(Names in brackets mean that these are separate programs. The others are sub-modules of MODIFY.)
1 PITCHSHIFT: shift the pitch of infile while retaining (more or less) the same duration
2 TIMESTRETCH: stretch or compress the infile in time, while retaining the same pitch
3 REVERB: use 3 parameters to create a kind of reverberant effect
pitch transposition factor (Range: -0.33 to 0.33)4 SCRAMBLE: random reordering of grains within a timeframeNB: To get a scramble effect, you do need to provide a timeframe (in ms) with -rrange to overcome the default setting of 0.5 GRANULATE: 'granulate' (put a grainy surface on) a source
A density of 1.0 will achieve this; < 1 will introduce gaps, and values > 1.1 (out of a range which ends at 2.0) begin to sound smooth again.6 BRASSAGE: powerful segmentation/fragmentation procedures using constants or time-varying breakpoint files
7 FULL MONTY: finely tuned granular textures using parameter ranges with upper and lower limits, which can be set as constants or as time-varying breakpoint files, or mixtures of the two. You are recommended to use the graphic program GrainMill for this, if it is available.
In Sound Loom, the upper limit parameters for Mode 7, which here begin with an 'h', are called 'limit' instead. A number of the parameters are also given different names. These are noted below, prefixed with 'SL:'.
infile soundfile to be processed, normally, but not necessarily, mono (see space)
outfile output soundfile written after processing
velocity (SL: timeshrink) speed of advance in infile, relative to outfile (Range: >= 0)This is the inverse of a timestretch (i.e., 1/n: higher values make the output shorter, lower values less than 1 make it longer). This permits an infinite timestretch. Remember this when using Mode 2.density amount of grain overlap (Range: > 0)Values < 1 leaves intergrain silence, i.e., gaps. Extremely small values will cease to perform predictably.grainsize size of the grains in milliseconds (Range: must be > 2 * the length of the splices; overall range: 2.0ms to 1997.12ms; Default: 50ms)
pitchshift transposition of grains in + or - (fractions of) semitones
amp (SL: grain loudness range) gain applied to the grains (Range: 0 to 1; Default: 1.0)Use only if you want amplitude to vary over a range &/or in time.bsplice (SL: start splice) length of start-splices on grains in ms (Default: 5)
esplice (SL: end splice) length of end-splices on grains in ms (Default: 5)
space (SL: spatial position) set stereo position in outfile 0 = L, 1 = R (Range: 0 to N) -rrange (SL: search range) of search for next grain, before infile 'now' (Default: 0 ms)
-jjitter (SL: scatter) randomisation of grain position; Range 0 to 1 (Default: 0.5)
-loutlength maximum outfile length (if end of data is not reached)Set to zero (the Default) for this parameter to be ignored. But if velocity is 0 anywhere, outlength must be given.-cchannel (SL: channel to extract) extract and work on just one channel of a stereo input (Range: 1 or 2) Set channel to 0 (the Default) for this parameter to be ignored.
-x do exponential splices (Default: linear)
-n no interpolation for pitch values (quick but dirty)
The presence of the upper limit parameters
hvelocity (SL: timeshrink limit),
hdensity (SL: density limit),
hgrainsize (SL: grainsize limit),
hpitchshift (SL: pitchshift limit),
hamp (SL: loudness range limit),
hbsplice (SL: startsplice limit),
hesplice (SL: endsplice limit), and
hspace (SL: spatial position limit)
make it possible to specify a range of values. When set, the program chooses random values for that parameter for each grain between the lower and upper limits. For example, if pitchshift is -3 and hpitchshift is 4, the pitch of each grain will be somewhere within the interval of a perfect 5th.
NB: Furthermore, any of these parameters may vary in time (provide the name of a time value breakpoint file).All parameters except outlength and channel may vary over time.
MODIFY BRASSAGE is a re-working of the original CDP program called GRANULA. Those with a PC system can (and are advised to) use the graphic version, GrainMill, with built-in graphic breakpoint editing. This command-line version offers greater flexibility with the shapes of the grains, handled with the bsplice and esplice parameters. The fact that velocity is an inverse function makes stretching time less intuitive, but allows scope for (even more) extreme values than the graphic version.
MODIFY BRASSAGE is mainly designed as a way of texturing pre-existing sample data. It is therefore not the same as a 'granular synthesis' program which synthesises sound using granula techniques. A search for 'granular synthesis' on the web will provide a great deal of informatiion on current work in this area.
This command line is exceptionally complex, especially because the low and high values for a number of (optional) parameters are not adjacent. (This was done in order to bring the command line into line with the requirements of a graphic interface.) It is recommended that you create a template batchfile which lines up the parameters as shown above, that is, with a newline which enables you to put the high values directly under the low values. But remember to delete the newline before saving (or it won't read the rest of the command line) and save to a new name so that you don't overwrite your template.
The flexibility with which the parameters can be handled is enormous. You can have:
- constant values
- random values selected (by the program) from a range specified by low and high limits
- time-varying values specified by breakpoint files
- a mixture of constant values and breakpoint files, such as a constant low limit and a time-varying high limit, or v.vs, or even breakpoint files as both lower and upper limits. (NB: the 'lines' created by joining up the points in the files for the upper and lower limits must not cross: at any given time location in the two files, a lower-file value must not be higher than a higher-file value, and v.vs.).
The different modes of the program now make it easier to realise a specific musical objective, in the manner of presets. Use Modes 6 and 7 for 'brassage' (serious fragmentation of soundfiles), for more control and for creating the more complex, granular types of texture.
The three key parameters are:
- grainsize how coarse or fine the granulation will be; 'granular synthesis' typically uses gransizes of less than 50ms
- velocity how the infile is read; it actually relates to grainsize:
- a value of 1 means that the read-steps are the same as the grainsize, and the length of the outfile will therefore be the same as that of the infile
- a value less than 1 means that the read-steps are shorter than grainsize, material will overlap and the length of the outfile will be longer than that of the infile: time-stretch
- a value greater than 1 means that the read- steps are longer than grainsize, material will be skipped, and the length of the outfile will be shorter than that of the infile: time-compression
- density how the outfile is constructed:
- a value of 1 means that the outfile lays out grainsize grains end to end; the infile and the outfile will be the same length, but the outfile will have a 'granulated' surface texture due to the action of the splices on each grain.
