Lecture 5:
The Psychophysics of Sound: 3
Slide 1
Recap
Two theories of Pitch Perception
Place Theory
Temporal Theory
Place Theory
Pattern of activation on basilar membrane disassembled into components
Pattern of harmonics ‘predicts’ a lower note - the fundamental of the series
Requires at least 2 resolvable harmonics to work
Temporal Theory
Pattern of activation on basilar membrane disassembled into components
Pattern of periodic information intrinsic to wave form coincides at period of
fundamental frequency
Uses the unresolved harmonics
Pattern of AM in beats of unresolved harmonics – phase locked
Phase locking drops away > 5KHz
Mixed models
Combination of Place and Temporal Information
Temporal Info refines initial rough analysis by place mechanism?
Fits with fineness of JND over Critical Bandwidth
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 2
Pitch Perception
Combination or DifferenceTones
H&A 5.5.1
JND – The Difference Limen
H&A 3.3.2
Loudness Perception
Calculating dB
Measurement & Range of Loudness
H&A 2.4
Measurement of Subjective Loudness
H&A 2.4
SPL, SL, Phons & Sons
Loudness and Critical Bandwidth
JND for loudness
Intensity, Frequency & Duration
Hearing Damage & Loss
H&A 2.5
Tinnitus
Ghost Harmonics
Adaptation
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 3
Combination or Difference Tones
A Combination Tone is a tone that is heard when two tones are
sounded together, but is not present when the tones are
presented separately.
Several Combination/Difference tones are produced for two
frequencies.
where
f2 > f1
Simple Difference Tone at f2 – f1
Cubic Difference Tone at
f1 – K(f2 – f1 ) or K(f1) – K–1(f2)
If the two tones are adjacent harmonics of a lower fundamental
then the combination tone is that fundamental and the other
difference tones are the harmonic series.
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 4
Combination or Difference Tones
One theory is that Combination tones arise from interference
between the two primary waves in the Basilar Membrane.
The effect is dependent on the intensity of the sounds and is quite
fragile.
The combination tone can be suppressed by a continuous tone
just below f1 and f2, or by masking with noise.
The phenomena of the Difference Tones is used in making
instruments, e.g. organs.
If we want a low pitch (v. long pipe) if can use two shorter pipes in
the ratio 2:3, this produces the lower frequency that is in the
relationship 1:2:3,
e.g. 55 Hz can be produced by playing two pipes of 110 & 165 Hz.
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 5
Just Noticeable Difference for Pitch
Pitch, & loudness & duration (also other senses: colour, shape,
size, texture, smell) are continuous, however we perceive the
world discretely i.e. we split the continuum into points or chunks.
The granularity of this discrimination is called a Difference Limen
or Just Noticeable Difference (JND) – it is the difference there
must be between two signals before we can recognise them as
different.
If two pure tones (of the same intensity) are alternated and the
difference between them is slowly reduced, we get to a point
where they are indistinguishable.
f
Or conversely if two pitches are gradually separated when the
difference between them exceeds the JND, we hear two pitches
instead of one
JND is much finer than Critical Bandwidth
JND is approx. 1/27th of a CB – Zwicker(1970)
JND is much finer than pitches used in musical scales
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 6
Measuring Frequency JND
There are 2 measures of frequency JND
DLF 2 successive steady different tones – 2nd higher
DLF is where 75% of correct discrimination into two tones
DLF is smallest for low frequencies – larger for higher.
FMDL – frequency modulation of a continuous tone – amount of
modulation required for subject to recognise variation.
FMDL smallest for middle frequencies larger for higher & lower
Also, Frequency JND is blurred. There are two thresholds in the
differences between pitches
Difference below which change is never detected.
This is called the Neural Quantum – NQ
Difference above which change is always detected,
This is (approx)  NQ x 2
JND varies from person to person and is affected by musical
training, It is also dependent on the frequency, intensity &
duration of pitches
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 7
Variation in JND by frequency
1000 Hz
ERB
100 Hz
JND
^
^
^
B
A
N
D
W
I
D
T
H
10 Hz
1 Hz
100 Hz
1000 Hz
10000 Hz
Frequency >>>>>
Plot of JND for pure tones from 100 – 10,000 Hz at 80 dB
< 1kHz, JND  2-4 Hz (at 30Hz – 10% = major 2nd)
 1 kHz JND  0.30% (0.05 semitones, 5 cents)
> 1kHz JND  relatively constant 0.25% of CB
>10 kHz discrimination of pitches deteriorates rapidly.
