Advisory U.S. Department of transportation Circular Federal Aviation Administration Subject: COCKPIT NOISE AND SPEECH Date: 3/22/89 AC No: 20-133 INTERFERENCE BETWEEN CREWMEMBERS Initiated by: AIR-120 Change:
PURPOSE. This advisory circular (AC) provides information about the relationship between flight crew cockpit voice communication and cockpit noise levels. Guidance, on speech interference levels, noise measurement and measurement systems, and methods to improve cockpit communication, is provided for those manufacturers, owners or operators who believe cockpit noise may be a problem on their aircraft. This guidance material is relevant to the operation of all types of civil aircraft.
2. BACKGROUND.
a. Many modern aircraft provide comfort, convenience, and excelient periormance. At the same time that the manutacturers have developed more powertul engines, they have tried to give the occupants better noise protection and control, so that many of today's aircraft are more powerful, yet quieter than ever. Still, the levels of sound associated with powered flight are high enough in some aircraft to raise concern about the effect these noise levels may have on direct voice communication between flight crew members.
The National Transportation Safety Board (NISB) investigation of an accident involving a twin-engine, small airplane, concluded that the cockpit noise levels of that particular airplane were loud enough to interfere with direct voice communication. In the opinion of the NISB, this communication interference could have affected crew coordination and contributed to the accident. The NTSB also believes that poor crew communication, because of high cockpit noise levels, may have contributed to other accidents.
3. DEFINITIONS.
a. Noise - Any sound which is undesirable because it interferes with speech and hearing.
b. Noise Spectra - The description of noise sound waves by resolution of their components, each of different frequency and (usually) different amplitude and phase.
c. Frequency(Hz) -- The number of oscillations per second of a sine-wave of sound.
d. Decibel (dB) -- The unit in which the relative levels of intensity of acoustical quantities, such as sound pressure levels, noise levels and power levels, are expressed on a scale from zero for the average least perceptible level to about 130 for the average pain level.
A-Weighted Sound Level (dB(A)) -- A single event sound level which has been filtered or weighted to discriminate against the low and high frequency extremes to approximate the auditory sensitivity of the human ear.
f. Octave Band - All of the components, in a sound spectrum, whose frequencies are between two sine wave (pure tone) components whose ratio of frequencies exactly two, ie. separated by an octave.
4. DISCuSSION. Today's large, jet-powered, air-transport airplanes present few speech-interference problems for flight crews. However, propeller or rotor driven aircraft, regardless of the power plant used, have noisier cockpits for several reasons.
frequencies, which are much more difficult to attenuate than high-frequency sounds. In nonpressurized aircraft, construction permits air leaks that are both sound transmitters and sound sources; propeller and rotor tips can travel at or near Mach l, which means, in some flight configurations, small sonic booms constantly bombard the aircraft. In addition, techniques for minimizing sound production or sound transmission require the addition of physical mass to the system, and where payload determines the value or utility of the aircraft, adding enough mass to reduce noise, can cost severely in payload. Streamlining can be very costly in new design costs (to remove air leaks) and it may also require major changes in production methods. Some of these methods require additional weight which reduces utility.
outside the aircraft, noise spectra vary greatly as a function of aircraft size and type and the variety of powerplant, but the interactions ot those spectra with the sound-insulation properties of the various airtrames generally lead to strikingly similar spectra on the inside. Cockpit noise studies have shown the spectral shapes of cockpit noises vary only slightly from one type of fixed-wing aircraft to another.
b. The primary energy in those noises lies in the low frequencies, ranging mostly from 100 to 300 Hz, with a rapid decrease as frequency increases.
This spectral configuration may peak at different sound levels for different airplanes. The overall sound intensity varies from about 70 dB(A) to more than 100 dB(A). Generally, the quietest cockpits are found in jet aircraft; the noisiest are found in open cockpit airplanes such as those used for aerial application in agriculture and in some small military jets that use afterburners.
c. Within a general class of aircraft (for example, light, single-engine airplanes), the variations in cockpit noise level among airplanes of a single type may be about as large as the variations found among all the types within the class. Age and history seem to be important determinants of the cockpit noise level as much as the original design. Therefore, little is to be gained by looking at a single sound spectrum from a single airplane as if it were typical of its type and would remain typical of its type.
d. The following sections present an overview of a means to assess the level of cockpit speech interference due to noise and methods to measure and improve cockpit communications.
