Agostino Di Scipio
MODES OF INTERFERENCE
2005-06
audio feedback system
with trumpet and electronics
commission ZKM Karlsruhe
for Marco Blaauw
1 performer with
trumpet trumpet (or flughorn) of any intonation, with at least 3 piston valves
(for M.B.: may be double-bell trumpet with additional valves for microtones)
electronics 2 miniature microphones
2 loudspeakers (minimum)
mixer
computer equipped with ADC/DAC converters, and running real-time digital-signal-processing software (a PD signal-patch can be requested from
the composer).
documentation includes :
- a short INTRODUCTION to the compositional concept and the process of performance
- description of the minimal TECHNICAL SET-UP
- instructions concerning NOTATION and PLAYING TECHNIQUES
- instructions on COMPUTER OPERATION AT PERFORMANCE TIME
- graphical SCORE for the performance (2 pages)
In addition, a decription of the DIGITAL SIGNAL PROCESSING METHODS is provided, which is needed only if you don't use the composer's own PD implementation.
Karlsruhe, October 2005
L'Aquila, January-March 2006
this document released March 8, 2006
INTRODUCTION
This work is a composed dynamical system. It is entirely based on an audio feedback loop.
The two ends of the loop are (a) a miniature microphone (inside the trumpet), and (b) two loudspeakers. In between are (c) the instrument (tube with its natural
resonances), and (d) a signal-processing computer. A very high feedback gain is requested, so the audio loop (a-b) builds-up until it results in a howl round, also called
Larsen effect. The latter usually represents a technical problem in sound systems, but in this composition it is the main sound source. The trumpet and the computer
are utilized to play with it, turning a problem into an opportunity. The computer dynamically adjusts the feedback gain, trying to keep the overall system in equilibrium at
all times, avoiding Larsen signal saturation (d.1). The overall system remains always subject to fluctuations and perturbations from the environment (e).
In performance, the trumpet player explores the sonic potential of the audio feedback loop by interfering with it in several manners. Actions on the trumpet are meant
(c.1) to modulate the resonances of the instrument (valve action makes the tube act like a variable filter internal to the audio feedback loop), and (c.2) to introduce
noise into the overall system (percussive effects, "breath" sounds, etc.). Both actions change the loop resonances, altering the Larsen tones. The computer transforms
and extends all sounds being produced (d.2), and the output of transformations in turn interferes in the feedback loop.
(c) trumpet :
(c.1) variable resonances
and (c.2) noises
(a) microphone
high gain feedback (Larsen) (b) speakers
inside trumpet
(d) computer :
(d.1) adaptive feedback gain
and (d.2) sound transformations
(e) environment
(room)
TECHNICAL SET-UP
Microphones
Two miniature microphones must be utilized (DPA…tech specs…), placed as illustrated. Use tape to firmly stick the cables onto the bell's inner surface.
mic1 (as far as possible into the tube, but before the turn)
mic2
Audio Connections
computer in2
computer out1 computer in1
computer out2
aux out1 out2 mixer
mic1 input1 aux send1 cable connection
mic2 input2 aux send2 mixer routing
input3 output1 L1 (left speaker) level control slider
input4 output2 R1 (right speaker) audio feedback
L2
dist1
Sound Diffusion
L1 dist2
L3
This work is meant for frontal stereo diffusion.
trumpet
If multiple pairs of speakers are available, positioned around the computer audience
audience, they can be used to replicate the front stereo pair, with and mixer
delays directly proportional to the distance from the stereo pair (delay
L1, R1 = stereo front
time = distance / 344). See illustration.
R1 stage R3 L2, R2 = optional, del time = (dist1 / 344)
L3, R3 = optional, del time = (dist2 / 344)
In any case, only the front speakers should be effectively involved in
the audio feedback loop.
R2
NOTATION AND PLAYING TECHNIQUES
General Information
The score consists in 32 (+ 1) "time-windows", presenting instructions and/or graphical annotations as to actions to perform in the given duration. Most time-windows
have a specific duration. Some don't (denoted by fermatas), and last as long as necessary to accomplish the specific task they require.