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 8
Variations in JND
JND Varies with intensity
Louder Pitches are easier to discriminate
Louder = finer JND
JND Varies with duration
Longer Duration Pitches are easier to discriminate
longer = finer JND (von Bekesy 1960), 2 Subjects; 800 Hz pure
tone at 40 dB SL
f/f (%)
1.0
0.9
0.6
0.4
0
0.1
0.2
0.3
0.4
( secs)
0.2
Demonstration ASA Demo 17
0.0
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 9
JND for Pitch
You hear 10 groups of 4 pairs of tones (a,b,c,d) each tone is 500
ms long and each pair is separated by 250ms
The base frequency f is 1000 Hz
Tone pairs are either rising or falling f  f +  or f +   f
 starts at 10 Hz and drops by 1 Hz in each subsequent group
Group
a
b
c
d
1
2
3
4
5
6
7
8
9
10
Instructions: As you listen to the pairs of pitches write down in
the box whether or not you hear the pitch rise or fall (R,F) or you
cannot tell (C) or you think they are the same (S)
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 10
Loudness
The loudness of sounds is measured in terms of pressure.
There are various measures – e.g. Pascals = Newtons per m2
(1 Newton is force needed to accelerate 1 kg to 1 Meter per sec)
This measures pressure or intensity - in sound called amplitude.
Intensity = Pressure2
The range of intensity of sounds is very large and so amplitude is
measured on a logarithmic scale.
The most commonly used scale is the deciBel Scale (dB)
Named after Alexander Graham Bell.
 dB SPL (SPL means Sound Pressure Level)
 dB SPL is the difference between a reference and a
measured sound.
The reference of 0 dB SPL = 20 micro pascals (20 Pa)
It is the average threshold for a child’s hearing at 1 kHz
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 11
Calculating dB
The difference in Decibels is calculated as 10  the log of the ratio
of intensity of the measured (m) and reference (r) sound
bel = log(Pm/Pr)
dB = 10 log (Pm/Pr)2
dB = 20 log (Pm/Pr)
(log x2 = 2 * log x)
The pressure ratio of 10 to 1 is 20 decibels
Pressure
Intensity
0 dB  20 dB =
10
100
20 dB  40 dB =
10
100
40 dB  60 dB =
10
100
0120 dB is 6 increments of 20, i.e. 106, = 1 to 1,000,000
Normal hearing is 20  100 dB = 104 = 1 to 10,000
0dB
=
120dB =
threshold of hearing
threshold of pain
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 12
Loudness Perception
The threshold of hearing varies with frequency.
Hearing is most sensitive in the frequency range - 1 to 4.5 kHz
Outside this range higher sounds are easier to hear than lower.
Two ears are more sensitive than one in the higher frequencies
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 13
dB SL
A second measure - dB SL – in which SL = Sound Level.
This is the sound level above the threshold of hearing at a
frequency
e.g. at 100 Hz the threshold of hearing is 32 dB SPL, therefore 10
dB SL equals 42 dB SPL.
dB SL tries to calibrate the subjective level of sound intensity – by
measuring relative to the threshold. But it is difficult to use,
because you need to know what the threshold of sound is at a
particular frequency.
A tone of 40 dB SPL is audible at 1 kHz but is inaudible at 50 Hz
Both SPL and SL are unsatisfactory because they are relative to
the sensation of loudness.
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 14
Subjective Loudness – The Phon Scale
One ‘objective’ scale of subjective loudness is the Phon Scale
People judge loudness relative to a reference tone
The reference tone is at 1 kHz at 0 dB SPL
The intensity of other frequencies are varied until they sound as
loud. (Fletcher & Munson, 1933)
The measurements define contours of equal-loudness
Subjective Loudness rises more slowly for frequencies below 500
Hz – (C4) – i.e. you need more energy to hear
This is important for music reproduction systems. Because lower
frequencies need more energy to be heard it follows that as
volume is decreased that relatively speaking bass levels need to
be boosted to remain at the same subjective level.
e.g. – if there is a balance across frequencies at 80 dB SPL – if
the sound level drops to 40 dB SPL frequencies < 100 Hz will be
inaudible those at 1 kHz are still audible. The answer: reduce the
intensity of lower less than higher frequencies.
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 15
Subjective Loudness – The Son Scale
Another scale for subjective loudness is the Son scale (Stevens
1936,1955).
Again this uses a reference tone – 1 kHz at 4 dB SPL = 1 Sone.
The sound was changed in intensity and listeners were asked to
judge how much louder the new sound was – e.g. 2, 1/3rd etc.
The results were averaged and plotted as follows…..
Stevens derived a Power Law for the relationship between the
change in power and perceived loudness
L = kI 0.3
Where k is a constant, and I is the Intensity of the sound.