- a value less than 1 means that the time between grains in the outfile will be longer than the grains themselves, introducing audible gaps ('pointillist' effects). E.g., a grainsize of 0.05 (50ms) divided by 0.5 (a value for density less than 1) = 0.1. Thus grains only 5/100ths of a second long will occur every tenth of a second in the outfile, leaving significant gaps of 95/100ths of a second.
- a value greater than 1 means that the time between grains in the outfile will be shorter than the grains themselves, causing the grains to overlap in time: the next grain will start before the previous has finished. Thus two (or more) grains will be mixed in the outfile, making the texture more dense. If there is no pitch transposition (of the grains), this will be heard as a more intense, thicker sound, with internal 'edges' determined by the nature of the grains and their splice slopes. If there is pitch transposition (of the grains), the grains will be scattered in a vertical space defined by the pitch range. E.g., a grainsize of 0.05 (50ms) divided by 2.0 (a value for density greater than 1) = 0.025. Thus grains which are 5/100ths of a second long will occur every 2.5 100ths of a second, overlapping therefore at each grain's half-way point. This is how the 'massive' textures characteristic of granular synthesis are produced.
The other parameters serve to create qualitative differences in the granular textures.
The most noticeable of these is pitchshift, which adds or subtracts semitones or fractions of semitones to/from each grain.
- Using constants, you can 'tune' the texture, perhaps creating a number of differently tuned texture streams which will then be mixed in some way to create textures with a harmonic flavour.
- Using ranges, you can create thick, churning textures by placing a high density output within a narrow pitch band, or conversely, widely scatter grains (thickly or thinly depending on density) over a wide range.
- Using a constant and a breakpoint file makes it possible to create, for example, triangular shaped ranges (i.e., pitch fields), the base being the constant, and the triangular shape affected by the breakpoint file.
- Using double-breakpoint ranges make it possible to design pitch fields which move through time in a variety of patterns. But bear in mind that the pitch shape of the original material forms the starting point for these transpositions. Also remember that the two breakpoint shapes (lower and upper) are not supposed to cross (lower rising above the line of the upper, or upper dipping down below the line of the lower).
The amplitude parameter is also very important, for variation in amplitude add a natural suppleness to the texture, and can also be used to create aurally directional movement: crescendi and decrescendi. In case of overload, which the program does its best to avoid, this parameter can be used to scale down the output (use as a constant). The maximum value for amplitude is 1.0, so it is not possible to scale the infile amplitude upwards. There is usually some loss of signal level due to all the splices, so it is generally wise to ensure that the infile is at or close to a maximum level.
The space parameter is another way to enrich the output by using left and right limits to scatter the grains in horizontal space or have the granular stream move in a directional manner. Using constants enable you to locate the output at a specific point in the horizontal plane.
Range and jitter are another way to loosen things up. The former provides a search range in the infile from within which grains are selected, and the latter adds a further degree of randomisation to the time placement of grains in the outfile. With short grainsize and high density, jitter will have a more a more subtle effect, noticeable in many cases as a modulation or softening of the comb filter effect arising from overlapping grains. When density is 1 (in which grains are normally end-to-end), it will have the effect of overlapping some grains, while leaving (very small) spaces between others.
One of the main pioneers in this area is Barry Truax. I recommend you try to get to listen to his famous compositions Riverrun, The Wings of Nike, and Pacific. Also see his articles:
- "Composing with real-time granular sound" (Perspectives of New Music 28(2) 120-134
- "Discovering Inner Complexity: Time Shifting and Transposition with Real-time Granulation Technique" (CMJ 18(2) 38-48
Also see The Computer Music Tutorial by Curtis Roads (et al.), MIT Press, 1996), pp. 168-184 and pp. 440-444.
It is also useful to study the role of texture in 20th century composition. The relationship of melody and textural accompaniment in the work of Bohuslav Martinu indicate how the 'background' often becomes aural 'foreground'. Pieces such as Stockhausen's Momente and Gruppen are classic examples of a full-blown collage/texture technique, as are the amazing collages in Berio's Laborintus II.
But 'brassage' and 'granular' textures really came into their own once the computer became a musical tool, when it became possible to create sound textures out of tiny fragments of sound, e.g., 0.035 sec long. MODIFY BRASSAGE provides an enormously powerful tool with which to explore the effects and side-effects of these processes.
If you have the CDP GrainMill program, you will be able to realise and study the HTML tutorial for it written by Philippos Theocharidis when he was at the University of Newcastle. The sounds can be made very easily by loading the settings files he provides.
For the moment, the best way into the program is first of all to run Modes 1-5 to explore the presets. Don't be afraid to try extreme values, for they will more easily reveal what this wonderful program can do. After that point, I recommend that you revise the section above about the 'three key parameters' and try out the different settings described for grainsize, velocity and density. (But do use template batch files for Modes 6 and 7.)
As a basic point of reference, consider the following combinations:
velocity density result in the outfile < 1 < 1 time-stretched, with sonic material repeated, but with gaps in the outfile < 1 > 1 time-stretched, with sonic material repeated and overlapping grains in the outfile > 1 < 1 time-compression, with sonic material skipped, and gaps in the outfile > 1 > 1 time-compression, with sonic material skipped, and overlapping grains in the outfile
The addition of a pitch range to the above may make the effects more observable. With time-varying parameters, you can move fluidly between these various basic configurations.
ALSO SEE: MODIFY SAUSAGE, which is able to apply the same functionality to several source files.
insndfile1 first input soundfile to be convolve with
insndfile2 the second input soundfile, which must not be longer the insndfile1; both soundfiles must have the same channel count
outsndfile output soundfile
transposfile textfile of time semitone-transposition transposition pairs
Convolution is based on complex mathematical operations between two source soundfiles. There is a great deal of information about convolution available in Wikipedia, but your editor does not know just which ones Trevor Wishart uses in his program.
The result of convolution is unpredictable but fascinating. The operation of this particular sub-module can take rather a long time, so you are recommended also to make use of FASTCONV.