The whole divides in two parts, which are identical under all respects but valve action (see below). Overall duration is 9-10 minutes.
Valve Action and Larsen
Operating the piston valves changes the length of the trumpet's tube and its resonances. Because of that, when the audio feedback loop (microphone and speakers)
generates Larsen, valve action directly affects pitch, amplitude and overall quality of Larsen tones (also the time it takes for the Larsen to arise may significantly vary).
That is the basic and most important playing technique utilized in this composition. No blowing is involved (exceptions are discussed later).
Ideally, every single piston combination would give rise to a different Larsen tone. However, due to several acoustical and technical circumstances which are beyond
control, different piston combinations may result in Larsen tones having the same pitch. Moreover, some piston combinations may be Larsen-effective, others may be
less effective or even totally idle, insufficent to generate Larsen at all. Beware that during the performance, any given piston combination may be Larsen-effective at
some times but less effective or even idle at other times : that depends on the very sound material recircultating in the loop, as well as on changes - however slight - in
the distance between microphone and speakers.
Position Relative to Speakers
On stage, find a position relative to the speakers which allows the feedback loop to create Larsen. Small displacements of the instrument - and of your own body, too -
affect the length in the feedback loop, and also alter the acoustical shadows and resonances in the surrounding of the instrument. That offers an additional chance to
play with the audio feedback loop, and may help you varying the pitches arising from any one single piston combination. All movements should be very subtle and
discreet, never overstated.
Piston Combinations and Labels
The notation prescribes what pistons you can play with to interfere with the feedback loop. It leaves to you the actual order and timing of all action, except that a
minimum rate of change is explicitly indicated.
The fundamental piston position ( = no piston depressed) represents a default Larsen-generating position that can be used any time throughout the performance.
When testing the technical set-up, make sure that position actually lets Larsen tones to arise.
The label FUND is used to refer to the fundamental position. Labels listed below are used for other piston combinations, using 1, 2 or 3 pistons. H means that pistons
must be half-depressed ( = any intermediate position). F means that pistons must be fully-depressed ( = all the way down).
The following labels are used in the first part of the work, and refer to the piston combinations graphically depicted :
1.H = any combination with 1 half-depressed piston, + FUND = TTT TTT TTT + TTT
2.H = any combination with 1 or 2 half-depressed pistons, + FUND = TTT TTT TTT TTT TTT TTT + TTT
3.H = any combination with 1, 2 or 3 half-depressed pistons, + FUND = TTT TTT TTT TTT TTT TTT TTT + TTT
1.F = any combination with 1 fully-depressed pistons, + FUND = _TT T_T TT_ + TTT
2.F = any combination with 1 or 2 fully-depressed pistons, + FUND = _TT T_T TT_ __T _T_ T__ + TTT
3.F = any combination with 1, 2 or 3 fully-depressed pistons, + FUND = _TT T_T TT_ __T _T_ T__ ___ + TTT
The following labels are used in the second part of the work :
M1(H or F) = valve action as relative to any musical excerpt where 1 piston at a time is used, + FUND. Pistons fully- or half-depressed.
M2(H or F) = valve action as relative to any musical excerpt where 1 or 2 pistons at a time are used, + FUND. Pistons fully- or half-depressed.
M3(H or F) = valve action as relative to any musical excerpt where 1, 2 or 3 pistons at a time are used, + FUND. Pistons fully- or half-depressed.
The phrase "any musical excerpt" means
- select any passage of existing trumpet music (of any style, from any time in history) where only the prescribed number of pistons is used; or
- select any musical scale (diatonic, chromatic, microtonal, or else) where only the prescribed number of pistons is used.
Only the pattern of piston combinations is relevant here (more details on that in the score, Part 2). Pitches actually heard have nothing to do with the musical excerpts
adopted, as they emerge from the interference of valve action with the audio feedback loop.
