1、INTERNATIONAL TELECOMMUNICATION UNION)45G134 0 TELECOMMUNICATIONSTANDARDIZATION SECTOROF ITU4%,%0(/.%G0G042!.3-)33)/.G0G015!,)496/#!“5,!29G0G0!.$G0G0%THRESHOLDS OF AUDIBILITY AND INTELLIGIBILITY(Geneva, 1972; amended at Geneva, 1976, 1980;Malaga-Torremolinos, 1984; Melbourne, 1988)1 Factors which af
2、fect the crosstalk thresholdThe degree of audibility and intelligibility of a crosstalk signal depends on a large number of factors.The main factors influencing the intelligibility of the vocal crosstalk signal are listed below.1.1 Quality of transmission of telephone apparatusThe sending and receiv
3、ing loudness ratings are decisive factors. The same is true of the sidetone rating whenroom noise is present. The use of modern telephone apparatus with smooth frequency response curves is assumed.1.2 Circuit noiseThe circuit noise on the connection of the disturbed call must be taken into account.
4、The noise level ismeasured by a psophometer equipped with a weighting network for telephone circuits, as described inRecommendation O.41.1.3 Room noiseRoom noise affects the ear directly through earcap leakage between the ear and the receiver and indirectly bysidetone. Sidetone also depends on opera
5、ting conditions. Unlike circuit noise, the effect of room noise can be reducedto some extent by the user of the telephone. For this reason, and to allow for unfavourable cases, measurements on theaudibility of crosstalk have been made with slight room noise as well as with negligible room noise. Bec
6、ause theaudibility threshold is very sensitive to masking effects, “negligible” room noise means a noise level well below 10dBA. The relatively low noise level of 40 dBA has a very marked masking effect and may therefore serve as anexample of “slight” room noise.1.4 Telephone set noiseIn addition to
7、 the masking effects on crosstalk by circuit noise and room noise, the internal noise of thetelephone set in the disturbed connection has to be considered. In modern telephone sets this noise is generated in theelectronic circuitry (amplifiers, etc.) while in older sets the origin is noise from the
8、carbon microphone. The internalnoise can be expressed and treated as an equivalent circuit noise.1.5 Conversation on the disturbed connectionDuring active speech on the disturbed connection, practical levels of crosstalk are inaudible. However, beforethe conversation starts or during long pauses in
9、the conversation, it is possible for crosstalk to be heard and perhapsunderstood. In general, it would be unwise to plan on the basis that the disturbed connection is always active;accordingly, the information given in this Recommendation assumes no conversation on the disturbed connection.1.6 Cross
10、talk couplingThe intelligibility of a vocal crosstalk signal depends largely on the nature of the crosstalk coupling, which isgenerally a function of frequency.The loudness rating of the crosstalk transmission path from the speech signal present on the disturbing line tothe subscribers set subject t
11、o the disturbance can be divided into the loudness loss of the crosstalk path from thedisturbing to the disturbed line and the receive loudness rating of the disturbed subscribers set. Figure 1/P.16 illustratesthis subdivision.2 Volume V - Rec. P.16For a given speech level Vc, the intelligibility of
12、 the crosstalk signal depends on the loudness rating d + r. InRecommendation G.111, A.4.4.4, the crosstalk receive loudness rating is defined as:XRLR = RLR(set) + Lxwhere RLR(set) refers to the disturbed telephone set.The crosstalk loudness Lxis computed as a loudness rating but with the exponent m
13、= 1, which is valid near theaudibility threshold.In the absence of further information, the value of Lxmay be approximately taken as the attenuation measuredor calculated at a frequency of 1020 Hz.2 Median listener threshold of the audibility and intelligibility of vocal crosstalkThe curves in Figur
14、e 2/P.16 represent the crosstalk receive loudness rating corresponding to the threshold ofaudibility and intelligibility (XRLRt) as a function of circuit noise. For planning purposes, it is recommended that roomnoise be regarded as negligible, which represents the most unfavourable condition.Volume
15、V - Rec. P.16 3The criterion for the threshold of audibility is that the presence of a speech signal is only just detectable but thatno part of the speech can be understood. The criterion for the threshold of intelligibility is that single words or phrasescan sometimes be understood while listening
