1. A Brief Look at VoIP QoS

2. Why is VoIP QoS important? PSTNs have played the first role in voice communications for a long time. But future is going to be different. It seems that VoIP will become the first choice in voice communications. Albeit on one condition: it must be able to provide a QoS as good as PSTNs’.

3. VoIP QoS Before establishing a call Setup Delay Blocking Probability After establishing a call Voice Quality Delay Echo Etc.

4. VoIP traffic VoIP traffic is one of those kinds that must be provided with real-time bandwidth. Thus in order to have good QoS, network manager must guarantee real-time bandwidth for VoIP.

5. Voice Quality is Subjective The concept of voice quality is some how subjective. It varies person to person and depends on many factors such as person’s mood, person’s definition of a good voice quality, etc.

6. Voice Quality Definition One of the important factors in VoIP QoS is Voice Quality (VQ). Despite the subjectivity of VQ, it can be quantified according to three parameters: Service quality Sound quality Conversation quality

7. Service Quality Service quality depends on services that company provides users with. These services include: Offered Services Availability of users in foreign countries Network availability, down time, Busy time Reliability Price

8. Sound Quality By the term sound quality we mean: Loudness Distortion Noise Crosstalk Fading

9. Conversation Quality Conversation quality manly depends on: Echo End-to-end delay Loudness distortion noise Fading Crosstalk Silence suppression performance

10. Primary Factors Among the parameters mentioned in previous slides, three parameters are more important than others: Clarity End-to-end delay Echo

11. Clarity By the term clarity we mean clearness and distortion-less. There are some parameters that affect clarity: Packet loss Speech codecs Noise Echo Voice Activity detectors (VADs) External factors

12. Packet Loss As the network, becomes congested, router buffers fill and start to drop packets. Route changes as a result of inoperative network links. Packet experiences a large delay in the network and arrives too late to be used in reconstructing the voice signal.

13. Solutions for Packet Loss Assure minimum throughput for selected applications. Prioritization (Classification) & Router flow control

14. Speech Codecs A speech codec transforms analog voice into digital bit streams, and vice versa. Some codecs perform compression as well. A codec’s effect on VQ is influenced by compression, packet size, packet loss, and any error-correction mechanisms used by the codec itself.

15. Other Factors Noise –Noise can originate from analog lines or from bit errors on data transmission lines. Voice Activity Detectors Echo External Environmental Factors Room noise, end-user mood, end-user expectations, etc. may cause the audio quality be perceived as unacceptable.

16. End-to-End Delay There are some factors that affect end-to-end delay: PSTN delay IP network delay Packet capture delay Switching/Routing delay Jitter Buffer Delay VoIP devices (gateways, terminals, etc.) delays

17. End-to-End Delay (contd.) Certain amount of delay is acceptable in voice communication. But there is an upper threshold for this value. Below 100 ms: not detected by user Between 100 and 300 ms: user notices the delay Above 300 ms: delay becomes annoying

18. Echo Echo is mainly caused by electrical mismatch between analog telephony equipment and transmission lines. Echoes from near end circuits does not degrade VQ so much. But echoes from far end circuits affect VQ in a high degree.

19. Echo (contd.) Relationship between echo levels, delay and perception is shown in figure below

20. Echo (contd.) Echo can easily be cancelled by the help of echo canceller blocks or by using digital circuits. Performance of these blocks depend on doubletalk and echo return loss.

21. Measuring QoS Two measurement units: MOS(ITU-T P.800) and E-Model Rating Value (R Value) (ITU-T G.107)

22. Measuring QoS (contd.) In order to measure QoS, operator must measure certain parameters: Clarity Delay Echo

23. Measuring Clarity The mostly used method for measuring clarity is to employ human listeners, measure mean opinion score (MOS) and the compare it with standards such as P.800. Another method is Perceptual Speech-Quality Measurement (PSQM) defined by ITU in P.861. The third method is Perceptual Analysis Measurement System (PAMS).

24. Measuring Delay There are tow methods for measuring end-to-end delay: Acoustic Ping MLS Normalized Cross-Correlation (MLS stands for Maximum Length Sequence)

25. Measuring Echo Measure Echo Delay Measure Echo Return Loss (ERL) According to ITU-T G.168 ERL should be greater than 55dB Measure echo canceller performance Convergence time Cancellation depth Doubletalk robustness Perceived Annoyance Caused by Echo (PACE)

26. Trade offs

27. Some Numerical Results Usually G.711 is used since it is less sensitive to packet loss. Moreover, there exist some Packet Loss Recovery (PLR) algorithms for G.711. Example of codec impact on rate

28. Example of packetization period (packet size) trade off

29. Delay Budget 150ms is often targeted for end-to- end delay, with around 100ms assigned to PSTN long-distance propagation delay. The result in this case is a remainder of 50ms for local VoIP path delay.

30. Delay Budget (contd.) Jitter buffer size, packetization period, and packetization-related grant uncertainty combined with other delays to form a VoIP path delay. The following figure provides an example of one-way VoIP path delay (upstream and downstream for an “on-net” call). The delay is broken down by component and sub-component contribution and assumes 10ms packetization and 15ms jitter buffer play-out delay.

32. Conclusion: Providing Better VoIP QoS Supporting dedicated bandwidth Improving loss characteristics Avoiding and managing network congestion Setting traffic priorities across the network Echo cancellation Minimizing delay

33. Thanks Any Questions?

34. References 1)Voice Quality in Converging Telephony and IP Networks, the International Engineering Consortium (available at www.iec.org/online/tutorials/acrobat/voice_qual.pdf) www.iec.org/online/tutorials/acrobat/voice_qual.pdf 2)Voice QoS, Jeremy Chan (available at www.twnic.net.tw/seminar2004/html/29d.ppt) www.twnic.net.tw/seminar2004/html/29d.ppt 3)Engineering CMTS and HFC for VoIP with Capital and Operating Expenses in Mind, Motorola Corporation, (available at http://broadband.motorola.com/ips/pdf/MTA-VoIP.pdf)http://broadband.motorola.com/ips/pdf/MTA-VoIP.pdf 4)Overcoming Barriers to High-Quality Voice over IP Deployments, Intel Corporation, 2003

35. Glossary CMTS: Cable Modem Termination System CNG: Comfort Noise Generator DOCSIS: Data Over Cable Service Interface Specification HFC: High Frequency Communication MG: Media Gateway MLS: Maximum Length Sequence MTA: Multimedia Terminal Adaptor PSTN: Public Switched Telephone Network QoS: Quality of Service VAD: Voice Activity Detectors VQ: Voice Quality