mirror of
https://github.com/LMMS/lmms.git
synced 2026-01-24 06:18:10 -05:00
Replaced remaining occurences of old constants and superfluous "void"
on empty argument list of functions.
(cherry picked from commit 166701f9f3)
453 lines
8.8 KiB
C++
453 lines
8.8 KiB
C++
/*
|
|
* effect_lib.h - library with template-based inline-effects
|
|
*
|
|
* Copyright (c) 2006-2008 Tobias Doerffel <tobydox/at/users.sourceforge.net>
|
|
*
|
|
* This file is part of Linux MultiMedia Studio - http://lmms.sourceforge.net
|
|
*
|
|
* This program is free software; you can redistribute it and/or
|
|
* modify it under the terms of the GNU General Public
|
|
* License as published by the Free Software Foundation; either
|
|
* version 2 of the License, or (at your option) any later version.
|
|
*
|
|
* This program is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
* General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU General Public
|
|
* License along with this program (see COPYING); if not, write to the
|
|
* Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
|
|
* Boston, MA 02110-1301 USA.
|
|
*
|
|
*/
|
|
|
|
|
|
#ifndef _EFFECT_LIB_H
|
|
#define _EFFECT_LIB_H
|
|
|
|
#include <math.h>
|
|
|
|
#include "templates.h"
|
|
#include "lmms_constants.h"
|
|
#include "lmms_basics.h"
|
|
|
|
|
|
namespace effectLib
|
|
{
|
|
|
|
template<typename T>
|
|
class monoBase
|
|
{
|
|
public:
|
|
typedef class monoBypass bypassType;
|
|
|
|
static void process( sample_t * * _buf, const f_cnt_t _frames )
|
|
{
|
|
for( f_cnt_t f = 0; f < _frames; ++f )
|
|
{
|
|
_buf[f][0] = T::nextSample( _buf[f][0] );
|
|
}
|
|
}
|
|
} ;
|
|
|
|
template<typename T>
|
|
class stereoBase
|
|
{
|
|
public:
|
|
typedef class stereoBypass bypassType;
|
|
|
|
static void process( sample_t * * _buf, const f_cnt_t _frames )
|
|
{
|
|
for( f_cnt_t f = 0; f < _frames; ++f )
|
|
{
|
|
T::nextSample( _buf[f][0], _buf[f][1] );
|
|
}
|
|
}
|
|
} ;
|
|
|
|
|
|
template<class FXL, class FXR = FXL>
|
|
class monoToStereoAdaptor : public stereoBase<monoToStereoAdaptor<FXL, FXR> >
|
|
{
|
|
public:
|
|
monoToStereoAdaptor( const FXL & _mono_fx ) :
|
|
m_leftFX( _mono_fx ),
|
|
m_rightFX( _mono_fx )
|
|
{
|
|
}
|
|
|
|
monoToStereoAdaptor( const FXL & _left_fx,
|
|
const FXR & _right_fx ) :
|
|
m_leftFX( _left_fx ),
|
|
m_rightFX( _right_fx )
|
|
{
|
|
}
|
|
|
|
void nextSample( sample_t & _in_left, sample_t & _in_right )
|
|
{
|
|
_in_left = m_leftFX.nextSample( _in_left );
|
|
_in_right = m_rightFX.nextSample( _in_right );
|
|
}
|
|
|
|
FXL & leftFX()
|
|
{
|
|
return( m_leftFX );
|
|
}
|
|
|
|
FXR & rightFX()
|
|
{
|
|
return( m_rightFX );
|
|
}
|
|
|
|
private:
|
|
FXL m_leftFX;
|
|
FXR m_rightFX;
|
|
} ;
|
|
|
|
|
|
template<class FX>
|
|
class stereoToMonoAdaptor : public monoBase<stereoToMonoAdaptor<FX> >
|
|
{
|
|
public:
|
|
stereoToMonoAdaptor( const FX & _stereo_fx ) :
|
|
m_FX( _stereo_fx )
|
|
{
|
|
}
|
|
|
|
sample_t nextSample( sample_t _in )
|
|
{
|
|
sample_t s[2] = { _in, _in };
|
|
m_FX.nextSample( s[0], s[1] );
