Andrea Szigetvári

Timbre Solfége Exercises

Introduction

This set of ear training exercises helps to develop aural skills related to those qualities
of sound known as "timbre". Timbre is not explored in the framework of traditional
music solfége. A series of structured exercises and tests will help identification,
classification and memorization, as well as the perception of relationships between
different acoustic and semantic timbre dimensions.

The title of the course “Timbre Solfége” originates from the Sound Engineering
Department at the Chopin Academy of Music in Warsaw, where a technical listening
course was designed for training sound engineers and producers in the ‘60's. The
original course is still being taught at the Chopin Academy mainly aimed at
developing listening skills related to timbre categorized in terms of filtered
resonances.
The current exercises attempt to extend the course developed at the Chopin Academy
in the context of sound processing and sound synthesis techniques applied by
electroacoustic musicians, composers, sound designers and multimedia artists.

The syllabus will explore six main fields:

1. harmonic sounds created by means of additive synthesis
2. inharmonic sounds created by means of additive synthesis
3. combining different timbre dimensions created from harmonic and inharmonic
sounds
4. beating and roughness created by means of additive synthesis
5. filtering white noise with Low Pass Filter, High Pass Filter and Band Filter
6. filtering white noise with multiband resonant filters
7. sounds built from grains using granular synthesis
1. Timbre dimensions

Including timbre (a secondary music parameter until the end of the 20th century) in the
process of musical composition has resulted in increasing interest in its research in the
fields of aesthetics, acoustics and psychoacoustics. Timbre is a substantially more
complicated parameter than pitch, rhythm or loudness, because it is multidimensional
and has an indefinable, theoretically infinite, range. Its methods of organization
cannot be described by the rules of traditional music theory. Since the 50's in the field
of electroacoustic and computer music research, the multidimensional nature of sound
and the study of acoustic and semantic sonic dimensions have become a central
concern. Separating individual timbre dimensions facilitated the creation of scales
and motifs based on the 'colours' of the tones. With the help of the method called
'multidimensional scaling', the perceptual correlates of acoustic parameters involved
in musical timbre were separated. As a result of ongoing research, we can name and
quantitatively describe some of the most relevant acoustic dimensions. An
understanding of these dimensions makes it possible to create reduced timbre spaces,
which can be understood as matrices containing a limited amount of sensory
dimensions of limited range. With the help of the concept of reduced timbre space,
one can explore in a manageable way, small parts of the theoretically infinite
multidimensional timbre space.
Our timbre solfége exercises are matrices representing different simpler or more
complex reduced timbre spaces. Sometimes the reduction makes some of the
sonorities presented in this set of exercises simplified, somewhat “computer-like”.
This simplification is needed to learn how separated timbre dimensions sound.
Further exercises will be built on those attributes containing more complex
dimensions.

Acoustic dimensions explored in this set of timbre solfege exercises:

Brightness
Brightness is considered as one of the perceptually strongest distinctions between
sounds. The perception of brightness can be formalized acoustically as an indication
of the amount of high-frequency content in a sound. To define the brightness of a
sound, two measurement methods are used widely:
–   spectral centroid (spectral weight): indicates where the center of gravity of the
spectrum is, by defining the weighted mean of the amplitudes of the partials
present in the spectrum. Since very similar spectral centroid values can be
obtained for spectra with different density and width, additional information is
needed to clarify the timbre.
 Figure TD01.: spectral centroid

–   “acum”: is used to calculate the sharpness (a close synonym of brightness) of

the spectra. The value of the “acum” is based on the distribution of loudness
between the critical bands, and is portrayed on a Bark scale. Its reference
value (1 acum) is the percept created by a narow noiseband at 1 KHz with a
bandwidth smaller than 150 Hz and a loudness of 60 dB (0 dB being the
threshold of hearing).

Resonances, formants
The presence of peaks on the amplitude envelope of the spectrum indicates places of
resonances in the timbre. Resonances tend to stay essentially the same when the
frequency of the fundamental is changed. Their position can be defined by their
center frequency and bandwidth. Very often the properties of the resonances will
determine what is the aurally perceived material of the sound source (i.e. wood, metal,
glass, etc.)

A
BW

f
fc
Figure TD02.: resonant frequency

The spectral peaks of a sound spectrum are often called formants. Formants in the
sound of the human voice are particularly important because they are essential
components of speech sounds. The meaning of vowel sounds can be attributed to the
differences in their first three formant frequencies. Hearing vowels in other timbres
than speech often helps to define maximum frequency regions in the spectra.

Harmonicity-inharmonicity-noiseness
The position of partials of the spectra on the frequency axis determines if a pitched
sound (note) is heard, or an inharmonic chord-like sonority appears, or as an extreme
state a rough, non-pitched noise is perceived. The different classes of sonorities can
be placed on a note-noise continuum. Denis Smalley classified “note” as one
endpoint of the axis and noise as the other. Note is divided into two more classes,
harmonic and inharmonic. There is a continuous transition between the different
states.