16、to a conversation.The threshold curves represent median values for the two criteria such that in each case 50% of subscribersopinions are respectively above and below the particular curve. The standard deviation for listeners has been observedto lie in the range 4 to 6 dB and a value of 5 dB is reco
17、mmended for planning purposes. Typical response curves for alarge sample of listeners for the threshold criteria are shown in Figure 3/P.16 (no circuit noise). The difference inXRLR between the two curves is about 12 dB.The results of the original experiments (from which the curves in Figure 2/P.16
18、were drawn) were expressed interms of speech level (e.g. in Volume Units (VU) and on that basis showed a satisfactory degree of coherence.However, earlier versions of Recommendation P.16 were based on the assumption that there is a fixedrelationship between the sending loudness rating and the speech
19、 level on the line. This assumption required acorrection in the range of 11 dB and is therefore not justified. Furthermore, speech levels expressed in Volume Unitsappear to differ systematically as measured in different countries on identical speech samples. Therefore, a fixedspeech level on the dis
20、turbing line is assumed, independent of the send loudness rating (SLR) of that circuit.The thresholds given in Figure 2/P.16 are based on the assumption that the speech level Vcunder normalconversational conditions is 18 dBV active speech level (measured according to Recommendation P.56) at thetermi
21、nal of the disturbing telephone set. This value is the estimated average of the conversational level in manycountries at the send end of a connection with fairly high overall loudness rating between the optimum and themaximum permitted (OLR).4 Volume V - Rec. P.16The standard deviation of talking le
22、vels is fairly high. For calculation purposes a value of = 5 dB should beused.To calculate the threshold value for a speech level different from 18 dBV, the XRLRtvalue should becorrected by the amount of the difference, with its sigh (higher levels require higher XRLR values, and vice versa).The val
23、ue XRLRtis the sum of the crosstalk path loudness loss and the receiving loudness loss on the disturbedline. In order to obtain the loudness loss of the crosstalk path, Lx, for a particular threshold value, the RLR(set) valuehas to be subtracted.In general, for any speech level and receiving loudnes
24、s rating, Lxis obtained from Figure 2/P.16 as:Lx= XRLRt RLR(set) + (18 + Vc)3 Effects of room noiseRoom noise reaches the listeners ear both by leakage under the earcap of the telephone handset and by thesidetone path. For a given sidetone the room noise can be converted to an equivalent circuit noi
25、se by means of atransmission model such as described in Supplement No. 3. A family of conversion curves with sidetone loss asparameter is found in Figure 2 of this Supplement.As an example, with a fairly high sidetone loss (the same as used in the previous version of RecommendationP.16) a level of 4
26、0 dBA room noise is equivalent to a circuit noise level of 85 dBmp. This noise level reduces thethreshold XRLR value by about 8 dB. An additional reduction will in most cases be caused by earcap leakage.However, the importance of this effect cannot be generally predicted, since it depends both on th
27、e shape of theearcap and on user habits.Volume V - Rec. P.16 54 Crosstalk probabilityWhile the curves in Figure 2/P.16 present the median values for various noise conditions, the curves ub Figure3/P.16 represent the probability of audible or intelligible crosstalk, in percent, for the negligible noi
28、se condition.Similar probability curves can be derived from the median values for any circuit noise condition by the use ofcumulative normal distributions with a standard deviation of 6 dB.In a more general case, the talker variance should also be added. The mean speech level used in the calculation
29、smay be chosen to be lower than the relatively high level assumed in Figure 2/P.16, e.g. 20 dBV, which is closer to theaverage level in the network. An example of such an overall probability calculation is given in Annex A.The threshold values of crosstalk loudness rating given in this Recommendatio
