790 7 - 22
 Optimization of
Optimization of Database
Database Parameters
Parameters

Power Control Execution:

For the power control execution parameter settings the following considerations should be taken
into account:
Since typically a power increase command is more urgent than a power reduction command, the
power increase step size should be greater than the power reduction step size.
The power increase and power reduction step sizes should be on the one hand small enough to

© TECHCOM Consulting 
enable an accurate power control, on the other hand large enough to reduce the number of
necessary power control commands and therefore the signalling load.

Example settings:
POW_INCR_STEP_SIZE = 4 dB
POW_RED_STEP_SIZE = 2 dB

MN 1790 7 - 23
 Optimization of
Optimization of Database
Database Parameters
Parameters

Quality triggered handover:

Depending on the measured and averaged RXQUAL_XX and RXLEV_XX values the system (MS
and/or BS) may increase or decrease the output power or may handover the call.
Remarks to the corresponding handover threshold settings:
L_RXQUAL_UL_H, L_RXQUAL_DL_H

© TECHCOM Consulting 
RXLEV_UL_IH, RXLEV_DL_IH
In case of bad quality (RXQUAL_XX > L_RXQUAL_XX_H) and high signal strength (RXLEV_XX >
RXLEV_XX_IH) at the same time, there is a high probability of the presence of:
cochannel interference, adjacent channel interference, intermodulation problems, intersystem
interference.
Temporary solution: intracell handover
Intracell handover doesn’t help: if frequency hopping is switched on, or if there is only 1 TRX in the
serving cell and the interference is continuous and not bursty.

Examples settings: L_RXQUAL_UL_H = 5, L_RXQUAL_DL_H = 5

RXLEV_UL_IH = -85 dBm , RXLEV_DL_IH = -78 dBm

MN 1790 7 - 24
 Optimization of
Optimization of Database
Database Parameters
Parameters

Quality triggered Power Control:

Depending on the measured and averaged RXQUAL_XX and RXLEV_XX values the system (MS
and/or BS) may increase or decrease the output power or may handover the call.
Power is increased if the received quality is bad:
RXQUAL_XX > L_RXQAUL_XX_P

© TECHCOM Consulting 
Power can be decreased if the received quality is very good:
RXQUAL_XX < U_RXQAUL_XX_P
However, often it is more suitable to control the power decrease by the level criteria and to set
U_RXQAUL_XX_P = 0 or a small value, i.e. to ‘disable’ the power decrease due to good quality.
To make ‘power up before handover’ possible, the following relation between power control and
handover thresholds should be taken into account:
U_RXQUAL_XX_P < L_RXQUAL_XX_P < L_RXQUAL_XX_H

Example settings: U_RXQUAL_XX_P = 0 (or 1)

L_RXQUAL_XX_P = 4
L_RXQUAL_XX_H = 5

MN 1790 7 - 25
 Optimization of
Optimization of Database
Database Parameters
Parameters

Handover Thresholds:

RXQUAL
L_RXLEV_XX_IH

© TECHCOM Consulting 
Intercell handover Intracell handover
due to quality due to quality
L_RXQUAL_XX_H

Intercell No handover action

handover due to quality or level,
due to level e.g. power budget handover

0
0 L_RXLEV_XX_H 63 RXLEV

MN 1790 7 - 26
 Optimization of
Optimization of Database
Database Parameters
Parameters

RXQUAL Power Control Thresholds:

7
PowerIncrease
Power Increase
L_RXQUAL_XX_P

No action due Power

© TECHCOM Consulting 
Power to power control
Increase Decrease
U_RXQUAL_XX_P

Power Decrease
0
0 L_RXLEV_XX_P U_RXLEV_XX_P 63 RXLEV

POW_RED_STEP_SIZE
Example settings

L_RXLEV_DL_P = 25 L_RXLEV_UL_P = 10 POW_RED_STEP_SIZE = 2 dB

U_RXLEV_DL_P = 35 U_RXLEV_UL_P = 15 POW_INCR_STEP_SIZE = 4 dB
L_RXQUAL_DL_P = 4 L_RXQUAL_UL_P = 5
U_RXQUAL_DL_P = 1 U_RXQUAL_UL_P = 4

MN 1790 7 - 27
 Optimization of
Optimization of Database
Database Parameters
Parameters

Handover triggered by power budget:

In an interference limited area (e.g. small cells in cities) most of the handovers should be power
budget handovers:
For this type of handover not the level, quality, or distance is the handover cause, since all the
corresponding thresholds are not exceeded in the serving cell, but a neighbour cell offers a better
service (a smaller path loss, see link budget).

