reverbsc2

Applies stereo reverberation using eight delay lines and simpler delay interpolation than reverbsc.

reverbsc2 uses the same feedback network, delay modulation and damping filters as reverbsc. It changes how the delay lines read samples between stored positions.

Syntax

ModernClassic

aoutL, aoutR = reverbsc2(ainL, ainR, kfblvl, kfco [, israte [, ipitchm [, iskip]]])

aoutL, aoutR reverbsc2 ainL, ainR, kfblvl, kfco [, israte [, ipitchm [, iskip]]]

Initialization

israte is the sample rate assumed by the delay and filter calculations, in Hz. An omitted or nonpositive value uses the current instrument's sample rate. The chosen rate must be between 5000 and 1000000 Hz. Normally, leave it at the default. Changing it alters delay lengths and damping without changing the rate at which Csound processes audio.

ipitchm sets the depth of random delay-time variation. It defaults to 1 and accepts values from 0 to 20. This is a multiplier, not a value in Hz or seconds. Lower values reduce pitch movement in the tail. Zero disables modulation and also disables delay interpolation in reverbsc2.

iskip defaults to 0, which clears the delay lines and filter state. A nonzero value preserves an already initialized instance during reinitialization when its buffer allocation size stays the same. Keep the other initialization arguments unchanged when retaining a tail.

Performance

ainL and ainR are the left and right audio inputs. Either input can feed both output channels through the reverb network.

aoutL and aoutR contain only the reverberated signal. Add the dry input separately if needed, and keep the instrument running after its input stops so the tail can decay.

kfblvl sets feedback gain. Use values from 0 up to, but below, 1 for a decaying tail. Higher values within this range give longer decay. Zero produces no wet output. This control is not a wet/dry mix or a decay time in seconds. Values above 1 can make the feedback grow without bound. A value of 1 does not guarantee an unchanged tail because filtering and interpolation can still remove energy.

kfco sets the cutoff of the first-order low-pass filters in the feedback paths, in Hz. Use a positive value no higher than half the effective israte. Lower values make high frequencies decay faster.

Differences from reverbsc

Both opcodes accept the same arguments and defaults. Their delay reads differ as follows.

Opcode	ipitchm	Delay read
reverbsc	Any valid value	Cubic interpolation using four samples
reverbsc2	Greater than 0	Linear interpolation using two samples
reverbsc2	0	One stored sample, with no interpolation

Linear interpolation does less work per delay read, but it also reduces high frequencies more than cubic interpolation. Since the signal passes through the delays repeatedly, this can make the tail darker or shorten its high-frequency decay even when both opcodes have the same kfco setting.

With ipitchm = 0, the delay lengths stay fixed. reverbsc2 ignores the fractional part of each read position, while reverbsc still interpolates it. The two outputs can therefore differ even with modulation off.

Examples

The example plays the same short sawtooth burst through four separate reverb instances. First it compares reverbsc and reverbsc2 with the default modulation depth. It then repeats the comparison with modulation off. Each note leaves time for the tail to decay.

Listen for changes in brightness and pitch movement. The dry burst stays at the same level in all four notes.

It uses reverbsc2.csd.

<CsoundSynthesizer>
<CsOptions>
-d -odac
</CsOptions>
<CsInstruments>
sr = 48000
ksmps = 32
nchnls = 2
0dbfs = 1

instr CompareReverbs
  iVersion = p4
  iModulation = p5

  // A short, bright burst leaves most of the note for the reverb tail.
  aEnvelope = linseg(0, 0.002, 1, 0.04, 0, p3 - 0.042, 0)
  aSource = vco2(0.2, 330) * aEnvelope

  if iVersion == 1 then
    aLeft, aRight = reverbsc(aSource, aSource, 0.88, 10000, sr, iModulation)
  else
    aLeft, aRight = reverbsc2(aSource, aSource, 0.88, 10000, sr, iModulation)
  endif

  // Fade the end of the tail before the instrument stops.
  aFade = linseg(1, p3 - 0.1, 1, 0.1, 0)
  out(0.2 * aSource + aLeft * aFade, 0.2 * aSource + aRight * aFade)
endin
</CsInstruments>
<CsScore>
// reverbsc, then reverbsc2, with delay modulation.
i "CompareReverbs" 0 5 1 1
i "CompareReverbs" 5.5 5 2 1
// Repeat with fixed delays.
i "CompareReverbs" 11 5 1 0
i "CompareReverbs" 16.5 5 2 0
e
</CsScore>
</CsoundSynthesizer>

See also

reverbsc, freeverb, Reverberation

Credits

Victor Lazzarini added reverbsc2 in 2025. It builds on Istvan Varga's 2005 C implementation of reverbsc and Sean Costello's 1999 Csound orchestra version.