HITHER GATE MUSIC

Starting Point (varied inputs)

In general, varied and rhythmic sounds work well. If the sound is too steady state, the result may not change enough during the course of playback to remain interesting. Also, if a sound is already very rough to start with, the distortion processing may go over the top, leading to an end result which remains pretty much the same throughout.

However, for this sound-builder sequence, we want to stress the 'ever-changing' aspect. For this we will input pairs of soundfiles into programs which put together selected features from both files. Three varied results from 3 different programs will then be spliced to create the changing sonic pattern. After some harmonic filtering, we then move on to the granulation procedures.

Because processes that use two different soundfiles are to be employed, we select a pair of contrasting inputs: a busy or rhythmic one and a more steady-state one. The Envelope Replace function will transfer the busy or rhythmic amplitude pattern onto the steady-state sound. Using sounds which contrast in this way provides the maximum opportunity to find interesting sonic interactions and varied outputs.

Step 1: Copy

Make copies of the input soundfiles, giving them a single letter name.

We have only to provide a name for the outfiles, in this case just 'c1' and 'c2'. The other parameters use the defaults – any other settings refer to multi-channel soundfiles.

Dialogue box:

Default parameters

New name is 'c1.wav'

We also make a copy of the second input, naming it 'c2.wav'.

Command:

rem function mode infile outfile
copysfx aklcdp.wav c1.wav
copysfx tomcdt.wav c2.wav

Step 2: Analyse

Analyse soundfile to create data needed by spectral operations.

The 'Time Domain' stream of time amplitude data is converted to a detailed frequency amplitude data file. In the graphic interface, the .ana extension is added automatically, but with the Command, we specify it.

We do this for each of the input soundfiles, naming the second one 'd'.

Dialogue box:

Create analysis data

Command:

rem function mode infile outfile
pvoc anal 1 c1 c1.ana
pvoc anal 1 c2 c2.ana

Step 3: Envelope Replace

Extract the amplitude profile from one sound and impose it on another.

The first sound opened is the one onto which the envelope is imposed. The second sound is the one from which the envelope is taken. In this case we take rhythm from c2.wav (2^nd input) and impose it onto c1.wav (1^st input).

The enveloping_window parameter refers to the time-resolution of the envelope extraction process (the time in milliseconds over which to average the envelope amplitude data). Smaller values will achieve a more accurate envelope. 10ms = 0.010 sec (1/100^th of a second.

Dialogue box:

Envelope replace

Inputs and output

Command:

rem function mode impose_infile extract_infile outfile enveloping_window
envel replace 1 c2 c1 cer 10

PLAY cer.mp3 (re-enveloped soundfile)
New name: 'c envelope replace'

Step 4: Interleave

Join two (or more) different analysis files by interleaving groups of analysis windows.

There is a way of shuffling cards in which the cards alternate (more or less) when the two decks are strummed adjacent to each other. The Interleave process is similar to this. Here we are dealing with analysis windows rather than cards, and we also set how many windows to group together. Then they alternate, a group from the 1^st soundfile followed by a group from the 2^nd soundfile. Small window groups result in a fine-grained interleave, while larger groups are more coarse-grained.

The rapid interleaving of two different sounds in this manner can result in acoustic serendipities – it's a very interesting process.

Dialogue box:

Window group size

Inputs and output

Command:

rem function inanalfile1 inanalfile2 outanalfile windows
combine interleave c1.ana c2.ana ci.ana 10

(Optionally) audition the analysis file with F10 in SoundShaper. Now go straight to Spectral/Convert to Resynthesise ci.ana to create ci.wav.

PLAY ci.mp3 (interleaved soundfile)
New name: 'c interleave'

Step 5: Random Mix

Put together the loudest components of 2 or more soundfiles.

This is a very unusual mixing process based on the amplitude of the frequencies in the analysis file. For each analysis window, the software compares the amplitude in each file, keeps the loudest window and discards the less loud one(s).

Dialogue box:

Keep loudest frequencies

Inputs and output

Command:

rem function inanalfile1 in analfile2 outanalfile
combine max c1.ana c2.ana cm.ana

(Optionally) audition the analysis file using F10 in Soundshaper. Now go straight to Spectral/Convert to Resynthesise cm.ana to create cm.wav.

PLAY cm.mp3 (randomly mixed soundfile)
New name: 'c envelope max'

Step 6: Splice

Splice together the 3 previous results.