|
|
return( ( s[0] + s[1] ) / 2.0f );
|
|
}
|
|
|
|
private:
|
|
FX m_FX;
|
|
} ;
|
|
|
|
|
|
class monoBypass : public monoBase<monoBypass>
|
|
{
|
|
public:
|
|
monoBypass()
|
|
{
|
|
}
|
|
|
|
sample_t nextSample( sample_t _in )
|
|
{
|
|
return( _in );
|
|
}
|
|
} ;
|
|
|
|
|
|
class stereoBypass : public stereoBase<stereoBypass>
|
|
{
|
|
public:
|
|
void nextSample( sample_t &, sample_t & )
|
|
{
|
|
}
|
|
} ;
|
|
|
|
|
|
/* convenient class to build up static FX-chains, for example
|
|
|
|
using namespace effectLib;
|
|
chain<monoToStereoAdaptor<bassBoost<> >,
|
|
chain<stereoEnhancer<>,
|
|
monoToStereoAdaptor<foldbackDistortion<> > > >
|
|
fxchain( bassBoost<>( 60.0, 1.0, 4.0f ),
|
|
chain<stereoEnhancer<>,
|
|
monoToStereoAdaptor<foldbackDistortion<> > >(
|
|
stereoEnhancer<>( 1.0 ),
|
|
foldbackDistortion<>( 1.0f, 1.0f ) ) );
|
|
|
|
// now you can do simple calls such as which will process a bass-boost-,
|
|
// stereo-enhancer- and foldback-distortion-effect on your buffer
|
|
fx_chain.process( (sample_t * *) buf, frames );
|
|
*/
|
|
|
|
template<class FX0, class FX1 = typename FX0::bypassType>
|
|
class chain : public FX0::bypassType
|
|
{
|
|
public:
|
|
typedef typename FX0::sample_t sample_t;
|
|
chain( const FX0 & _fx0, const FX1 & _fx1 = FX1() ) :
|
|
m_FX0( _fx0 ),
|
|
m_FX1( _fx1 )
|
|
{
|
|
}
|
|
|
|
void process( sample_t * * _buf, const f_cnt_t _frames )
|
|
{
|
|
m_FX0.process( _buf, _frames );
|
|
m_FX1.process( _buf, _frames );
|
|
}
|
|
|
|
private:
|
|
FX0 m_FX0;
|
|
FX1 m_FX1;
|
|
} ;
|
|
|
|
|
|
|
|
template<typename sample_t>
|
|
inline sample_t saturate( sample_t _x )
|
|
{
|
|
return( qMin<sample_t>( qMax<sample_t>( -1.0f, _x ), 1.0f ) );
|
|
}
|
|
|
|
|
|
class fastBassBoost : public monoBase<fastBassBoost>
|
|
{
|
|
public:
|
|
fastBassBoost( const sample_t _frequency,
|
|
const sample_t _gain,
|
|
const sample_t _ratio,
|
|
const fastBassBoost & _orig =
|
|
fastBassBoost() ) :
|
|
m_frequency( qMax<sample_t>( _frequency, 10.0 ) ),
|
|
m_gain1( 1.0 / ( m_frequency + 1.0 ) ),
|
|
m_gain2( _gain ),
|
|
m_ratio( _ratio ),
|
|
m_cap( _orig.m_cap )
|
|
{
|
|
}
|
|
|
|
inline sample_t nextSample( sample_t _in )
|
|
{
|
|
// TODO: somehow remove these horrible aliases...
|
|
m_cap = ( _in + m_cap*m_frequency ) * m_gain1;
|
|
return( /*saturate<sample_t>(*/ ( _in + m_cap*m_ratio ) *
|
|
m_gain2/* )*/ );
|
|
}
|
|
|
|
void setFrequency( const sample_t _frequency )
|
|
{
|
|
m_frequency = _frequency;
|
|
m_gain1 = 1.0 / ( m_frequency + 1.0 );
|
|
}
|
|
|
|
void setGain( const sample_t _gain )
|
|
{
|
|
m_gain2 = _gain;
|
|
}
|
|
|
|
void setRatio( const sample_t _ratio )
|
|
{
|
|
m_ratio = _ratio;
|
|
}
|
|
|
|
private:
|
|
fastBassBoost() :
|
|
m_cap( 0.0 )
|
|
{
|
|
}
|
|
|
|
sample_t m_frequency;
|
|
sample_t m_gain1;
|
|
sample_t m_gain2;
|
|
sample_t m_ratio;
|
|
sample_t m_cap;
|
|
} ;
|
|
|
|
// for some reason this effect doesn't work... (in=out)
|
|
class bassBoost : public monoBase<bassBoost>
|
|
{
|
|
public:
|
|
