Swarm of Echoes was influenced by the short novel The Invention of Morel by Adolfo Bioy Casares in which a fugitive escapes to a mysterious deserted island only to find a group of tourists appear seemingly out of nowhere.  Full of “reasoned imagination” and the fantastic, the novel questions “the boundaries between fantasy and reality.” (Casares, Introduction, viii).  The fugitive continually questions his own senses and at times concludes that the appearance of the vacationers on the island must be hallucinations.  Swarm of Echoes attempts to create real environments consisting of rain and waves that evolve and are juxtaposed with synthetic environments of strange bird-like sounds and impossible crashing waves.  The piece thus takes influence from the imagery and metaphysical ideas of the real and the immaterial.  The listener is taken on a journey into worlds that are believable, yet strange and mysterious.

            In one scene in The Invention of Morel, before the appearance of the people, the fugitive searches an abandoned museum on the island for medicine and food.  He stumbles upon a secret basement, in which he “saw the same room duplicated eight times in eight directions as if it were reflected in mirrors.  Then I heard the sound of many footsteps…” (Casares, 18).  Then as he continues to explore, he is “intermittently escorted by the diligent swarm of echoes, many dimensions of the same echo.” (Casares, 19). The idea of swarms of sounds surrounding the listener is an occurring theme in Swarm of Echoes.

 

Pacific Slope

            Pacific Slope by Joseph Anderson also influenced the aesthetics of this piece.  In Pacific Slope, Anderson uses sounds of waves and trees cracking recorded in the Pacific Northwest to create a piece with acousmatic ideas, a piece in which sounds are recognizable, but obviously not seen.  Acousmatic music, according to Denis Smalley, is largely concerned with space and while not entirely acousmatic, Swarm of Echoes concentrates on spatial experience and somewhat recognizable sounds.

Under the influence of Pacific Slope, this piece uses ambisonic technology to create a three-dimensional sound experience, which aids in creating believable sound environments and allows for the complex layering of sounds to be spread out across space.  For the listener, it will be difficult to experience with less than four speakers.

Anderson has a way of getting the listener “inside” the sounds he records, exhausting them spectrally and spatially in order to reveal their innermost qualities.  He then uses these qualities to influence other sounds in the piece.  The metaphor of the wave is prevalent even when the waves themselves are absent from the piece; bells surround the listener and rumble and crash mimicking the movement in spectrum of a wave crashing on the Pacific coast.  In Swarm of Echoes, similar techniques are employed to manipulate rain to sound like booming thunder, and both the sound of snow crackling under a snowboard and whispering pitches to rise and fall with the rhythm of the ocean.

 

            Swarm of Echoes is deeply influenced by my home in the islands of Hawaii.  In Hawaii, the ocean plays a crucial role in almost every aspect of life, from the abundant nutrients and food it produces, the rain it feeds that makes the islands teeming with life, to the currents that bring life to its shores from far away lands.   The constant sound of the ocean can also be therapeutic and meditative.   Nonetheless, the ocean has played a large part in my life, and was a major influence on this piece.  In March 2010, I was fortunate enough to bring a 3D microphone and recorder from the University of Washington to the island of Oahu to record the sounds of the ocean.  These recordings became the basis for this thesis. 

            The piece opens with the calming sound of the rain and soon after, the waves.  As the waves increase in volume and number, the waves begin to transform into a more synthetic sound, which eventually takes over the environment.  From this point, the listener is transported into a different world, which is at once somewhat believable and strange.  Strange bird-like sounds begin to saturate the sound field and are continually at odds with the whispering waves and a more dominant “flapping” that cuts through the space in rhythmic intervals.  Synthesized birds and frogs soon take over the piece only to be silenced by a large storm that begins to swirl around and leads into the climax of the piece, waves that have been transformed into granular melodies.  The pitched grains are fleeting as the ocean takes back the space one final time before the entire environment dissipates into nothing.

            Wavelets are small finite mathematical waves that can be used to extract features from a signal.  Having studied the use of spectral analysis of sound using the Fourier transform, I became interested in the use of wavelets to represent a sonic signal that has a strong noise component.  Swarm of Echoes is a first major attempt to explore the use of wavelets to resynthesize a 4-channel ambisonic signal.  Using the mathematical programming language Matlab (http://www.mathworks.com/products/matlab/) along with the wavelab toolkit (http://www-stat.stanford.edu/~wavelab/), I experimented with wavelets ability to reproduce each of the three sets of recordings.  There are many different wavelets each with different properties, and after many experiments, the Daubechies wavelet yielded the best representation of the sound files. 

