Calibrating Audiometers

This article will deal only with those aspects of ANSI S3.6-1996 that are most relevant to the practicing clinician.  Those topics pertaining to bone transducers, sound field testing etc can be found elsewhere within this site.  Please refer to the table of contents or enter a search query above to locate your topic of interest.   

The current standard for the calibration of audiometers is ANSI S3.6 1996.  In order for an audiometer to meet the current ANSI standard it must meet the following general requirements of: 

a.      electrical safety

b.     warm-up time

c.     variation in the power supply and environment

d.     generation of unwanted acoustic signals

e.      subject response system

f.       monitoring earphone or loudspeaker system

g.     talk back system  

Audiometer calibration should be performed at least annually according to the procedures described below. The equipment necessary to perform these measurements is a sound level meter, octave-band filter set, and a National Bureau of Standards 9A 6 cc coupler.  Three styles of earphones are described ANSI S3.6 1996.  They are supra-aural, circumaural, and insert. 

The NBS 9A 6 cc coupler is standardized for the Telephonics MX41-AR cushions only.  If circumaural phones are used, an aluminum plate must be added to the coupler between the top and bottom portions to increase the relevant surface areas so that a seal between the circumaural cushion and the coupler can be formed. 

This standard also includes information on how to calibrate insert receivers.  The end of the receiver is inserted into an occluded ear simulator or into an HA-2 acoustic coupler.  The RETSPL values are listed in Table 3 below.  According to ANSI S3.6 - 1996 section 3.6, the RETSPL (reference equivalent sound pressure level), is the mean equivalent threshold sound pressure level at a specified frequency, as measured in a specified acoustic coupler, artificial ear, or ear simulator based upon hearing threshold data from a sufficiently large number of otologically normal individuals of both genders ranging in age from 18 - 30 years.

The RETSPL values are relevant in audiometer calibration for several reasons.  Without these figures, we would not know how much sound pressure is needed to reach 0 dB HL (threshold of hearing for an otologically normal hearing population) at a specific frequency.  By choosing a standard coupler (i.e. HA-2), we can be assured that a consistent (standardized) value can be applied to a common coupler to deliver a value of 0 db HL.          

In making these measurements, the accuracy of the calibrating equipment shall be sufficient to determine that the audiometer is within the tolerances permitted by ANSI.

Briefly, this is what needs to be done:

(1) "Sound Pressure Output Check"

A. Fit the coupler (i.e. HA-2) over the microphone of the sound level meter and place the earphone on the other end of the coupler.

B. Set the audiometer's hearing threshold level (HTL) dial to 70 dB.

C. Using the sound level meter, measure the sound pressure level of the tones at each test frequency from 500 Hz through 6000 Hz for each earphone.

Note:  In order to measure the sound pressure level, the mean square pressure must be averaged over a certain period of time.  Standard sound level meters normally incorporate "fast" and "slow" response settings corresponding to averaging times of 125 ms and 1 s, respectively (IEC 651, 1979, Berglund & Lindvall, 1995).  

When measuring steady state sounds, it does not really matter what averaging time you use, as long as the time you use is long enough compared with the time period of sound pressure fluctuations (Berglund & Lindvall 1995 pg 27.  For calibration purposes a slow setting will suffice.  Only when examining impulse sounds, will a fast averaging time be necessary, however it should be noted that "fast" (<10 ms) isn't fast enough for most impulse sounds.      

One other feature of the sound level meter that needs to be addressed is the weighting.  Your sound level meter may have a setting for A-, B- and C- filters, or perhaps just A- or linear.  The A- weighting curve is used to weight sound pressure levels as a function of frequency, approximately in accordance with the frequency response characteristics of the human auditory system.  For this reason, we will set the sound level meter to A- for our audiometer calibration purposes.   

D. At each frequency the readout on the sound level meter should correspond to the levels in Table 1 or Table 2 (listed below), as appropriate for the type of earphone, in the column.

(2) “Linearity Check”

An audiometer is linear if for every increase on the dial, an equal increase in output is measured.  By taking readings from a sound level meter, you can ensure that the dial step sizes are linear and accurate. 

A. With the earphone in place, set the frequency to 1000 Hz and the HTL dial on the audiometer to 70 dB.  

B. Measure the sound levels in the coupler at each 10-dB decrement from 70 dB to 10 dB, noting the sound level meter reading at each setting.

