Dr.-Ing. H. W. Gierlich Head of Telecom Division

1. CITEL (PCC.I)/ ITU Forum on Information and Communication Technology Service: Quality, Control and Surveillance (Cartagena de Indias, Colombia, 23-24 September 2013) The importance of models and procedures for planning, monitoring and control in the provision of communications services Dr.-Ing. H. W. Gierlich Head of Telecom Division

2. Outline • Introduction • Communication Services – Underlying System configurations • Planning: The ITU-T E-model (G.107 & G.108) • Network Monitoring: P.862, P.863 & P.563 • Wideband – The new Challenge • Summary

3. ITU-T: QoS and QoE • Quality of Service (QoS): • Totality of characteristics of a telecommunications service that bear on its ability to satisfy stated and implied needs of the user of the service. • Quality of Experience (QoE): • The overall acceptability of an application or service, as perceived subjectively by the end-user. • Quality of experience includes the complete end-to-end system effects (client, terminal, network, services infrastructure, etc.). • Overall acceptability may be influenced by user expectations and context.

4. Speech Quality – what we would like to have 1m “orthotelefonic reference position”

5. Speech Quality… … from the user’s perspective talking situation listening situation speech quality conversational situation

6. End to End Conifguration add. 1 - 100 ms ! IP- GW MSC BSS DSL IP- GW ISC ISC PBX GW PSTN Netw. 1 Netw. 2 10 - 400 ms 10 - 400 ms 240 ms 1-15 ms 90-120 ms • Impact on speech quality: • delay now time variant • echo much more dominant • Coding distortion • background noise problems • insufficient quality of the analog network components (att., noise, distortion…) • insufficient quality of acoustical components A big problem of today‘s networks, interconnection & terminals : delay – unpredictable, load dependant, time variant

7. Contributions to Quality • Network planning • Network monitoring • Laboratory terminal testing • Interoperability testing • The networks • Type of networks • Interconnection • QoS management • The endpoints • Types of terminals • Interoperability • Terminal – Network • Terminal – Terminal • The users‘ location • Environmental conditions

8. Outline • Introduction • Communication Services – Underlying System configurations • Planning: The ITU-T E-model (G.107 & G.108) • Network Monitoring: P.862, P.863 & P.563 • Wideband – The new Challenge • Summary

9. Reference Connection in the E-model

10. Transmission Rating in the E-model Basic assumption: Psychological factors on the psychological scale are additive • Rating factor R: • With: • R0 – Basic signal to noise ratio (takes into account circuit noise, room noises) • Ix – Impairment factors (see next slide) • A – Advantage factor (takes into account a potential advantage for a user for a specific transmission in a specific situation)

11. Network planning & E-model (ITU-T G.107 & G.108) • Basis of the E-model: Impairment factors • Simultaneous impairment factor Is (non optimum loudness rating, non optimum sidetone, PCM coding distortion) • Delayed impairment factor Id (impact of delay, talker- or listener echo) • Equipment impairment factor Ie (all types of impairments in equipment such as coding distortion including the effect of packet loss)

12. How to derive Ie? • Ie is based on subjective tests and can be derived as follows: • Equipment impairment factors for the most popular codecs in ITU-T G.113 • From subjective listening tests described in ITU-T P.833 • From objective models (e.g. ITU-T P.863) following the procedure in ITU-T P. 834

13. Some Ie factors from G.113

14. The E-model Prediction Range • R – value range: prediction on a scale • Mapping to MOS: • For R 0: • For 0 R 100: • For R 100:

15. Guide for User Satisfaction based on R-values

16. Guide on Transmission Planning: ITU-T G.108 Demonstrates the application of the E-model for end-to-end transmission planning in a wide range of networks Keep in mind: G.107 and G.108 are applicable for narrowband networks

17. Outline • Introduction • Communication Services – Underlying System configurations • Planning: The ITU-T E-model (G.107 & G.108) • Network Monitoring: P.862, P.863 & P.563 • Wideband – The new Challenge • Summary

18. Networks & Network Monitoring • Main network related impairments: • Delay • Delay variation (jitter) • Listening speech impairment due to • Coding • Transcoding • Interconnection • Packet loss/jitter

19. ITU-T Models for Perceptual based Speech Quality Measurement Results of Typical Processing Steps (Schematic): Listening Tests! processed signal Test Object Hearing Model Speech Signal Comparison, Reference, S index Q Reduction, reference signal Signal Value Adaptation Hearing Model ITU-T P.862 (2001): PESQ ITU-T P.863 (2011): POLQA The basic principle:

20. Principle • Intrusive test procedure • Objective prediction of MOS (mean opinion score) as achieved in listening test subjectively • Prediction on a MOS-scale: • MOS 5 – excellent • MOS 4 – good • MOS 3 – fair • MOS 2 – poor • MOS 1 – bad • Substitution of subjective tests for known codecs and impairments

21. Application • Intrusive testing: • Insertion of a reference speech sequence • Acquiring the transmitted, distorted speech sequence • Calculation of the speech quality by comparing to the reference • Output: MOS LQOn – mean opinion score, objective, narrowband

22. Typical setups • Mobile Probe • automated call setup • generation of test sequences Mobile Network • Central server • automated call setup • generation of test sequences • analysis & statistics • Mobile Probe • automated call setup • generation of test sequences Fixed Network 2 Fixed Network Very difficult in monitoring: network one way transmission delay • Probe @ NTP • automated call setup • generation of test sequences • Probe @ NTP • automated call setup • generation of test sequences

23. Not Intrusive Testing - ITU-T P.563 Listening speech quality prediction based on speech signal in a call No reference file insertion required Average listening speech quality prediction on statistical evaluation of a high number of calls on the same connection Not recommended for test on a call by call basis Much less reliable than intrusive testing based on P.862 and P.863

24. The wideband challenge High quality expectation by the user Wideband must be significantly better in all quality aspects Significant contribution to quality by the terminals First attemt to certify high quality wideband speech transmission: GSMA

25. Wideband listening examples Narrowband Wideband speech noise Fullband Narrowband Wideband …in noise air interface Mobile Phone Mobile signal processing Speech Transcoder RF-Interface

26. Wideband Network Planning – ITU-T G.107.1 • The principle of the model is identical to the narrowband version G.107 • The model reflects the improvements in quality in wideband: • Maximum R = 129 (instead of 100 for narrowband) • Model provides new wideband R calculation mapping R to the range of 0-100 • All factors – R0 and Impairment factors are adapted accordingly

27. Wideband Network Monitoring Setup as in narrowband Only intrusive testing available Network monitoring is based on ITU-T P.863 POLQA ITU-T P.862 PESQ should not be used in wideband

28. Conclusions and Recommendations ITU-T provides models for planning and monitoring Network planning is essential for achieving high network QoS Network monitoring is essential to detect potential issues – e.g. changing radio conditions, changing network conditions (jitter, packet loss) Both is not sufficient to guarantee a good QoE – terminals play a major role for the overall speech quality