Re-Sampling by Electro Mass

~ 32 ~

Re-sampling and Re-pitch Looping Techniques Brief History of Re-sampling Re-sampling can mean a couple of things in music technology, one of the meanings is used when changing the sample rate on an audio file. A good example would be changing a wav from a 48k sampling rate to 44.1k so the audio can be using with a compact disk but this is not the type of re-sampling that this part of the project will be exploring. The re-sampling process this paper will be covering is about re-recording a sound several times, layering a given sound for different purposes. Re-sampling is a destructive process meaning once you have rendered your audio then it cannot be reversed unless you have the original copy still saved on your computer's memory device. Re-sampling is a technique that many people have heard of, but generally is not well documented; this part of the paper will hopefully bridge that gap and help producers explore the power and flexibility of this technique. The ideals of re-sampling have been around for many decades, a look into some of history that's connected to the technique. Re-sampling has been used as a production tool since the early days of recording technology, mixing engineers would record many instruments/voices on one track due to the restrictions of two or four track tape. This type of method was pioneered in 1949 by Les Paul and Mary Ford as Benson (1988) explains: "In the absence of wartime restrictions, applications of the new technology spread quickly. In 1949 performers Les Paul and Mary Ford pioneered the technique of recording multiple parts performed by one person" Benson (1988 p10.2) Although this can be seen as a creative thought pattern, other areas in music were showing more creative processes relating to re-sampling. The technique is about changing audio through re-recording sounds. Some of the fundamentals of the re-sampling method were created thanks to the minimalist scene as Demers (2010) describes: "The inspiration behind such layering owes a great deal to minimalist works of Steve Reich's Drumming or Terry Riley's In C, in which the scaffolding of a basic rhythm serves as the foundation for extenuated figuration in higher registers." Demers (2010 p96) This is a great early example of the re-sampling although Terry Riley's 'In C' did not use re-sampling directly but the methodology of layer sounds in creative ways was a fundamental part of his publication. If we come closer to our time, it shows how technology has helped mould this technique. Digital samplers were a mile stone for re-sampling and re-synthesis techniques being introduced in the early 1980's. One of the first basic samplers was the QASAR M8 designed in 1978 which progressed to the fully developed digital synthesiser known as the Fairlight's 'Computer Music Instrument' which became available in 1979, Manning (1993).

~ 33 ~

These early samplers had the ability to store digital audio and re-synthesis as Manning (1993) explains "In the Fairlight, synthesis from first principles using mathematically generated wave data is replaced by sampling technology, where externally generated sounds are digitized and their patterns stored in memory for processing and re-synthesis." Manning (1993 p260) With the advent of digital technologies came a new world exploration, giving the producer the chance to save audio clips, affect them in creative ways and re-record the result and even re-synthesis if desired. In the late 1980's cheaper versions of samplers were created by Akai, E-mu Systems, Synclavier, Fairlight Instruments and others, overcoming some of the early problems sampler design encountered as Manning (1993) explains "By the end of the late 1980's the basic problems of fidelity in MIDI samplers had finally been surmounted, leaving the flexibility of the hardware and its operating software as the primary features distinguishing one sampler from the other" Manning (1993 p308) With the great success and flexibility of the sampler instruments more advanced DSP systems were created with bigger memory for storing data. This gave the capabilities of re-recording or re-sampling several times, each with its own effect adding to and sculpting the sound. With this came a more competitive market driving down prices so the average mid class person could afford samplers, this gave birth to new music styles, sampling old vinyl and composing new music from bits of other artists records, although this history is beyond the scope of the paper. This was the birth of the bedroom producer being able to buy equipment to produce music on, having such wide spread technology that can facilitate re-sampling, it can only promote experimentation with the technique. Samplers were not the only thing on the market that can do such processes and workflows, the digital audio workstation (DAW) is a graphical interface to record, edit and mix audio as Roads (2001) explains "A graphical sound editing and mixing program provides a two-dimentional view of sonic structure. The vertical axis presents multiply rows of tracks, while the horizontal axis presents a time line." Roads (2001 p183) DAW's our the modern day composers tool for recording, production/editing and mixing. Even the most modest computers can run several third party software in one of many forms like VSTs, with this enhanced flexibility compared to even just a decade ago meaning that there is so many different ways to create sound design and produce tracks in the digital domain. This paper is going to explain the findings of the projects research in the techniques of re-sampling and re-pitch looping and explain some of the advances in technology.