HARMONICITY
NOTE
INHARMONICITY
 NOISE
Figure TD03.: the harmonicity-inharmonicity-noise axes

The difference between the harmonic and non-harmonic categories is essential, since
the basic condition of accurate pitch perception is harmonicity. Also, closer to
saturated noise, where no pitched fragments occur in the perceived sonority, is a
substantial change in character. For the measurement of the harmonic and non-
harmonic characters there are two important acoustic attributes - sensory consonance
and tonalness.

Sensory consonance and dissonance (the degree of smoothness or roughness): a major

component of sensory (or psychoacoustic) dissonance is the roughness caused by
beating partials of the spectra. Sensory consonance is the smoothness associated with
the absence of such beats.

Sensory dissonance occurs when partials in a spectum produce amplitude fluctuations

in a certain frequency range. Depending on the frequency of the fluctuation, the
sensation of beating or roughness occurs. Sensations of 'beating' and 'roughness' are
fluctuations of amplitude that occur due to interference of sound waves. The term
'roughness' is analogous to a tactile sensation in the sense of small irregularities in the
perception of sound.
The rate of amplitude fluctuation formed by two overlapping sound waves is equal to
their difference in Hz. Sound waves with a difference lower than 20Hz produce slow
fluctuations perceived as tremolo or 'beating'. Faster fluctuations are responsible for
sensations of roughness at differences of frequencies (or musical intervals) that
depend on the pitch register of hearing.

Tonalness: closeness of the partials to a harmonic series; tonalness is thus decreased

as partials deviate from harmonicity.

Spectral fluctuation: a measure of how quickly and to what extent the spectrum of a
signal is changing. The change of the shape and structure of spectra in time is a very
important factor in perceiving timbre. The patterns of the spectral changes are
responsible for those sonic attributes which create the special distinctive
characteristics of the perceived sonic entities, and are recognizable even in poor
acoustical conditions. Spectral fluctuation is calculated by comparing the spectrum for
one “timeslice” (or timewindow) against the spectrum from the previous “timeslice”.
It can be represented as the variation of instantaneous spectral centroid around the
mean spectral centroid or as the average of correlations between the amplitude spectra
in adjacent timewindows.

Amplitude envelope: the evolution of the energy of the signal over time is a very
important factor in timbre perception. This continuous change is usually represented
using an attack, decay, sustain and release part. Because very often some parts are
missing from most of the real sounds, and because the attack release part was proved
to be the most significant factors, a simpler representation, the attack/rest envelope
can be more practical.
2. Hearing changes in the harmonic spectrum

In this exercise the timbre of harmonic spectra is explored. The sound examples are
created through additive synthesis, where up to 1-48 sinewaves are added together.
A scale of sonorities close to the spectra of common harmonic waveforms (sinewaves,
triangle/square, saw) is presented in a matrix with different fundamental frequencies.
The matrices forming reduced timbre spaces demonstrate how the number and
amplitudes of the sinewaves present in the spectra affect what kind of sound classes
one can hear. Dimensions of bright-dark, sharp-blunt, hollow-nasal are examined.

2.1 Theory

Additive synthesis

Additive synthesis generates sound by changing the frequency and amplitude of each
individual sinusoidal partial. Often, each harmonic generator has its own
customizable volume envelope, creating a realistic, dynamic sound that changes over
time.
Additive synthesis is directly related to Joseph Fourier’s concept. Fourier, a French
mathematician discovered that all periodic signals, when represented as a
mathematical function can be composed of a sum of sinusoidal functions of various
frequencies.

Harmonic spectra
A harmonic spectrum contains only partials that are integer multiples of the
fundamental frequency. Periodic waveforms always have harmonic spectra.
The series of integer ratio frequencies is also known as the harmonic overtones of the
fundamental. The characteristic sounds of many musical instruments are built from
overtones. Since the overtones are integer multiples of the fundamental frequency,
the distance between the overtones is constant, but, while the neighbouring frequency
values have the same difference, the distance between the perceived pitches gets
narrower as we climb up the register. This convergence and increasingly dissonant
interval is very important to hear and identify while listening to, and analyzing,
harmonic spectra.

Harmonic partials tend to fuse together, meaning, we do not hear them as separate
pitches, but as a sound with only one pitch with a characteristic timbre. The partials of
harmonic sounds can be made audible depending on many factors such as the pitch
ratio of the partials, the amplitude of the partials, the amplitude envelope of the
partials, the overall envelope of the sound, fundamental frequency, the overall length
of the sound.