Examples :
2.F [< 6"] = explore combinations with 1 or 2 fully-depressed pistons; hold each piston combination for 6" or shorter
M2.H [< 4"] = explore combinations with 1 or 2 pistons as found in any passage of extant trumpet music or any musical scale: depress pistons half-way down,
and hold each combination for 4" or shorter
1.H+3.F [< 8"] = explore combinations with 1 half-depressed piston, and combinations with 1, 2 or 3 fully-depressed pistons; hold each for 8" or shorter
Extra Valves
When using an instrument equipped with more than three valves, all of the above applies as is. However, you can operate one or two extra pistons at any given time
during the performance, at your will.
OTHER PLAYING TECHNIQUES
Valve Clicks
In usual trumpet playing, valve action (depressing and releasing pistons) is typically done silently, avoiding or minimizing any audible by-products of the valve
mechanics. That's the case with long passages of the present work, too - but not with the whole work. In many time-windows valve action has to be made clearly
audible, by deliberately causing or emphasizing the click-noise when operating pistons. Clicks eventually affect the resonances, both inside the tube and in the overall
feedback system, with repercussions on Larsen tones. Also, clicks are transformed and/or extended by the signal-processing into textural or rhythmical materials.
Trills
Should possibly be introduced only with Larsen-effective piston positions. Use any number of pistons to make a trill, not necessarily depressing the piston(s) all the
way down, but only to the point where it audibly affects the audio feedback loop. Speed is free.
You may freely introduce trills, even if not requested, whenever very loud Larsen tones arise, as a way to control and decrease their amplitude.
these require playing the mouthpiece with fingers
Mouthpiece Taps
- one or more fingers gently tapping on the edges of the mouthpiece's cup, covering the cup either entirely or partially. The sound has a clear pitch, depending on tube
length. It may reinforce harmonic frequencies in the audio feedback loop, with repercussions on Larsen tones. It is transformed and/or prolonged by the signal-
processing into textural or rhythmical materials.
Mouthpiece Scratch
- one or more fingers scratch hard against the edges of the mouthpiece's cup. The sound is an intermittent noise, somewhat pitched at the frequency relative to the
tube length. It may reinforce harmonic frequencies in the audio feedback loop, with repercussions on Larsen tones. It is transformed and/or prolonged by the signal-
processing into textural or rhythmical materials.
these require that you bring your lips to the mouthpiece. Do not press lips hard against the cup (pressing them hard, may alter the feedback loop).
Breath ("breathing" sounds)
- blowing-out with minimal air-pressure, always with lips apart from one another (to avoid reed-like action).
Whistle ("whistling sounds", random pitch, but very high), coming in either of two variants :
- blowing-out, with heavy air-pressure and somewhat tensed lips
- blowing-in or -out, with variable air-pressure, while leaning the upper teeth against lower lip.
TongueSlap (single)
- the tongue impacts hard against teeth and palate. Blowing-out.
PulseTongue ("tonguing", multiple slaps)
- short sequence(s) of rhythmically-delivered small tongue-slaps, against upper lip and/or teeth. As fast as possible. Blowing-out, or no blowing at all.
TK (or TKT)
- usual "double" (or "triple") tongue action. The one you can play the fastest should be preferred. Blowing-in or -out, or no blowing at all.
Do never clean the tube during performance : allow the sound of tiny water-drops to enter the audio feedback loop.
DYNAMICS
The notation of musical dynamics is relative to specific playing techniques, so it does not refer to the overall sounding result. Total amplitude may vary, depending on
interactions between overlapping playing techniques, as well as on Larsen tones arising and overall feedback system behaviour.
ppp should be understood as pianissimo as possible
mf and f should be understood almost mezzoforte and almost forte
Beware that - because of the microphone inside the trumpet - the amplification gain level is alwyas rather high : every tiny event of sound inside the trumpet will be
largely boosted. You must keep yourself very attentive and refrain from playing loud. That is especially the case with techniques needing some blowing ("breath",
"whistle") and with tongue action (slaps, pulsed tonguing) : these will be very loud even when you play ppp.
