1、Technical Report No. 56January 1999Performance Guidelinesfor Voiceband Services over HybridInternet/PSTN ConnectionsPrepared by T1A1.7Working Group onSignal Processing andNetwork Performance forVoiceband ServicesCopyright 1999 by Alliance for Telecommunications IndustrySolutionsAll rights reserved.C
2、ommittee T1 is sponsored by the Alliance for Telecommunications Industry Solutions (ATIS) and accredited by the American National Standards Institute (ANSI)No part of this publication may be reproduced in any form, in an electronic retrieval system or otherwise, without the prior written permission
3、of the publisher.A Technical Report onPerformance Guidelines for Voiceband Services overHybrid Internet/PSTN ConnectionsAbstractThis Technical Report provides transmission performance guidelines for voiceband services overnetworks that are an interconnection of an IP network (e.g., “the Internet”) a
4、nd the PSTN. Asmodern telecommunications systems continue to evolve, hybrid networks of various kinds willbecome more prevalent. This tendency, together with the reality that voiceband communicationswill be an important source of traffic on such networks, has led T1A1 to develop this TechnicalReport
5、. The intent is to provide to a wide audience a set of principles on transmission performanceaspects of hybrid Internet/PSTN connections.Document T1A1.7/98-002r6Prepared byT1A1.7Working Group on Signal Processing and Network Performance forVoiceband ServicesiForewordThis Technical Report was initiat
6、ed in the T1A1.7 Working Group on Signal Processing andNetwork Performance for Voiceband Services. T1A1.7 recognized the need for guidance intransmission performance for voiceband services on evolving telecommunications networks. Inparticular, the need for such guidance when traditional voiceband se
7、rvices are carried on hybridconnections consisting of IP networks and the PSTN was noted. At the time this report wascompleted, T1A1.7 had the following members:D.J. Atkinson Maurice Givens Max J. Roesch, Jr.Chuck Balogh Ricky B. Grams John RoquetDick Bobilin Eric Hauch Glenn J. ScheydLorence F. Bro
8、wn Ralph E. Jensen Ernest SchmidtE.Mel C Celi Michael Kalb Dan ThomasGregory Cermak Gary Knippelmier Pat TweedyAntony Crossman Mustafa Kocaturk Stephen VoranChuck Dvorak Gary Koerner Arthur WebsterGeorge Fawcett, Jr. Barry Lerich Bernard WorneMichael Floyd A.C. Morton William R. WycoffSuhas S. Gandh
9、i Tom Oshidari Sangamesh VinayagamurtyTom Geary Mark E. Perkins J.A. ZebarthiiiContents1 Scope, Purpose, and Application . 11.1 Scope 11.2 Purpose . 11.3 Application . 12 Terminology . 33 Acronyms now the subject of draftITUT Recommendation G.107equipment impairmentfactora parameter of the E-Model,
10、used to capture the impact of impairmentscaused by a specific piece of transmission equipment (e.g., a speechcodec)hybrid PSTN/Internet a connection that includes at least one segment where traffic is carriedon the PSTN and at least one segment where traffic is carried on anetwork that uses the Inte
11、rnet Protocol suitegateway/IWF a connection element that interconnects different networks and performsthe necessary translation between the protocols used on those networksinterwork the ability of two networks to be connected and transfer traffic from oneto the other3 Acronyms percentage of speech c
12、lipped; frequency of clipping; and overall speechactivity. Based on the results of detailed subjective tests, two guidelines ( J. Gruber and L.Strawczynski, “Subjective effects of variable delay and speech loss in dynamically managed voicesystems,” IEEE GLOBECOM 82, Vol. 2: F.7.3.1-F.7.3.5, Miami, F
13、L, Nov.-Dec. 1982) tomaintain good speech quality are:clipping of speech segments 64 ms should always be avoided, andclipped segments 64 ms should be kept below 0.2 percent of active speech.These guidelines have recently been included in T1 Technical Report 19 (Wireless PersonalCommunications) and i
14、n Draft ITUT Recommendation G.116.ECH.323 Internet PSTN EndOffice PhoneTechnical Report No. 56157.6 Environmental (Acoustic) NoiseAcoustic noise at the transmitting end of a connection will have negative impact on theperformance of speech coders. The speech coders identified in Table 2 have been tes
