MantissaMask reduce precision


MantissaMask.ar(in, bits, mul, add)


Masks off bits in the mantissa of the floating point sample value. This introduces a quantization noise, but is less severe than linearly quantizing the signal.


in - input signal

bits - the number of mantissa bits to preserve. a number from 0 to 23.



// preserve only 3 bits of mantissa.

{ MantissaMask.ar(SinOsc.ar(SinOsc.kr(0.2,0,400,500), 0, 0.4), 3) }.play


// the original

{ SinOsc.ar(SinOsc.kr(0.2,0,400,500), 0, 0.4) }.play


// the difference.

(

{ 

var in;

in = SinOsc.ar(SinOsc.kr(0.2,0,400,500), 0, 0.4);

Out.ar(0, in - MantissaMask.ar(in, 3));

}.play

)



// preserve 7 bits of mantissa. 

// This makes the lower 16 bits of the floating point number become zero.

{ MantissaMask.ar(SinOsc.ar(SinOsc.kr(0.2,0,400,500), 0, 0.4), 7) }.play


// the original

{ SinOsc.ar(SinOsc.kr(0.2,0,400,500), 0, 0.4) }.play


// the difference.

(

{ 

var in;

in = SinOsc.ar(SinOsc.kr(0.2,0,400,500), 0, 0.4);

Out.ar(0, in - MantissaMask.ar(in, 7));

}.play

)



Berlin: clubs bars cafes nightlife going out