diff options
Diffstat (limited to 'plugins/WinVST/VariMu/VariMuProc.cpp')
| -rwxr-xr-x | plugins/WinVST/VariMu/VariMuProc.cpp | 54 |
1 files changed, 18 insertions, 36 deletions
diff --git a/plugins/WinVST/VariMu/VariMuProc.cpp b/plugins/WinVST/VariMu/VariMuProc.cpp index b966cd8..323b5e4 100755 --- a/plugins/WinVST/VariMu/VariMuProc.cpp +++ b/plugins/WinVST/VariMu/VariMuProc.cpp @@ -233,18 +233,14 @@ void VariMu::processReplacing(float **inputs, float **outputs, VstInt32 sampleFr //nice little output stage template: if we have another scale of floating point //number, we really don't want to meaninglessly multiply that by 1.0. - //noise shaping to 32-bit floating point - float fpTemp = inputSampleL; - fpNShapeL += (inputSampleL-fpTemp); - inputSampleL += fpNShapeL; - //if this confuses you look at the wordlength for fpTemp :) - fpTemp = inputSampleR; - fpNShapeR += (inputSampleR-fpTemp); - inputSampleR += fpNShapeR; - //for deeper space and warmth, we try a non-oscillating noise shaping - //that is kind of ruthless: it will forever retain the rounding errors - //except we'll dial it back a hair at the end of every buffer processed - //end noise shaping on 32 bit output + //stereo 32 bit dither, made small and tidy. + int expon; frexpf((float)inputSampleL, &expon); + long double dither = (rand()/(RAND_MAX*7.737125245533627e+25))*pow(2,expon+62); + inputSampleL += (dither-fpNShapeL); fpNShapeL = dither; + frexpf((float)inputSampleR, &expon); + dither = (rand()/(RAND_MAX*7.737125245533627e+25))*pow(2,expon+62); + inputSampleR += (dither-fpNShapeR); fpNShapeR = dither; + //end 32 bit dither *out1 = inputSampleL; *out2 = inputSampleR; @@ -254,12 +250,6 @@ void VariMu::processReplacing(float **inputs, float **outputs, VstInt32 sampleFr *out1++; *out2++; } - fpNShapeL *= 0.999999; - fpNShapeR *= 0.999999; - //we will just delicately dial back the FP noise shaping, not even every sample - //this is a good place to put subtle 'no runaway' calculations, though bear in mind - //that it will be called more often when you use shorter sample buffers in the DAW. - //So, very low latency operation will call these calculations more often. } void VariMu::processDoubleReplacing(double **inputs, double **outputs, VstInt32 sampleFrames) @@ -488,18 +478,16 @@ void VariMu::processDoubleReplacing(double **inputs, double **outputs, VstInt32 //nice little output stage template: if we have another scale of floating point //number, we really don't want to meaninglessly multiply that by 1.0. - //noise shaping to 64-bit floating point - double fpTemp = inputSampleL; - fpNShapeL += (inputSampleL-fpTemp); - inputSampleL += fpNShapeL; - //if this confuses you look at the wordlength for fpTemp :) - fpTemp = inputSampleR; - fpNShapeR += (inputSampleR-fpTemp); - inputSampleR += fpNShapeR; - //for deeper space and warmth, we try a non-oscillating noise shaping - //that is kind of ruthless: it will forever retain the rounding errors - //except we'll dial it back a hair at the end of every buffer processed - //end noise shaping on 64 bit output + //stereo 64 bit dither, made small and tidy. + int expon; frexp((double)inputSampleL, &expon); + long double dither = (rand()/(RAND_MAX*7.737125245533627e+25))*pow(2,expon+62); + dither /= 536870912.0; //needs this to scale to 64 bit zone + inputSampleL += (dither-fpNShapeL); fpNShapeL = dither; + frexp((double)inputSampleR, &expon); + dither = (rand()/(RAND_MAX*7.737125245533627e+25))*pow(2,expon+62); + dither /= 536870912.0; //needs this to scale to 64 bit zone + inputSampleR += (dither-fpNShapeR); fpNShapeR = dither; + //end 64 bit dither *out1 = inputSampleL; *out2 = inputSampleR; @@ -509,10 +497,4 @@ void VariMu::processDoubleReplacing(double **inputs, double **outputs, VstInt32 *out1++; *out2++; } - fpNShapeL *= 0.999999; - fpNShapeR *= 0.999999; - //we will just delicately dial back the FP noise shaping, not even every sample - //this is a good place to put subtle 'no runaway' calculations, though bear in mind - //that it will be called more often when you use shorter sample buffers in the DAW. - //So, very low latency operation will call these calculations more often. } |