1、 Rec. ITU-R BT.1720 1 RECOMMENDATION ITU-R BT.1720*Quality of service ranking and measurement methods for digital video broadcasting services delivered over broadband Internet protocol networks (Question ITU-R 100/6) (2005) Summary This Recommendation specifies performance requirements and objective
2、 measuring methods of quality of service (QoS) for the delivery of digital video broadcasting services over broadband Internet protocol (IP) networks. The specified performance requirements are based on an IP QoS ranking at various levels, from “excellent” to “out-of-service”. They rely on the objec
3、tive end-to-end measurement of the values of a small number of parameters on the delivered IP streams, performed at the consumer premises equipment and relayed back to the head-end. The recommended objective measurement methods and parameters are known to influence the QoS delivered to the user. The
4、 ITU Radiocommunication Assembly, considering a) that the development of broadcast and non-broadcast television systems is being widely undertaken, and, with their development, new levels of potential image quality are available; b) that with the development of new image transmission technologies in
5、 broadcast and non-broadcast television, television system parameters can be chosen on the basis of compromises between image quality and the cost of image; c) that for the definition of the requirements for television systems and for the various sections of the service delivery chain, the potential
6、 level of image quality to be provided by these systems is an important element; d) that ISO/IEC MPEG-2 have standardized the encoding and transport mechanisms for audio, video and accompanying data which are adopted for digital video services, further considering a) that digital television services
7、 have begun to be distributed in IP broadband networks through IP multicast technologies and protocols (IP multicast distribution is analogous to broadcasting techniques in the radio transmission world); b) that, in an IP network, interactive television services, such as video-on-demand (VoD), usual
8、ly associated with a unicast content distribution method, are now available to end-users; c) that, in an IP network, video receivers decode IP delivered television channels to the TV display, *This Recommendation should be brought to the attention of Telecommunication Standardization Study Group 9.
9、2 Rec. ITU-R BT.1720 noting 1 that packet loss ratio, latency and jitter are crucial IP transport requirements for end-to-end IP network performance assessment, recommends 1 that methods for quality of service (QoS) measurements for digital television broadcasting services streamed in a broadband IP
10、 network should be tailored to the specific features of the transport services provided by an IP communications network; 2 that, for video services, the requirements in noting 1 should be measured and used for IP end-to-end network performance as described in Annex 1; 3 that end-to-end measurements
11、should be performed on the video stream after its IP packetized structure is removed as described in Annex 2; 4 that QoS should be measured end-to-end in order to provide a close approximation to the quality offered to the end-user, taking into account the influence of the IP network on the video st
12、ream; Annex 3 shows a system measurement model of a chain for IP transmission of television services. Annex 1 IP layer 1 IP transport requirements IP networks are multi-hop, may be complex and different transmission technologies are usually employed along the network paths. The transmission control
13、protocol (TCP)/IP protocol stack sees all these as “below layer 3” layers. Measurements and quality parameters at the IP layer make it possible to define reference values for network requirements that are agnostic of the underlying transmission technologies and are suitable for use in end-to-end qua
14、lity assessment. The noise introduced in an IP packet network is described by the following parameters: Packet loss ratio (PLR): The ratio between the number of the packets lost in the network and the total number of transmitted packets1. 1According to the measurement scheme and the methodology prop
15、osed in this Recommendation, the total number of lost packets in the PLR parameter is the sum of IP packet loss ratio (IPLR) and IP packet error ratio (IPER) as defined in ITU-T Recommendation Y.1541. A more complete definition of this parameter is given in ITU-T Recommendation G.1020 where 7.7.1 de
16、fines “Overall (frame/packet) loss ratio” for frames or packets. Being the measurement header on top of the transport layer, if, for an IP packet, the IP or user datagram protocol (UDP) checksum fails, this packet will not be presented to the measurement (or real time protocol (RTP) layer. Rec. ITU-
17、R BT.1720 3 Latency: The time interval between initial transmission and final reception time of a packet. Jitter: The latency variation. The quality of the video streams will impose a minimum value for the downstream throughput requirements; upstream end-to-end throughput requirements depend on appl