15、ted for theeffects of environmental noise at the sending end of the connection. However, some codecs havebeen tested more extensively than others. All have been shown to be fairly robust under conditionsthat included addition of circuit noise or speech babble. If a particular type of background nois
16、ewill be dominant in a given application, it is advisable to verify that performance of the speechcodec is satisfactory under those conditions.7.7 Idle Channel NoiseIdle channel noise in VoIP applications should be negligible under most circumstances. If present,however, background idle channel nois
17、e should be less than 22 dBrnC0, a value consistent withANSI/EIA/TIA-579.7.8 Noise Contrast and Comfort NoiseNoise contrast occurs when background noise is interrupted due to digital speech processing, suchas echo cancellation using center clippers, and voice activity detection (silence removal). Co
18、mfortnoise is noise that can be introduced to mask the negative effects of noise contrast.Recommendations on noise contrast limits, and comfort noise values, are for future study.For comfort noise insertion, some digital cellular systems (e.g., GSM) use an approach where noiseparameters are extracte
19、d at the sending end and transmitted to the receiving end at a low bit rate. Itis then possible to reconstruct (to good approximation) the background noise. This approachshould provide superior subjective performance for voice users of circuits using voice activitydetection and comfort noise inserti
20、on. The voice activity detectors and comfort noise generatorsdescribed in Annex B to Rec. G.729 and Annex A to Rec. G.723.1 both operate in this fashion.The best (subjective) performance will be realized when the noise inserted at the receiving endmatches, as closely as possible, the background nois
21、e at the sending end. The following commentson CNGs can be made: the noise used should match the background noise, both in frequency content and level; level of the inserted noise should match that of the background noise; appropriate levelmeasurements and adjustments should be done using dBm0p; the
22、 time course of changes in the level of the inserted noise should match, as closely aspossible, the level changes that occur in the background noise.7.9 BandwidthTo maintain good speech quality and intelligibility, a minimum passband of 300-3400 Hz (3 dBpoints) should be delivered. For non-waveform
23、coders, traditional measurement methods usingsingle-frequency sine waves may not be adequate to evaluate effective bandwidth and levelstability. At the date of this Technical Report, there are no industry-accepted methods to assessbandwidth of these nonlinear systems.Technical Report No. 56167.10 St
24、ability LossFor VoIP systems interfacing digitally to the PSTN, a minimum loss of 6 dB is recommendedbetween the digital input and output paths of the VoIP system at the access port of the terminal.This guideline is provided to assure that singing does not occur when the handset terminal is usedunde
25、r conditions different from those to which the TCL W measurement applies (e.g., placing ahandset on a hard surface should not cause singing).7.11 DistortionDistortion in packetized systems such as Internet Telephony will be due primarily to the operationof speech codecs. It is essential that high qu
26、ality speech codecs are used and that they have beenthoroughly tested (subjectively) to ensure that there are no annoying effects.7.12 Overall Listening QualityOverall assessment of the speech transmission quality of a hybrid Internet/PSTN service isadvisable. While it is important to assess the spe
27、ech quality of various system components (e.g.,speech coders, echo cancellers), the cumulative effects of impairments from multiple speech signalprocessing devices will be the limiting factor in determining user acceptability of a new service.8 Non- Voice Transmission (End-to-end)This section provid
28、es an indication of transmission performance needs for voiceband data and othernon-voice applications in VoIP services.8.1 Applicable StandardsPerformance of the voiceband data channel is currently stated in terms of the prevalenttransmission metrics defined at the analog interfaces on a connection.