C. For each 10-dB decrement on the audiometer the sound level meter should indicate a corresponding 10 dB decrease.

D. This measurement may be made electrically with a voltmeter connected to the earphone terminals.

The standard requires that the output level of the audiometer always be within 1 dB of the stated value for attenuators marked in 5 dB intervals or 0.3 times the interval of attenuators that are marked in smaller intervals. 

The electrical noise floor must be accounted for when performing a linearity check at low levels.  Measure the noise floor with the signal on, then off, and then compare the two values.  This is done at a 60 dB dial setting, and the electrical signal in any 1/3 octave band must be 10 dB below the signal RETSPL at all test frequencies.  The RETSPL value is important here for indicating whether or not the electrical noise floor meets ANSI standards.

For example:

The RETSPL value for a HA-2 coupler at 250 Hz is 14 dB.  This means that the electrical noise floor must be 4 dB or less at 250 Hz in order for the audiometer to meet ANSI 3.6 standards.  However, because human thresholds varies according to frequency, a 4 dB noise floor will not be acceptable at all frequencies.  At 1000 Hz, the RETSPL value is 0 dB, therefore the electrical noise floor must be -10 dB or less.  

(3) “Tolerances”

When any of the measured sound levels deviate from the levels in Table 1 or Table 2 by +/ - 3 dB at any test frequency between 500 and 3000 Hz, 4 dB at 4000 Hz, or 5 dB at 6000 Hz, an exhaustive calibration is advised. An exhaustive calibration is required if the deviations are 15 dB or greater at any test frequency.

Table 1- Reference  Threshold Levels For Telephonics – TDH-39 Earphones 

 from ANSI S3.6-1996

 Table 2 – Reference Threshold Levels for Telephonics – TDH 49 Earphones

   from ANSI S3.6-1996

Table 3 Reference Pressures for 0 dB HL

   from ANSI S3.6-1996

ANSI S3.6 1996 classifies audiometers into five categories, 1 through 5 depending on the features and capabilities it has for pure-tone testing.  There are also three classes (A through C) for audiometers, which help indicate the number of features for speech testing with speech signals.  In general, you can say that a type 1A is the fully loaded model and a type 5 is the basic model.  For a view of a chart featuring the requirements of each audiometer type, click here.   

Type 1 audiometers must be able to present octave tones  of 125 – 8000 Hz, and 750, 1500, 3000, and 6000 Hz, with a frequency accuracy within 1% of the frequency in question.  In fact, the numeric type of audiometer corresponds to the accuracy (in percentage) by which it must meet for a given test frequency  (this differs from the previous revision of ANSI S3.6 1989.  Previously, the frequency tolerance was +/- 3 % of the intended frequency).  

Harmonic Distortion

Harmonic distortion requirements have changed since ANSI S3.6 1989.  The 1996 revision requires that the maximum level of the harmonics of the test tone relative to the level of the fundamental not exceed values in the table below (taken from ANSI S3.6 1996).

click to enlarge

The measurement of harmonic distortion is different for both air and bone conducted signals.  For air conducted signals, the transducer needs to be mounted on an acoustic coupler, an artificial ear, or an ear simulator similar to the one used for the specification of the RETSPL.  For measurements above 5 kHz, errors are more likely due to limitations in the acoustic coupler, artifical ear, or ear simulator.   

Rise/Fall Time of Tone  

The following figure (from ANSI S3.6 1996) will help illustrate the rise and fall time requirements for audiometers.

above figure taken from ANSI S3.6 1996

The rise time is described as A to C on the figure above.  From the onset of the signal/tone, A to C should not exceed 200 ms and B to C should be no less than 20 ms.  The sound pressure level must also rise progressively without discontinuities.       

The fall time requirements are basically the reverse of the rise time.  D to H should not be more than 200 ms with E to G being no less than 20 ms with the sound pressure level falling progressively without discontinuities.  

Overshoot requirement in ANSI S3.6 1996 states that, "the sound pressure level produced by the transducer shall not exceed +1 dB relative to its steady state level during either its rise or fall" (pg 16-17).

More detailed topics which are discussed in the standard include the following:

a.      Tone signal source frequency and sound pressure level ranges, accuracy, and distortion characteristics; signal rise and decay time

b.     speech signal source sound pressure level, distortion, frequency response, and monitoring capability  

c.     properties of masking sounds 

d.     resolution, linearity, and accuracy of signal and masker level controls  

e.      calibration procedures for supra-aural circumaural and insert earphones, bone conduction vibrators, and loudspeakers

Note:  This site is best viewed using IE 4.X or higher under 800x600 resolution