Calibration of Speech Signals

ANSI Requirements for Speech Audiometers (ANSI S3.1 section 5.2 )

When speech test material is used, the presentation level must meet a standard reference level according to ANSI S3.1.  This reference level can be monitored using a VU meter.  The speech audiometer VU meter pointer should reach 99% of the input signal value in 300 msec and then overshoot by 1 to 1.5%.  The best way to check this is to use a calibration tape recording, which contains a 1000 Hz tone.  With this tone, the Audiologist can visually monitor on the VU meter that the pointer or needle accurately points to 0 dB.    

Sound Pressure Level of Speech (ANSI S3.1 section 5.3)

According to ANSI, the sound pressure level of a speech signal at the earphone is defined as the rms sound pressure level of a 1000 Hz signal adjusted so the at the VU meter deflection of this signal is equal to the average peak VU meter deflection produced by the speech signal.  Usually, pre-recorded speech signals and the 1000 Hz calibration tone will be at the same level.  If however they are not, then the VU meter sensitivity can be adjusted as needed during testing. 

Standard Reference Threshold Sound Pressure Level for Speech (ANSI S3.1 section 5.4)

The standard reference threshold SPL for speech through an earphone is 10 dB re 20 u Pa.  What this means is that on the audiometer, a reading of 0 dB HL will correspond to 19 dB SPL.  This figure was attained using a Western Electric, type 705-A earphone on a NBS 9-A coupler.  When referenced to other earphones, the proper correction factors must be used.  The correction factors for other popular earphone manufacturers are listed in table 1.        

Reference-level calibration can be done by setting the 1000 Hz signal at 0 V through the microphone and measuring the output SPL of the 1000 Hz signal at a HL dial setting of 60 dB.  ANSI standards allow a 3 dB tolerance around this value.  Therefore, according to the above table, a TDH-39 earphone can be considered within acceptable calibration range if it falls within 77 – 83 dB SPL (80 +/- 3dB). 

The calibration equipment used for the reference-level calibration can be the same as that used for SPL measurements with tones.  The 1000 Hz reference signal needs to be from the same source as that for speech audiometry.  This can either be a microphone (live voice), or prerecorded material (tape or CD).

If you chose to use a microphone, which most Audiologists tend to use, the external calibration tone needs to be applied in an anechoic space.  The reason for this is to ensure that the measurement of the audiometer characteristics is independent of the acoustic environment in which the microphone is placed. 

In order to calibrate the microphone in an anechoic condition, you may place the microphone in a hearing aid test box (FONIX BOX or AUDIOSCAN).  The advantage of this approach is that the control microphone of the test box can ensure that the level of the test tone at the audiometer mic is correct.  Another means of calibration the audiometer mic is to use a portable audiometer or external oscillator connected to an earphone or preferably a speaker can be used to deliver the calibration tone.  When this approach is used the mic and the transducer should be at the same distance that the talker will be from it when used in the clinic.  The disadvantage of this method of calibration is that the acoustic environment (test room) can influence the measurements taken. 

On two-channel audiometer, you can set up pure tone testing on one channel and speech testing on the other channel.  What you can do is to use one channel present the calibration tone to calibrate the other channel.  Then you can reverse this and calibrate the other channel in the same fashion.  To do this, you can set one channel to live voice testing and the other channel to present a continuous 1000 Hz tone. 

Acoustic Fidelity (ANSI S3.1 Section 5.6)

Live Voice Audiometer (ANSI S3.1 Section 5.6.1) 

The frequency response characteristic of a live-voice audiometer shall be such that with the microphone in an acoustic sine-wave field of a given sound pressure level (approximately 74 dB), and in the relationship to the acoustic source which is recommended by the manufacturer as to distance and orientation, the sound pressure level developed by the audiometer at each of the frequencies 200, 300, 400, 700, 1500, 2000, 3000, 4000 Hz does not differ from that at 1000 Hz by more than +/- 5 dB. 