After some experimentation with the inputs used while writing this, I decided to place the Random Mix first, the Interleaving second and the Envelope Replace third – that order seemed to sound best. Another order might be better when different inputs are used. Open the first file and then add the next two by selecting them in the splice Select box.

The default splice slope of 15ms is used, so it doesn't need to be specified. The output filename is made up of the deliberately concise names of the 3 inputs. This reminds us of what went into the splice and in what order.

Dialogue box:

3 file splice

Command:

rem function infile1 infile2 infile3 outfile
sfedit join cm ci cer cmcicer

PLAY cmcicer.mp3 (spliced soundfiles)
New name: 'c max interleave envelope replace'

Step 7: Filter

Pass the sound through a harmonic filter bank.

The aim here is to add a harmonic dimension to the sound by passing it through a filter bank focused on the harmonics. This means that the frequency foci of the filter frequency bands are determined by multiplying a root frequency by successive integers: 2, 3, 4 etc. This creates the natural overtone series.

We can define the bank of filters in a text file we write ourselves. It contains pitch data, either as frequencies in Herz or as MIDI pitch values. In this case we choose to use the harmonic overtone series option of the BANKFREQS program to help us on our way. This creates a set of frequencies related in this way. We then have to edit the file to add amplitude data in dB (decibels), either to cut or boost the amplitudes. In this way we create cmcicer1h.txt. Note that part of it is shown in the file window of the dialogue box below. [The screenshot mistakenly shows it as cmcicer3h.txt – R.F.]

The Q parameter controls how tightly the filter band is restricted to its frequency focus. If very tight, there can be resonance which needs to be reduced by lowering the amplitude values. If broader, the filtering process can reduce the volume too much, and the amplitude values may need to be increased. A broader Q (lower values with USERBANK) allows more of the original sound through. The value of 100 selected here is high enough to produce the harmonic effect desired, but low enough to retain some of the original sound and rhythm.

Dialogue box:

Harmonic filter

Command:

rem function mode infile outfile filterbank_file Q Gain
filter userbank 1 cmcicer cmcicerflth cmcicer1h.txt 100 5

PLAY cmcicerflth.mp3 (filtered soundfile)
New name: 'c max interleaved envelope_replace filter_harmonic'

Step 8: Granulate

Break up the sound into a series of adjacent tiny segments.

Here we call a simple form of granulation, which only uses a density parameter, for which we select the maximum value, 2. This parameter controls the degree to which the grains overlap (a default grainsize is employed). If 1, the grains will be adjacent, i.e., virtually no change. When greater than 1, they overlap. A value of 2 will overlap them half-way: i.e., 1 ÷ 2, or 0.5. A value of less than one will create gaps between grains.

The result in this case adds a bit of light flapping to the soundfile.

Dialogue box:

Grain overlap density

Command:

rem function mode infile outfile overlap_density
modify brassage 5 cmcicerflth cmcicerflthgrn 2

PLAY cmcicerflthgrn.mp3 (filtered soundfile)
New name: 'c max interleaved envelope_replace filter_harmonic granulate'

Step 9: Analyse

Analyse cmcicerflthgrn.wav to prepare it for a spectral process.

This now a familar single-step data conversion operation.

Dialogue box:

Create analysis data

Command:

rem function mode infile outfile
pvoc anal 1 cmcicerflthgrn cmcicerflthgrn.ana

Step 10: Linear Frequency Shift

Shift the whole spectrum of frequencies with a time contour.

With the following time contour (breakpoint file shift2.brk), we push the frequencies into the stratosphere, and then have them drift downwards.

0	1500
1	1000
1.1	1200
3	 700
3.1	 900
4.4	  50

The shift that is made here is a linear one, meaning that the same value (at each time period) is added to each partial value. This tends to squeeze the frequencies together, frequencies which normally are in logarithmic relationship: i.e., doublings are replaced by constants.

Dialogue box:

Linear shift

Command:

rem function mode infile outfile shift_file
strange shift 1 cmcicerflthgrn.ana cmcicerflthgrnshift.ana shift2.brk

(Optionally) audition the analysis file using F10 in Soundshaper. Now go straight to Spectral/Convert to Resynthesise cmcicerflthgrnshift.ana to create cmcicerflthgrnshift.wav.

PLAY cmcicerflthgrnshift.mp3. (shifted soundfile)
New name: 'c max interleave envelope_replace filter_harmonic granulate shift'

Step 11: Granulate with more features

Use GrainMill to process cmcicerflth.wav.