Example command line to create a Doppler effect:
dshift -d15 pan.brk dshift.brk 12.5
-dN a flag to set the time it takes for a sound to change direction. N is the time in milliseconds. Default: 10ms. This flag is useful in order to avoid clicks.
inpanfile input PAN breakpoint file to use when adding Doppler effect
outtransposefile output transposition file (in semitones) to use with MODIFY SPEED Mode 2 (on the soundfile panned with the PAN breakpoint file) to create the Doppler effect
distance_between_speakers optional parameter to specify the space in meters between the speaker pair that will play back the sound
The Doppler effect is the phenomenon of pitch change that we hear when something emitting a loud noise, such as a horn or siren, goes quickly past us. This program was written by Rajmil Fischman. It has been implemented in Soundshaper, but not yet in Sound Loom, either on the PC or on the MAC. It can be run from the Command Line on the PC, but has not yet been ported to the MAC.
There are 3 main steps to creating this effect with DSHIFT:
- Start with a pan breakpoint file and use it with PAN (MODIFY SPACE 1) to create a sound that moves between the speakers in a stereo field.
- Then use DSHIFT to use the pan breakpoint file again as a guide to creating the Doppler pitch change effect. The output of DSHIFT is a pitch transposition file in semitones.
- Run MODIFY SPEED with the panned soundfile and the pitch transposition file as inputs, to create the final, Doppler-shifted, soundfile.
To operate DSHIFT in Soundshaper:
- Pan a mono sound as normal, using a breakpoint pan file. The resulting panned soundfile is the output of this process.
- Select SPARE FILE and load the pan breakpoint file.
- Run DSHIFT from the Data Menu: "Doppler-Shift Transposition".
- Name the output breakpoint transposition file (in semitones); the default name is 'dshift.brk'.
- Deselect SPARE FILE (the pan file) and run MODIFY SPEED Mode 2 (semitones), with the panned soundfile you made in Step 1 as the soundfile input and 'dshift.brk' (or your own name) as the time-varying transposition file. The output of SPEED is a panned soundfile that also changes in pitch.
This is a 'real life' acoustic effect that can be used to enhance literal recreations of a natural environment. The case that most easily comes to mind is that of a vehicle passing by, especially if it is blasting a horn or siren as it moves. More abstract sounds can be used, with a 'real life' effect adding verisimilitude to a sound coming from a strange or unfamiliar source.
infile input stereo soundfile, which has been panned over time in the stereo field
time the time-point in the sound at which you wish to query the pan position
If you are carefully positioning sounds in stereo-space it might be useful to know exactly where on the stereo-stage a panning sound is at any particular time. This process assumes the input is an (originally mono) sound which you then panned so that it moves around the stereo space.
The process uses the file loudness data on the two channels to calculate where in the stereo space the sound appears to be. It assumes that the infile contains a sound that has previously been panned to a position in the stereo field. If this is not the case, the results will be misleading. For example, it could be a stereo file which is not in fact moving across the spatial field.
It is important that the input soundfile was originally mono.
- If it had always been a stereo file, you could not find out where anything is in the stereo stage by comparing the left and right channels they could, for example, carry completely different sounds.
- If the file were originally mono and then panned, the left and right channels will have the same signal, but at different levels.
- Thus, by comparing those levels, you can tell where the sound is on the stereo stage.
Suppose you have made a sound that moves about in space, such as across the horizontal field from left to right. Now you want to place another sound such that it is in exactly the same spatial location as the moving sound at a specific time point. The question is: where is that spatial location?
MODIFY FINDPAN asks you to supply a stereo input soundfile and a time. It then returns the pan location of the input soundfile at this time-point. You then use this information to specify the placement of the sound you want to add to the mix.
1 GAIN: adjust level by factor gain
2 dBGAIN: adjust level by dB gain
3 NORMALISE: force level (if necessary) to the maximum possible, or to the level given (Range: 0 to 1)
4 FORCE LEVEL: force level to maximum possible, or to the level given (Range: 0 to 1)
5 BALANCE: force the maxmimum level of infile to the maximum level of infile2
6 INVERT PHASE: invert phase of the sound
7 FIND LOUDEST: find loudest of 2 or more files
8 EQUALISE: force all of 2 or more files to level of loudest file. The input files are rescaled in input order, with output names as outfile, outfile1, outfile2, outfile3 etc.
infile soundfile to have its level altered
infile2 soundfile to which infile is to be balanced
outfile resulting soundfile, with readjusted level. When there are 2 or more input files, outfile is a generic name to which are added '1', '2' etc. to identify the matching output files.
gain in Mode 1, floating point multiplier: < 1.0 reduces level, > 1 increases level;
OR, in Mode 2, level expressed in dB (Range: -96dB to 96dB). NB: every drop of -6dB halves the previous level.NB: a gain value of -1 inverts the phase of the sound. This can sometimes be used to maintain high signal levels without clipping. See SUBMIX GETLEVEL.-llevel amplitude level (Range: 0 to 1)
Information about the current level of a soundfile is given by SNDINFO MAXSAMP. Information about the units used to express loudness is summarised in the chart Getting a Handle on dB values. This chart shows both gain multipliers and dB values. (Click on 'Back' on your Browser to return from these locations.)
It is best to adjust level earlier rather than later in a set of processes. Also, it is important to start with a source sound which has a reasonable signal level. Artefacts will appear if a low signal is boosted too much.
Level overflows reported by Csound are usually due to a fault in the orchestra file.
This will be one of the 'frequently used' functions, e.g., to:
- readjust a level which has either dropped a little too much or overflowed due to processing
- to boost a source signal which is a little too low to enter into a processing sequence
- to reduce the level of a soundfile after PVOC (the Phase Vocoder) has reported overflow. You can use the gain factor recommended by PVOC, or double-check levels with MAXSAMP before applying gain.
- When filters of time-varying Q are used, the output level tends to drop as the Q increases. This can be compensated for by using MODIFY LOUDNESS in Mode 5 (Balance sources), submitting the original file and the filtered file as the two sources. The resultant soundfile will have the sound of the filtered file with the amplitude contour of the original sound.
- And it can also be used to impose an envelope on a sound (just like ENVEL IMPOSE), if you use a time-varying gain value in Mode 1.