~ 34 ~

Introduction of Re-sampling and Re-pitch Looping This part of the paper is to explore modern techniques in sound design, using the techniques surrounding re-sampling to help aid understanding with these relatively new techniques and testing some of the myths about the subject. Analysis on hardware, testing total harmonic distortion (THD) and inter modulation distortion (IMD) after several generations of re-sampling. Also ideas about each step or generation in the re-sampling process and how you fill and shape the sound with software and hardware tools. After re-sampling is explained a look into re-pitching audio, describing the sound design possibilities and compositional methods surrounding both re-sampling and re-pitch looping. Resampling Testing In this section equipment will be tested for THC and IMD. First the test track was made up including a 1000Hz sine wave for THD, one 19kHz and one 20kHz for IMD. The test tone will be recording once through the system to give a reference. The results taken from this will then be crossed referenced with the internally recorded KP3 tests. The signal will leave the laptop, into the Korg Zero8 mixer via firewire, then get recorded into the KP3. Once in the KP3 it will be internally re-sampled 3, 5 and 7 times and recorded back into the Ableton Live 8 on each. The only D/A to A/D conversions are thought to be to and from the Zero8 to the KP3. Once the test track has been recorded 4 times it will then be analysed by Izotope‟s “RX Advanced”. The highest 5 distortion peeks will be noted and used to record the THD and IMD. The maths for these calculations will be done in this example. The 1000Hz sine wave is at -6 dB, the 8000Hz distortion is at -95 dB. First you have to add the difference of the 1000Hz sine wave to 0 db to the distortion which would be -89 dB. Then with your -89 dB you have to get a percentage of that distortion.

The percentages for each of the 5 highest distortion peeks then get squared and added. Then the answer is square rooted to get the root mean squared value of THD.

1000hz 8000hz

(Fig 31) Example THD

~ 35 ~

All of the harmonics will be recorded using Izotope‟s RX Advanced spectral analyser. All of the test tones where made with Adobe Audition in 48kHz sample rate and 24 Bit. Here are the results of the 4 recorded generations on THD.

Frequency

214Hz

2000Hz

3000Hz

6000Hz

8000Hz

THD%

Gen 1 (TT)

-98.8 dB

-104.3 dB

-104 dB

-107.8 dB

-103.6 dB

.0072%

.0038%

.0039%

.0025%

.0041%

0.01%

Frequency

214Hz

2000Hz

3000Hz

6000Hz

7000Hz

THD%

Gen 3

-100 dB

-107 dB

-107 dB

-106 dB

-100 dB

.0056%

.0028%

.0028%

.0031%

.0063%

.0098%

Frequency

111Hz

2000Hz

3000Hz

6000Hz

7000Hz

THD%

Gen 5

-98.8 dB

-101 dB

-101.8 dB

-103.9 dB

-103 dB

.0057%

.0044%

.004%

.0031%

.0035%

.0097%

Frequency

214Hz

2000Hz

3000Hz

6000Hz

7000Hz

THD%

Gen 7

-99 dB

-101.8 dB

-101.9 dB

-103.9 dB

-96.7 dB

.0063%

.0045%

.0045%

.0035%

.0082%

.012%

The results show that there is no cause to worry about constantly re-sampling when considering THD. The main differences between values can be put down to the inconsistencies of the preamps in the Zero8 Mixer as well as undefined amplitude adjustments due to the interface. KP3 is digital so the only places it could possibly gain THD is the D/A conversion from the Zero8 mixer with an A/D to the KP3 and back. The number of Gen‟s has nothing to do with this process and remains digital.

~ 36 ~

Here are the results of the 4 recorded generations on IMD.

