JBL Technical Note - Vol.1, No.26 电路原理图.pdf

1. Introduction: The purpose of a sound system is to trans- mit information. In the case of public address, paging, voice alarm and speech reinforcement systems the object is to transmit intelligible speech to listeners and intended message recipients. This aspect is far more important than sound quality per se, since there is no point in designing a system if it can not be understood or is incapable of getting the mes- sage across. Although sound quality and speech intelligibility are inextricably linked, they are not the same. It is possible to have a poor sounding system that is highly intelligible (e. g., a frequency limited re-entrant horn with uneven response) and a high quality loud- speaker that is virtually unintelligible (an expensive hi-fi system in the center of an air- craft hangar). Many factors important to speech intelligibili- ty are well understood and can be used to help develop guidelines for successful system design. The importance of high system intelli- gibility is ever increasing, not only as the pub- lics expectation of sound quality continues to grow, but also as the need to make intelligible emergency announcements at public facilities and sports venues takes on greater impor- tance. The information presented in this broaa overview of sound reinforcement has been assembled from many sources. Through an understanding of these essential principles, users will be better able to design, install and troubleshoot sound systems for speech 2. Clarity and Audibility: A common mistake often made when dis- cussing intelligibility is to confuse audibility with clarity. Just because a sound is audible does not mean it will be intelligible. Audibility relates to hearing sound, either from a physio- logical point of view or in terms of signal-to- noise ratio. Clarity describes the ability to detect the structure of a sound and, in the case of speech, to hear consonants and vow- els and to identify words correctly A speech signal involves the dimensions of sound pressure, time and frequency. Figure 1 shows a typical speech waveform for the sylla- bles J, B and L. Each syllable has a duration of about 300 - 400 ms, and complete words are about 600 - 900 ms in length, dependent on their complexity and the rate of speech. A spectrographic analysis of the phrase JBL is shown in Figure 2 In this display the left (y) axis shows frequency, the bottom (x) axis shows time, and the intensity of the display shows amplitude. The lower horizontal bars in the display represent the fundamental voice Technical Notes Volume 1, Number 26 Speech Intelligibility - A JBL Professional Technical Note frequencies at approximately 150, 300, 450 and 600 Hz for the letters J and B. For the letter L the fundamentals are at approximate- ly 190, 370 and 730 Hz. Figure 1. A typical speech waveform: J-B-U, 1 1 4iU - vtltz Figure 2. Time / frequency spectrograph of JBL. Figure 3 shows a spectrum analysis of the vowel sound a and consonant sound V . The vowel is made up of a series of reso- nances produced by the vocal cord-larynx sys- tem. The s sound has a different spectrum and is continuous over a wide, high frequency range extending beyond 12 kHz. Figure 3. Spectral response of typical vowel (a) and consonant (s) sounds. Unvoiced speech sound: consonant V 3. Factors Determining or Affecting Sound System Intelligibility: Primary factors are: * Sound system bandwidth and frequency response * Loudness and signal-to-noise ratio (S/N) * Room reverberation time * Room volume, size and shape of the space * Distance from listener to loudspeaker * Directivity of the loudspeaker * Number of loudspeakers in operation * Direct to reverberant ratio (directly depend- ent upon the last five factors) * Talker annunciation/rate of delivery * Listener acuity Secondary factors include: * Gender of talker * System distortion * System equalization * Uniformity of coverage * Sound focusing and presence of any discrete reflections * Direction of sound arriving at listener * Direction of interfering noise * Vocabulary and context of speech information * Talker microphone technique The parameters marked are building or system related, while those marked relate to human factors outside the control of the physi- cal system. It should be noted however that two of the primary factors (talker annuncia- tion/rate of delivery and listener acuity) are outside the control of both the system and building designer. Each of the above factors will now be dis- cussed. Voiced speech sound: vowel a Tilt - IS! 4. Frequency Response and Bandwidth: Speech covers the frequency range from approximately 100 Hz to 8 kHz, although there are higher harmonics affecting overall sound quality and timbre extending to 12 kHz, as seen in Figure 3. Figure 4 shows an averaged speech spectrum and the relative frequency contributions in octave band levels. Maximum speech power is in the 250 and 500 Hz bands, falling off fairly