Question 17/XII - Actual and preferred speech levels in telephone connections (New Question) Considering (a) that a new speech level measurement technique has been recommended (Recommendation P.56); (b) that the speech levels assumed as a design rule for FDM carrier systems (see Recommendation G.223) may not represent the actual speech levels found in most telephone networks of today; (c) that PCM equipment reacts to the peak values of signals, and that the relation between peak and RMS values may vary with the different pre- emphasis used for the telephone set response curves as well as the different subscriber line attenuation distortions encountered; (d) that subscribers expect received speech in recorded announcements to be within a range of levels they have found to be comfortable and giving good intelligibility for normal telephone speech; (e) that systems for reproducing recorded announcements are used to provide network services; (f) that systems for recording and reproducing voice messages left by one customer for receipt by another at a later time are used or will be used to provide network services such as voice mail; (g) that the level of the speech signals produced by these systems is adjustable and must be set to the proper level; (h) that customer has a preferred listening level that depends on, among other things, room and circuit noise; (i) that the speech announcement or playback systems may be located some distance (electrically) from the receiving customers and the loudness loss of the electrical path may change from call to call, The CCITT has decided to study: 1. What are the actual speech levels (RMS and peak values), activity factors and distributions found in typical telephone networks? 2. What is the preferred speech level for announcements, recorded messages and synthetic speech systems? 3. What is the preferred speech level for normal telephone connections? 4. What reference points are used to establish the level? 5. What measurement techniques and devices are recommended for measuring the speech level? Note 1 - Annex 1 provides guidance for this study. Note 2 - Attention is drawn to Supplement No. 5, Volume III.2 for information on measurements conducted in study periods 1968-72 and 1973- 76. Note 3 - Attention is drawn to Question 5/XII with respect to synthetic speech systems. ANNEX 1 (to Question 17/XII) To simplify calculations when designing carrier systems the CCITT has adopted a conventional value to represent the mean absolute power level (at a zero relative level point) of the speech plus signalling currents of -15 dBmA of which -16 dBmO is assumed for the speech signals. With an assumed speech activity of approximately 25 per cent the corresponding active speech power could be -10 dBmO. However, actual speech power distribution have a standard deviation of several dB and some information suggests that the mean of the actual distribution and the preferred value of speech level based on user opinion may be substantially lower. For example, information in contributions from Bellcore permit the following estimates: a) Preferred active speech level Preferred active speech level at 0 dB RLR -23 dBm (for noise level of -56 dBmp) (TD WP XII/3) Short-term objectives for RLR referred to 1-6 dB 0 dBr international switching point (CCITT Recommendation G.129) Preferred active speech level at 0 dBr -17 to -22 dBm international switching point b) Actual speech level Mean active speech power at 0 dB RLR -9-OLR dBm Supplement 3, Volume V Short-term objective for SLR referred to 7 to 15 dB O dBr international switching point Corresponding mean active speech level at-16 to -24 dBm 0 dBr international switching point These results would suggest that speech announcements should be in the range from -16 to -24 dBm at a 0 dBr international switching point. To facilitate the study of this Question administrations are invited to submit contributions on the following points: The preferred speech level at 0 dB RLR as a function of the circuit noise level. The actual mean active speech level and standard deviation at a 0 dBr international switching point. The mean active speech levels used for speech announcements in their networks referred to a 0 dBr switching point.