Frequency

1000Hz

2000Hz

3000Hz

18000Hz

21000Hz

IMD%

Gen 1 (TT)

-34 dB

-85.5 dB

-62.8 dB

-45 dB

-46 dB

22.3%

.059%

.81%

6.3%

5.62%

23.85%

Frequency

1000Hz

2000Hz

3000Hz

18000Hz

21000Hz

IMD%

Gen 3

-35.5 dB

-47 dB

-62.8 dB

-45 dB

-46 dB

21.1%

5.62%

.91%

7.07%

6.3%

23.81%

Frequency

1000Hz

2000Hz

3000Hz

18000Hz

21000Hz

IMD%

Gen 5

-32 dB

-47.5 dB

-51.8 dB

-40.2 dB

-41 dB

28.1%

4.73%

2.88%

10.96%

10%

32.25%

Frequency

1000Hz

2000Hz

3000Hz

18000Hz

21000Hz

IMD%

Gen 7

-34.2 dB

-47.2 dB

-54 dB

-41.7 dB

-42.2 dB

21.8%

4.89%

2.23%

9.22%

8.7%

%25.78

The IMD results are somewhat puzzling, at first thoughts it seemed that the Zero8 could be distorting or creating a time delay/feedback loop that was creating the severe distorts. This was then crossed referenced with the test track running though the Zero8 mixer and not the KP3. (Fig 32) shows the recording through only the Zero8 mixer, (Fig 33) shows the recording through both the Zero8 Mixer and the KP3. This shows that the KP3 A/D and D/A conversions are flawed. The point of the experiments were to see the effects of re-sampling when done several times which in fact have came out positive, only seeing a 2% rise from Gen 1 to Gen 7 which would be hard audible if at all. Gen 5 has an anomaly at 1kHz that makes the THD very high, this in fact did not affect Gen 7 so has been put down as a isolated problem with the Zero8 mixer when recording.

~ 37 ~

The results have also brought up worries about the quality of the KP3 dealing with IMD and high frequency. One of the reasons could be a brickwall antialiasing filtering as Roads (2002) explains. “These steep filters can cause significant time-delays (phase distortion) in midrange and high audio frequencies.” Roads (2002 p32) This could explain the severity of the IMD in (Fig 32).

(Fig 32) Zero8 Mixer Only

(Fig 33) Zero8 Mixer and KP3

~ 38 ~

DSP in the Time and Frequency Domain Time domain DSP are techniques such as fixed time delay effects, creating phase modulation from variable delay lines, or granulation where you repeat or vary very small splices of audio from a piece of recorded audio. On the other hand you have frequency based effects which is a process of analysing an inputted signal and then processing the frequency data and re-synthesising. Some of the earliest effects were time based like chorus and flanger, these are created using variable delay lines as Lazzarini et al. (2008) explains. "A well-known side-effect of variable delays is the phase modulation of the delay-line input. This is the basis for all classic variable-delay effects such as flanging, chorusing, pitch shifting, and vibrato." Lazzarini et al. (2008 p10) These types of effects are very common and originally came from analogue roots being common in popular music in the 1960s and 1970s and still being used to this day in all types of music from rock to electronic. These analogue effects have now been digitalised into software like VST's which most people use today with their digital audio workstation (DAW). Another category of time based effect are fix delay effects, the difference being as Roads (2002) states "In fixed time delay unit, the delay time delay does not change while the sound is passing through it. In a variable delay unit, the delay time is constantly changing" Roads (2002 p435) These effects can create depth like an echo or be rhythmically synced to create interesting patterns. Although fixed time delay are simple in nature, when combined with variable time effects it can create several options and be a creative tool for any producer. Distortion is another common effect in all genres of music, waveshaping synthesis is a well known type of distortion where it modulates the input signal with a shaping function or signal similar to frequency modulation synthesis. There are many types of waveshaping like movable and fractional each having their own characteristics. Waveshaping can harmonically enhance your source input, expand low and clip high amplitudes giving warm fuller sound all the way up to a distorted crisp transformation of the input signal through heavy clipping and expansion. In recent years multiband waveshaping distortion has been made available. This effect splits the input signal into frequency bands, giving the user the opportunity to affect each band separately as Fernandez-Cid et al. (2001) states "The "Multiband Decomposition" block is a filterbank that produces a set of reduced bandwidth signals from the original full-band one. There signals are processed with WaveShaping elements whose "overtone pattern" can be independently adjusted in a per band basis" Fernandez-Cid et al. (2001 p282) Anyone can see with the added bands, new flexibility to waveshaping can enhance audio with more precision and artistic expression with automation. Different approaches are found with frequency domain effects, or spectral

