1. Key QoS parameters for Voice Services Workshop on “Monitoring Quality of Service and Quality of Experience of Multimedia Services in Broadband/Internet Networks” (Maputo, Mozambique, 14-16 April 2014) Martin Brand Vice Chair ITU-T Study Group 11 Joachim Pomy Consultant@joachimpomy.de OPTICOM, Germany

2. Maputo - Mozambique - 14 - 16 April 2013 2 Components and related QoS category with voice quality functions 

3. Maputo - Mozambique - 14 - 16 April 2013 3 Applicability of existing QoS metrics and measurement to PSTN-IP-PSTN and IP network impacts

4. Maputo - Mozambique - 14 - 16 April 2013 4 ITU-T G.114 ITU-T G.107  Highly interactive tasks (e.g., some speech, video conferencing and interactive data applications) may be affected by delays below 100 ms.  For many intra-regional (e.g., within Africa, Europe, North America) routes in the range of 5000 km or less, users of VoIP connections are likely to experience mouth-to-ear delays < 150 ms.  Absolute delay does not impair the intelligibility of speech but if the total delay exceeds around 100 ms from mouth to ear, it begins to affect the interactivity of conversations. Therefore, if possible, large delays should be avoided.

5. Maputo - Mozambique - 14 - 16 April 2013 5 Contributing Factors Delay - Codec - Packet frame duration - DSP/CPU processing time - Play out buffer - Link speed (serialization) - Propagation delay - De-jitter buffer delay - PLC - Codec - Jitter - Use of VAD - Lost of packets - Transcoding - Sound level - Echo level Distorsion

6. Maputo - Mozambique - 14 - 16 April 2013 6 E-Model Transmission Rating (R) ITU-T Rec. G.107  R summarizes the effects of network impairments including delay, distortions signal level & echo level R= Ro- Is - Id- Ie + A Simplified E-Model for codec and packet loss/burst impairments R = 93.4 – Ie Ro= basic signal to noise ratio Is= level impairments and distortion impairments Id= delay impairments Ie= codec and packet loss/burst impairments A= advantage factor

7. Maputo - Mozambique - 14 - 16 April 2013 7 Provisional planning values for the equipment impairment factor (Ie) Codec typeReferenceOperating rate kbit/sIe value ADPCMG.726, G.727402 G.721(1988), G.726, G.727327 G.726, G.7272425 G.726, G.7271650 LD-CELPG.728167 12.820 CS-ACELPG.729810 G.729-A  VAD 811 VSELPIS-54820 ACELPIS-6417.410 QCELPIS-96-A819 RCELPIS-12786 VSELPJapanese PDC6.724 RPE-LTP GSM 06.10, Full ‑ rate 1320 VSELP GSM 06.20, Half ‑ rate 5.623 ACELPGSM 06.60, Enhanced Full Rate12.25 ACELPG.723.15.319 MP-MLQG.723.16.315

8. Maputo - Mozambique - 14 - 16 April 2013 8 Codec combination

9. Maputo - Mozambique - 14 - 16 April 2013 9 E-Model Prediction of Echo and Delay Impairment

10. Maputo - Mozambique - 14 - 16 April 2013 10 Relation between R-value and user satisfaction R ValueMOS CQEN Value Categories of User Satisfaction 944,42 934,40 924,38Very satisfied (Best) 914,36 904,34 874,195 854,18 824,09 814,06Satisfied (High) 804,03 773,85 733,74 703,60Some users dissatisfied (Medium) 683,50 603,10Many users dissatisfied (Low) 502,58Nearly all users dissatisfied (Poor) MOS = 1 + (0,035) × R + (000 007) × R (R - 60) (100 - R) NOTE 1:Connections with R-values below 50 are not recommended. NOTE 2:Although the trend in transmission planning is to use R-values, equations to convert R ‑ values into other metrics e.g. MOS, % GoB, % PoW, can be found in ITU-T Recommendation G.107 [i.4], annex B.