Example command line to create a pitch-specified delay :
newdelay newdelay in.wav out.wav 67 0.75 0.1
insndfile input soundfile
outsndfile output soundfile
midipitch pitch of the output, expressed as a MIDI Pitch Value. MIDI Range: -76 to +136. (See Notechart)
mix the amount of delayed signal in the final mix. 0 gives a 'dry' result, 1 a 'wet' result
feedback produces resonance related to delay time (with short times)midipitch may vary over time.
CDP already has a time-variable delay in MODIFY REVECHO, Mode 2. This program specifies the delay time as a MIDI (or pseudo-MIDI) pitch value. In a delay line, with sufficient feedback, very short delays of typically less than 40ms (25Hz) create a pitched resonance. Midipitch allows you to set this delay time according to the required pitch. However, discrete echoes can also be expressed as a pseudo-MIDI value (e.g., -12 produces delays of about 1/4 sec.), thereby making an important connection between pitch and rhythm.
Example command line to create phased signal:
phase phase 1 four.wav fourinv.wav
1 Invert phase of a sound.Parameters
2 Enhance stereo separation by phase shifting. Input file must be stereo.
infile input soundfile; must be stereo for Mode 2
outfile output soundfile, stereo in Mode 2.
-ttransfer amount of signal to be used in phase-cancellation. Range: 0 to 1. 1 = all.
In Mode 1 playback of the output inverted soundfile shows that there is no audible difference from the input soundfile. However, if you mix the original sound with the phase-inverted sound, you will generate a silent output: positive and negative amplitude values cancel out. However, if you mix the input and output with an offset, you get signal and hear an overlap of the two soundfiles (the cancelling out does not fully happen). You can try this with SUBMIX MERGE.
Also see: MODIFY LOUDNESS Mode 6, which likewise inverts the phase.
Mode 2 is supposed to be a means of creating an enhanced stereo image by suppressing any aspects of the Left image which have bled to the Right, by adding (part of) the phase inverted Left signal on the Right, and vice versa. T Wishart writes: I learned of this idea from a mixing engineer I met in France, but I'm still not convinced that this achieves anything except in very special circumstances (with particular types of material). The output may sound the same as the input, except perhaps on the largest PA systems.
Inverting the phase of signals might be appropriate in some mixing or texturing procedures. See the technical literature.
1 REVERSE: Sound plays backwards
2 SHRED: Sound is shredded, within its existing duration
3 SCRUB BACK & FORTH: As if hand-winding tape spools over a tape head
4 LOSE RESOLUTION: Sound is converted to a lower sample rate or bit-resolution
5 RING MODULATE: Against input modulating frequency, creating sidebands
6 CROSS MODULATE: Two input soundfiles are multiplied, creating complex sidebands
7 QUANTISE: Sound is converted to specific bit-resolution (mid-rise).
infile input soundfile
infile2 2nd input soundfile, for cross modulation
outfile output soundfile written after processing
repeats number of repeats of shredding process
chunklen average length of chunks to cut and permutate
-sscatter randomisation of cuts (Range: 0 to K, where K = duration of infile/chunklen Default = 1)
-n use this flag for a smoother output
- If scatter = 0, reorders without shredding
- NB1: chunklen * scatter must be less than the length of the program's sound buffer (Default is 1 Mb or the setting for CDP_MEMORY_BBSIZE)
- Scatter results in chunks of variable length.
- NB2: If the input sound is greater than the internal buffer length, each buffer of sound is shredded independently
dur minimum length of outfile required
-ldown lowest downward tranposition in semitones
-hup highest upward transposition in semitones
-sstart scrubs start before time start seconds
-eend scrubs end after time end
bit_resolution Range: 1 - 16 Default 16-bit.
srate_division divide the sample-rate. Range: 1 to 256 Default 1 (normal)
modulating-frq number of cycles per second
- The value entered will be rounded to a power of 2
- Works only on Mono files
Let's take this Mode by Mode:
- Mode 1 REVERSE: The soundfile is re-written back to front: starting at the end and ending at the beginning.
- Mode 2 SHRED: The soundfile is (randomly) segmented, and these segments reordered by means of a permutation process. As the number of repeats increases, it gets more and more jumbled, literally 'reducing it to shreds'. The -n adds splicing that results in a smoother output.
- Mode 3 SCRUB: This is an acceleration/deceleration process which models an editing procedure used in the 'classical' tape studio: the desired edit point was found by turning the tape spools by hand so that the tape moved (very slowly!) across the tape head. You could hear locate exactly where silence began or ended, where clicks came etc., although the sound was very low because of the slow speed.
The SCRUB function could also be used creatively, to create extreme speed modifications of the source sound on the tape, e.g. an abrupt acceleration-deceleration as the tape went from not-moving, to fast-moving, to stopped as the hands jerked the tape across the heads.
- Mode 4 LOSE RESOLUTION: Reducing the sample rate reduces the level of the Nyquist frequency (sample_rate/2), thereby lowering the frequency level which can be safely handling during processing. Lowering the bit-resolution reduces the precision of the numerical expression of the data, making the digital 'quantisation' of the sonic material coarser. This means that time-varying information is lost.
- Mode 5 RING MODULATION: multiplies two (bipolar) signals. In this case, one signal is a soundfile and the other is a modulating_frq. This creates two 'sidebands' which are the sum and the difference of the two signals, while the carrier signal disappears. The result is a timbrally 'hollow' sound. (See Curtis Roads, The Computer Music Tutorial, pp. 215-220).
- Mode 6 CROSS MODULATION: multiplies two different soundfiles, producing a very strange mixture of the two. Each frequency of the second sound ring-modulates the first sound.
- Mode 7 QUANTISE: This lowers the bit-resolution, like the second parameter of LOSE RESOLUTION.
Again, Mode by Mode:
- Mode 1 REVERSE: Reversing a sound can be used simply as a first stage in moving a familiar sound towards more abstract sonic material. Also, sounds can have a very characteristic 'signature' when played backwards (e.g., piano tones sound like crescendoing chords), so reversal can be used to achieve these effects.