We now have several parameters which we can set:

• timestretch: 4 – factor by which to multiply the length of the output soundfile
• density: 10.2105 – also grain overlap, but in this case it can be very much larger than 2!
• grainsize: 22.03 – now this can be specified, and we are making it very small, only 0.02203 sec.
• pitch: 0 – this is a transposition value in semitones; in the first version it is zero (no change), and in the second version we give it a range of 1 octave.
• scatter: 75% – this randomly displaces the time of each grain, here set at 75% of the maximum allowed displacement
• loudness: 0dB to -10.5dB – a volume setting which here randomly selects a volume level for each grain between the upper and lower limits we have set
• spatial 0 (L) to 1 (R) – pan (horizontal placement between the speakers), here set to full Left-Right range; thus each grain will be randomly placed anywhere between the speakers, including full Left or Right.
• wander: 0.5 (½sec) – this creates a 'searchfield' from somewhere within which each grain is taken, i.e., it displaces from where in the infile a given grain is taken.

In running GrainMill we have first to ensure that the 'Working Directory' is set (under Options), then 'Open' a soundfile to granulate. Having done this, the dialogue box is displayed or can be called up by clicking on 'Show Dialog' under 'View'. We can then fill in the values we want. Alternatively, we can close the dialogue box, go to 'Load Settings' under 'File' and select grnflowgm1.grn. However, this will only work if the Working Directory is c:\sdbats. I had to put it somewhere, and selected the C: drive as the one most likely to be available. If your files are elsewere, fill in the dialogue box and then 'Save Settings' under 'File' to the same name, overwriting the existing preset. The preset will then be available for future use.

After making the sound(s), Save it to the name cmcicerflthgm1.wav (and cmcicerflthgm2.wav) so that they will play from this HTML file.

Dialogue box:

Granulate without a pitch range

An alternative version creates a pitch range of an octave in which to place (transpose) the grains. This is done by ticking 'Range' next to the pitch parameter. Boxes for specifying upper and lower transposition values appear, which we set as 7 and -5 respectively. The 7 goes up seven semitones (a Perfect 5^th) and the -5 goes down five semitones (a Perfect 4^th). 7 + 5 = 12, so the total range is 12 semitones, or one octave.

Here is the dialogue box showing this change:

Granulate with a pitch range

Much more can be done with GrainMill to shape the granulated texture: by making the parameters vary over time. The means of doing this come to life when the 'Time Contours' boxes to the right of each parameter are ticked. This enables you to create, load and edit breakpoint files, i.e., automated time contours.

Command:

The command format for this program is so complex that it is very unwieldy, so I only use the GrainMill graphic version.

PLAY cmcicerflthgm1.mp3 (GrainMill soundfile)
PLAY (if made) cmcicerflthgm2.mp3 (GrainMill soundfile)
New name: 'c max interleaved envelope_replace filter_harmonic grainmill'

Step 12: Run GrainMill again

We can now process cmcicerfltagrnshift.wav with GrainMill as in Step 11, loading this new soundfile and re-using the same Preset: grnflowgm1.grn. My initial results weren't too exciting – everything too smoothed out – so it will be work exploring changes to the density, grainsize and pitch parameters.

The output soundfile can be named cmcicerflthgrnshiftgm3.wav and, if more changes are made, cmcicerflthgrnshift4.wav, etc.

PLAY cmcicerflthgrnshiftgm3.mp3 (GrainMill soundfile)
PLAY (if made) cmcicerflthgrnshiftgm4.mp3 (GrainMill soundfile)
New name: 'c max interleaved envelope_replace filter_harmonic granulate shift grainmill'

Complete Sequence: Batch File grnflow.bat

Use grnfldel.bat to remove the outputs

rem grnflow.bat - 'GRAINFLOW' batch file for third set of functions
rem Description: multiple file operations combined, filtered & granulated
rem Last updated: 28 March 2003

rem Use grnfldel.bat to clear outputs before using again

rem Source Pairs (with comments on results after processing):
rem   aklcdp.wav  (1.38 sec) & tomcdt.wav  (1.3 sec) - rhythms good
rem   frogcdt.wav (1.8 sec)  & bashdt.wav  (1.5 sec) - frogs good
rem   touccdt.wav (3.0 sec)  & hoggdt.wav  (3.7 sec) - toucan audible
rem   touccdt.wav (3.0 sec)  & whdtm.wav   (2.4 sec) - steadier high pitch