Common harmonic waveforms: sinewave, triangle/square, saw

Adding together sine waves that are in a harmonic relation to each other, more and
more complex periodic waveforms will result (e.g. triangle or sawtooth). The spectra
of these waveforms can be defined by simple rules. Their timbres are very
characteristic, that's why they were chosen as the basic building blocks of many
synthesizers.
A sinewave is the simplest building block to describe and approximate any periodic
waveform.
A Sawtooth wave is a complex waveform that contains all the integer harmonics (f, 2f,
3f, 4f, etc.) with descending intensity (A, A/2, A/3 etc.).
Triangle and square waves are complex waveforms in which the spectra contain only
the odd numbered partials (f, 3f, 5f, 7f, etc.). The higher harmonics of a triangle wave
have lower amplitudes than those of a square wave (proportional to the inverse square
of the harmonic number as opposed to just the inverse).
Spectra based on common waveforms (sinewave, all harmonics, odd harmonics)
display very characteristic, distinct sonorities. A Sinewave sounds pure and empty.
The sound of a saw-type spectra, depending on the fundamental pitch, is rich, harsh,
and sometimes nasal, bright. Sounds based on the spectrum of the square and triangle
waveform sound hollow, and are very similar to the sound of a clarinet. Those
characteristics persist even when changing the amplitudes of the partials. That’s why
one can create interpolations between waveforms by changing the amplitudes of their
partials. In this way a continuous timbre scale can be built up between the sine and
the other waveforms (triangle, sawtooth, square).

2.2 Practice

Open the patch called “HarmonicSpectra.maxpat”!

In the panel in the bottom left corner you can choose from 2 Demonstrations
(Demonstration01 and Demontration02) and 4 tests.

Demonstration01 presents a matrix, where, by pressing the buttons in the ‘Amplitude

Presets” panel you can hear and become familiar with the sound of the sinewave, the
sawtooth-wave (All Partials) and the triangle-wave (Odd Partials) and the transitions
between them. The button marked 0 in the panel triggers the fundamental sinewave.
Buttons 1 thru 7 add more partials and build up spectra containing more and more
harmonics with increasingly higher amplitudes at higher frequencies.

Test01_Level01 allows you to check your ability to hear the transition between the
sinewave and the sounds containing all partials and odd partials. In the TEST panel
press the button ’!NEW SOUND!’ to hear one of the fifteen spectra. Listen to it and
click on the numbered button (0-7) you think is correct. If you need to hear the sound
again, press the ’!REPEAT!’ button.
Your answer will be evaluated in the TEST panel. The correct answer consists of 2
"hits". Type: “All Partials or Odd Partials and Step (how many partials ). A green led
in the middle of the TEST panel indicates both type and number of partials are
correct. If it is red, neither choice was right. . Answers close to the correct solution
will be indicated with amber, yellow and light green.

Test01_Level02 allows you to guess 10 spectra in a row.

Demonstration02 presents a matrix, with three different fundamental frequencies

(100, 200, 400 Hz).
In Test02_Leve01 and Test02_Level02 you can check if you can differentiate between
spectra containing all partials, odd partials at the different fundamental frequencies
(100, 200, 400 Hz).
3. Hearing changes in the inharmonic spectrum

In this exercise, the timbres of inharmonic spectra are explored. The sound examples
are created through additive synthesis, where 1- 48 sinewaves are added together.
A scale of inharmonic sonorities is presented, where different spectral densities are
the result of different amplitude and frequency ratio settings. The dimensions of
rough-beating-smooth, dense-sparse and of bright-dark are explored.

3.1 Theory

An inharmonic spectrum contains partials whose frequencies are not integer multiples
of the fundamental.
The closeness of the partials, their frequency ratio, duration and amplitude all define
the degree of fusion and the characteristics of the sound produced. Whilst it is obvious
what pitch is heard when a harmonic sound is played, with inharmonic sounds it is not
so clear.
The difference between harmonic and inharmonic sounds can be perceived as an
increase in sensory dissonance. The level of inharmonicity is dependent upon the ratio
of the frequencies of the partials and the distance between the partials of a spectrum.
There are different degrees of inharmonicity, ranging from beating through roughness
to distinguishable intervals.
When the difference between the frequencies of the sinewaves is larger than the so
called critical band, they will be heard separately and a percept of an interval or
fusion will occur.

According to Beauchamp1 inharmonic sounds are categorized in three ways:

- sounds with nearly harmonic partials

The partials of this spectrum diverge slightly, but increasingly, from the harmonic
partials so the higher the partial, the greater the divergence. Beating or plucking a
string will produce a spectrum divergent from a clear harmonic one. This
phenomenon results in a clearer, colder, more tense and brighter sound.

- sounds with widely spaced (parse) partials

Since the partials are dispersed and are separate from each other, there is no
”roughness” between them. At the same time, some partials could be in harmonic
relation so they fuse better than others. This is typical for certain acoustic
instruments (e.g. percussion with wooden or metal plates).

- sounds with closely spaced (dense) partials

Since the partials are very close to each other, the roughness of the spectrum is very
evident. Therefore no pitch content can be heard. Most metal plate or stretched
membrane instruments sound this way. Their spectra might sometimes resemble
that of white-noise (see Chapter 7). The sound of a suspended cymbal is a good

                                                                                                               
1  Beauchamp,
James W.: „Analysis and Synthesis of Musical Instrumental Sounds”. In: James W.
Beauchamp (ed.): Analysis and Synthesis, and Perception of Musical Sounds, The Sound of Music.
New York: Springer Science+Business Media, 2007. 1-89.
 example of this, since the frequencies of its partials are only 20 Hz from each other,
which is a tiny interval especially in the higher register.