~ 39 ~

processing. Spectral processing programs work on an analysis and re-synthesis basis. This analysis locates relevant particles of the inputted audio and allows them to be resynthesised at specified frequencies. There are certain methods to spectral processing as Sethares et al. (2009) described "This enables a variety of routines including spectral mapping (changing all partials of a source sound to fixed destination frequencies), spectral morphing (continuously interpolating between the particles of a source sound and those of a destination sound), and what we call Dynamic Tonality (a navel way of organizing the relationship between a family of tunings and a set of related spectra)." Sethares et al. (2009 p71) As you can realise spectral processing has a very different frame work compared to time domain effects, giving exciting new way to access and effect data reaped from a chosen source material. A lot of time domain effects have been made with added spectral analysis for extra features or improved quality. Other effects like reverb have had a rethink on how to emulate a space. Convolution reverb takes a recorded impulse of the room which provides the information of the envelope of the room, by obtaining the impulse response (IR) which is a type of complicated filter. The arbitrary input signal is then convolved with the IR. This then gets summed with the original input signal in a varied quantity to taste, making the audio sound like it has been played in that room of the recorded impulse. This technique surpasses many of the other techniques and shows true quality of frequency domain effects compared to parrelled time based ones. Time based effects like granulation, and similar processes like in the micromontage section of the paper, take time based effects in an interesting direction to the ones mentioned above. Granulation is at the forefront of modern time domain effects being able to granulate and modulate with huge sound design capabilities. Granulation is solely in time domain as Roads (2001) describes "Granulation is a purely time-domain operation. This stands in contrast to techniques such as the Fourier, wavelet, and Gaber transforms, which analyze a signal in both time and frequency." Roads (2001 p188) Granulation can bring new life to old sounds, reordering and adding interesting rhythms to input signals. A form of time-stretching is granulation, whereby you can change the length of a given sound but not affect the pitch, unlike a standard transpose function, through repeating very small sections of audio in equal measure to fit the new lengthen of time, instead of slowing it down or speeding up. As you can imagine this opens up synced tempo changes with ease and can add interesting textures and variation to a musical element once recorded.

~ 40 ~

Granulation in the time domain is not the only way it can be done, huge scope and possibilities come from analysis found with frequency domain operations. Exciting new technologies have been made available using particle synthesis as explained in Roads (2001) like Glisson, Grainlet, Trainlet and Pulsar synthesis. The synthesis particles are dissimilar to the physical kind as Roads (2001) explains with an insight to how the sounds are formed. "Unlike particles probed by physicists, synthetic particles inhabit a virtual world. This world is invented, and the waveforms produced by the particles are algorithmically derived. They may be simple and regular in structure, forming smooth pitched tones, or complex and irregular, forming crackling noisy masses" Roads (2001 p121) This opens up an exciting new world of sound design and manipulation which highlights the advance possibilities of frequency domain processing which will be taking us into the future.

~ 41 ~

Effects Chains and Re-sampling Generations A re-sampling generation (Gen) is every step or time you render your audio, each step can incorporate, mixing effecting audio files from previous steps as well as the most recent audio generation. Each step can have some sort of DSP affecting the audio, below is an explanation of the three categories of DSP and techniques that the project has sectioned for the purpose of re-sampling only. This is a guideline and is not technical fact or set in stone in some cases, only to aid in the re-sampling process. DSP Categories Additive - The additive category adds extra frequencies to your audio, some like bit-crushing are destructive in some senses but still produce a lot of extra frequencies. Associated DSP -- Bit-crushing, Re-sampling Frequency, Distortions,

Formant Talk Filters, Phase Vocoding, Subtractive/Destructive - This category is you're sculpting equipment with filters being the main tools. Other interesting methods are granular synthesis and re-pitch looping which can be very creative tools.