29、 These metrics include signal-to-noise ratio, attenuation distortion, envelope delay distortion, intermodulation distortion, phasejitter, etc. It should be noted that IP networks will probably introduce additional impairments totransmission that are not covered by the above parameters. Section 8.3 ,
30、 below, indicatesapplications where the operation of the gateway contributes to these end-to-end considerations. Thefollowing documents provide good general guidance for voiceband channels from an end-to-endperspective:Network Performance Criteria (TR30.3/86-11-109; also presented in Table B-3of App
31、endix B in ANSI T1.506-1997) . This list of telephone channel impairmentparameters and limits was produced by TIA TR-30.3 and sets down end-to-endrequirements for voiceband data transmission performance needed for applicationsrunning at speeds up to 9.6 kbit/s.Network Performance - Switched Exchange
32、 Access Network TransmissionSpecifications (ANSI T1.506-1997). This document contains information on levelsof voiceband data impairments that occur on end-office (EO) to point oftermination (POT) connection segments. This standard also lists service affectinglimits (SALs) for voiceband data impairme
33、nt parameters on exchange access.Technical Report No. 5617Tandem Encoding Limits for 32 kbit/s ADPCM (ANSI T1.501-1994). Thisstandard discusses implementation guidelines for 32 kbit/s ADPCM in end-to-endtelephone connections and channel segments.ITUT Recommendation G.113 (02/96). This international
34、standard has a numberof annexes that report on voiceband data (VBD) transmission performance interms of impairment parameters. Definitions and some typical values for VBDimpairments, as well as a discussion of 32 kbit/s ADPCM as related to VBDperformance, are provided. Annex D makes general statemen
35、ts about the ability ofvarious low bit-rate speech codecs to support different bit rates of voiceband data.8.2 Application RequirementsUser applications must be considered to determine desirable performance levels for hybridInternet/PSTN connections. It is possible to classify most applications in a
36、 few broad categoriesdepending on their accuracy performance requirements. Required modem performance, as afunction of application, maps into related network performance requirements. Table 3 gives aclassification of applications by accuracy parameter and required limit.Table 3 : Modem accuracy perf
37、ormance as a function of application.Application Parameter LimitStringentTypicalForgivingBERBER1000-bit BLER10 -610 -510 -2Stringent applications, such as data that are transmitted without error protection, need the lowesterror ratios for the bit error ratio (BER), the most critical parameter. Typic
38、al applications, such asfacsimile, can accept a more relaxed limit than stringent applications, but still need expression interms of the BER parameter. Lastly, forgiving applications, such as bulk data (with errorprotection) using a 1000 bit block, need a block error ratio (BLER) limit derived from
39、throughputconsiderations.8.3 Applications SupportThere are a variety of user applications used on the PSTN that must continue to operate properlyon hybrid connections. These include facsimile, encryption of voice and data (e.g., STU-III),ASCII file transfer, and use of special terminals. Packet loss
40、 and/or effects of low bit rate speechcoding may limit the success of many of the more stringent applications. Difficulties associatedwith low bit rate coding require special care to ensure that these applications continue to operate atlevel that is satisfactory to end-users. These applications incl
41、ude:Technical Report No. 5618 FaxProper fax transmission over hybrid Internet/PSTN connections will require specialconsideration. The role of the gateway will be central to this process. Each of the scenariosshown in Figure 2 must be considered. DTMFDTMF signals from the VoIP terminals may be used t
42、o interact with DTMF-basedservices such as message retrieval. Thus, the VoIP system should support good end-to-endtransmission of DTMF signals. Since some speech codecs will corrupt DTMF signals, specialconsiderations may be necessary to ensure acceptable DTMF transmission. For an example oftypical
43、DTMF requirements, see ANSI/TIA/EIA/464-B-96. Call Progress SignalsIt is also expected that call progress signals, such as audible ringbackand busy, not be seriously degraded by the VoIP system. Detailed guidelines are for furtherstudy. TTY Devices and TDD Devicesthe very low bit rate terminals are
44、used on the PSTN andmay find use on hybrid connections.9 Future Work and Request for User/Industry FeedbackT1A1 plans to continue work on transmission of voiceband services on hybrid Internet/PSTNnetworks in other areas of performance (e.g., traffic, availability), in addition to doing further worko
45、n transmission performance. In view of this ongoing work, members of the telecommunicationsindustry and members of the user community are encouraged to provide information and feedbackto T1A1 so that we can include your needs in formulating the direction and content of our work. Inparticular, inform
46、ation on user expectations regarding the performance of VoIP systems isespecially solicited.Technical Report No. 5619Appendix A Example E-Model CalculationsTable 4 shows sample E-Model calculations for several IP telephony scenarios. An “impairment-free” connection with 50 ms one-way delay is includ
47、ed as a reference. Such a connection meets thedelay allocation for a national segment as defined in ITUT Rec. G.114.The only E-Model parameters that vary are codec and delay. All other parameters are set to thedefault values shown in Table 5. In particular, effects of channel errors, packet loss and
48、 delayvariability are not included.One-way delay for a given IP telephony system is determined by the network delay and the numberof encoded frames that are included in each IP packet. Since a lost packet will result in speechclipping, the number of coded frames that are allowed in a packet is assum
49、ed to be less than 64 msof speech (see section 7.5 ). Hence, at most two frames of G.723.1 or six frames of G.729A areassembled into a packet. In Table 4 , the “Packet Delay” column is the time required to assemble apacket, and is computed as,)1( LF TNT +where FT is the frame size for the codec, N is the number of frames in a packet, and LT is thelook-ahead time for the codec. For G.723.1 (both bit rates), the frame size is 30 ms and the look-ahead is 7.5 ms. For G.729A, the frame size is 10 ms and the look-ahead is 5 ms. This is anoptimistic, yet realistic, estima