The above statement holds a lot of information.  Basically, the above condition can be met by following the steps from section 5.4.  That is, instead of using just a 1000 Hz tone to calibrate your audiometer, you use all the above frequencies (200 - 4000 Hz), and then determine if it is within +/- 5 dB of the 1000 Hz tone.  To make sure that the input at each frequency is close to 74 dB, a monitor microphone should be placed near the audiometer test microphone. 

Recorded Speech Audiometer (ANSI S3.1 Section 5.6.2)

The frequency response of a recorded speech audiometer through the frequency range of 200 – 4000 Hz, shall not differ from that at 1000 Hz by more than +/- 5 dB.  Any frequencies above 4000 or below 200 Hz cannot be greater than 10 dB above the values between 200- 4000 Hz.  In order to ensure this, the measurement device must be set to measure different frequency regions.  Such a device would include a sound level meter with one octave or one-third octave filters.   

According to the standard, speech audiometers used with recorded materials must have materials that comply with the Record Industry Association of America Standards, “Standard Recording and Reproducing Characteristics and Dimensional Standards – Disc Phonograph Records for Home Use,” Bulletin No. E-4.  The problem with this stipulation is that it is extremely difficult to determine if prerecorded speech material on discs complies with the standard, especially since ANSI-1969; R-1973 does not include a standard for recorded materials.  The recommended approach to this situation is to use tapes, which contain calibration tones at the appropriate frequencies. 

Overall Distortion (ANSI S3.1 Section 5.6.4)

Speech audiometers, like pure tone audiometers must be within tolerable limits of overall distortion present in the circuit of the device.  Distortion measurements should be made independently for each speech source (microphone, taped materials).  The input signal used to measure distortion through a speech audiometer must be free of any distortion.  According to ANSI this means that the pure-tone input must have no harmonic greater than 40 dB below the fundamental.

Distortion measurements are measured at very high levels, i.e. 120 dB.  When measuring the overall distortion through a speech audiometer, ANSI requires the fundamental frequency be at least 25 dB or more above the level of any higher harmonic.

Pre-recorded discs or tapes can be used to test for distortion.  When testing the sound path through the microphone, a pure tone signal with a 74 dB SPL input is used.  A sound level meter with one or one-third octave filters with filter roll-off greater than 25 dB per octave can be used to make the distortion measurements.  In order to do so, one must adjust the HL dial so that each tone (200, 400, 700, 1000, 2000 and 4000 Hz or 250, 500, 1000, 2000 and 4000 Hz) is produced at 120 dB SPL.  After this, read the levels at each of the harmonic intervals above the test frequency to make sure that they fall at or below 95 dB SPL (120 – 25).

Sound Pressure Levels for Speech Audiometers (ANSI S3.1 Section 5.7). 

Range and Intervals of Hearing Threshold Levels for Speech (ANSI S3.1 Section 5.7.1)

Accuracy of Sound Pressure Level (ANSI S3.1 5.7.2)

On the physical audiometer, the dial readings should be able to span a range of 0 – 100 dB in steps of 2.5 dB or less.  The measured difference between to successive intervals of the audiometer dial and the actual output should not differ by more than 1 dB.  

The SPL produced by the earphones should not differ from the audiometer’s indicated value by more than 3 dB.  This measurement is made at 1000 Hz at 60 dB. 

Since most audiometers use the same attenuator dial for both pure tone and speech signals, it is not necessary to recheck the attenuator for speech signals if the calibration for pure-tones has been conducted. 

Noise (ANSI S3.1 Section 5.8)

The noise floor according to ANSI must be at least 50 dB below the level of the signal.  To make this measurement, the presence and absence of a 1000 Hz signal is used.  On the audiometer, the gain control is set so that the VU meter indicates a reference level when the input is 1000 Hz.  This input can be a signal from recorded material or a signal through a microphone, in which case the signal needs to be at 85 dB SPL.  For the microphone signal path, the level is set to 85 dB SPL.  The output of the audiometer is measured with the above input signals set to 100 dB on the audiometer dial.  In addition, the output will also be measured with no signal input to the audiometer.  Signal to noise ratios are then determined by comparing the output levels with the input and with no input. 

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