© TECHCOM Consulting 
Since the power budget hanodver looks for the serving cell with the smallest path loss, this kind of
handover will:
Reduce interference
Prolong MS battery time

The power budget is defined as the difference between the path loss in the serving cell and the
path loss in the neighbour cell:

PBGT(n) = (BS_TXPWR – RXLEV_DL) – ( BS_TXPWR_MAX(n) – RXLEV_DL_NCELL(n))

MN 1790 7 - 28
 Optimization of
Optimization of Database
Database Parameters
Parameters

Handover triggered by power budget:

Assumption:
BS_TXPWR_MAX – BS_TXPWR_MAX(n) = MS_TXPWR_MAX – MS_TXPWR_MAX(n)

PBGT(n) = RXLEV_DL_NCELL(n) – RXLEV_DL – PWR_C_D + min (MS_TXPWR_MAX,P) – min

© TECHCOM Consulting 
(MS_TXPWR_MAX(n),P)

Where PWR_C_D is defined as: BS_TXPWR_MAX – BS_TXPWR

If PBGT(n) > HO_MARGIN(n) the path loss in the serving cell is greater than the path loss in the
neighbour cell + HO_MARGIN so that the neighbour cell is considered as the much better cell.

MN 1790 7 - 29
 Optimization of
Optimization of Database
Database Parameters
Parameters

Remarks to the setting of the Handover Margin:

• The HO_MARGIN setting should be a compromise between ideal power budget handover (which
requires a small HO_MARGIN value) and a setting to reduce the risk of ping-pong handovers
(which requires a greater HO_MARGIN value).
• A small handover zone increases the risk of ping-pong handovers.

© TECHCOM Consulting 
• Usually HO_MARGIN is set symmetrically.
• Asymmetrical HO_MARGIN can be used to influence the size of the handover area and/or to
move the handover area, i.e. to move the cell boundaries.
• Adjusting HO_MARGIN values can therefore also be used to adapt the cell area to the traffic load
or to avoid local interference.
• RXLEV_MIN(n) should be set to such a value that RXLEV_DL_NCELL(n) > RXLEV_MIN(n) in
those areas where PBGT(n) > HO_MARGIN(n) to really allow the power budget handover as soon
as the power budget condition is fulfilled.

MN 1790 7 - 30
 Optimization of
Optimization of Database
Database Parameters
Parameters

HOM: Handover margin for

adjacent cell
HOM = 0 HOM = 0

© TECHCOM Consulting 
HOM = + 6 dB HOM = +6 dB

MN 1790 7 - 31
 Optimization of
Optimization of Database
Database Parameters
Parameters

HOM = +6 dB HOM = 0

© TECHCOM Consulting 
HOM = -6 dB HOM = 0 dB

HOM = 0 dB HOM = +6 dB

MN 1790 7 - 32
 Optimization of
Optimization of Database
Database Parameters
Parameters

Remarks to the pre-processing (averaging) of the measurements needed for power control
and handover decisions:

In general:
• Many measurements should be averaged in case that reliable decisions are necessary (better

© TECHCOM Consulting 
statistics).
• Only a few measurements should be averaged in case that fast decisions are necessary.

For level / quality triggered handover / power control decisions:

• To allow the system to ‘power up before handover’ usually the averaging process for the handover
decisions should include more measurements than for power control decisions.

• Usually for level triggered handover decisions more measurement values should be averaged
than for quality triggered handover decisions since quality handovers must be executed quickly if
sudden interference appears.

MN 1790 7 - 33
 Optimization of
Optimization of Database
Database Parameters
Parameters

BCCH allocation:
Also neighbor cell list is target of optimization process:
Missing neighbor cell ⇒ perhaps call drop
Too many neighbors ⇒ bad statistics, unprecise measurement values, perhaps wrong decisions
In practice: ≈ 6-8 neighbors

© TECHCOM Consulting 
Number of BCCH carriers in BA Number of samples per carrier and
SACCH multiframe
32 3-4

16 6-7

10 10-11

8 12-13

MN 1790 7 - 34
 Optimization of
Optimization of Database
Database Parameters
Parameters

Fast handover: Street corner effect: e.g. 20 dB loss

Fast handover mechanism necessary:
- trigger: uplink measurement receive level below threshold
- short averaging period of measurements
- predefined target cell lists

© TECHCOM Consulting 
- small handover margins
- short timer settings
- allow back handover

MN 1790 7 - 35
 Optimization of
Optimization of Database
Database Parameters
Parameters

Location Area:
The size of the location area must always be a compromise:
Too big ⇒ perhaps paging overload (PCH overload)
(MS is paged in the whole location area)
Too small ⇒ perhaps too many location updates (AGCH overload)

© TECHCOM Consulting 
(MS has to perform location update if location
area is changed)

MN 1790 7 - 36
 Example Drive
Example Drive Test
Test

© TECHCOM Consulting 

MN 1790 7 - 37
 Example Drive
Example Drive Test:
Test: Exercise
Exercise

Exercise:

Discuss the drive test given above:

a) Localize the problem area(s)

b) Suggest counter measures to solve the problem(s)

© TECHCOM Consulting 

MN 1790 7 - 38
 Example Drive
Example Drive Test
Test

© TECHCOM Consulting 

MN 1790 7 - 39
 Example Drive
Example Drive Test
Test

© TECHCOM Consulting 

MN 1790 7 - 40
 Example Drive
Example Drive Test
Test

© TECHCOM Consulting 

MN 1790 7 - 41
 Example Drive
Example Drive Test:
Test: Exercise
Exercise

Exercise:

Discuss the drive test given above:

a) Localize the problem area(s)

b) Suggest counter measures to solve the problem(s)

© TECHCOM Consulting 

MN 1790 7 - 42