Inverse spectral density

The spectral density of a sound can be expressed by the number and proximity of
partials. According to Beauchamp2 "spectral density can be measured by counting
the number of spectral peaks above a certain threshold and dividing by the
frequency range of these peaks". A more convenient measure is the inverse spectral
density, which will define the average frequency distance between [the – szerintem
nem kell] adjacent partials.

3.2 Practice

Open the patch called “INHARMONICS.maxpat”

In the panel in the bottom left corner you can choose from 2 Demonstrations
(Demonstration01 and Demontration02) and 4 tests.

In this exercise we explore inharmonic spectra with different spectral densities and
nodes. The spectral density is the result of the combinations of the frequency ratios
and amplitude of the partials.

The ratios of the partials to the fundamental can be seen in the lower central panel.
There are five different presets of inharmonic spectra:
– Very narrow: 1 – 1.91
– Narrow: 1–3.4
– Medium: 1-10.72
– Wide: 1-25.02
– Very Wide: 1-59.7

The six amplitude preset buttons can be seen in the upper central panel. The presets
are as follows:
1: 5 partials representing 5 nodes of the possible 48 frequencies are present evenly
distributed with maximum amplitude
2-5: more and more partials [are growing out – szerintem nem kell] around the 5 nods
6: all 48 partials present with maximum amplitude

Demonstration01 presents six Amplitude preset buttons. By pressing buttons 1 to 6

you can hear and become familiar with the sound of the inharmonic spectrum
becoming more and more dense… The default frequency ratio setup is “Medium”.
Next, choose a different frequency setup (Very Narrow, Narrow, Wide or Very Wide)
and repeat the amplitude steps.
Try changing the amplitude envelope and the length of the partials to hear how it
affects the sound.

                                                                                                               
2  Beauchamp, James W.: „Analysis and Synthesis of Musical Instrumental Sounds”. In: James W.

Beauchamp (ed.): Analysis and Synthesis, and Perception of Musical Sounds, The Sound of Music.
New York: Springer Science+Business Media, 2007. 1-89.
Test01_Level01 allows you to check your ability to hear the different spectral
densities created by the adjacent Amplitude presets. In the TEST panel press the
button ’!NEW SOUND!’ to hear one of the six spectra. Listen to them and click on
the numbered button (1-6) that you think is correct. If you need to hear the sound
again, press the ’!REPEAT!’ button.
Your answer will be evaluated in the TEST panel. The correct answer consists of 1
"hit". A green led in the middle of the TEST panel indicates the answer is correct. If
it is red, the choice was not right. Answers close to the correct solution will be
indicated with yellow.
Test01_Level02 allows you to guess 10 spectra in a row. To begin the test, press the
“START” button. At the end of the 10 guesses a number box will light up and
indicate in percents (%) how many good answers you hit.

Demonstration02 presents a combination of the Amplitude presets with three of the

Frequency Ratio presets (Narrow, Medium and Wide).
In Test02_Leve01 and Test02_Level02 you can check if you can identify the spectra
of different spectral densities.
4. The smooth-beating-rough dimension

In this exercise we examine the continuum between beating and roughness by adding
sinusoid partials whose frequencies are close to each other. The sound examples are
created by combining 144 sinewaves, i.e. by additive synthesis. The default “main”
spectrum is harmonic and contains 48 partials. When transforming the source
spectrum in order to introduce beat or roughness, 2 sinewaves are added above and
below each partial. To control the extent of beating, the frequency difference and the
amplitude of the added sinewaves are changed.
A scale of smooth-beating-rough sonorities is presented.
The dimension of smooth-beating-rough is examined.

4.1 Theory

Beating occurs between two sinewaves whose frequencies are close together. When
the two frequencies are between 1 and a few Hertz apart, our ears hear only one
frequency with a periodically changing amplitude. The speed of the beating is
determined by the difference between the frequencies of the two sinewaves.
If the sinewaves are further away from each other (approx. 10 Hz or more), we can no
longer perceive the beating as separate amplitude peaks, they occur so fast, and we
hear them as ”roughness”.
When the difference between the frequencies of the sinewaves is larger than the so-
called critical band, they will be heard separately and a percept of an interval or
fusion will occur.
In harmonic complex tones, the occurrence of beats will be clearly perceivable as a
new dimension. In the case of inharmonic spectra introducing beat to the original
partials will strengthen already existing beats or roughness, and the sound will be
perceived as being denser.
To create organically sounding beating or roughness it is not enough to add regular
beats, in real sounds beating and roughness change randomly.

4.2 Practice

Open the patch called “Beating_Roughness.maxpat”!

In the panel in the bottom left corner you can choose from 1 Demonstration
(Demonstration01) and 2 Tests.