Steven’s Power Law gives (approx.) a two fold increase in
perceived loudness for a 10 dB SPL change in pressure,
1/2 intensity means approx 0.81 of subjective loudness
1/3rd Intensity means approx 0.72 of subjective loudness
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 16
Perceived Loudness & Critical Bands
Complex tones comprise pure tones components.
The Critical Bandwidth contributes to what is heard. The general
principles are:
If a sound has components whose energy is constant and the
components lie within one CB then it sounds softer than if the
components lie in more than one CB. (Zwicker, Flottorn &
Stevens 1957)
Loudness is proportional to the total intensity of the sound in a CB
If the intensity is constant then the perceived loudness is
independent of the number of components and their bandwidths
within the CB
i.e. the sound is judged to be as loud as a pure tone of the same
intensity at the Centre Frequency of the CB.
If the bandwidth is > than a CB then the sound is louder
(measured in phons). This is generally the case for noise and
complex tones)
e.g. loudness in phones for 1
kHz sounds of various
bandwidths
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 17
Perceived Loudness & Critical Bands
This can be understood using Steven’s Power Law
Say we have a sound of Intensity 1 in one CB the loudness = 1
If a sound is distributed over 2 CB’s, each CB receives
approximately 1/2 the energy intensity.
1/2 the intensity  80% of loudness.
2 x 0.81 = 1.62
rather than 1
or over 3 CB’s – 1/3rd intensity = 0.72% loudness
3 x 0.72 = 2.16
rather than 1
The physiological reason for this is thought to be that neuron firing
rates at moderate to high intensities change relatively slowly as a
function of intensity (they are saturating). So when all the energy
is in one CB it is integrated and its net effect is decreased.
Halving the energy delivered to a CB is offset by the increased
mass of neuronal activity across multiple CBs.
Overall - perceived loudness depends on the bandwidth of the
sound
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 18
JND for loudness
Our perception of changes in energy levels is measured in 2 ways
Amplitude Modulation – when change is noticed = JND
Ordered Pulses – L-S, S-L etc – judgement of order = JND
Wideband Noise
JND is a constant ratio of the difference in intensity and intensity
I/I = constant
This is Weber’s Fraction – the smallest detectable increment is
proportional to the magnitude of amplitude.
JND is 0.5 – 1 dB from 20 – 100 dB. Slightly higher when < 20 dB
Pure Tones
Weber’s Law is approximated
JND for loudness varies slightly at different frequencies
Discrimination is slightly better at higher energy levels, i.e. JND is
finer with higher dB.
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 19
Frequency and Duration
JND for Intensity becomes coarser as duration of pitches shortens
L/L
12
10
500 Hz, 40 dB
1000 Hz, 40
8
6
500 Hz, 70 dB
0
250
500
750
duration ms
Also, the perceived intensity of a sound is affected by its duration.
4
> 500 ms (1/2 sec) loudness appears constant
< 2200 ms decreasing perception of loudness as length drops.
This implies that energy is measured over time and if the time
0
period
during which the sound occurs is less than the period used
by the nervous system to ‘integrate’ the energy of an ongoing
stimulus then it sounds softer.
Other points
 Increase in dB = more nerve firing
 Phase locking increases with energy
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 20
Hearing Damage and Loss
Hearing is permanently damaged by listening to sounds that are
too loud and/or by listening to them for too long.
It occurs most easily in the range that we are most sensitive to –1
to 4.5 kHz. Centred around 4 kHz
Prolonged exposure to sounds of greater than 90 dB SPL leads to
permanent damage. The damage potential of a sound level is
calculated in terms of a ‘consolidated dose’ (borrowed from
measuring exposure to radiation). The measurement is relative to
90 dB SPL for 8 hours. This is equivalent to 93 dB SPL for 4
hours, 96 dB SPL for 2 hours, etc….
Hearing loss takes two forms
 Loss of sensitivity
 Loss of acuity (accuracy)
Loss of sensitivity
Due to mechanical damage to hair cell – flattening
Loss of acuity
Probably due to an increase in Critical Bandwidth caused by
“The Evoked Cochlear Mechanical Response” (Kemp 1978).
This is feedback from the hair cells to the Basilar Membrane
that amplifies the wave in the Cochlea – a self sustaining loop.
If CB is larger then separation of components is cruder, hence
degradation of ability to analyse and discriminate sounds.
Implies Place mechanism
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 21
Hearing Loss
The degrees of hearing loss induced by a series of intensities
over time, measured in dB – i.e. the amount of additional energy
required for sounds to be audible
Reduction in
Sensitivity
106 dB
dB
40
100 dB
94 dB
30
88 dB
20
1
10
100
mins
10
0
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 22
Tinnitus
Tinnitus is where the ear spontaneously produces noise – hisses,
pure tones, clicks etc.