11. Maputo - Mozambique - 14 - 16 April 2013 11 Provisional Planning Values for the Equipment Impairment Factor Ie under Conditions of Packet Loss for Codecs G.711, G.729A + VAD and G.732.1 + VAD

12. Maputo - Mozambique - 14 - 16 April 2013 12 Subjective measurement  Based on subjective experiments  Mean opinion score  Costly and time consuming Objective measurement  Good correlation with subjective measurement  Highly repeatable  Real-time 5 Excellent 4 Good 3 Fair 2 Poor 1 Bad MOS MOS - Speech Quality Measure

13. Maputo - Mozambique - 14 - 16 April 2013 13  Active voice quality measurement  Active voice quality measurement techniques, also known as intrusive models, require a test call to be made over the network. A comparison of reference and degraded signals produces a quality score based on MOS. This type of measurement is useful for the preinstallation testing of a system as well as network optimization.  Passive voice quality measurement  Common to all communication technologies are Passive voice quality measurement. Passive voice quality measurement is also known as a non-intrusive measurement.  These models monitor live network traffic and again measure the voice quality on the MOS scale. Passive measurement is ideal for the testing of systems already in use QoS Voice Testing (1)

14. Maputo - Mozambique - 14 - 16 April 2013 14  PSQM ITU-T Rec. P.861 - designed for codec evaluation. Assesses error loudness, noise disturbance and asymmetry to predict a PSQM value Withdrawn February 2001  PAMS - developed for real world networks Assesses time aligned, level aligned, spectrally weighted error surface  PESQ © - ITU-T Rec. P.862 Perceptual Evaluation of Speech Quality Designed for network avaluation, using specific scale  PESQ© - ITU-T Rec. P862.1 (PESQ extension from Nov.2003)‏ Linear mapping to P.800 MOS scale QoS Voice Testing (2)

15. Maputo - Mozambique - 14 - 16 April 2013 15  PESQ© - ITU-T Rec. P862.3 [2007] (Implementation guide, Methods for objective and subjective assessment of quality)‏  POLQA – ITU-T Rec. P.863 [2011]  3SQM, perceptual single-sided speech quality measure (non-intrusive) according to ITU-T rec. P.563 [2004] QoS Voice Testing (3)

16. Maputo - Mozambique - 14 - 16 April 2013 16 The instance of media-information quality assessment methods comparison Accuracy Cost HighGoodAverageMethod type HighPESQ, POLQA  Active Average  P.563  Passive Low  E-model Passive/ Modelling

17. Maputo - Mozambique - 14 - 16 April 2013 17 GUIDANCE ON OBJECTIVES FOR QUALITY RELATED PARAMETERS AT VOIP SEGMENT- CONNECTION POINTS RESULTS PUBLISHED IN ETSI TR 102 775 V.1.6.1 17

18. Maputo - Mozambique - 14 - 16 April 2013 18 Scope of the development  To provide guidance on the quality parameters that need to be considered at the Segment-connection of Voice over IP (VoIP) services and provides guidance on objectives for these parameters.

