Handbook for Sound Engineers

48 Chapter 3

audio professionals choose to monitor at very high
levels? There could be many reasons. Loud levels may
be more exciting. It may simply be a matter of habit.
For instance, an audio engineer normally turns the
volume to his or her customary level fairly accurately.
Moreover, because frequency selectivity is different at
different levels, an audio engineer might choose to
make a recording while listening at a “realistic” or
“performance” level rather than monitoring at a level
that is demonstratedly more accurate. Finally, of course,
there are some audio professionals who have lost some

hearing already, and in order to pick up certain

frequency bands they keep on boosting the level, which

unfortunately further damages their hearing.

3.3.2 Masking and Its Application in Audio

Encoding

Suppose a listener can barely hear a given acoustical sig-

nal under quiet conditions. When the signal is playing in

presence of another sound (called “a masker”), the signal

usually has to be stronger so that the listener can hear it.^17

The masker does not have to include the frequency com-

ponents of the original signal for the masking effect to

take place, and a masked signal can already be heard

when it is still weaker than the masker.^18

Masking can happen when a signal and a masker are

played simultaneously (simultaneous masking), but it

can also happen when a masker starts and ends before a

signal is played. This is known as forward masking.

Although it is hard to believe, masking can also happen

when a masker starts after a signal stops playing! In

general, the effect of this backward masking is much

weaker than forward masking. Forward masking can

happen even when the signal starts more than 100 ms

after the masker stops,^19 but backward masking disap-

pears when the masker starts 20 ms after the signal.^20

The masking effect has been widely used in psycho-

acoustical research. For example, Fig. 3-10 shows the

tuning curve for a chinchilla. For safety reasons,

performing such experiments on human subjects is not

permitted. However, with masking effect, one can vary

the level of a masker, measure the threshold (i.e., the

minimum sound that the listener can hear), and create a

diagram of a psychophysical tuning curve that reveals

similar features.

Besides scientific research, masking effects are also

widely used in areas such as audio encoding. Now, with

distribution of digital recordings, it is desirable to

reduce the sizes of audio files. There are lossless

encoders, which is an algorithm to encode the audio file

into a smaller file that can be completely reconstructed

with another algorithm (decoder). However, the file

sizes of the lossless encoders are still relatively large.

To further reduce the size, some less important informa-

tion has to be eliminated. For example, one might elimi-

nate high frequencies, which is not too bad for speech

communication. However, for music, some important

quality might be lost. Fortunately, because of the

masking effect, one can eliminate some weak sounds

that are masked so that listeners hardly notice the differ-

ence. This technique has been widely used in audio

encoders, such as MP3.

Figure 3-9. Tuning curve with (solid) and without (dashed)
functioning outer hair cells. (Liberman and Dodds,
Reference 14.)

Figure 3-10. Tuning curve at various levels at a particular
location of the basilar membrane of a chinchilla. (Plack, Ref-
erence 15, p90, Ruggero et al., Reference 16.)

0.1 1.0 4.0

100

80

60

40

20

0

Treshold—dB SPL

Frequency—kHz

70 dB SPL

2 4 6 8 10 12 14 16

Frequency—Hz

Basilar Membrane velocity—dB

90

80

70

60

50

40

30

20

10

0

20 dB SPL

40 dB SPL

60 dB SPL

80 dB SPL

30 dB SPL

50 dB SPL

90 dB SPL