Demonstration01 presents a matrix. By pressing buttons 0 to 7 in the BEATING

panel you can hear and become familiar with the sound of a harmonic spectrum
(default!) with beating frequencies.
The button marked with 0 in the panel triggers a smooth spectrum without any
beating added (in the case of inharmonic source spectra, the original beating will
remain the same). Buttons 1 - 7 add beating with increasing frequency. In each step,
the value of the beating frequency varies randomly in a fixed range. The random
ranges are:
Nr. of Frequency range Percept
button
1 1-5 slow beat
2 4-8 moderate beat
3 7-12 fast beat
4 11-16 state between beat and roughness
5 15-23 roughness
6 22-32 State between roughness and pitched modulation
7 31-51 Added pitch

In the upper row, the added sinewaves have the same amplitude as the waves in the
original spectrum.

In this exercise you will explore a continuum from a pure harmonic spectrum (0)
thru beating to roughness (7) as discussed earlier. The source spectrum is built from
48 partials with integer ratios. Transforming the source spectrum there are 2
sinewaves are added above and below each partial. The difference between the
frequency and amplitude of the partials and those of the added sinusoids will
determine if we hear beating or roughness.
In the panel marked "Beating" there are two rows of buttons 1-7. In the top row
the amplitude of the added sinewaves is equal to the amplitude of the partials of the
main spectrum. In the lower row the amplitude of the added sinewaves is one third of
the amplitude of the partials of the main spectrum. Pressing the buttons of the upper
row will trigger beats with stronger amplitude changes than the buttons of the lower
row.
The green panel under the buttons indicates both the frequency range and the
amplitude of the beating.

After listening and learning the sonorities of the beating/roughness axes with the
default settings (f0=200 Hz, harmonic spectrum, amplitudes descending along higher
frequencies) try listening to the beating/roughness with different settings. The
settings can be changed in the “Amplitudes of the Partials”, and in the “Frequency
Ratio of the Partials” panels by choosing between the presets.

Test01_Level01 allows you to check your ability to hear the different levels of
beating and roughness. In the TEST panel press the button ’!NEW SOUND!’ to hear
one of the fifteen sonorities. Listen to them and click on the numbered button (0, 1-7
in the upper or the lower row) you think is correct. If you need to hear the sound
again, press the ’!REPEAT!’ button.
Your answer will be evaluated in the TEST panel. The correct answer consists of 2
"hits", 1. For the correct Amplitude and 1 for the correct Step (range of frequency). A
green led in the middle of the TEST panel indicates both amplitude and range of
frequencies are correct. If it is red, neither choice was right. Answers close to the
correct solution will be indicated with amber, yellow and light green.

Test01_Level02 allows you to guess 10 spectra in a row. To begin the test, press the
“START” button. At the end of the 10 guesses a number box will light up and
indicate in percents (%) how many good answers you hit.
5. Combining timbre dimensions of harmonic and inharmonic spectra

This exercise [looks at – nekem inkább discusses] the combination of timbre

dimensions of harmonic and inharmonic spectra presented in the previous chapters.
The sound examples are created through additive synthesis, where 1-16 sinewaves are
added together.
Four independent acoustic dimensions are combined.

5.1 Theory

The four dimensions controlled in the exercise are:

– Harmonicity- inharmonicity
Six categories were created using different sets of frequency ratios of the partials in
relation to the fundamental frequency. Three harmonic and three inharmonic spectra
can be selected. The frequency ratios of the partials can be seen below the matrix
of categories.
– Amplitude relations
The four categories include simple amplitude envelopes along the frequency axes.
The amplitudes ascend or descend gradually, reaching a peak in the spectra, (a) at
the lowest partial, ( b) at the highest partial, and (c) in the middle of the spectrum.
In the fourth category the amplitudes of all partials are equal .
– Length ratios
The length parameter-set is introduced here as a new dimension, it has not [yet –
nem inkább ide kellene???] been [yet] explored in the previous chapters. Varying
the lengths of partials provides spectral fluctuation in the spectra. Five sets of
relative partial length[s] are made available.
– Amplitude envelope
The shape of the amplitude envelope is similar for each partial, but their temporal
sequences are dependent on the lengths of the partials. If the amplitude-peak of the
envelope is moved away from the starting point of the sound, and the lengths of the
partials will not be equal, the pitches of the individual partials will be heard at
different times resulting in a non-fused, melody-like sonic event.
Five different exponential attack-release amplitude envelopes were created with
incremental attack times.

5.2 Practice

Open the patch called “HARM_INHARM_DIMENSIONS.maxpat”!

In the Tab at the bottom left corner you can choose from 1 Demonstration
(Demonstration) and 2 Tests.

The upper part of Demonstration page contains four colunms of presets

(HARMONICITY, AMPLITUDES, LENGTHS, AMPLITUDE ENVELOPES).
Below the tabs you can see representations in form of number-boxes and sliders of the
values of the parameter-presets.
To create a sound you should select a preset from each of the four dimensions by
clicking on the appropriate tab and hitting the PLAY button in the TEST panel. There
are 6x4x5x5=600 different combinations (sonorities) available.