Filters, Spectral Band Isolators, Granular Synthesis, Re-synthesis, Re-pitch Looping,

Spatial - The last category is where you control your sounds three dimensions, depth, width and height Reverb, Delay, Chorus, Phaser, Flanger,

~ 42 ~

Not all the effects are dedicated to one of the categories. (Fig 34) is the Venn Diagram to help showing this and have as a reference when processing a sound.

(Fig 34) Venn Diagram

Re-sampling Generations Each sound can have as many generations (Gen) as any given person wants, each generation can be a mixture of previous ones or completely separate. Below is family tree style of looking at re-sampling. Basic re-sampling model (Fig 34) (Audio Example R1) Gen1 has no effects; Gen2 has a lowpass (LP) filter with frequency cut-off modulation. Then both Gen1 and Gen2 then get summed on Gen3. Gen3 then also has separate lowpass filter modulation on the cut –off to create both Gen 4 and 5. Both Gen4 and Gen5 are summed on the same channel while been affected by the

Source Sound

2 Oct Square

wave

End Product

Gen 1

Gen 3

Gen 5

Gen 4

Gen 2

(Fig 35) Audio Example R1 Model

~ 43 ~

distortion, highlighting the LP filtering creates a very complex sound. This form of re-sampling can be used by a producer which has already written and recorded a melody. The producer can then make several re-sampling models affecting the audio and slicing the best bits from four re-sampling variations for instance. This next example will demonstrate this technique (Audio Example R2)

Source Sound

2 Oct saw wave and

sub sine wave

Gen 1 Gen 1 Gen 1

Gen 2 Gen 2 Gen 2

Gen 4

Gen 3

Gen 5

Finish

Gen 3

Gen 4

Gen 6

Finish

Gen 3

Gen 4 Gen 5

Gen 6 Gen 7

Finish

Variation 2 Variation 1 Variation 3

Variation 1 Gen 2 - bit-crushing Gen 5 - LP filter cut-off modulation. Gen 6 - LP filter, cut-off modulation.

Variation 2 Gen 2 - a locked LP filter

cut-off separating the bass

and sub freq.

Gen 3 - a highpass filter

separating the midrange

and high frequencies

Gen 4 – fuss distortion

Gen 5 - LP filter, cut-off modulation. Gen 6 – multi band delay

Variation 3 Gen 2 has a locked LP filter

cut-off separating the bass

and sub freq.

Gen 3 has a highpass filter

separating the midrange

and high frequencies

Gen 4 – fuzz distortion

Gen 5 - flanger

Gen 6 - LP filter, cut-off modulation. Gen 7 – Talk Delay

Gen 5

(Fig 36) Audio Example R2 Model

~ 44 ~

Variation 1, 2 and 3 are all different re-sampling techniques of processing the audio. Variation 1 is the most basic, affecting all of the frequencies at once on every stage, using mixing to get the desired levels. Variation 2 gives the producer a chance to separate a low band and a high band, only placing effects on the high band so that the bass will be unaffected by effects that could take power away from the lower frequencies. Variation 3 is far superior to the previous variations, it has three layers of frequency separation, the first band is for bass, second is for mid range frequencies and the third is for spatial processing. Variation 3 give you more control as a producer, provokes more thought into processing bands of frequencies and strengthens good practice, keeping the bass frequencies pure and creating movement and depth with the mid and highs. Re-sampling processing are experiments that requires musical intellect, each of the generations needs a good level of understanding and ability to mixing audio, knowledge of effects and there parameters as well as a strong ideals of what their trying to achieve. Re-sampling is cold in the fact that it is destructive editing, unless you save every generation and keep it aside, you can easily lose your work if something goes wrong later on in the process. The producer does have tools to prevent and detect mistakes. A spectral analyser should always be monitoring the output to detect a loss of sub frequencies and high peaks in the frequency range. A phase meter will also help to detect phase errors. This can be created by time delays between Gens or from repeatedly layering certain effects. This is not good for the sound as Roads (2002) explains “frequency-dependent phasing shifts distort musical signals audibly and interfere with loudspeaker imaging” Roads (2002 p20) With these tools and experience, any producer has the ability to use these techniques successfully but what would be the point compared to a sting of effects chains? Apart from the obvious CPU workload effects chains have on a system, compared to playing a finished re-sampled wave file, it gives you a different way to approach sound design. All of the examples use basic wave forms but all of them have unique qualities, qualities that cannot be detected just by listening to the finished piece of audio in some cases. Layering effects to such an extent lends itself well to giving audio character, such is the holy gravel for an producer to develop there „sound‟ which defines them, re-sampling can be a gate way to move away from traditional production techniques. The audio examples (RS Model 01, 02, 03, and 04) are explained in the next section, the audio examples have undergone no mastering or any editing after the re-sampling process, this was to show the raw state in which is audio is left like after re-sampling. Re-sampling is a sound design method and should undergo move processing when mixed in a piece of music.