echo on

rem Step 1. Copy to simple name
copysfx aklcdp.wav c1.wav
copysfx tomcdt.wav c2.wav

rem Step 2. Prepare analysis data files
pvoc anal 1 c1 c1.ana
pvoc anal 1 c2 c2.ana

rem Step 3. Replace d's envelope with c's
envel replace 1 c2 c1 cer 10

rem Step 4. Interleave c & d
combine interleave c1.ana c2.ana ci.ana 10
pvoc synth ci.ana ci

rem Step 5. Random mix based on max amplitude, on window-to-window basis
combine max c1.ana c2.ana cm.ana
pvoc synth cm.ana cm

rem Step 6. Splice the three results together: cm+ci+cer in that order
sfedit join cm ci cer cmcicer

rem Step 7. Filter, using pre-edited & existing filterbank data file
rem (Otherwise create with FILTER BANKFREQS, 100-2000Hz)
rem cmcicer1h.txt (HARMONIC) cmcicer2a.txt (ALT HARMONICS) or 
rem    cmcicer3s.txt (SUBHARMONIC).
filter userbank 1 cmcicer cmcicerflth cmcicer1h.txt 100 5

rem Step 8. Granulate (only using density parameter)
modify brassage 5 cmcicerflth cmcicerflthgrn 2

rem Step 9. Prepare ...grn sound for shift function
pvoc anal 1 cmcicerflthgrn.wav cmcicerflthgrn.ana

rem Step 10. Time-varying spectral shift
rem (NB: check that timings in shift2 .brk match infile length)
strange shift 1 cmcicerflthgrn.ana cmcicerflthgrnshift.ana shift2.brk
rem Resynthesise the shifted analysis file
pvoc synth cmcicerflthgrnshift.ana cmcicerflthgrnshift.wav

rem THE FOLLOWING ARE DONE WITH THE GRAPHIC GRAINMILL PROGRAM
rem Step 11. Granulate with GrainMill, opening cmcicerflth.wav (or another 
rem   soundfile) and loading grnflowgm1.grn (if able to do so). 
rem   See grnflowgm1.txt for ascii version.
rem   Outfile(s) should be named cmcicerflthgm1.wav (cmcicerflthgm2.wav 
rem   with pitch range settings, e.g., +7 & -5 to give octave spread)

rem Step 12. Granulate with GrainMill opening cmcicerflthgrnshift.wav and 
rem   re-loading (if possible) grnflowgm1.grn.
rem   Outfile(s) should be named cmcicerflthgrnshiftgm3.wav and 
rem   cmcicerflthgrnshiftgm4.wav

echo off

---

SCREENSHOT UPDATES

(by Robert Fraser, March 2016)

The Soundshaper screenshots used in this tutorial are out of date in two respects:

• Output files are no longer named on Parameter Pages: instead, the output is saved to a temporary file and appears on the Main Page patchgrid in a new cell named after the process.
  If you want to save the output permanently, highlight the cell (if necessary) and click File/Save or CTRL+S to obtain a standard SAVE dialog:
  
  
  
  
  There is normally no need to do this until the end of the sequence.


• FFT conversion (PVOC Analysis / Resynthesis) now uses current settings and the FFT page is only used to set these, not to perform a conversion, as shown. As the tutorial examples use default values, there is nothing to set!
  Furthermore, Soundshaper does these conversions in the background, so you do not even have to select ANALYSE or SYNTH – just select the next process.

To run the GRAINFLOW chain of CDP processes in Soundshaper:

[After each step, you can play the result in the transport panel.]

• STEP 1 (COPYSFX): Open the tutorial folder. Drag and drop the source aklcdp.wav onto Soundshaper's Main Page.
  Program runs COPYSFX and outputs to cell A_0.
  Click on cell B_0. Drag and drop the source tomcdt.wav onto the page.
  Program runs COPYSFX and outputs to cell B_0.


• STEP 2 (PVOC ANALYSE): You can skip this step!
  The settings used are the defaults, so there is nothing to set.
  Soundshaper will run PVOC automatically before performing a spectral function.


• STEP 3 (ENVEL REPLACE): Click on Cell B_0, if not highlighted. Select Soundfiles/Envelope/Replace.
  'Add Input' dialog appears: click on cell A_0 to select this as the 2nd input. Click dialog's 'OK' button.
  Parameter page: set parameters as in tutorial (Window=10).
  Click OK to run; output appears in cell B_1 ("EnvReplace").