19. Maputo - Mozambique - 14 - 16 April 2013 19 Reference Configuration

20. Maputo - Mozambique - 14 - 16 April 2013 20 Generic Segment-connection Points

21. Maputo - Mozambique - 14 - 16 April 2013 21 End-to-End delay values between originating and terminating Service Provider premises West Europe Vienna North Europe RU – Vladimir RU – Moskwa SE - Östersund East Europe RO – Bucharest HU – Budapest South Europe GR–Athens IT - Roma East Asia (CN- Dongguan) South Asia Malaysia - MY Oceania Australia -AU Sydney N. America East Cost US Washington DC N. America West Cost US Vancouver Central America Panama - PA Mexiko- MEX South America Brasilia - BR Africa ZA - Cap town Burkina Fasso West Europe Bern Paris Frankfurt DE - AT (Frankfurt – Vienna) 7 ms CH – DE (Bern – Frankfurt) 8 ms CH - AT (Bern – Vienna) 10 ms FR- AT (Paris -. Vienna) 15 ms CH - RU (Bern – Vladimir) 23 ms CH - RU (Bern – Moskwa) 22 ms DE- RU (Frankfurt- Moskwa) 21 ms DE-SE (Frankfurt – Östersund) 12 ms CH-RO (Bern – Bucharest) 25 ms DE-RO (Frankfurt – Bucharest) 18 ms DE-HU (Frankfurt – Budapest) 12 ms CH-GR (Bern – Athens) 30 ms DE-GR (Frankfurt – Athens) 24 ms IT-CH (Roma – Cern) 22 ms Frankfurt – Roma 13 ms CH – CN (Bern- (Dongguan) 150 ms DE-CN (Frakfurt – Dongguan) 160 ms CH- MY 110 ms DE – MY 115- 164 ms CH-AU (Bern – Sydney) 170 ms DE-AU (Frankfurt – Melburne) 165 ms DE-AU (Frankfurt – Sydney) 146 ms CH- US (Bern – Washington) 46 ms DE-US (Frankfurt – Washington) 55 ms CH-US (Bern – Vancouver) 75 ms DE-US (Frankfurt – Vancouver) 75 ms CH-PA (Bern – Panama) 103 ms DE-MEX (Frankfurt – Mexiko City) 75 ms CH-BR (Bern – Sao Paulo) 136 ms DE-BR (Frankfurt – Sao Paulo) 125 ms CH-ZA (Bern - Cape Town) 108 ms DE-ZA (Frankfurt – Cape Town) 90 ms BE-BF 90 – 130 ms North Europe (SE- Östersund) SE - RU (Östersund – Moskwa) 10 ms SE-RO (Östersund – Bucharest) 30 ms SE-GR (Östersund- Athens) 42 ms SE-CN 150 ms SE-MY 124 ms SE-AU 171 ms SE-US 62 ms SE-US 83 ms SE- PA 84 ms SE-MEX 78 ms SE-BR 155 ms SE-ZA 87 ms East Europe (HU- Budapest) HU- RU (Budapest – Moskwa) 33 ms HU – RO (Budapest- Bucharest) 14 ms HU – GR (Budapest – Athens) 20 ms HU-CN 166 ms HU-MY 125 ms HU-AU 170 ms HU-US 60 ms HU-US 100 ms HU-PA 95 ms HU- MEX 96 ms HU-BR 152 ms HU-ZA 100 ms South Europe (IT- Roma) IT- RU (Roma – Moskwa) 38 ms IT-RO (Roma – Bucharest) 32 ms (IT-GR) Roma – Athens 23 ms IT-CN 160 ms IT-MY 150 ms IT-AU 160 ms IT-US 55 ms IT-US 95 ms IT – PA 104 ms IT- MEX 113 ms IR-BR 108 ms IT-ZA 95 ms

22. Maputo - Mozambique - 14 - 16 April 2013 22 Reference Configurations (1) Backbone NGN PSTN/ISDN access Configuration PSTN/ISDN classic access Configuration ADMAdd-Drop-Multiplexer DLrouter Distribution Layer CLrouter Core Layer TRAUTranscoder and Rate Adaption ETHEthernet Unit

23. Maputo - Mozambique - 14 - 16 April 2013 23 Reference Configurations (2) Access DSL Configuration Access configuration from UMTS Release 4