Test_Level01 allows you to check your ability in recognizing the effect of the
different combinations of acoustic dimensions. In the TEST panel press the button
’!NEW SOUND!’ to hear a new timbre. Select a category from each of the four
dimensions you think are correct and press the !CHECK! button in the TEST panel. If
you need to hear the sound again, press the ’!REPEAT!’ button.
Your answer will be evaluated in the TEST panel. The correct answer consists of 4
"hits": harmonics, amplitude, length and envelope. A green led in the middle of the
TEST panel indicates all dimensions are correct. If it is red, neither choice was right.
Answers close to the correct solution will be indicated with amber, yellow and light
green. Also, you can see which parameters are correct in the corresponding number
boxes (1.0 indicates a correct choice and 0 indicates not [yet szerintem nem kell ide]
correct).

Test_Level02 allows you to guess 10 spectra in a row. To begin the test, press the
“START” button. At the end of the 10 guesses a number box will light up and
indicate in percents (%) how many good answers you hit.
6. Filtering White Noise

This chapter explores what happens when we put white noise through simple filters, a
Low Pass Filter, a High Pass Filter and a Band Pass Filter. Filtering is also known as
subtractive synthesis. This type of synthesis is the opposite of additive synthesis,
(where complex sound is created by adding together simple waveforms ). Conversely,
in subtractive synthesis, the starting point is a complex waveform with a complex
spectrum, from which parts are removed by a process known as filtering. White noise
is a useful starting point because it contains all frequencies of the audible range.
Filtering white noise is a very effective way of creating new timbres, it produces a
surprising variety of sounds along many timbre dimensions such as bright-dark,
sharp-blunt, open-covered, noisy-pitched, rough-smooth, voiced-unvoiced.

6.1 Theory

Subtractive Synthesis
A synthesis technique in which parts of the spectrum are subtracted by filtering. A
filter is a device that performs some sort of transformation on the spectrum of a
signal.
The most common types of filters are:
– Low-pass filter: passes the partials below the given cutoff frequency
– High-pass filter: passes the partials above a given cutoff frequency
– Band-pass filter: passes the partials around a given centre frequency in the range of
a given bandwidth.

To filter a sound, we need to specify the following parameters:

– the cutoff frequency (for LP and HP),
– the center frequency (for BP),
– the bandwidth (for BP).

White noise
Audible white noise is a signal that contains random sinewaves of all frequencies with
constant density between 20 Hz and 20 KHz. The term ’white noise’ originates from
its visual equivalent, where white light contains all visible frequencies; red, orange,
yellow, etc.
The waveform of white noise is made up of random amplitude values with no regular
repetition. White noise has a saturated and well-balanced spectrum containing all
frequencies. It is possible to think of white noise as a block of marble from which a
sculpture can be carved. Although, one can filter or subtract parts from any sound
spectrum, the most productive filtering will occur with a source sound that is rich in
partials. Since white noise contains all possible frequencies, we chose it as the starting
point for our subtractive synthesis. It can be thought of as sculpting in sound.
Filtering will completely change the timbre of white noise.

White noise bands

Bandpass-filtering white noise will result in very different sounds depending on the
center frequency and the bandwidth of the filter. In some cases we hear clear pitches,
in others, more pitches or even cluster-like chords are heard, and in other cases we
simply hear noisebands with no pitch content. You will find that the octave (8) band-
pass filter is much noisier than the third (3) filter, but the noisiness of both filters
increases, the higher the center frequency.
It is important to note that even though we are discussing pitched sound versus
noisiness as though they were two separate things, you will hear many states in
between. There is no clear boundary that separates pitch and noise. We could say,
band-pass filtering will add pitch to white noise: it augments the noisiness instead of
replacing it. We still do not know why it is that at certain bandwidths and certain
center frequencies we hear individual or grouped pitches or what causes certain
pitches (the center frequency or the upper or lower cutoff frequencies or even others)
to be heard. Much more research is needed to answer these questions.

6.2 Practice

Open the patch called “WNFilter_MAIN.maxpat”! There are two exercises related to
the subject of filtering white noise:
- filtering sounds separately (see “FILTER” tab)
- building up timbre motifs (Klangfarben melodies) from filtered sounds (see
MOTIFS tab)

FILTER
First press the “FILTER” button on the top of the page. In the tab at the bottom left
corner you can choose from a Demonstration and 4 Tests.

Demonstration presents a matrix, where, by pressing the buttons representing the

filtering parameters, you can hear and become familiar with the sound of white noise
filter[ed] by low-pass filter (LPF), high-pass filter (HPF) and 2 band-pass filters with
two fixed bandwidths, the perfect octave (BPF 8) and the just third (BPF 3). The
buttons of a row represent 7 cutoff/center frequencies at: 125 Hz, 250 Hz, 500 Hz,
1000 Hz, 2000 Hz, 4000 Hz and 8000 Hz. On the top of the panel the button will
trigger the original, non-filtered white noise.