- Mode 2 SHRED: Jumbling segments can be used for humorous effect, to texture a sound, or to create a series of randomly placed impulses (if the sound has sharp attacks in it).
- Mode 3 SCRUB: Scrub can be used for crazily improvised changes of speed within a sound.
- Mode 4 LOSE RESOLUTION: These processes will produce a deliberately coarse sound. 'Rough' sound material is sometimes useful as a serendipitous input to other processes.
- Mode 5 RING MODULATION: achieves a timbral alteration which makes it sound thin and hollow, depending on the modulating_frq.
- Mode 6 CROSS MODULATION: another wierd and wonderful effect which the CDP System allows you to explore!
1 STANDARD DELAY: with feedback & mix (0 = dry) of original and delayed signal
2 VARYING DELAY: with low frequency oscillator varying the delay time
3 STADIUM ECHO: create stadium P.A. type echoes
infile soundfile to process
outfile output soundfile
delay delay time, in milliseconds
mix amount of delayed signal in final mix: 0 gives 'dry' result (Range: 0 to 1)
feedback produces resonances related to delay time (with short times) (Range: -1.0 to 1.0)
tail time to allow decayed signal to decay to zero (Range: -1.0 to 1.0)
-pprescale prescales input level, to avoid overload
-i inverts the dry signal (for phasing effects)
lfomod depth of the delay-variation sweep (Range: 0 to 1)
lfofreq frequency of the delay-variation sweep (negative values give random oscillations)
lfophase start-phase of the delay-variation sweep (Range: 0 to 1)
lfodelay time in seconds before the delay-variation sweep begins
-sseed non-zero value gives reproducible output (with the same seed) where random oscillations are used (see lfofreq)
-ggain to apply to input signal (Default: 0.645654)
-rroll-off rate of loss of level across stadium (Default: 1)
-ssize multiplies average time between echoes (the Default time between echoes of 0.1 sec)
-ecount number of stadium echoes (Default and max: 23)
Mode 1 provides a fixed delay time. If a smooth echoey effect is wanted, be careful to keep mix on the low side. Remember that delay is given in milliseconds; after 60ms or so, one begins to hear a strong reverberation; after 100 ms the reverberation begins to 'bounce', and by 200 ms one begins to hear distinct echoes. Delay times greater than 1000 will cause repetitions of (all or much of) the sound. Don't forget to lengthen the tail to match the delay time.
The mix parameter, in adding in the delayed signal, increases the reverberant effect. With a short delay time, this can sound like the reverberation which occurs in an enclosed space. Feedback has a similar effect, becoming very pronounced towards a value of 1.0 rather like the 'feedback' which occurs when a microphone is placed facing a loudspeaker.
REVECHO Mode 1 Mapping out the Parameter Space Name & Mode infile & outfile delay mix feedback tail
modify revecho 1
modify revecho 1
modify revecho 1
modify revecho 1
modify revecho 1
modify revecho 1
The operation of a standard delay line in Mode 1 ranges from modest reverberation to echo effects. The reverberations, however, tend easily towards rough edges, so for smoother reverbs, look to the programs listed below.
In Mode 2, lower values for lfofreq produce a more noticeable wave motion, while higher values add to the 'bounce' of reverberation.
Mode 3 creates a stereo outfile, bouncing the signal between speakers with a very prominent delay factor. The idea is that you hear the signal bouncing around the 'stadium'. The defaults work quite nicely, but you can intensify the effect by multiplying the delay time with the size parameter. Note that this is a multiple, so values less than 0 (multiplied with the default time of 0.1 sec) will decrease the delay time. Beyond size = 2, you are likely to run into insufficient buffer space, so if you really want longer delay times, you will have to increase the buffer size with 'set CDP_MEMORY_BBSIZE=...' (bear in mind the RAM capacity of your machine). (The default buffer size is 1 Mbyte and the units are in 'k': e.g., a BBSIZE of 3000 units is 3 Mbyte).
In Release 6, various multi-channel options became available in this update. See the space and channel parameters.
In Release 7, also see NEWTEX in the GRAIN function group, which also has multi-channel output options.
infile soundfile to be processed, normally, but not necessarily, mono (see space)
infile2 ... additional input soundfiles to process
outfile output soundfile written after processing
velocity speed of advance in infile, relative to outfile (Range: >= 0)This is the inverse of a timestretch (i.e., 1/n: higher values make the output shorter, lower values make it longer). This permits an infinite timestretch.density amount of grain overlap (Range: > 0)Values < 1 leaves intergrain silence, i.e., gaps. Extremely small values will cease to perform predictably.grainsize size of the grains in milliseconds (Range: must be > 2 * the length of the splices; Default: 50ms)
pitchshift transposition of grains in + or - (fractions of) semitones
amp gain applied to the grains (Range: 0 to 1; Default: 1.0)Use only if you want amplitude to vary over a range &/or in time.bsplice length of start-splices on grains in ms (Default: 5)
esplice length of end-splices on grains in ms (Default: 5)
space set stereo position in outfile 0 = L, 1 = R (Range: 0 to N)Various multi-channel options are available:range of search for next grain, before infile 'now' (Default: 0 ms)
- If the space flag is used on a stereo input, the program mixes it to mono before acting.
- In order to spread the output of the brassage process over the multi-channel stage, both the space and hspace parameters should be set (use Mode 7), with different values. These can now take values from zero to N, the maximum channel count specified for the output file. The output of the BRASSAGE process will be spread out between the two output channels specified. (Values below 1 correspond to positions between channel 1 and the highest output channel).
- For audibly equivalent results to a similar process with normal stereo output, a higher density value must be set.
jitter randomisation of grain position; Range 0 to 1 (Default: 0.5)
outlength maximum outfile length (if end of data is not reached)Set to zero (the Default) for this parameter to be ignored. But if velocity is 0 anywhere, outlength must be given.channel extract and work on just one channel of a multi-channel input (Range: 1 to N)Various multi-channel options are available:
Also see the space parameter above. -d
- Set channel to 0 (the Default) for this parameter to be ignored, or for a multichannel infile to be mixed to MONO before processing.
- Set channel to a positive integer (N), e.g., 3, the Nth channel of the multi-channel file is selected for processing.
