A frequency-domain convolution UGen using two linearly interpolated fixed kernels. When a trigger is received, a linear fade will be performed from the previously used kernel (internally stored by the UGen) towards the snapshot of the current kernel content upon receiving the trigger.

The delay caused by the convolution when the kernel is a dirac impulse is equal to frameSize - controlBlockSize , so for a frameSize of 2048 and a control-block size of 64, this is 1984 sample frames.

Note: If a trigger is received before the previous fade is complete, the interpolation is broken and the kernel instead jumps straight to one of the two buffers.

Examples

// three example kernels
def mkBuf(n: Int, amp: => Double): Buffer = {
  val v = Vector.tabulate[FillValue](n) { i =>
    (i.linLin(0, n, 0, 2048).toInt, amp)
  }
  val b = Buffer(s)
  b.alloc(2048, completion = b.zeroMsg(b.setMsg(v: _*)))
  b
}

val b = mkBuf(3, 1)
val c = mkBuf(50, math.random)
val d = mkBuf(20, 1)

val x = play {
  val z     = Impulse.ar(16)
  val buf   = "kernel".kr(b.id)
  val tr    = "trig"  .tr
  val dur   = 4.0          // fade-time in seconds
  val n     = 2048
  val block = SampleRate.ir / n
  val p     = dur * block  // ... in periods
  Convolution2L.ar(z, buf, tr, 2048, p) * 0.5
}

x.set("kernel" -> b.id, "trig" -> 1)
x.set("kernel" -> c.id, "trig" -> 1)
x.set("kernel" -> d.id, "trig" -> 1)

x.free(); b.free(); c.free(); d.free()