The Discrete Wavelet Transform (DWT) essentially acts as a series of filters that results in a number of signals corresponding to octave bands in frequency based on the sample rate of the input signal.  Each recording was converted to a sample rate of 48K using to make sure that the bands from each sound file would match.  

For each 4-channel recording, I performed the DWT on each of the 4 channels individually.  Then, from Matlab, I imported the wavelet transform output (wavelet coefficients) into SuperCollider (SC), the audio programming language (http://www.audiosynth.com/) used to create this piece.  In SC, I convolved the wavelet coefficients with white noise filtered to the corresponding frequency band.  Wavelets offer a few advantages over working directly with the sound file.  Not all bands have to be synthesized, nor do they have to be played back at the same rate.  Using wavelet coefficients convolved with noise allows for variable play back rates without significant changes in pitch.  Before using the convolution method, I had done experiments by simply multiplying the coefficients by filtered white noise, which is ring modulation.  The problem arises when the play back rate of the coefficients is modulated.  This causes the pitch of the coefficients to change and thus the side bands that are a result of ring modulation change as well.  This is a perceivable pitch change in the overall sound.  Using convolution eliminates the side bands and reduces the perceived pitch change.

The main goal for the use of wavelets was to reproduce entire 3D sound fields and to attempt to create new sounds that retain some of the spatial properties of the original sound field.  Using wavelets to represent 3D sound fields has much potential for future work.  The original recordings of the rain and waves are presented relatively unprocessed at the very beginning of the piece and the remaining sounds are all comprised of wavelet coefficients convolved with white noise.

 

Recordings

 

There are five sound recordings used for the creation of Swarm of Echoes.   Each one of the first three, purposely, involves the sound of water.  One consists of the sound of the Pacific Ocean crashing against the northwest coast of the island of Oahu at La’ie Point.  These recordings were taken with a three-dimensional microphone, the ST250 by SoundField, and an Edirol R-4 4-channel portable recorder.  In March of 2010 I traveled to my home of Hawaii and took a day of recordings at La’ie Point.  Using the 3D microphone, I tried to capture the movement of the waves from multiple angles in order to have many choices in the compositional process. 

A second set of recordings was taken at the Canyons Resort in Park City, Utah the first week of April 2010.  These recordings were taken using the Zoom H2 recorder, which captures a two-dimensional surround field.  I choreographed the movement of snowboarders coming close to the microphone, and then getting further away in hopes of attaining interesting spatial movements in the recording.  These recordings make up the last section of the piece. 

 The third and fourth recordings were found at Sound of Space  (www.soundofspace.com), which is a website devoted to spreading ambisonic recordings.  The two recordings I chose were the sound of rain hitting the roof of a tent and the sound of birds.

The last recording consists of frogs.  The frogs were recorded in Bellevue, WA in May 2011 using the ST350 3D microphone and DEVA IV 8-channel digital recorder.  I chose to record the frogs for two reasons.  First, they fit the theme of environmental recordings and go along with the birds. Second, with the microphone positioned at the center of a group of frogs, there are interesting spatial aspects including natural call and response across space.

Each recording was preprocessed to eliminate background noise using the Waves  (http://www.waves.com/) plug-in z-noise, which analyzes a recording and attempts to reduce any unwanted noise, such as the sound of the wind in the case of the recording of the waves.

 

SuperCollider Work Flow

            Swarm of Echoes was created with the audio programming language SuperCollider.  Some example code is included on the DVD.  There are several layers of sound happening at any time during the piece and similarly there are a few layers of code. 

The first layer of code contains all the wavelet synthesis, some spectral enveloping and placing individual wavelet bands into space.  Once an individual sound was satisfactory, it was saved to a sound file and then used in the next layer.  The wavelet coefficients were imported from Matlab into SuperCollider as WAV sound files.  Since there are nine bands used per mono source and 4 channels per surround sound signal, there are thirty-six sound files for each recording.  These sound files were loaded into buffers, scaled to the appropriate playback rate (the wavelet coefficients contain 1/(2^J) samples where J is the level number starting with one, which corresponds to highest band in frequency), moved around in space (or not depending on the application), scaled in amplitude, delayed at different rates, then convolved with filtered white noise at the corresponding frequency band and mixed together in the output.

This second layer consists of all the different sections, which are comprised of all the different sound files from the first layer.  This section of code provides critical timing issues, amplitude enveloping, some minimal filtering and most of the spatialization techniques.  Each section contains about more than five sound files from the previous layer.

The final layer is simple and puts all the sections together.  There is still some minimal spatialization and filtering in this layer.  Also, some sound files from layer one were not used in any section and were added in this final layer.