• STEP 4 (INTERLEAVE): Click on Cell A_0. Select Spectral/Combine/Interleave.
  'Add Input' dialog appears: click on cell B_0 to select this as the 2nd input. Click dialog's 'OK' button.
  Parameter page: set splice slope to 15ms. Click OK to run; output appears in cell A_1 ("SpecInterl").


• Copy cell A_0 to C_0: Click on cell A_0; click the Copy Cell START button; click on cell C_0.
  The copy is needed so that the first sound can be treated in STEP 5.


• STEP 5 (RANDOM MIX [combine max]): Click on Cell C_0, if not highlighted. Select Spectral/Combine/Maximum.
  Add Input' dialog appears: click on cell B_0 to select this as the 2nd input. Click dialog's 'OK' button.
  There are no parameters. Output appears in cell C_1 ("SpecMax"):





• STEP 6 (SPLICE): With cell C_1 (Max) highlighted, select Soundfiles/Edit /Join.
  'Add Input' dialog appears: click on cell A_1 (Interl) to select this as the 2nd input.
  (This is a spectral file, but it is converted to .wav on selection.)
  Now click on cell B_1 (EnvReplace) to select this as the 3rd input. Click dialog's 'OK' button.
  Parameter page: nothing to set as the default 15ms splice is used. Click OK to run.
  Output appears in cell C_2 ("Join").


• STEP 7 (FILTER): Select Soundfiles/Filter/Userbank.
  Parameter page: Select Mode 1 (bands as freq.). Drag and drop the datafile cmcicer1h.txt onto the Parameter page (or click the Open button and select it).
  Parameters: Q=100; Gain=5. Click OK to run; output appears in cell C_3 ("Userbank").


• STEP 8 (GRANULATE): Select Soundfiles/Grain/Granulate (Selects Mode 5).
  Parameter page: set Density=2. Click OK to run; output appears in cell C_4 ("Granulate").


• STEP 9 (PVOC ANALYSE): You can skip this step!
  The settings used are the defaults, so there is nothing to set.
  Soundshaper will run PVOC automatically before performing a spectral function..


• STEP 10 (FREQ. SHIFT): Select Spectral/Freq.Pitch/Shift Spectrum.
  Parameter page (Mode 1): Click T-V checkbox.Uncheck Auto-scale .brks at the top.
  Drag and drop the datafile shift2.brk. Click OK to run. Output appears in cell C_5 ("Shift").

---

The next step has been done in GrainMill, which, like Soundshaper, uses the CDP program MODIFY BRASSAGE, but it interprets some of the parameter names and value-ranges differently, making a direct "translation" difficult. Here is an approximation to the Grainmill file cmcicerflthgrnshiftgm4.mp3, mentioned in Step 12:

• STEP 11 (GRAIN): Select Soundfiles/Grain/Grain.
  Parameter page (Mode: Whole Grain). This mode sets High-Low ranges, so most parameters have to be entered twice to achieve a single value. Set parameters: Velocity=0.25 (=Timestretch 4); Density=10.2105; Grainsize=22.03; Pitchshift -5 to +7 (Pitchshift Hi); Amp 0.297; Space 0 to 1 (Space Hi); (Start/End splices: use defaults); Range=500ms; Jitter=0.75.
  Click OK to run; output appears in cell C_6 ("Grain"):






---

• SAVE FILE (OPTIONAL): Save the result to a permanent file. Hit CTRL+S or File / Save(As) to obtain a standard SAVE dialog.
  Enter a file name, e.g. grainflow.wav (or cmcicerflthgrnshiftgm4.wav if you prefer).


• SAVE PATCH (OPTIONAL): Uncheck "Include Sources", unless you want to save the copied source (~A_0.wav) with the patch.
  Click the patch SAVE button (central panel). Select a save name (e.g. churbles.ssp); click OK. Optionally enter a description; click OK.
  


• RUN PATCH (OPTIONAL): To run the patch with a different source (assuming the patchgrid has not been cleared), click on cell A_0, select a source (e.g. by drag+drop) and click the central RUN button, with the "Whole Patch" checkbox checked.


• RELOAD PATCH (OPTIONAL): If the patch is not on the grid, click the central LOAD button. Select the .ssp file (e.g. churbles.ssp). If you have not saved the source (~A_0.wav), the patch will still run with it, provided it still exists in the same location. Otherwise, you will be prompted to select a source. To run the patch with a different source, see RUN PATCH above.