Test_Level01 allows you to check your ability to hear and identify the effects of the
different filters have on white noise. In the TEST panel press the button ’!NEW
SOUND!’ to hear one of the spectra. At the top of the Test panel you can choose,
which types of filtering you want include in the test. Pressing the LP_only tab, for
example, will trigger only low-pass filtered sounds. The LP+HP+BP tab will include
all filter types in the test. Listen to the triggered sound and click on the button you
think is correct. If you need to hear the sound again, press the ’!REPEAT!’ button.
Your answer will be evaluated in the TEST panel. The correct answer consists of 2
parts or "hits": filter type and frequency. A green led in the middle of the TEST panel
indicates both filter type and frequency are correct. If it is red, neither choice was
right. Answers close to the correct solution will be indicated with amber, yellow and
light green.

Test_Level02 is almost the same like Test_Level01, but without any audio feedback
when selecting your answer.

Test_Level03 asks you to guess 10 sounds in a row. To begin the test, press the
“START” button. At the end of the 10 guesses a number box will light up and
indicate, as a percentage (%), how many correct answers you selected.
Test_Level04 is same as Test_Level03 with the difference that
– there is no audio feeedback
– there is no color feedback
– no possibility of repeating the sound to be guessed
so you have to rely purely on you memory of the sounds

MOTIFS
Press the “MOTIFS” button on the top of the page. In the tab at the bottom left corner
you can choose from a Demonstration and 2 Tests.

Demonstration presents a set of matrices, which are reduced versions of the matrix
from the FILTER page. The default is 3 matrices, but up to 6 matrices can be selected
in the number box “nr of parts” in the upper left corner.
In this page, a timbre motif can be created by choosing one filtered noise from each
matrix and hitting the PLAY button in the TEST panel in the center of the page. A
motif is a sequence of 3 - 6 sounds depending on your choice (the default is 3) You
can select the number of parts at the top left conrner of the panel. The tempo, duration
and amplitude of the filtered sounds can be changed in the bottom right corner of the
panel, as in previous patches

Test_Level01 allows you to check your ability to hear and remember motifs of filtered
white noise. In the TEST panel press the button ’!NEW MOTIF!’ to hear a motif. At
the top of the Test panel you can choose, which filter type you want to include in the
test. Pressing the LP_will select only low-pass filtered sounds. The LP+HP+BP tab
will include all filter types in the test. Listen to the triggered motif and choose the
filter[ing – szerintem nem kell]s you think are correct in the matrices and press the
!CHECK! button in the TEST panel! If you need to hear the motive again, press the
’!REPEAT!’ button.
Your answer will be evaluated in the TEST panel. The correct answer consists of 2
"hits": filter type and frequency for each matrix. The led and the percentage
numberbox in the TEST panel indicates the average of the hits. If the led is green, all
answers were perfect. If it is red, no choice was right. Answers close to the correct
solution will be indicated with amber, yellow and light green.

Test_Level02 allows you to guess 10 spectra in a row. To begin the test, press the
“START” button. At the end of 10 guesses, a number box will light up and indicate as
a percentage (%), how many correct answers you selected.
7.  Multichannel  resonant  Filter  
 
In this exercise we examine the phenomenon of resonance. The sounds you will hear,
are created by summing the output of 9 resonant filters from a white noise source. To
control the attributes of the resonance, the center frequency, the bandwidth and the
amplitude of each resonant filter are manipulated.
Dimensions of resonant-blocked, bright-dark, sharp-blunt, pitched-noisy are explored.
 
7.1 Theory
 
Resonance is common throughout physics, mechanics, electricity and magnetism. It
is a tendency for certain frequencies to vibrate or oscillate with higher amplitudes
than others. Resonance is the property that all musical instruments and the human
voice exploit.

Our resonance-creating audio engine is modeled with the help of 9 resonant filters. A
resonant filter is a very narrow band bandpass filter, which creates a ringing sound
thru the addition of a feedback loop. As the bandwidth of the filter gets narrower, the
resonance or ringing will become longer.
The basic resonant filter in this exercise is controlled by three parameters:
– center frequency
– bandwidth: the range of frequencies affected by the filter (in Hz) above and below
the center frequency.
– amplitude of the filtered frequency band.

The outputs of the 9 resonant filters are added together, so that in this way we add
together, or "sum" several instances of subtractive synthesis, i.e. we create sound
through a combination of additive and subtractive synthesis.
 
7.2 Practice
 
Open the patch called “ResFilter_MAIN.maxpat”!
 
In the panel in the upper right corner you can choose from 2 demonstrations
(Demonstration01 and Demonstration02) and 2 tests (Test01 and Test02) from the
pop-down menu.
 
Demonstration01 presents a matrix, where, by pressing the buttons from 0 to 9 in 5
rows in the “Number of filtered bands” panel you can hear and become familiar with
the sound of resonant filters added together. The numbers in the rows represent the
number of the filter-bands added together. Number 1 triggers the lowest filter-band
alone. Buttons 2 - 9 will add each consecutive partial. Clicking the columns to the
left or right changes the spectral centroid of the whole spectrum, modifying the
percept of brightness.
The five rows differ in bandwidth. Filter-bands triggered in the first row (BW 5) are
the narrowest, having the most resonant character. In descending order, the bands
become wider and their noisy character gets stronger, such as in the bottom row
where the BW is 500.
The frequency ratios of the 9 bands are fixed, when clicking the buttons in the matrix.
However, different sets of ratios can be explored by choosing from the preset buttons
located above the “Filter Parameters” panel. Here the fundamental frequency,
duration and amplitude envelope of the sounds can be changed, too. It is advised to
try the resonance matrix with different frequency ratios, fundamental frequencies,
durations and envelopes to get an idea of the range of sounds possible.