24. Maputo - Mozambique - 14 - 16 April 2013 24 Delay with regional propagation delay (1 400 km / 7 ms) Uplink / Downlink 256 kbit/s 384 kbit/s 512 kbit/s 768 kbit/s 1024 kbit/s 1152 kbit/s 1536 kbit/s 2048 kbit/s 2304 kbit/s 3072 kbit/s 6144 kbit/s POTS -----------55 ms R=91 256 kbit/s128 ms R=90 118 ms R=90 113 ms R=90 108 ms R=90 104 ms R=90 103 ms R=91 102 ms R=911 101 ms R=91 101 ms R=91 100 ms R=91 97 ms R=91 82 ms R=91 384 kbit/s-108 ms R=90 102 ms R=91 97 ms R=91 95 ms R=91 93 ms R=91 94 ms R=91 92 ms R=80 92 ms R=91 91 ms R=91 91 ms R=91 75 ms R=91

25. Maputo - Mozambique - 14 - 16 April 2013 25 Delay with regional propagation delay (1 400 km / 7 ms) Uplink / Downlink 256 kbit/s 384 kbit/s 512 kbit/s 768 kbit/s 1024 kbit/s 1152 kbit/s 1536 kbit/s 2048 kbit/s 2304 kbit/s 3072 kbit/s 6144 kbit/s POTS -----------55 ms R=91 256 kbit/s128 ms R=90 118 ms R=90 113 ms R=90 108 ms R=90 104 ms R=90 103 ms R=91 102 ms R=911 101 ms R=91 101 ms R=91 100 ms R=91 97 ms R=91 82 ms R=91 384 kbit/s-108 ms R=90 102 ms R=91 97 ms R=91 95 ms R=91 93 ms R=91 94 ms R=91 92 ms R=80 92 ms R=91 91 ms R=91 91 ms R=91 75 ms R=91 G.711/20 Uplink / Downlink 256 kbit/s 384 kbit/s 512 kbit/s 768 kbit/s 1024 kbit/s 1152 kbit/s 1536 kbit/s 2048 kbit/s 2304 kbit/s 3072 kbit/s 6144 kbit/s POTS 512 kbit/s --97 ms R=91 92 ms R=91 91 ms R=91 91 ms R=91 91 ms R=91 91 ms R=91 90 ms R=91 90 ms R=91 90 ms R=91 74 ms R=91 768 kbit/s ---91 ms R=91 90 ms R=91 90 ms R=91 90 ms R=91 89 ms R=91 89 ms R=91 89 ms R=91 89 ms R=91 73 ms R=91 1024 kbit/s ----90 ms R=91 89 ms R=91 89 ms R=91 89 ms R=91 89 ms R=91 88 ms R=91 88 ms R=91 73 ms R=91 2048 kbit/s -------88 ms R=91 88 ms R=91 88 ms R=91 88 ms R=91 88 ms R=91 GSM202 ms R=81 196 ms R=81 194ms R=82 193 ms R=82 193 ms R=82 192 ms R=82 192 ms R=82 192 ms R=82 192 ms R=82 191 ms R=82 191 ms R=82 175 ms R=83 UMTS Rel.4207 ms R=80 201 ms R=81 199 ms R=81 198 ms R=81 198 ms R=81 197 ms R=81 197 ms R=81 197 ms R=81 197 ms R=81 196 ms R=81 196 ms R=81 180 ms R=82 LTE163ms157 ms 155 ms 154 ms 153 ms 152 ms 136 ms

26. Maputo - Mozambique - 14 - 16 April 2013 26 Relation between R-value and user satisfaction R ValueMOS CQEN Value Categories of User Satisfaction 944,42 934,40 924,38Very satisfied (Best) 914,36 904,34 874,195 854,18 824,09 814,06Satisfied (High) 804,03 773,85 733,74 703,60Some users dissatisfied (Medium) 683,50 603,10Many users dissatisfied (Low) 502,58Nearly all users dissatisfied (Poor) MOS = 1 + (0,035) × R + (000 007) × R (R - 60) (100 - R) NOTE 1:Connections with R-values below 50 are not recommended. NOTE 2:Although the trend in transmission planning is to use R-values, equations to convert R ‑ values into other metrics e.g. MOS, % GoB, % PoW, can be found in ITU-T Recommendation G.107 [i.4], annex B.