- Set channel to a negative integer value (-N), to produce a soundfile. For example with the setting -c-8, an 8-channel output soundfile will be produced.
-x do exponential splices (Default: linear)
-n no interpolation for pitch values (quick but dirty)
The presence of the upper limit parameters hvelocity, hdensity, hgrainsize, hpitchshift, hamp, hbsplice, hesplice, and hspace make it possible to specify a range of values. When set, the program chooses random values for that parameter for each grain between the lower and upper limits. For example, if pitchshift is -3 and hpitchshift is 4, the pitch of each grain will be somewhere within the interval of a perfect 5th. NB: Furthermore, any of these parameters may vary in time (provide the name of a time value breakpoint file).All parameters except outlength and channel may vary over time.
MODIFY SAUSAGE is in fact a slight variant on MODIFY BRASSAGE in which the latter is adapted to accept more than one infile. This means that the 'brassage' operations will cycle around the input soundfiles, producing a more complex sonic texture.
For Reference manual details, see MODIFY BRASSAGE above.
NB: This program is not the same as the pre-Release 4 CDP program, SAUSAGE. The older SAUSAGE has the useful option to have the grains from the different sources appear in regular rotation or in random order. It can also cycle round a list of pitch (i.e., grain transposition) values. Because of these unique features, the older SAUSAGE program has been recompiled for Release 4 and included under 'CDP EXTRAS'. It is invoked on the command line simply by 'sausage' and has its own HTML manual of the same name.
MODIFY SAUSAGE should be used when granular transformations need to work with several input soundfiles.
infile input soundfile, MONO only
outfile output soundfile
pan pan data breakpoint file: positions sound in a stereo field, from -1 (Left) to 1 (Right), or beyond
-pprescale gain reduction factor applied to input level in order to avoid clipping (Default: 0.7)
Sometimes it is useful to apply a process to many sounds, slightly modifying that process to suit the different durations of the source files. This is especially useful when using BULK PROCESSING on the Sound Loom, so that new data files do not need to be written for every single input source.
MODIFY SCALEDPAN enables you to take an existing pan file, for example with a panning pattern you use often, and apply it to a new sound. With bulk processing you could apply it to hundreds of sounds of slightly different duration.
In breakpoint files, it can be important that the final value come at or near the end of the input soundfile. This can be particularly significant in PAN operations. Besides enabling you to reuse a favourite pan file with new sounds easily, the scaling also ensures (automatically) that the panning is properly timed with the duration of the infile.
infile1 input stereo soundfile
outfilename note that a '0' will replace the last character of your given outfile name
starttime time when the shuddering will begin
frq the average frequency of the shuddering
scatter randomises the shudder events in time. (Range: 0 to 1)
stereo_spread positions the shudder events in space. (Range: 0 to 1)
mindepth maxdepth amplitude of the shudders each gets a random value between MIN and MAX
minwidth maxwidth the duration of the shudder events in milliseconds each gets a random value between MIN and MAX
Name of outfile must not end with a '1'
MODIFY SHUDDER imposes tremulations on a stereo file, randomised both in time and space, so that it appears to shudder. It is used in the shuddering noise band which then 'speaks' in Trevor Wishart's Imago.
Note the relationship between frq and width: the frequency per second of the shudders and the width of soundfile segment that is affected. Thus there could be one shudder per second involving 500ms (0.5 sec) of sound, or one shudder per second involving 100ms (0.1 sec), etc. Depth and width MIN and MAX can be the same value.
Although not unlike tremolo and vibrato, these 'shudders' are designed to be more gestural, more dramatic in character.
Suggestion: experiment with larger values for count (e.g., 10 or more), high values for (amplitude) depth (e.g., 0.9 - 1.0), and smaller values for width (e.g., 0.07 - 0.1). Then try slower shudders, such as count = 2 and width (both MIN and MAX) = 0.5.
1 PAN: Position or move mono sound in a stereo field
2 MIRROR: Invert stereo positions in a stereo field (i.e., the two channels swap sides, left going to right, and right to left)
3 MIRRORPAN: Invert stereo positions in a pan data file
4 NARROW: Narrow the stereo image of a sound
infile input soundfile to modify; Mode1 accepts a mono input and produces a stereo output. The other modes accept only a stereo input.
inpantextfile a text file containing time pan_position data
outfile resultant spatialised soundfile
outpantextfile a text file containing time pan_position data produced by the program, with the pan positions reversed
pan floating point value to specify location between the pair of stereo speakers, or breakpoint file of time pan pairs. Range: -1.0 to 1.0; 0.0 is centre.
-pprescale gain factor multiplier with which to adjust the input level (Default: 0.7)
narrowing make the stereo image less wide:
- 1 leaves the stereo image where it is
- .5 narrows the stereo image by half
- 0 converts the stereo image to mono
- negative values work similarly, but also invert the stereo image
Each Mode performs different but related operations on a sound:
- Mode 1 PAN: Adjusts the amplitude levels of the two channels so as to create spatial illusions. A time pan breakpoint file is used to create movement.
- Mode 2 MIRROR: the data in the two channels swaps sides
- Mode 3 MIRRORPAN: this Mode acts on a breakpoint file, reversing the pan data. Thus, if the pan was from Left to Right, (-1 to 1), it would become Right to Left (1 to -1).
- Mode 4 NARROW: a way to restrict the spatial location of a sound in the horizontal plane
These functions are part of the basic toolbox.
- Mode 1 PAN: control of sound in 3-D space has been a basic procedure when composing with 'sound' material, especially when these are drawn from nature/the environment. The assumption here is that you are thinking about the placement and movement of sound as an integral part of your compositional process. In this case, we are dealing with building these procedures into the sound itself. There is a further aspect external to the sound itself, achieved by 'mixing' (superimposing sounds and writing a new soundfile) and 'diffusion' (sound placement during playback achieved by manipulating the pan controls on a mixer).