18、ication interactivity requirements. This Annex 1 does not guarantee that the classification that it provides is sufficient for assessing the perceived quality on a TV broadcasting over IP system, since IP end-to-end network performance is measured before forward error correction (FEC) is applied. 2
19、Video streaming IP service class Video services, such as VoD or TV services, are classified also as streaming services. In a high-quality television environment they have the following high-level requirements: good audio/video quality; high availability; medium interactivity. These high-level requir
20、ements should be translated into values for transport requirements for an IP network. As specified in Annex 3, it is up to the head-end to introduce good quality video content into the network according to the maximum end-to-end bandwidth and packet rate available for video services. Any packet loss
21、 will reduce the quality of the video. To preserve good quality of the image, a low value of packet loss is required. 3 IP transport measurements The IP network layer should be unaware if the video signal, or any upper layer, is employing forward-error correction (FEC) or any error-correction techni
22、ques, and it should only guarantee the performance needed before any error-correction scheme is applied at any of the above layers. 3.1 Parameters Table 1 lists IP network measurement parameters. All measurements should be taken from points B to point C in the system measurement model described in A
23、nnex 3. 4 Rec. ITU-R BT.1720 TABLE 1 Parameter Equipment Motivation Monitoring method PLR Customer premises equipment (CPE) (set top box (STB) Image quality, video information loss estimation In service or through test streams with RTP/real-time control protocol (RTCP) or sequence numbers available
24、on packet header Periodic PLR summary: Reports with one-minute resolution Measurements of PLR requires analysis of a number of packets at least ten times greater than the number related to the target PLR value This determines the rate at which the PLR is reported Network latency Test probe at user s
25、ide, within CPE (STB) or as closest as possible to user access link Smooth playout Test stream Jitter CPE (STB) Smooth playout In service or through test streams with RTP/RTCP or time stamps available on packet header Downstream throughput CPE (STB) Service qualification, monitoring Test signal repr
26、esentative of worst-case encoding scenario, throughput test Upstream throughput CPE (STB) Service qualification, monitoring Throughput test 3.2 Values Before giving reference values for transport requirements, it is important to note that, in video services delivery architecture, a receiver buffer i
27、s employed at the CPE (STB) end to eliminate (to some extent) the jitter introduced by the network and to have a continuous video frame reproduction. Values that should be achieved in the network are outlined and motivated in the next paragraphs. 3.2.1 PLR value It is preferable to a specify PLR val
28、ue that is “codec independent” and dimensioned on a worst-case scenario. The PLR value needed to guarantee that an IP network seamlessly delivers video services is 105. This requirement on PLR is considerably more stringent than the IPLR objectives currently specified in ITU-T Recommendation Y.15412
29、.2There are plans to support digital video transport with some new QoS classes with values of IPLR PLR_out, then service may be considered unavailable. A value of 0.01 is proposed for PLR_out3. 5 IP network service classification In relation to video services, the performance of an IP network can be
30、 classified based on the value of PLR offered to the end-user. The PLR must be measured between points B and C of the system measurement model described in Annex 3. NOTE 1 In relation to the delivery of video services, the inclusion of the effect of latency and jitter for IP network classification p
31、urposes, as well as the evaluation of the impact of the definition of an FEC system needs further study. 3This value refers to a system where no FEC is employed; further study defining the FEC scheme may, in the future, result in defining a different value for PLR_out. 6 Rec. ITU-R BT.1720 Appendix
32、1 to Annex 1 Example of an IP network service classification This Appendix provides an example of an IP network service classification. The classification used for digital television services is given below: PLR 105excellent service quality (ESQ) PLR intermediate service quality (ISQ) PLR poor servi
33、ce quality (PSQ) PLR IP end-to-end service not available. Table 2 shows IP layer service classes that are related to the QoS service perceived by the end-user. The picture quality also depends on encoding conditions (bit rate, picture size, intra-refreshing method, etc.) and transmission parameters
