«digital lowercase from floating point errors» by eli.rosenkim

on 01 Jun'22 03:36 in lowercasefloating point

https://scsynth.org/t/why-the-plot-result-of-summing-2-same-frequency-sinosc-with-phase-reversed-is-not-just-zeros/5805/6

quick proof of concept

Results of a subtracting two identical out of phase signals from each-other, which should be silent but isn't because of the imprecision of floating point arithmetic... The quiet signals are then boosted to be audible, kind of like digital lowercase music

log fork json 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
(//floating point errors
{ var pink;
	pink = PinkNoise.ar().range(0.08, 0.12)!2;
	0.5*LeakDC.ar(Median.ar(2, Limiter.ar(99999999*(SawDPW.ar(pink, 0) + SawDPW.ar(pink, pi)), 0.9)))}.play;
{ var pink;
	pink = PinkNoise.ar().range(0.0999/3, 0.1001/3)!2;
	0.5*LeakDC.ar(Median.ar(5, Limiter.ar(99999999*(SawDPW.ar(pink, 0) + SawDPW.ar(pink, pi)), 0.9)))}.play;
{var pink;
	pink = PinkNoise.ar().range(0.00999, 0.01001)!2;
	(0.5*Median.ar(20,
	Limiter.ar(999999*LeakDC.ar(SawDPW.ar(pink, 0+(pi/2)) + SawDPW.ar(pink, pi+(pi/2))), 0.1)))}.play;
{0.5*Median.ar(30, Limiter.ar(99999999*LeakDC.ar((SinOsc.ar(0.005, 0) + SinOsc.ar(0.005, pi)), 0.05)))!2}.play;
{0.5*SVF.ar((Median.ar(10, Limiter.ar(999999*LeakDC.ar(LFPulse.ar(0.5, 0) + LFPulse.ar(0.5, pi)), 0.6))), 800)!2}.play;
)
raw 784 chars (focus & ctrl+a+c to copy)
reception
comments
eli.rosenkim ↪ user 02 Jun'22 07:39

I initially misunderstood what I was doing here. Only one of 4 voices truly should cancel (the sine voice), the other 3 are due to running oscillators well below the intended rate, removing dc, and then amplifying the artifacts