- Mode 2 MIRROR: a utility for a quick Left/Right swap
- Mode 3 MIRRORPAN: a utility to achieve Left/Right swaps within breakpoint files
- Mode 4 NARROW: this program allows you to narrow the spread of a stereo field
Eventually, multi-channel facilities will open up the possibilities of spatial movement enormously. Until that point, everything has to be done in stereo and mixed to create more complicated, e.g., 4-channel effects. For example, if working with a 4-channels and 4 speakers positioned 1 -2 in front and 3 - 4 behind, oblique movement from 2 to 4 can be achieved by a Left to Right pan within the sound, and then using the mixing desk to move that sound from speaker 2 to speaker 4.
outpanfile output pan datafile produced by the program
cyclelen the duration of one complete sinusoidal pan cycle
width the width of the pan (from 0 to 1, full width)
dur the duration of the output file
quantisation the time step between successive space-position specifications
phase the angular position at which the pan starts. 0 is full left and 360 is full right.
cyclelen and phase may vary over time
MODIFY SPACEFORM produces a pan breakpoint file (.brk would be the correct extension) that can be used with PAN or MODIFY SCALEDPAN. Note that the duration of the file, i.e., the last time in the file, can be specified.
This function in effect swings the sound back and forth between the speakers in a stereo field. The speed at which it does so is controlled by cyclelen and the location between the speakers is handled by width. If width is:
- 1, the full stereo field is used
- 0, no panning takes place and the sound is heard equally (without movement) from both Left and Right speakers.
- 0.5, the swing is from halfway between the Left speaker and Center to halfway between the Right speaker and Center, i.e., from -0.5 to + 0.5
The quantisation parameter determines the number of intermediate pan positions there will be during each swing around the loop. 0.1 appeared to produce a good result. The phase sets where in the stereo field the sound will be heard at the beginning of each repeat cycle, from whence the movement will emerge.
Note the sophistication that can be brought to the spatial movement when breakpoint files for cyclelen and phase are employed.
A slowly cycling panning motion can be created, generating a sense of an object rotating in space, e.g. a marble rolling around the rim of a huger glass bowl in front of you. This spatial image can be enhanced by making the level increase on the left-to-right motion, and decrease on the right-to-left motion (or vice cersa), suggesting that the sound moves towards, then away from you, as well as from right to left. This perception can be enhanced by adding treble-cut filtering and/or subtle reverb to the 'more distant' part of the motion. More generally speaking, sound streams in a piece can be differentiated and characterised by their spatial motion.
Release 6: The multi-channel equivalent of MODIFY SPEED is STRANS MULTI.
1 Vary speed & pitch of a sound, constant or time-varying
2 Vary speed & pitch by a constant or time-varying (fractional) number of semitones
3 Get information on varying speed in a time-changing manner
4 Get information on time-variable speed change in semitones
5 Accelerate or decelerate a sound
6 Add vibrato to a sound
infile input soundfile to process
outfile resultant soundfile
speed transposition value (ratio) expressed as a floating point multiplier. See Chart of Ratios covering up and down 2 octaves. NB: use 1.0 for no transposition.
semitone-transpos transposition value in positive or negative number of semitones; e.g., 12 raises the sound by an octave, and -12 lowers it by an octave. NB: use 0 (semitones) for no transposition.Both speed and semitone-transpos may vary over time.accel multiplication of speed to be reached by goaltime i.e., a transposition ratio
goaltime time in outfile at which the accelerated speed is to be reached.
starttime time in infile / outfile at which the acceleration begins
- If the infile does not end there, it continues to accelerate.
- If the infile finishes before goaltime is reached, the outfile won't reach the specified acceleration value.
vibrate the rate of vibrato shaking in cyles-per-second (Range: 0.0 to 120.0)
vibdepth vibrato depth (pitch shift from centre) in [possibly fractional] semitones (Range: 0.0 to 96.0)Vibrate and vibdepth can vary over time.-o breakpoint times are read as times in the outfile. The Default is to read them as times in the infile
Speed modification processes change the duration and the pitch of the sound together. Thus a faster speed causes a higher pitch, a slower speed a lower pitch.
MODIFY SPEED offers a range of functions which affect the speed of the soundfile. Perhaps it will be most often used for transposition. Modes 1 and 2 both accept either single values or the names of time-varying breakpoint files with time transposition pairs.
The single values act as constants and transpose the whole soundfile up or down by the given amount. In the breakpoint files, transposition can be almost instantaneous (almost same time, different transposition value), or gradual, creating glissandi (different time, different transposition value). No transposition between times is a third possibility (different time, same transposition value). These three possibilities are illustrated in the table below:
Time-varying transposition (using ratios) time speed Comments 0.0 1.0 No transposition: start at the original pitch 1.0 1.498 Over 1 sec., gliss upwards through a Pefect 5th 3.0 1.498 Hold this new level for 2 sec. 3.0001 0.5 At 3 sec., (almost) instantly drop one 8ve below the original pitch,
i.e., to 19 semitones below the previous position
6.0 1.0 Spend the next 3 sec. glissing back to the original pitch
The program will not accept exactly simultaneous values, giving a message to the effect that the times 'are not in increasing order'. To get around this, add a tiny bit to the second time as in the example so that the 2nd is nominally later than the 1st, but virtually simultaneous.
IMPORTANT! Note that transposition changes are always relative to the original speed of the soundfile, not its current output speed. Thus, in the example above, the soundfile glisses up a perfect 5th (from speed 1.0 to speed 1.498). With the next ratio, 0.5, the soundfile will drop to half of its original speed, and consequently to the octave below its original pitch.
Sometimes, you may need certain transpositions to occur at specific times in the outfile. The -o flag makes this possible. Thus, changing the speed of a file will alter it's duration, and, especially when the speed change itself varies in time (using a breakpoint file), it will be difficult to determine at what time events in the output sound will appear. For example, if you specify a speed of 0.5 at time 2.0 with the default mode, the speed of the file will reach 0.5 after 2 seconds of the INPUT file have passed but because the speed of the file has been changed, this will not be at 2.0 seconds in the OUTPUT file. Therefore, if you want the speed to reach 0.5 at time 2.0 in the OUTPUT file, you should use the -o flag.
Transposition which also changes the speed, and therefore the pitch, of the soundfile greatly alters the character of the sound. It is often very interesting to hear what a sound will be like 1, 2 or even 3 octaves below its original pitch. Deep, rich tones can be achieved in this way. These tones can slowly rise or descend if created with a time-varying breakpoint file e.g., moving an octave up or down over the time of the whole sound (airplane takeoff sounds, etc.).