~ 45 ~

Practical Examples and Descriptions Percussive Bass (RS Model 01) Ripple Bass (RS Model 02)

Source Sound

2 Oct square wave

and sub sine wave

Gen 1

Gen 3

Gen 2

Gen 4

Gen 5 Gen 6

Gen 7

Gen 8 Gen 9

Finish Clean undistorted Bass

Percussive Bass

Gen 3 - Highpass filter

Gen 4 - LP filtering, cut-off mod

Gen 5 - LP filtering, cut-off mod

Gen 6 - Sum on Gen 4 - 5

Gen 7 - Talk mod delay

Gen 8 - Fuss distortion

Gen 9 – LP filter, cut-off LFO

mod

Gen 2 Finish Clean undistorted Bass

Source Sound

2 Oct saw wave and

sub sine wave

Gen 1

Gen 3

Gen 4

Gen 5

Gen 6

Gen 7

Gen 8

Gen 9

Gen 10

Ripple Bass

Gen 4 - LP filtering, cut-off mod

Gen 5 - LP filtering, cut-off mod

Gen 6 - LP filtering, cut-off mod

Gen 7 - LP filtering, cut-off mod

Gen 8 - LP filtering, cut-off mod

Gen 9 - LP filtering, cut-off mod

Gen 10 – Highpass filter

RS Model 01 - This model demonstrates a complex sound

from the basic DPS building blocks, using LP filters to create

movement, distortion to excite that and a talk modulated

delay to create width. Then as a final process a tempo synced

LFO on the LP filters cuff-off creates the percussive, rhythmic

elements. This model would be impossible to achieve with

just an effects chain on one channel, this gives the reason to

do these processes which promote individualism and express

that would be almost impossible replicate.

RS Model 02 – This model is a complex arrangement of LP

filtering. This was designed to show that these processes cannot

be done in a VST or hardware synthesiser. Native Instruments

‘Massive’ only has 2 filters, and cannot achieve some of the

mixing functions like Gen 6 and 7. One problem with this design

was not having a highpass filtering at Gen 3. Due to all the filter

movement being concentrated around 100Hz to 1000Hz some of

the LP filter swells crossed around 200-500Hz and because of

the resonance created undesired peeks

~ 46 ~

Messy Bass (RS Model 03)

Heavy Distorted Bass (RS Model 04)