27. Maputo - Mozambique - 14 - 16 April 2013 27 Guidance on Access Segment Objectives

28. Maputo - Mozambique - 14 - 16 April 2013 28 Maximal IPDV values for xDSL and ETH Access Segment ParameterValue Access Network (sending side)< 35 ms Access Network (receiving side); < 10 ms (note) NOTE:10 ms are recommended, the maximum IPDV value is 40 ms.

29. Maputo - Mozambique - 14 - 16 April 2013 29 Guidance on Objectives for Total Transit Segments ParameterValue IPDV Intra-continent Jitter Value -5 ms per Provider (maximum of 2 involved in the service delivery chain) (see note) 10 ms IPDV Inter-continent Jitter Value -10 ms per Provider (maximum of 2 involved in the service delivery chain) (see note) 20 ms IPLR3,0 × 10 -4 IPER3 × 10 -5 Ie0 NOTE:The Jitter Values are based on values contained in the GSMA document IR.3445 [i.15].

30. Maputo - Mozambique - 14 - 16 April 2013 30 Dependence of DSP/CPU processing time TriplePlay-Results IAD_2_IAD Test requerments : routed IAD-Konfiguration, worst casetwo TV Streams, TestlaborHSI- Connection and activated WLAN, VP-voice upstream Bandwith 384 kBit/s. Supplier : 1234 IAD_2_IAD MOS Value QoS two wire (PSTN) 10 ms packetization 20 ms packetization 4,2 4,0 4,2 4,1 Delay / Delay Range/ Std-Dev (each direction max values) 10 ms packetization 20 ms packetization 58/29/10 66/35/11 86/10/4 95/0/0 111/10/4 114/13/6 63/17/6 66/10/4

31. Maputo - Mozambique - 14 - 16 April 2013 31 Delay Objectives for BEST (G.109) voice communication quality (R > 90) and Access Network Jitter < 35 ms

32. Maputo - Mozambique - 14 - 16 April 2013 32 Number of channels on a 2.048 Mbit / s system

33. Maputo - Mozambique - 14 - 16 April 2013 33 Bandwidth calculations and prioritization in VoIP systems The IP delay of VoIP packets over a link transporting Voice and Data depends on following factors: the instantaneous system load (not the average system load), the size of the packets (both the size of the RTP packets and IP data packets on the same links), the manner in which QoS is implemented in the system (a system with priority, such as Diff-Serv, behaves differently than a system with best effort). The delay and jitter increase linearly up to a limit of parallel VoIP channels. Above this limit, the bandwidth utilization and the delay increase exponentially and the VoIP transmissions become unstable.

34. Maputo - Mozambique - 14 - 16 April 2013 34 RTP is transferred preferentially

35. Maputo - Mozambique - 14 - 16 April 2013 35 The maximum link capacity The maximum capacity with Diff-Serv is bandwidth dependent. For example, the utilization for a 386 kbit / s link should not be higher than 85%, a 4 Mbit/s link can be utilized with 96 %. Utilization (ρ1+ ρ2)Link capacity RHO < 0.96> 4 Mbit/s RHO < 0,953 Mbit/s RHO < 0,892 - 3 Mbit/s RHO < 0,891 - 2 Mbit/s RHO < 0,850,386 - 1 Mbit/s Tests conditions: Data utilization ρ2= 62 %; Data Packet size = 1500 bytes; Codec: G.711

36. Maputo - Mozambique - 14 - 16 April 2013 36 QoS is implemented by over-dimensioning of the system Without prioritization only 40 % to 80% of the bandwidth can be used depending on the flow characteristics, as can be seen in the following figure:

37. Maputo - Mozambique - 14 - 16 April 2013 37 Any questions ? Contact: Consultant@joachimpomy.de