It is associated with hearing loss
It is triggered by loud sounds – which also damage hearing
People with hearing loss suffer from it
People with it are more susceptible to hearing loss
Tinnitus seems to be the result of combinations of two or three
components
1) Activity in the Cochlea
2) Increase Auditory Nerve activity
3) Higher centres (brain stem and cortex)
These three components operate differently in different people –
including people who are deaf!
Activity in the Cochlea
Used to be thought cause was middle ear muscle contractions.
Now it is thought to be due to “The Evoked Cochlear
Mechanical Response”. In Tinnitus it occurs in the absence of a
sound – i.e. no wave.
This may be prominent in Basilar Membrane areas surrounding
places for frequencies that have suffered damage
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 23
Tinnitus
Tinnitus is not just an increase in Auditory Nerve activity
Some hearing deficits decrease nerve activity
This implies some contribution by higher centres (brain stem &
cortex)
If the auditory nerve is cut in some cases it does not reduce
Tinnitus
In some cases if there is increased stimulation of Auditory nerve
then Tinnitus is reduced.
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 24
Ghost Harmonics
When a pure tone is sounded (loudly) people can hear harmonics,
as if the pure tone was complex tone – these overtones are
produced by the auditory system.
How much we hear them depends on how loud the pure tone is
The louder the tone is the more harmonics we hear and the louder
each harmonic sounds.
The effect is that the pure tone sounds as if it is the fundamental
of a complex tone.
E
n
e
r
g
y
3
9
Harmonics
This may be related to Difference Tones
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 25
Hearing - Adaptation and Fatigue
Our hearing changes as we listen to something over time. Two
closely related phenomena are:
Adaptation – the subjective magnitude of a stimulus decreases –
it seems softer – most adaptation within a couple of minutes
affects both loud and soft sounds, some studies indicate that
adaptation is greatest for higher frequencies
Fatigue – the energy needed to produce nerve firing is increased,
i.e. the threshold to register a stimulus is raised.
‘Temporary Threshold Shift’ (TTS) increases markedly above 90100 dB SPL – also the point above which permanent damage
occurs
Recovery time can be as long as 16 hours (or never)
TTS is difficult to measure – it seems to be affected by the
intensity of the tone, its duration, its frequency, the frequency of
the test stimulus and how long after the test tone occurs
At low intensities only close frequencies are affected.
At high intensities spread can be as large as ½ an octave
e.g. levels above fatiguing tone for I kHz tone – greater for
frequencies higher than fatiguing tone
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 26
Adaptation and Fatigue
TTS
dB
7
1400 Hz
6
5
1000 Hz
4
3
20
40
60
80
100
dB
2
Fatigue is highest 4-6 kHz – region of greatest hearing damage
1
0
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 27
Adaptation
The nervous system is designed to deal with change.
In hearing auditory nerve neurones respond to the onset of stimuli
and then their response declines rapidly over 2-5ms and more
gradually until 40-50 ms.
The drop is stronger for higher dB
Onset and rapid adaptation tracks intensity.
This is reflected in the ratio of max. firing to start firing.
-5dB flat response: no
initial high peak
47 dB variable
response: high onset rapid decline
For natural sounds
onset is often gradual -up to 40 msecs.
Thus, the net effect of
this is that when a
neurone fires it is less
sensitive for a period of
time – it becomes
adapted, and is less
able to respond to
renewed stimulation.
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick
Lecture 5:
The Psychophysics of Sound: 3
Slide 28
Adaptation
Adaptation is important when two tones are very close or identical
in frequency and are simultaneous or follow each other rapidly.
If we have two tones
Background tone is presented
Test tone is presented
This causes Adaptation
Firing rate drops - Background
adapts - firing decreases
Response to the test tone
suppressed
There is asymmetry in how long the adaptation takes to build up
and dissipate.
45 msecs to drop by 63% of initial firing rate
90 msecs to recover from 63% reduction
The implication is that in an ongoing musical or sonic universe
firing rates are being continually modified by the effect of other
preceding and simultaneous events adapting neuron activity.
This means that over short time scales onset timing and stimulus
intensity affect the way that we can distinguish different tones:
Adaptation is also thought to contribute to Masking and interact
with Critical Bandwidth
e.g. for figure and ground, a slight dislocation of onsets is enough
to separate sounds (Western & Japanese Music both use this).
CS5611 - Psychoacoustics - Niall Griffith – Semester 1
Computer Science and Information Systems, University of Limerick