30、n can be used in different ways.One possible interpretation is to require all normal telephone connections (i.e. faulty connections excluded) to havecrosstalk conditions between the two threshold criteria. This means that, on the one hand, there is no point in requiringa higher crosstalk attenuation
31、 than the one corresponding to the audibility threshold and, on the other hand, that theintelligibility threshold should not be exceeded.Another interpretation is to set the requirement so that there is a given small probability (e.g. 5%) thatintelligible crosstalk can be encountered with negligible
32、 room noise and with the lowest circuit noise level found in thenetwork. In practice, noise conditions are more favourable in the sense that crosstalk quite often is masked by room andcircuit noise to the extent of becoming inaudible. For the average of all connections the risk of intelligible cross
33、talk willtherefore be much smaller than the given percentage for the most unfavourable condition.Crosstalk requirements may not necessarily be the same for all parts of the network. Although the maintenanceof telephone secrecy is primordial, the subscriber is more likely to make a severe judgement o
34、n crosstalk in a local calltaking place in his immediate environment and in which indiscretion due to crosstalk may have unfortunate socialconsequences. The problem of “social crosstalk” is dealt with in 1.In practice, simultaneity of speaking on the disturbing line and listening on the disturbed li
35、ne (duringconversation pauses) is not present in all cases. Information concerning this topic and showing how ti calculate theprobabilities concerned will be found in 2.As guidelines, the probabilities of subscribers encountering potentially intelligible crosstalk should not beworse (i.e. higher) th
36、an the following: own exchange calls: 1 in 1000, other calls: 1 in 100.Note The fundamentals of calculating crosstalk probability in general are considered in RecommendationG.105.ANNEX A(to Recommendation P.16)Example of probability calculationThe probability of understanding single words of a conve
37、rsation overheard by crosstalk may be calculated for alistener chosen at random from a population of subscribers. The result of such a calculation can be used as a basis forestablishing rules for, inter alia, the minimum required crosstalk attenuation between subscriber lines in a nationalnetwork.In
38、 order to demonstrate the method of using the information given in this Recommendation to calculate theprobability of encountering (intelligible) crosstalk, the following assumptions may be made:Mean speech level Vc= 20 dBV;Receive loudness rating of telephone sets RLR(set) = 6 dB;No room or circuit
39、 noise;Standard deviation of talking levels T= 5 dB;Standard deviation of listener response distribution L= 6 dB;6 Volume V - Rec. P.16Standard deviation of RLR(set) s= 1 dB.The threshold value for crosstalk intelligibility without noise, taken from Figure 2/P.16 is XRLRt= 67 dB.According to the for
40、mula at the end of 2 and with the given assumptions, the required median crosstalk pathloudness loss becomes:Lx= 67 + 6 2 = 71 dB.The total standard deviation of the probability function is:TLS222787+=. dBWith these values of Lxand , a cumulative normal distribution function as in Figure A-1/P.16 ca
41、n be drawn.The function indicates the probability that a listener can understand single words if crosstalk for a specific value of thecrosstalk path loudness loss. For example, for Lx= 75 the probability is 30%. On the other hand, to obtain only 5%probability a crosstalk path loudness loss of 84 dB
42、would be necessary. For 1% probability, 89 dB would be required,as well as 95 dB for 0.1% probability.This calculation was based on some typical values of speech level and receiving sensitivity under noise-freeconditions. Similar calculations can easily be made with other data, also including the ef
43、fects of noise. For a realisticestimation of the probability of intelligible crosstalk for subscribers in general, some statistical distribution of circuitnoise (and possibly of room noise at the subscribers locations) will have to be assumed.Volume V - Rec. P.16 7References1 WILLIAMS (H.), SILOCOCK (W. W.), SIBBALD (D.): Social crosstalk in the local area network, El. Comm.,Vol. 49, No. 4, London, 1974.2 LAPSA (P. M.): Calculation of multidisturber crosstalk probabilities, BSTJ, Vol. 55, No. 7, New York, 1976.