2 Par 3
(1) Speech Interference Level. This AC utilizes a noise interference metric known as the perferred-frequency speech interference level (PSIL). The PSIL is an average of the unweighted noise sound pressure level of three octave bands at 500,1000 and 2000 Hz and relates to an "A" weighted decibel measurement (AB(A)). The PSIL has been accepted as a suitable predictor for a much more complex measure of speech intelligibility known as the articulation index (AI).
The AI ranges from 0.0 to 1.0 with an increasing value indicating a more perfect communication. The Armed Forces maintain that for communications approximately intelligibility level. The maximum PSIL for AI=0.2 is 83 and for AI=0.3 is 78.
efforts should be made to aid communications by use of one or more of the methods discussed in this AC. The evolution of speech intelligibility research and the development ot criteria regarding speech interference is covered in some detail in appendix 1.
(2) Cockpit Noise Measurement. A portable sound level meter (SIM) which indicates the sound output in "A" weighted decibels (dB(A)) is recommended for the measurement of cockpit noise.
(a) A quick noise survey of the cockpit can be made by observing the sound level for approximately 20 seconds while the aircraft is in stabilized flight. One or two repeat readings are recommended to average the data.
Readings should be taken in the takeoff, approach, cruise and descent modes of flight so that a comprehensive noise picture is obtained.
(b) If the above tests indicate a noise problem or a borderline noise problem exists, additional noise measurements should be taken and recorded, as discussed in appendices 2 and 3. Recording noise levels is desirable as this will allow a more complete noise analysis to be made. In addition, a sample calculation of PSIL is shown in appendix 4.
(3) Methods to Improve/Aid Cockpit Communication. When the noise level in the cockpit, exceeds 88 dB(A) (PSII=78), the noise will be of sufficient magnitude as to interfere with normal cockpit communications, i.e.
voice and radio. Therefore, efforts should be made to aid communications.
following methods are suggested to improve the signal (voice)-to-noise ratio, which will enhance the intelligibility of cockpit communications. Appropriate FAA approvals must be obtained for any type design changes resulting from any of the following methods employed:
(a) Decrease the cockpit noise level.
(i) Use of door seals
(ii) Acoustical insulation.
(b) Increase the voice signal levels or modify the signal-to-noise ratio.
(i) Increase the gain of intervening audio amplifiers.
(reference TSO-C50c, Aircraft Audio and Interphone Amplifiers) Par 4 3
(ii) Use of electronic headsets, noise cancelling or boom microphones and intercom systems. (reference TSO-C57b, Aircraft Headsets and Speakers (for Air Carrier Aircraft) and TSO-C58a, Aircraft Microphones (for Air Carrier Aircraft))
(ili) Appropriate use of hearing protectors.
(iv) Move the flight crewmembers closer together.
e. Appendix 5 discusses in detail the advantages and disadvantages of the methods described above to improve cockpit communications. The overall objective of the modification should be to improve the intelligibility of communications. The minimum goal should be to achieve an articulation index (AI) of 0.3., identifiable by a PSIL of 78 or a measured noise level of 88 aB(A) £. Regardless of the method used to aid communications care should be taken to assure that aural warnings (i.e. overspeed, stall, and landing gear) can be heard with or without the communications aid in place.
M.C. Beard Director, Aircraft Certification Service, AIR-1 4 Par 4
Appendix 1 APPENDIX 1
QUANTIFYING SPEECH INTERFERENCE Several researchers have contributed landmark studies of the ways in which noise can interfere with the understandability or intelligibility of speech.
It has been demonstrated that the frequencies necessary for 100 percent intelligibility of a speech signal cover the range from about 300 Hz to about A measure of that portion of the speech intelligibility range that is available in a specific communication situation is known as the articulation index (AI).
The AI was developed by French and Steinberg and is a number falling between 0 and 1.0.* AI accounts for the level and spectra of ambient noise, and describes the relative ease or difficulty of a particular communication situation. An AI of 1.0 is considered perfect, with lower values indicating communications of lesser quality.
* French, N.R. and Steinberg J., 'Factors governing the intelligibility of speech sounds, Journal of Acoustical Society of America, 19,90-119, 1947.
Researchers have devised a set of relationships between AI and speech intelligibility for several sorts of speech test materials ranging from nonsense syllables, in which the content is quite unpredictable, to sentences, which are, comparatively, perceptually redundant--if you hear part of a sentence, you have a reasonably good chance to guess correctly what the rest of it is.
