GSM voice quality

Final report

Tunis, 15th of July 2004

Customer: Tunisiana

Telecommunications and IT Know-How in: Training Commissioning and Maintenance (OAM) Network Planing

Radio coverage tests Traffic measurements and analysis

Company profile (MCI)

MCI: Management Conseil et Ingénierie

Traditional approach for GSM radio link test:

• Protocol analysers (RXQUAL)Data useful in network maintenance but not for the estimation of the Quality of Service (QoS) as viewed by the mobile phone user (perception!!).

RXQUAL does not represent “real life” conditions. The following slides discuss inherent problems of RXQUAL and demonstrate that the measurement cannot be related to perceived speech quality in an accurate and reliable way.

Why voice QoS

In today’s competitive environment, it is

important for GSM operators to be able to

accurately assess performance and

voice quality of their networks in order to

maintain or extend market position.

From a subscriber’s point of view, the

relevant performance measure are:

• Call success rate

• Speech quality

QoS is critical to success

• RXQUAL does not consider the varying efficiency of interleaving and bit error correction under different environmental conditions (fading or frequency hopping)

• RXQUAL does not consider quality degradation caused by stolen speech frames (Handover: TCH becomes FACCH for 0,2s)

• RXQUAL does not detect any PSTN quality impairing effects (echo, crosstalk or noise in the analogue part of the network)

• Speech quality is the only parameter to detect defects within voice processing circuits

Arguments for Speech Quality

• RXQUAL does not recognise whether bad speech frame indication (BFI) failed

• Downlink RXQUAL may Fail in Detecting QoS degradation in Uplink Direction

• RXQUAL does not consider acoustical power level variations

Arguments for Speech Quality

Objectives of the study

Evaluation of the quality of service for the telecom operator:

Orascom Telecom TunisiaOrascom Telecom Tunisia

Location: Tunis

Training of the agents

• Identification of the typical noise (echo, metallic or robot voice, lost syllables, etc…) ;• Evaluation of the voice quality (recorded audio samples)

MCI used a software tool for recording and editing of voice messages

Training of the agents

The agents were trained on :

• Use of MCI Hotline

• Numbers to be called

• Itineraries to follow

• Manipulation of topographical maps

• Use of the test forms

Equipment used

Test types

• Outdoor : Outside buildings (pedestrian)• Indoor : Inside buildings, underground

excluded:• Close to doors or windows• Far from doors and windows

• Incar : Inside a car also called drive test

Measured parameters

• Call success rate / Call failure:• Due to network saturation• Due to missed coverage

• Call drop rate / Call completion rate

• Voice quality noted 1 to 4: • Excellent quality• Some noise but still good quality• Some words are not understood but the conversation is still OK• Bad quality. The conversation is not understood

Speech quality relates to:Intelligibility: Quality of perception of the meaning or information content of what the talker said;Naturalness: Degree of fidelity to the talker’s voice;Loudness: Absolute loudness level at the listener’s side.

Sample of the study

Totally 2706 calls:

- Outdoor: 1236- Indoor: 1226- Incar: 244

Tunis was divided into:- 6 regions- 79 sectors with different sizes

Incar calls: all main roads tested, also bidirectionally

R E S U L T S

O F T H E

S T U D Y

Major Indoor points

Liste of the indoor points (365 points):Libellé NombreMunicipalités 18Aéroport de Tunis Carthage 1Banques 5Centres commerciaux 19Souks et marchés 7Petits commerces 20Cafés et salons de thé 17Délégations 7Hôpitaux, cliniques et dispensaires 24Hôtels 4Immeubles 101Ecoles, lycées et collèges 18Habitations 17Magasins et supermarchés 7Autres locaux 2Postes 21Sociétés privées 18Universités 23Autres administrations 36

Constitution of the sample

Distribution of the calls by regions

Sector Region Total Nr.of calls Note 1 Note 2 Note 3 Note 4

407 75 7 782.06% 15.12% 1.41% 1.41%

222 48 2 2

81.02% 17.52% 0.73% 0.73%

313 78 6 2

78.45% 19.55% 1.50% 0.50%

91 23 6

75.83% 19.17% 5.00%

251 79 8 174.04% 23.30% 2.36% 0.30%

685 109 7 6

84.88% 13.51% 0.87% 0.74%

Frequency of notation

A Tunis Centre 496

B La Marsa, SidiBou Saïd,Carthage LaGoulette

274

C Ben Arous,Megrine, Radès,El Mourouj

399

D Ezzahra, Hammam Lif,Hamma Chatt

120 -

E Manouba 339

F Ariana/El Menzah/ Ennasr,Bardo

807

Speech quality by regions (1)

Speech quality by regions (2)

Distribution of the the notes 3 and 4 by regions

DATA ON GISArcView

Sample on ArcView (1)

Habib Bourguiba Avenue

Sample on ArcView (2)

Ariana City Center

Data for a sample on ArcView

Data stored on GIS for each measurement

End