Gen 2 Finish Clean undistorted Bass

Source Sound

2 Oct saw wave and

sub sine wave

Gen 1

Messy Bass

Gen 3 – Highpass

Gen 4 – LP filtering, cut-off mod

Gen 5 – Talk mod delay

Gen 6 - LP filtering, cut-off mod

Gen 7 – Multi tap delay

Gen 8 – Fuzz distortion

Gen 9 - LP filtering, cut-off mod

Gen 10 – Multi tap delay

Gen 3

Gen 4

Gen 5 Gen 7

Gen 6

Gen 10

Gen 8 Gen 9

Gen 2 Finish Clean undistorted Bass

Source Sound

2 Oct square wave

and sub sine wave

Gen 1

Heavy Distorted Bass

Gen 3 - Highpass

Gen 4 - LP filtering, cut-off mod

Gen 5 - LP filtering, cut-off mod

Gen 6 - Distortion

Gen 7 - LP filtering, cut-off mod

Gen 8 - LP filtering, cut-off mod

Gen 9 - Multi band pass delay

Gen 10 - LP filtering, cut-off mod

Gen 11 – Distortion

Gen 3

Gen 4

Gen 5

Gen 6

Gen 7

Gen 8

Gen 9

Gen 10

Gen 11

RS Model 03 – this model shows the intuitive methods in

which you can process the sound. Having two lines, one for

low mid range LP filtering, creating the belly of the sound

(Gen 4-6) and the other for stereo effects creating width and

depth( Gen 5-6). Then these separate lines get mixed

together and have a fuss distortion to excite the quiet delays

in the background before undergoing LP filtering and a Multi

tap delay. This re-sampling model gives constantly changing

complex timbral patterns which again would be hard to

replicate with software synthesiser.

RS Model 04 – This model is another example of using two

lines to create separate parts of the sound, then using

distortion at the end to make them gel together. The line with

even number Gens is creating powerful mid range using

distortion and LP filter sweeps. The line with Odd numbers is

creating the spatial elements for the final distortion to effect.

All of the models have the same Gen 2 which is a stationary

LP filter, this creates a clean bass track which will be

unaffected by any of the distortions and delays that could

‘muddy’ or create unwanted phasing.

~ 47 ~

Re-pitch Looping

Re-pitch looping is a technique developed by the project. This technique was

realised through the KP3 re-pitch looper. Through a relatively small section of audio,

a huge amount of sound design can be yielded. The producer first has to create a

landscape; this audio landscape needs to have interesting morphing qualities with

vast amounts of movement. Then with 8 bars of source material, you can loop and

re-pitch the audio in musical measurements. When you loop small measures, like a

semi-quaver, you completely change the context of the original 8 bars of audio. In 8

bars of music you have 32 crochets, 64 semi-quavers and not forgetting dotted

rhythms. By this point you are into hundreds of variations, all of which can be at

different pitches. This type of modulation is not widely used and can be an

alternative to low frequency oscillators modulating LP filters cuff-off frequency that is

flooding the electronic music scene today. Interesting area of this technique is the

seamless way you can change the loop length and pitch. The integration between

each pitch and loop length can create some interesting and unique modulations

which really sets this technique apart. The legato style effect between pitches as well

as loop length changes and modulations imposed on the audio before re-pitch

looping create interesting timbre changes which is a positive outcome as Roads

(2002) explains

“a time-varying timbre is usually more tantalizing to the ear than a constant

spectrum” Roads (2002 p139)

With a multi layered sound to begin with, there are several reason why this technique

can be used and explored by the modern producer.

Re-looping requires an interesting 8 to 16 bars of audio busy with modulating effects

and huge timbral changes. The source audio does not have to sound good when

played by its self; you need a variety of pitches and several layers of effects for the

best results. The techniques from the previous re-sampling section are a perfect

example of audio that would reap the best rewards, due to the type of processes and

complexity woven in the audio. For this reason the 4 re-sampling audio examples

(RS Model 01, 02, 03, 04) will be used to create 4 new examples (RS Model 01, 02,

03, 04 (re-pitch looper)).

~ 48 ~

Practical Descriptions and Compositional Methods

RS Model 01, 02, 03 and 04 (re-pitch looping)

As you can hear from the examples, there are loads of variations and sounds with no

obvious context to the original RS Models. Each of the loops captures a small piece

of audio; this capsulation retains a small piece of the modulating effects. This

repetition makes the audio have profile and structure in contrast to its heavily

modulated source sound. The less cluttered and tonally dynamic source sound will

not give you as much variation in the re-pitch looping and thus hinder the amount of

output.

This type of workflow when writing music requires big quantities of audio to select

the memorable or diverse sounds that fit into the songs „narrative‟ to craft a melody

from. Once the sounds have been selected, further effects and processes can be

applied in context to the music your writing. This creates several layers of thought at

each stage from synthesising a sound, re-sampling, then re-pitching and finally

piecing together a melody and applying them final edits and effects which make it

into a professionally finished track.

An alternative method of producing, mentioned in the re-sampling section (Audio

Example R2), is to have a melody already written and rendered. This type of

composition allows for more errors, needing to perfect re-sampling to get fantastic

results of a predefined melody. Using several re-sampling models can help improve

the amount of usable material, add variation on repeated phrases and improve your

re-sampling technique by repetition trial and error.