In 1947, Beranek published a report that serves as a further basis for determining how noise interferes with speech.* The speech interference level (SIL) is an average of the octave-band noise levels at some preselected set of center frequencies. In his original proposal, Beranek used the three octaves running from 600-4800 Hz. Later work, primarily by Webster and by Klumpp and Webster, showed that the inclusion of different frequency bands in the averages leads to AI predictions that are accurate for different communication conditions.** Thus, an average of the octave band levels at 500, 1000, and 2000 Hz seems well suited for predicting an AT of 0.2; i.e. a minimal communication environment. An average of 500, 1000, 2000, and 4000 Hz corresponds fairly well with an AI of 0.5 and an average of 1000, 2000, and 4000 Hz seems to go with an AI of 0.8. The 500, 1000, and 2000 Hz SIL has come to be known as the preferred-frequency SIL (PSIL), and it is often closely related to a dB(A) measurement of the same noise, though the relationship is not perfect.
*L. I. Beranek, "The design of speech communication systems," Proceedings of the Institute of Radio Engineers, 35, 880, 1947.
**J. C. Webster, Relations between speech-interference contours and idealized articulation-index contours, Journal of the Acoustical Society of America, 36, 1662, 1964; J. C. Webster, "Noise and Communication," in D. Jones and T.
Chapman (editors), Noise and Society, London: Wiley in preparation; R. G.
Klumpp and J.C. Webster, "Physical measurements of equal speech-interfering navy noises," Journal of the Acoustical Society of America, 35, 1328, 1963.
Appendix 1 The maximum PSIL for communications approximately 3 feet apart for an AI of 0.2 is 83. The maximum for an AI of 0.3 is 78. As will be shown below, these two AI's represent the range of acceptable minimum intelligibility levels.
Therefore, when a cockpit has a noise level above a PSIL level of 78, talkers and listeners can be expected to have some voice-communication problems. This prediction can be modified slightly by the fact that, in many cockpits, the pilot and copilot can be more or less than 3 feet apart. However, in the cockpits of aircraft likely to be relatively noisy, i.e. small aircraft, crewmembers would probably be seated at distances between 2 and 3 feet apart.
The messages that are expected to be transmitted in aviation communications come crom a prescribed vocabulary. However, even when that vocabulary is ignored, he messages are spoken in context, which usually means that they are more intelligible. The Armed Forces have set acceptable levels of performance for communications equipment, and those performance levels can be transformed into AI values: they range from 0.25 to 0.3. The Air Force, for example, defines an 80 percent score on a rhyme test as passing and a 70 percent score as In figure 1, it can be seen that the 80 percent criterion is almost exactly 0.3 and that the 70 percent criterion is very close to 0.25.
Navy and Army limits of acceptability are approximately the same as the Air Force's. Webster and Allen specified an 80 percent rhyme test score as (the Navy fence) the lowest acceptable value*. They reasoned that "95 percent of standard test sentences will be understood over a system that will pass 80 percent of rhyme test words. Following identical reasoning, the FAA believes that, short of measuring human performance on rhyme tests in cockpit-noise environments, the choice of AI=0.3 is both reasonable and acceptable. This AI equates to a PSIL of approximately 78 at a distance of 3 feet.
*J. C. Webster and C. R. Allen, Speech intelligibility in naval aircraft radios, Naval Electronics Laboratory Center Report, TR 1830, 1972.
Relation Bewteen Al and Varlous Measures of Speech Intelligibillty From ANSI S3.5-1968 100
TEST VOCABULARY
PERCENT OF SYLLABLES, WORDS, OR SENTENCES SENTENCES
PRESENTATION (D LISTENERS) 80 UNDERSTOOD CORRECTLY 70 50
HYME TESTS 40
EST VOCABULARY LI 30 OTE: THESE RELATIONS ARE APPROXIMATE. THEY DEPEND UPON 10 THE OF MATERIAL AND SKILL OF
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
ARTICULATION INDEX FIGURE 1
The following table also corroborates the relationship between the various test results and Articulation Index:
Table 1. Expected Word or Sentence Scores for Various Articulation Indices (AI)
PERCENT INTELLIGIBILITY Articulation Phonetically* Modified** Sentence* Index Balanced Test Rhyme Tests Test 0,2 22 54 77
0.3 41 72 92
0.35 50 78 95
0.40 62 86 96
0.50 77 91 98
0.60 85 94 98
0.80 92 98 99 *From Kryter and Whitman (1963) **From Webster and Allen (1972) Assuming that pilots can communicate visually with each other, an AI of 0.3 actually can be elevated to 0.47 as indicated by the following chart (figure 2).