rapidly at higher frequencies. Lower frequencies correspond to vowel sounds and the weaker upper frequencies to consonants The contributions to intelligibility do not follow the same pattern. In Figure 5 we can clearly see that upper frequencies con- tribute most to intelligibility, with the octave band centered on 2 kHz contributing approxi- mately 30%, and the 4 and 1 kHz bands 25% and 20% respectively. Figure 6 shows this in a different manner. Here the cumulative effect of increasing system bandwidth is shown, and 100% intelligibility is achieved at just over 6 kHz bandwidth. This graph is useful in that it allows the effect of limiting bandwidth to be evaluated. For example, restricting the higher frequencies to around 3.5 kHz will result in a loss of about 20% of the potential intelligibility. Figure 4. Long-term speech spectrum. Figure 5. Octave-band contributions to speech intelligibility. Figure 6. Cumulative effect of high frequency band- width on intelligibility. R i q u i n-ti,1 Hi Data with respect to bandwidth and intelligi- bility may vary according to underlying experi- mental methods. For example, Figure 7 con- trasts well known early data relating to tele- phone monophonic) measurements that do not include any room effects with a recent experiment carried out in a reverberant space with Tw = 1.5 s. The upper curve Fletcher, 1929) shows that the contribution to intelligibili- ty flattens out above 4 kHz, with little further improvement above that frequency. The lower curve, made with a sound system in a real space, shows that intelligibility improvements continue up to 10 kHz. The importance of achieving extended high frequency response is immediately seen and points up the need to ensure an adequate S/N ratio in the important intelligibility bands of 2 and 4 kHz. Figure 7. Effect of bandwidth on intelligibility. Limited bandwidth these days should gener- ally not be a problem, since most modern sound equipment can cover the spectrum important to speech intelligibility. There are however some exceptions: Imlligiailiy (*4 dB Frequency ( H i ) lr:r- in h : i I ( % ) Frequency ( H i ) * Inexpensive, poor quality microphones * Some re-entrant horn loudspeakers 1 Some inexpensive digital message storage systems * Miniature special purpose loudspeakers By far the most common problems in fre- quency response are caused by loudspeaker and boundary/room interactions and interac- tions between multiple loudspeakers. Figure 8 shows the effect of wall mounting a nominally flat response loudspeaker system, significantty affecting its perceived sound quality and clari- ty. These conditions will be discussed in later sections dealing with system equalization and optimization. Figure 8. Loudspeaker/boundary interaction. 5. Loudness and Signal to Noise Ratio: The sound pressure level produced by a sound system must be adequate for listeners to be able to hear it comfortably. If the level is too low, many people, particularly the elderly or those suffering even mild hearing loss, may miss certain words or strain to listen, even under quiet conditions. The levels preferred by listeners may be surprising; although informal face to face communication often takes place about 65 dB-A, levels of 70 - 75 dB-A are often demanded at conferences and similar meetings - even under quiet ambient condi- tions. In noisy situations it is essential that a good S/N ratio be achieved. As shown in Figure 9, at a negative S/N ratio the noise is louder than the signal, completely masking it and resulting in virtually zero intelligibility. At a zero dB nom- inal S/N ratio, occasional speech peaks will exceed the noise and some intelligibility will result. As the S/N ratio increases so does the intelligibility. Over the years various rules of thumb have been developed regarding required S/N ratios. As a minimum, 6 dB-A is required, and at least 10 dB-A should be aimed for. Above 15 dB-A there is some improvement still to be had, but the law of diminishing returns sets in. Figure 9. Effect of S/N ratios on speech intelligibility. There is also some contradiction within the body of accepted reference data. Figure 10 shows the general relationship between S/N ratio and intelligibility. As we can see, this is effectively a linear relationship. In practice however the improvement curve flattens out at high S/N ratios - though this is highly depend- ent on test conditions. This is shown in Figure 11, which compares results of a number of intelligibility studies using different test signals. We can see that, for more difficult listening Against wall 2 m from wall tasks, the greater the S/N ratio has to be in order to achieve good intelligibility. Figure 12 shows the Al0 3 r s percentage loss of consonants scale, which will be discussed in later sec- tions. Here again we see a linear relationship leveling off at 25 dB S/N. Figure 10. S/N ratio and intelligibility. Figure 11. Articulation Index versus intelligibility word