The overall crescendo forks (bottom line of notation) only illustrate how the average total sound level increases, following average density of events and complexity of
sound texture, across time spans of approximately one minute : when a peak is reached, density starts again from a lower degree and starts increasing again.
COMPUTER OPERATION AT PERFORMANCE TIME
Graphical User Interface (GUI)
The following refers to the GUI displayed on the computer screen at performance time, and illustrates how you operate on that. The particular GUI layout is specific
to the composer's own implementation of the internal digital signal processing. Any implementation other than the composer's own will have a different GUI layout,
still the featured controls should be of the same kind and should anyway have the same functions as described here.
Once all audio connections are in place and the computer is ready for use, you launch PD (the open-source programming language for sound design). Select and
open the MODES.pd signal patch. The following GUI appears on the computer screen :
Operations Required
All action is limited to the top-most icons (orange-coloured), and is clearly
invoked in the score. Here a short description is provided of the musical
function of actions requested.
(1) Set inputGain (just before performance begins)
You have to set the inputGain value, replacing the default value (0) with a
value high enough to let the feedback loop generate a Larsen tone, heard as a
continuing tone of rather constant amplitude, or at least without significant
fluctuations.
Testing and fixing the effective value for inputGain must be done at rehearsal
time, with all pistons released (fundamental position) and the performer
standing in the same position where he or she will be performing. Effective
value may vary depending on many acoustical and technical circumstances.
(2) Click StartProcess then click FadeIn/Out (time-window #0)
The audio engine gets started. Buttons proc1, proc2 and proc3 become
automatically active. Trumpet performance can start now.
(3) Click Part1 (time-window #1)
Starts the audio processing specific to the first part of the work.
(4) Click Part2 (time-window #17)
Starts the audio processing specific to the second part of the work.
(5) Switch-off the FadeIn/Out toggle (time-window #32)
The total sound fades-out (this may take some seconds).
Although not meant for direct operation in performance, proc1, proc2 and proc3
may be also operated anytime during rehearsals or sound-check, to switch
three different signal-processing transformations on and off. Similarly, you may
want to click FadeIn/Out to switch the audio on and off, or you may need to
change the inputGain (very cautiosly) anytime that seems to be the case.
Information Available in the GUI
Two timers are featured (cyan-coloured). One ("total_time") starts automatically when you click StartProcess. It shows the time elapsed since beginning of
performance. The other ("section_time") starts and resets when you click either Part1 or Part2.
Two sliders are also featured (cyan-coloured). They start moving as soon as you click StartProcess and keep doing so, incessantly, all through the performance. Their
current value reflects the current amplitude of input sound material. The right slider is roughly directly proportional to input amplitude, the left is inversely proportional.
The left slider is specially helpful. It represents the dynamical factor that scales down the input as the input's own amplitude increases. Because of that, feeble input
sounds will be boosted and, conversely, powerful ones will be scaled down and weakened. This adaptive balancing over the input level influences the total output sound
too (not only the Larsen sounds, but also the signal-processing tranformations) : the output becomes weaker as the amount and density of sound material recirculating
into the loop gets larger - and viceversa.
If the left slider moves closer to the bottom and enters the red ( = input sound is really loud), the output gets so weak it almost disappears. You will learn to avoid that,
avoiding feeding too loud sound materials, or temporarioly avoiding to add any sound at all (e.g. you may temporarily refrain from delivering new sounds - if you were so
doing - or you may shift to a piston combination not particularly good at generating Larsen - if Larsen sound are currently peaking).
A Note on the Computer-operated Sound Transformations and their Relationship to the Trumpet Performance
The real-time sonic transformations include three different layers of re-sampling of the input sound materials, also changing their pitch and stretching their duration.
The results may vary considerably, The result of tansformation is heard sometimes as a prolongation of input sounds, sometimes as strands of pulsed materials
randomly scattered in intricate textures, or more rhythmically patterned.
It is relevant for you, that some transformations (not all) are driven by some nuances in the trumpet performance - namely, the dynamics and density of sounds
(density = rate of events delivered).