EFFECTIVE AI WITH VISUAL CUES 1.0 0.9
o.
• 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
CALCULATED ARTICULATION INDEX Relation between calculated Al and effective Al for a communication system wherein the listener can see the lips and face of the talker (Sumby and Pollack, 1954).
FIGURE 2 3
Appendix l Thus, an AI of 0.3, if aided by visual cues, can raise the intelligibility level to approximately 98 percent (as shown figures 1 & 2). However, visual communication, while it can improve intelligibility, requires the persons to look directly at each other. This full-face orientation in the cockpit between rotorcraft and propeller-driven airplanes, especially the piston-engine types, can possibly exceed the maximum practical PSIL values noted above.
If one considers the distance between the heads of a pilot and copilot to be 3 feet, then in a noise field whose intensity exceeds a PSIL of about 90 (about 97 aB(A)), vocal effort cannot overcome the intelligibility problem created by the noise. First, shouted speech is not as intelligible as speech produced with less effort (se figure 3). Second, in that much noise, human vocal systems are, on the average, just about at the limit of their loudness. (Reflexively, talkers raise their voices in order to be heard above the background noise.
this instance, though, where noise levels are quite high, the reflex cannot lead to more intense speaking levels: the vocal system has already reached its physiological end point.) When PSIL = 90, AI approaches zero as does intelligibility--that PSIL condition is unacceptable at a 3-foot distance.
FROM: Pickett, J.M.: Limits of Direct Speech Communication in Noise. J.
Acoust. Soc. America, vol. 30, no. 4, Apr. 1958, pp. 278-281.
100 % WORD INTELLIGIBILITY 50 40
NAXIMUN
VERY SHOUT
NORMAL LOUD Ow FORCE FORCE 50 80 70
SPEECH INTENSITY, DB Relations between speech intelligibility in noise and vocal force. Vocal force measured as speech intensity one m from lips in a free field. Parameter, over-all signal-to-noise ratio, db.
Noise, 70 db, flat spectrum.
FIGURE 3
Appendix 2 APPENDIX 2
COCKPIT NOISE MEASUREMENT
TEST SETUP Measurement in the cockpit should be made at the typical head location of each flight crewmember. The microphone should be placed at the representative ear position on the side where speech communication is normally received and moved around slightly to obtain a spatial average of noise at the head position.
Whenever possible, the measurement shall be made with the crewmember absent from his location so as to minimize interference and shielding effects. During the measurements care should be taken not to hold the microphone close to a sound-reflecting or sound-refracting surface. A common recommendation is to stay at least one foot away; in practical use, a 6-inch distance is probably adequate.
TEST CONDITIONS The aircraft interior should be in a fully furnished configuration for its intended use (passenger, cargo, other) with tie downs, carpets, seats, curtains, interior trim panels, etc., installed. Systems used for providing conditioned air (i.e., pressurization, cooling, heating,) should be operational. Cabin pressure should be noted so that adjustments for differences in air pressure may be made, if necessary. Cabin pressure can affect noise measurements taken on the ground or inflight. The difference between these measurements is about
0.25 AB(A). On some aircraft, windows can be open during flight and could adversely affect the noise level in the cockpit. If this case exists this condition should also be tested.
If a tape recorder is used, the acoustic sensitivity calibration can be recorded during flight to establish the reference acoustic level for subsequent data processing and for comparison with the preflight recording of acoustic-sensitivity signals. Recorded noise levels should be measured on the ground and inflight to establish the proper gain to be used for recording above the background noise levels. At least one reel of tape used during the test should have a recording of acoustic-sensitivity calibration signals.
Where possible measurements should be made when all aircraft operating conditions (such as altitude, airspeed and engine power settings) are stabilized.
The aircraft cockpit noise should be tested in take-off, approach, landing, cruise, and descent at high speed.
On multi-engine aircraft the use of engine synchronization is optional depending on the test objectives. Installation and operation of engine synchronizers or propeller synchrophasers is frequently desirable for increased passenger comfort.
Operation of such devices during acoustical testing is advisable if the test objective is to measure the optimum cabin environment. However, imperfect synchronizer operation may introduce very low frequency beats which compromise
3/22/89 the data, so that intentional operation out of sync may be necessary. In such cases, the engines should be set to produce a known beat frequency high enough to allow reasonable length data records and minimize amplitude effects.