These forms of composition are different to working in MIDI inside a DAW where you

have complete control of your music with automation tracks. Re-sampling requires

intuition when affecting the sound, similar to a good chef evaluating each stage

before moving on, knowing that if a mistake was made the destructive nature of the

process means you can never go back. A well trained ear for mixing and trial and

error is the best method to learn this type of processes and the big names in the

electronic music industry most have been through hundreds of combinations before

mastering this form of composition.

~ 49 ~

Discussion of Re-sampling

and Re-pitch Looping

This part of the paper has explained and demonstrated interesting forms of sound

design and composition which is relatively unheard of in the academic world.

Tackling some of the myths relating to quality of these processes, which were

positive, only strengthening the case and integrity of the techniques. Audio examples

and descriptions of re-sampling models and re-pitch looping having been another

successful part to the project with explanations of compositional methods to

incorporate the sound design created.

Each of the re-sampling models explained in this part of the paper was at a basic

level. Improvement of these techniques can be found in numerous places. Going

right from the start only sine, saw and square waves were used to in the demos, in

today‟s hardware and software synthesiser market there are so many programs that

could invite more interesting sound sources like additive, FM, granular, phase

modulation and particle synthesis. „Found sound‟ could also be incorporated in both

source sound and layering Gen if desired. Each of the Gen‟s could undergo more

advanced effects then the basic LP filtering, delays and distortions which were used

in the demo, this could open up completely different set of results.

The re-pitch looping has also illustrated another approach to sound design, working

with a generative process. Transforming the four 20 second clips (RS Model, 01, 02,

03 and 04) into minutes worth of audio. Unfortunately due to KP3 effect, some parts

of the audio have created slight clipping laced in the output. This could be eradicated

using restoration software like Izotope RX Advanced.

The aims was to create understanding in these techniques, the models presented

show good practice with isolating bass and sub frequencies and ideals on frequency

band processing instead of affecting the whole range. Both the testing and examples

were positive, giving clear understanding of techniques with audio to aid the

literature. Compositional advice in the later sections gave a connection to written

music and sound design bridging the gap and making the processes seem more

accessible. All of the aims were meeting with positive reaction and will hopefully help

producers of all levels of expertise.

~ 50 ~

Conclusion

This paper was designed to explore and illuminate some exciting new methods of

sound design and composition with digital technology. This project focused on

interesting modulations and processes to affect the audio with both time and

frequency domain effects in real-time and offline processes. This view point of the

paper can be summed up by this quote from Roads (2001).

“Transformation has now become a central aspect of our sonic art, taking place on

many time scales, from global modifications of macrostructure to micro operations on

individual particles and samples.” Roads (2001 p181)

The papers three subject areas: spectral processing, micromontage/granulation and

re-sampling all have complete unique qualities in the transformation of sound in both

offline and real-time processes.

The research of the three subject areas could facilitate a brief history in all and both

software design in spectral processing and micromontage as well as re-sampling

models and techniques. The required level of understanding was meet and

executed. The references were all from reliable sources like Computer Music Journal

which were both modern in most cases and books from some of the leading names

in music history and technology like Curtis Roads and Peter Mann.

Professional software and equipment were reviewed, tested and evaluated. The

results were positive in proving the integrity of the processes but maybe not the KP3.

The software reviews also made some good distinctions between the projects

spectral processer and professional software, outlining the importance of frequency

domain processing in the future.

The software designed, although not professional in quality, could perform some

interesting transforms of sound. The spectral processing software was a success,

performing several effects when working in the frequency domain. The spectral

processor was moderate in quality but realistically meet the requirements of the

paper. The micromontage patch is a working piece of software but did not meet the

original expectation. Due to a more automated approach, the patch has more

qualities in generative sound design other than a composition method. Although this

was not the initial ideal, the software does have some thought provoking functions

and touches on granulation which is an exciting area of music technology.

The audio examples for each of the re-sampling models give a physical insight to the

creation of the sound as well as a step by step guide describing techniques for

different sound design options.