34、(packet size, FEC, etc.). The evaluation interval for end-to-end service availability is from 1 to 5 min. The network service classification is based on an evaluation interval of 30 min. The end-to-end performance of an IP network can then be calculated adding up the time intervals in which the meas
35、ured PLR was within the above thresholds during the reported time-slot. This is shown in the following example: TABLE 2 Class Time ESQ % Time ISQ % Time PSQ % Note A 99.8% Between 0 and 0.2 Between 0 and 0.1 To be computed in serviceB 99.8% Between 0 and 0.1 Between 0.1 and 0.2 To be computed in ser
36、viceD 99.8% To be computed in serviceThe end-to-end unavailable service time is not included in the above example. Annex 2 End-to-end measurements An IP network allows each CPE (STB) to also behave as a measurement end-point. This offers the valuable opportunity to have a measurement probe at each i
37、nstalled video CPE. Measurements and monitoring taken at the CPE are the ones closest to the users real experience of the service. Using a CPE as a measurement probe raises some point of attention since the CPE is not under the physical control of the network operator and measurements may be affecte
38、d by the users equipment (cable not well plugged, vertical cabling issues, improper use of the home network). The STB should have the capability to give additional information about the quality of the video signal Rec. ITU-R BT.1720 7 that is being decoded; receiver buffer fullness and frame rate ar
39、e two important indicators of service availability and overall performance. CPEs measurements should be used to: measure the end-to-end IP network performance; measure the network performance at any hierarchical level or aggregation point through statistical analysis and data processing exploiting c
40、orrelation among data; estimate the video quality offered to the end-user of the service; perform dedicated test sessions using test signals for qualification and troubleshooting. As an example, some network operators currently perform end-to-end measurements at all the STBs available in their resid
41、ential network, in order to evaluate end-to-end video service quality and network performance; STBs periodically send back frame rate and packet loss reports to provide a continuous quality feedback about the service in progress. 1 Video receiver measurements Table 3 shows the parameters that should
42、 be measured at video receivers to estimate video quality, as described in the system measurement model. These measurements can be used for all the assessments outlined above. TABLE 3 Parameter Value Equipment Purpose Monitoring method Measurement path(1)Video frame rate As required by the video sta
43、ndards STB Image quality In service through codec specific methods. Sampling From A to D Buffer underflows Not applicable STB Image quality, smooth playout In service, while playing video. Sampling. Measure underflows events and percentage of service time spent by the STB in an “underflow” state D B
44、uffer overflows Not applicable STB Image quality, smooth playout In service while playing video. Sampling. Measure underflows events and percentage of service time spent by the STB in an “overflow” state D Coding specific parameters Not applicable STB Image/service quality In service, while playing
45、video. Sampling Not applicable (1)See Fig. 2.8 Rec. ITU-R BT.1720 NOTE 1 Further studies should address video quality parameters which can be returned by the STB decoder and that may help in better evaluating the video reproduction process that takes place at the decoder. 2 Frame rate analysis Telev
46、ision standards may use different frame rates. The output of the decoder will produce exactly the original frame rate, except in the presence of video information loss. Measure of the frame rate at the output of the decoder, gives a rough estimate of the continuity of the service. Figure 1 shows, as
47、 an example for a 25 frame/s video stream, possible information that can be retrieved through frame rate analysis: Rec. ITU-R BT.1720 9 Annex 3 System measurement model In its simplest form, the television services distribution model, in an IP network, consists of three parts: Head-end: This include
48、s all the devices and applications needed to produce the video signals that are sent into the network. Transport network: This transports the video signal to the end-user CPEs. The CPE: This is an IP end-point (usually an STB) that decodes the video signal and displays it on a television set normall
49、y connected to it. Explicit service level agreements (SLAs) need to be established for the transport of the video streams between the head-end and the transport network (in particular between the service provider and the network operator if they are not the same). Audio, video, data and interactive services can be delivered in the IP transport network if the head-end and the STBs provide the necessary compliance. All the services and standards are compatible with the TCP/IP stack; the IP network should guarantee the required performance level and