The following flight data should be observed and noted while the acoustic data is being obtained:
a. Flight Regime - takeoff, cruise, approach, landing, descent etc.
b. Airplane pressure altitude.
c. Airplane indicated airspeed and/or Mach number.
d. Propeller RPM (if applicable).
e. Engine power settings.
f. Synchronizer or synchrophaser operation.
g. External ambient air temperature.
h. Cabin pressure and temperature.
i. Cabin system operation modes.
DATA ACQUISITION If tape recording is used, the record length at each location should be at least 2 1/2 times the data reduction integration period, but in no case less than 20 seconds. If audible beats are present the record shall include at least 3 complete beats. Sufficient precautions should be taken to ensure the data signals are not compromised by inappropriate tape recorded gain settings.
should be recorded with the sound level meter in the flat mode (unweighted).
When portable sound level meters are used for direct measurement of sound pressure levels, (use the A-weighting network with Siow response setting) the data to be reported shall be the maximum reading noted on the meter. When audible beats are present the meter should be observed for a period of time long enough to include at least three beats, and the maximum meter reading noted shall be reported. If the sound level meter has integrating capability where the time period is operator-controlled, the time period used shall be at least 10-20 seconds. If audible beats are present, the time period shall be sufficient to include at least 3 complete beats, but not less than 20 seconds.
DATA REDUCTION Data reduction, from the recording, when employed, should be performed by time averaging data samples of at least 8 seconds duration. When audible beats are present, the integration period should be extended to include at least a three-beat period.
Appendix 2 Sound pressure levels should be obtained for the eight-octave bands center frequencies from 63 Hz to 8 KHz. Overall sound pressures should be obtained by summing antilogarithmically the octave band data. Preferred speech interference level (PSIL) should be calculated by algebraically averaging the unweighted levels in the 500, 1,000, and 2,000 Hz octave bands.
Frequency weighting may be added to octave band sound pressure level data. The weighting function should correspond to that referenced in International Electromechanical Commission (IEC) 651. Frequency weighted overall sound pressure levels are obtained by antilogarithmically summing the octave-band data after weighting is applied.
Presentation of the acoustical data should include at least the following information:
1. Overall A-weighted sound pressure levels at each measurement location.
2. Preferred speech interference levels at each measurement location.
3/22/89 AC 20-133 Appendix 3 APPENDIX 3
MEASUREMENT SYSTEM A portable sound level meter (SIM) and a portable battery-powered IM recorder are recommended to measure cockpit noise. The SIM includes the microphone, amplifier, rectifier and a meter which gives a sound output directly in decibels.
A connecting jack is provided so the amplifier output can also be recorded on a magnetic recorder for further study.
Most sound level meters also include weighting networks selected by a panel switch. The "flat" position sums all frequencies evenly. The "C" position is almost the same as "flat" and one or the other may be omitted on cheaper instruments. The "A" and "B" weightings are designed to approximate the ear's response and to give a truer approach to loudness of complex sounds. (The "B" scale is little used today, while the "A" weighting is used extensively. The designation "JB(A)" or, less properly, "aBA", indicates the reading with the "A" weighting.)
More expensive meters include, either as an attachment or internally, a series of band pass filters, usually of one octave width. Eight such bands will cover the usual measurement range of 50 to 10,000 Hz. Such filters provide a convenient means for a quick evaluation of the frequency structure of a complex In order that sound level meters made by different manufacturers will agree adequately when measuring various sources, their characteristics are specified by the International Standards Organization (ISO) and American National Standards Institute (ANSI). This includes the characteristics of the weighting networks and the meter damping, as well as the overall accuracy. Sound level meters are divided by ANSI standards into several groups: Type l or "Precision" meters; Type 2, or "General Purpose," Type 3, or "Survey," and Type S or "Special Purpose." Type 1 meters meet the rigid tolerances for Precision meters and provide filtering and impulse measuring options. A Type l meter is recommended for evaluating cockpit noise.
A high quality EM tape recorder should be used to record the noise in the cockpit. Good results can be obtained from a portable battery-powered system. Several manufacturers now advertise high quality cassette recorders for instrumentation use.
The sound level meter or the recording system, if recordings are made should be calibrated using a PISTON-PHONE, or other calibration instruments, betore and after the test data is recorded. These calibration devices are available from manufacturers of sound level meters and measurement microphones. It is designed to fit tightly on the microphone, with adapters for various microphone sizes, and it produces a tone of accurately known sound pressure at the microphone diaphragm at one or more standard frequencies. A set-screw is usually provided in the sound level meter to standardize its output.