scores. Figure 12. Effect to Signal to Noise ration on %AIC intelligibility scale. POSITIVE SIGNAL-TO-NOISE RATIO POOR SIGNAL-TO-NOISE RATIO Intelligibility s , N R a t i o ( d 6 ) Intelligibility %) Sound Pressure Level (dB) Figure 14. Comparison of speech and noise levels. Good S/N (a); poor S/N (b). a Frequency Hz) b Frequency ( M i ) Under high noise conditions, such S/N ratios would normally require excessive signal levels. At high sound levels, the intelligibility of speech actually decreases, achieving a maxi- mum value at about BO dB, as shown in Figure 13. Where noise is a problem, a full spectrum analysis should be carried out, as shown in Figure 14. Such analysis will determine just where the problems lie and where most bene- fit can be obtained. Recall the frequency con- tributions to intelligibility shown in Figure 5; from this information the Articulation Index (Al) can be calculated. Figure 13, Effect of sound pressure level on speech intelligibility. In many situations the background noise may vary over time. This is particularly true in industrial situations, transportation terminals and particularly in sports and other spectator venues where crowd noise is highly dependent upon the activity. Figure 15 shows the time dependence of noise level in an underground train station. As the train approaches the plat- form, the noise level increases, reaching a maximum as the engine passes by. The doors then open and the people exit, with the noise level dropping appreciably. Announcements in competition with the 90 to 100 dB-A levels of the train arrival are difficult to understand- A better plan would be either to make announce- ments just before the train arrives or to wait until the doors are open. Figure 15. Subway noise versus time. Linear (upper- curve). A-woiglued (lower curve). Figure 16 shows the noise level pattern for a football game in a large stadium. The level varies rapidly, depending on field action, and during goal attempts and touchdowns the noise level is maximum. Figure 16. Football game noise analysis over a 12- second period. An aspect of S/N ratio often forgotten is the noise environment at the microphone itself. In many cases paging microphones are located in noisy areas, and the speech S/N ratio is fur- ther degraded by noise passing through the system. Directional microphones can often provide useful attenuation of interfering sounds - but this potential gain may be lost in reverberant spaces or by local reflections from the desk, ceiling or other surroundings. When the microphone has to be located in a particu- larly noisy environment, a good quality noise- canceling microphone should be used. There may also be the opportunity of providing a local noise refuge in the form of an acoustic hood or enclosure to produce a quieter local zone at the microphone. At least 20 dB-A, preferably 25 dB-A S/N should be targeted for the microphone zone. 6. Reverberation Time, Early Reflections and Direct to Reverberant Ratio: Just as noise can mask speech levels, so too can excessive reverberation. However, unlike the simpler the S/N ratio, the way in which the D/R ratio affects speech intelligibility is not constant but depends on the room reverberation time and reverberant field level. Figure 17 shows a simplified temporal enve- lope of the word back. The word starts sud- denly with the relatively loud ba sound. This is followed some 300 ms later by the lower level consonant ck sound. Typically the ck sound will be 20 - 25 dB lower than the ba sound. With a short reverberation time of 0.6 s, which would be typical of many well-furnished domestic rooms, the ba sound has time to die away before the onset of the ck sound. Assuming a 300 ms gap, the ba will have decayed around 30 dB and will not mask the later ck. However, if the reverberation time increases to 1 second and if the reverberant level in the room is sufficiently high, the ba sound will only have decayed by 18 dB and will completely mask the ck sound by some 8 to 13 dB. It will therefore not be possible to understand the word back or distinguish it from similar words such as bat, bad, ban, bath or bass, since the important consonant region will be lost. However, when used in the context of a sen- tence or phrase, it well may be deciphered by the listener or worked out from the context. Further increase of Jm to 1.5 s will produce 12 -13 dB of masking. Not all reverberation should be considered a bad thing since some degree of reverberation is essential to aid speech transmission and to provide a subjec- tively acceptable acoustic atmosphere, No one would want to live in an anechoic chamber. Figure 17. Reverberant masking. Waveform of word back (a); amplitude envelope (b); envelope with reverberant decay (c). The sound field in a large space is complex. Statistically it may be broken into two compo- nents: the direct and the reverberant sound fields. However, from the point of view of speech intelligibility, we can id