The graphic program Brkedit (PC systems) can create exponential or logarithmic breakpoint data, so glissandi in MODIFY SPEED can increase or decrease in speed as well as move in a steady (linear) manner. Alternatively, this can be done with Mode 5.
Vocal material is very sensitive to pitch changes, so upwards transposition of this type will produce fast, squeaky voices, like the mice trio in Babe, and downwards transposition will produce slow, drawn-out ponderous voices.
The vibrato created in Mode 6 is a frequency modulation. Given the very wide ranges allowed, this function is immensely powerful. A slow vibrate with a large vibdepth will swing the original sound wildly increase vibrate and it really 'flaps in the breeze' (like a flag in the wind). A fast vibrate with a reasonably tight vibdepth, e.g., a minor 3rd, will produce a fluttering effect.
Altogether, a great program to explore and use to push beyond accepted conventions.
Also see: STRANS MULTI, the multi-channel version of this function.
infile soundfile to stack
outfile resultant stacked soundfile
transpos when numeric, transpos is the transposition in semitones between successive copies, OR, as a file, it is a set of transposition values, one for each stack component
count the number of copies in the stack
lean the loudness of the highest component, relative to the lowest (may be > 1)
attack-offset adjusts the time at which the attack of each sound occurs; called attack-time in Sound Loom.
gain an overall amplifying gain factor on the output sound; some reduction (less than 1) may be needed. (It may also be > 1.) Range: 0.1 to 10
dur how much of the output to make (a proportion, from 0 to 1)
-s see the relative levels of the layers in the stack
New sounds can be developed from existing sources by stacking transposed copes of the original on top of each other. This is particularly successful where the source sound has a clearly defined attack at or near the start of the sound and then dies away gradually. If the copies are lined up in such a way that their attacks coincide exactly, the new sound will psycho-acoustically merge into a single sound event with harmonic content (rather than a chord made up of several sounds). The meaning of this is very precise: one is lining up the soundfiles at their perceived beginning, i.e., its attack, not the start of each soundfile. This is achieved by adjusting the start times of the transposed soundfiles.
Trevor Wishart has clarified what happens with this program as follows:It is important, therefore, that you accurately identify the time-location of the attack of the sound and work with this information, either completely fusing the sounds by making their attacks coincide, or introducing slight timing differences for various musical reasons, such as a sense of acceleration within a sound. Thus you can intentionally misinform the program about where the attack is in order to get different kinds of output. If you give an attack time that is only slightly wrong, this can produce interesting new attack characteristics in the original. If the attack time given is completely incorrect, the outcome could be anything.
"For example, suppose you start with a source with its attack at 0.5 sec, and your stack transpositions are 0, 12, 24 semitones (i.e. zero, 1 octave up, and 2 octaves up). If you tell the process that the attack is at zero (which it isn't!!!) then the transposed sounds in the output will be lined up to synchronise at zero time, but you will actually hear the event attacks at .125 (24 semitones up), .25 (12 semitones up), and 0.5 seconds: because the transposed-up sounds get shortened in time, the attacks of the higher transpositions occur earlier in the output sound. If, however, you declare, correctly, that the attack is at 0.5 seconds in the source, then all the sounds will line up with their attacks at 0.5 seconds in the output.
"In this example, where all the transpositions are upwards, the attack in the output is still at 0.5 secs as described.
"On the other hand, if we used a stack with 2 DOWNWARD transpositions by octaves(0, -12, -24), in order to REACH the attack of the most-slowed-down sound we have now to wait 0.5 x 4 = 2 seconds. Telling the program that the attack (in the original sound) is at 0.5 seconds will STILL line up all the attacks correctly, but they won't happen until 2 seconds into the output sound, as you can't reach the attack of the slowest sound before 2 seconds have elapsed.
"So this parameter is intended to tell the process where the attack in the original sound really is. However, you can get interesting effects by giving a slightly incorrect value, sometimes producing rapid pitch-slide attacks in the output."
In the illustration below, the attack comes at 1 second after the start of the sound. A downwards transposition by 1 octave doubles the length of the sound, causing the attack to come at 2. Downwards transposition by 2 octaves quadruples the length of the sound, causing the attack to come at 4.
- If we tell MODIFY STACK that the attack comes at the beginning of the sound (attack = 0), we get the upper diagram, with the time between attacks slowing down.
- If we tell the program (correctly) that the attack is at 1 second, then all the attacks are aligned at the 4 second mark, as in the lower diagram.
The dur (duration) parameter is the proportion of the sound you want to keep. 1 keeps all of it, 0, none of it, and 0.5, half of it. This is to enable you to truncate the output when very low transpositions would make the output excessively long.
Octave transpositions are particularly effective at 'strengthening' a sound source, but any harmonic structure may be used. Alternatively, the attacks may be very slightly staggered to generate downward (or upward) sweeping glissandi, or very tiny transpositions may be used introducing delay effects into the output.
MODIFY STACK is a very important program. It enables you to construct 'larger' versions of sounds by superimposing transposed versions on top of one another. Furthermore, it does this in such a way as to maximise the fusion of the layered sound by synchronising the attacks. This relates to a basic technique of orchestral scoring (where it is called 'doubling'). It is a form of automatic mixing involving one soundfile.
Because of the way the program functions, this is, however, quite a different process from transposing and mixing sounds to produce chords. (Doing this in the Spectral Domain, where transposition can be done without affecting duration, still has a place: see, for example, the duplication possibilites in SPECTWIN.) Here, if the attack points are made to coincide, the new sound-object is an integral whole, rather than a 'chord' made up of the transposed stack components: that is, the components fuse and cannot be separated in perception, as they can with a chord played, for example, on a piano keyboard.
Also note that, as a Time Domain process, the higher transpositions end sooner than the lower ones. Use Spectral Domain transpositon along with SUBMIX MIX to assemble transposed versions of the same sound when equal length is important.
Very low transpositions could produce an extremely long output. The dur parameter enables you to specify a much shorter outfile length. You can usually get an idea of the harmonic content of the output from the first few seconds.
Also see: SPECTWIN