Test01_Level01 allows you to check your ability to hear the different resonances
offered by the matrix. In the TEST panel press the button ’!NEW SOUND!’ to hear
one of the forty-five possibilities. Listen to them and click on the numbered button (1-
9) in the appropriate row you think is correct. If you need to hear the sound again,
press the ’REPEAT!’ button. Your answer will be evaluated in the TEST panel with
the help of a led. A green/red led indicates the correct/incorrect answer.

Demonstration02 demonstrates the effect of emphasizing two resonant bands close to

each other. These two bands are more resonant (having narrower bandwidth) and are
noticeably louder than the others. Clicking along the rows will select the position of
the two strongest filtered bands, clicking up or down the columns will change the
bandwidth of each of these two bands.
Explore different frequency ratios, fundamental frequencies, durations and envelopes.
 
Test01_Level01 allows you to check your ability to hear the different resonant areas.
In the TEST panel press the button ’!NEW SOUND!’ to hear one of the twenty-four
possibilities. Listen to them and click on the numbered button (1-8) in the appropriate
row you think is correct. If you need to hear the sound again, press the ’REPEAT!’
button. Your answer will be evaluated in the TEST panel with the help of a led. A
green/red led indicates the correct/incorrect answer.
 
8. Parameters of Granular Synthesis

Granular synthesis can   be   used   to   generate   sonorities   evolving   thru   many   timbre  
dimensions.   It   has   many   realizations   with   many   parameters.   A   very   simple,  
elementary   form   of   granularization   is   presented   here,   exploring   its   basic  
parameters.  
Dimensions of smooth-rough, rough-grainy, dark-bright, pitched-noisy, wide-narrow
are explored.

8.1 Theory

Granular synthesis is a term for systems that work by using very small pieces of
sounds that can be manipulated individually and are redistributed and re-organized to
form other sounds. The grains can be synthesized or cut from a prerecorded soundfile,
each usually between 5 ms and 200 ms long. Each slice is known as a 'grain', and a
sequence of grains is called a 'graintable'. To avoid clicks, each grain is faded in and
out with an amplitude envelope.

The main parameters of the synthesis technique are:

– the waveform of the grain: this will be determined by the source soundfile
– length of the grain (in ms): the length of the grain will determine whether we hear
the timbre of the source sound or the process of granulation in the form of
noisiness.
if the grain is:
- shorter than 2 ms: we can not hear the original sound, we hear noisy clicks
- at about 5 ms: the percept of pitch starts to emerge
- above 25 ms: we hear the pitch and timbre of the source sound.
Between these grain-lengths there are different transitional phases which contain
combinations of noise and pitches and timbres recognizable from the source.
– grainspeed: the rate at which one grain follows another. Certain grainspeeds will
influence the sound according to their frequencies. Very high speeds will add a
pitch to the overall sound. Slowing down the grainspeed significantly, will
transform the sound from a continuous stream to a succession of separate entities.
– position in the soundfile (ms): the grain can be taken from anywhere in the source
sound. If the grain is taken from one point, this part of the soundfile will be just
repeated or looped, resulting in a frozen kind of sound. Taking each successive
grain from random points will result in a rapidly changing sound, provided that the
source sound changes over time.
– transposition (there are two types)
- general transposition where all grains are transposed up or down by an equal
amount
- random transposition: where all grains are transposed randomly within a
specified range

8.2 Practice
 
Open the patch called “Granular_Main.maxpatch”!
 
In the panel in the bottom left corner you can choose from one Demonstration and
two Tests.
Demonstration presents a matrix, where one can explore how the duration of grains
and random transposition will affect the granulated sound.
The thirty granulated sounds are arranged in five rows and six columns. Rows
represent different levels of random transposition. In the top row, there is no
transposition (0). In the bottom row, there is a maximum of 2 octaves of
transposition.
Columns correspond to the lengths of grains. As you move from left to right, the
grainlength will increase from 5 ms on the left side to 200 ms on the right.

Test_Level01 allows you to check your ability to hear the different transformations of
sound that occur thru granular synthesis. Press the button ’!NEW SOUND!’ to hear
one of the thirty presets. To answer, click on one of the preset buttons.
Your answer will be evaluated in the TEST panel. The correct answer consists of 2
"hits", 1 for the correct random transposition and 1 for the correct grainlength. A
green led in the middle of the TEST panel indicates both parameters are correct. If it
is red, neither choice was right. Answers close to the correct solution will be
indicated with amber, yellow and light green.
Test_Level02 allows you to guess 10 spectra in a row. To begin the test, press the
“START” button. At the end of the 10 guesses a number box will light up and
indicate as a percentage (%) how many correct answers you have achieved.