Skip to content

pvsgendy

Adds random changes to the frequencies in a spectral stream.

Use pvsgendy to turn a steady tone into a less pitched, noisier sound. It applies a Gendy-style transformation to an existing spectrum. Unlike gendy, it needs an input signal rather than generating a waveform on its own.

Syntax

fout = pvsgendy(fin, kmrate, kfrate)

fout pvsgendy fin, kmrate, kfrate

Performance

fin is an amplitude-frequency spectral stream, such as the output of pvsanal. fout has the same analysis settings as fin. Use different f-variables for input and output.

The opcode processes every bin, including DC and Nyquist. Set both controls to 0 to copy the full input spectrum. In sliding analysis, it clears the output samples before the note starts and after it ends within each control block.

Reinitialize pvsgendy if the input's analysis settings change.

kmrate controls the size of random amplitude changes in sliding analysis. Each bin receives an independent offset of roughly -kmrate/2 to kmrate/2 in linear amplitude units. In ordinary frame-based analysis, this parameter has no effect and the opcode copies the input amplitudes. Use 0 to leave amplitudes unchanged.

kfrate controls the size of random frequency changes in Hz. Larger values give more detuning, and lower bins receive larger offsets than higher bins. Use 0 to leave the processed bin frequencies unchanged. Neither control sets the speed of the random changes. Use nonnegative values for both controls.

The two analysis modes use different frequency ranges. In the table below, b is the bin index starting at 0 for DC. The bounds are approximate.

Analysis mode Frequency offset in each bin When offsets change
Ordinary frames ±kfrate / (2 * (2*b + 1)) Hz On each new input frame
Sliding analysis ±kfrate / (2 * (b + 1)) Hz On each audio sample

For example, in ordinary frame-based analysis, kfrate = 1000 gives bin 1 an offset of roughly -167 to 167 Hz. Each update adds a fresh offset to the input value. The changes do not build up from one update to the next.

Each instance has its own random sequence. Set seed before the opcode initializes to make the result repeatable. Use the same seed and initialization order for repeat runs.

The opcode does not limit the resulting amplitudes or frequencies. Large settings can produce negative amplitudes in sliding mode or frequencies outside the range from 0 to sr/2 in either mode.

Examples

The example makes a stereo comparison from a sawtooth tone. The left channel resynthesizes the input spectrum. The right channel passes it through pvsgendy before resynthesis. Both channels use the same analysis, so their timing matches.

For the first two seconds, the frequency control stays at 0. It then rises to 3000 over two seconds and holds there. Listen for the right channel to lose its steady pitch. This can add a rough texture to a sustained note without changing its source oscillator.

It uses pvsgendy.csd.

Compare a steady spectrum with random frequency changes
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
<CsoundSynthesizer>
<CsOptions>
-d -odac
</CsOptions>
<CsInstruments>
sr = 48000
ksmps = 32
nchnls = 2
0dbfs = 1

instr SpectralTexture
  aSource = vco2(0.15, 220)
  fSource = pvsanal(aSource, 1024, 256, 1024, 1)

  // Keep the first two seconds steady, then add frequency changes.
  kSpread = linseg(0, 2, 0, 2, 3000, 2, 3000)

  // Ordinary analysis ignores the amplitude control, so leave it at 0.
  fTexture = pvsgendy(fSource, 0, kSpread)
  aOriginal = pvsynth(fSource)
  aTexture = pvsynth(fTexture)

  // Fade both channels after resynthesis to avoid clicks at note edges.
  aFade = linseg(0, 0.1, 1, p3 - 0.2, 1, 0.1, 0)
  outs(aOriginal * aFade, aTexture * aFade)
endin
</CsInstruments>
<CsScore>
i "SpectralTexture" 0 6
e
</CsScore>
</CsoundSynthesizer>

See also

pvsanal, pvsynth, gendy, Real-time spectral processing

Credits

Author John ffitch, 2009.