bassBoost( const float _frequency,
|
|
const float _gain,
|
|
const float _ratio,
|
|
const float _sample_rate ) :
|
|
m_gain( _gain ),
|
|
m_beta( -1.0f ),
|
|
m_shape( 1.0f ),
|
|
m_sampleRate( _sample_rate )
|
|
{
|
|
xn1=xn2=yn1=yn2=0.0f;
|
|
setFrequency( _frequency );
|
|
setRatio( _ratio );
|
|
}
|
|
|
|
sample_t nextSample( sample_t _in )
|
|
{
|
|
const float out = ( m_b0*_in + m_b1*xn1 + m_b2*xn2 -
|
|
m_a1*yn1 - m_a2*yn2 ) / m_a0;
|
|
xn2 = xn1;
|
|
xn1 = _in;
|
|
yn2 = yn1;
|
|
yn1 = out;
|
|
return out*m_gain;
|
|
}
|
|
|
|
void setFrequency( const float _frequency )
|
|
{
|
|
const float omega = 2*F_PI*_frequency/m_sampleRate;
|
|
m_sin = sinf( omega );
|
|
m_cos = cosf( omega );
|
|
if( m_beta > 0 )
|
|
{
|
|
updateFilter();
|
|
}
|
|
}
|
|
|
|
void setGain( const float _gain )
|
|
{
|
|
m_gain = _gain;
|
|
}
|
|
|
|
void setRatio( const float _ratio )
|
|
{
|
|
m_a = expf( logf( 1.0f ) * _ratio / 40 );
|
|
updateFilter();
|
|
}
|
|
|
|
private:
|
|
void updateFilter()
|
|
{
|
|
m_beta = sqrtf( ( m_a*m_a + 1 ) / m_shape -
|
|
powf( m_a - 1, 2 ) );
|
|
m_b0 = m_a*((m_a+1)-(m_a-1)*m_cos + m_beta*m_sin);
|
|
m_b1 = 2*m_a*((m_a-1) - (m_a+1) * m_cos);
|
|
m_b2 = m_a*((m_a+1) - (m_a-1)*m_cos - m_beta*m_sin);
|
|
m_a0 = ((m_a+1) + (m_a-1)*m_cos + m_beta*m_sin);
|
|
m_a1 = -2*((m_a-1) + (m_a+1) * m_cos);
|
|
m_a2 = (m_a+1) + (m_a-1)*m_cos-m_beta*m_sin;
|
|
}
|
|
float m_gain;
|
|
float m_a;
|
|
float m_sin;
|
|
float m_cos;
|
|
float m_beta;
|
|
const float m_shape;
|
|
const float m_sampleRate;
|
|
float m_b0, m_b1, m_b2, m_a0, m_a1, m_a2;
|
|
float xn1, xn2, yn1, yn2;
|
|
} ;
|
|
|
|
|
|
class foldbackDistortion : public monoBase<foldbackDistortion>
|
|
{
|
|
public:
|
|
foldbackDistortion( const float _threshold,
|
|
const float _gain ) :
|
|
m_threshold( _threshold ),
|
|
m_gain( _gain )
|
|
{
|
|
}
|
|
|
|
sample_t nextSample( sample_t _in )
|
|
{
|
|
if( _in >= m_threshold || _in < -m_threshold )
|
|
{
|
|
return ( fabsf( fabsf(
|
|
fmodf( _in - m_threshold,
|
|
m_threshold*4 ) ) -
|
|
m_threshold*2 ) -
|
|
m_threshold ) *
|
|
m_gain;
|
|
}
|
|
return _in * m_gain;
|
|
}
|
|
|
|
void setThreshold( const float _threshold )
|
|
{
|
|
m_threshold = _threshold;
|
|
}
|
|
|
|
void setGain( const float _gain )
|
|
{
|
|
m_gain = _gain;
|
|
}
|
|
|
|
|
|
private:
|
|
float m_threshold;
|
|
float m_gain;
|
|
|
|
} ;
|
|
|
|
|
|
class distortion : public monoBase<distortion>
|
|
{
|
|
public:
|
|
distortion( float _threshold, float _gain ) :
|
|
m_threshold( _threshold ),
|
|
m_gain( _gain )
|
|
{
|
|
}
|
|
|
|
sample_t nextSample( sample_t _in )
|
|
{
|
|
return( m_gain * ( _in * ( fabsf( _in )+m_threshold ) /
|
|
( _in*_in +( m_threshold-1 )*
|
|
fabsf( _in ) + 1 ) ) );
|
|
}
|
|
|
|
void setThreshold( const float _threshold )
|
|
{
|
|
m_threshold = _threshold;
|
|
}
|
|
|
|
void setGain( const float _gain )
|
|
{
|
|
m_gain = _gain;
|
|
}
|
|
|
|
|
|
private:
|
|
float m_threshold;
|
|
float m_gain;
|
|
|
|
} ;
|
|
|
|
|
|
class stereoEnhancer : public stereoBase<stereoEnhancer>
|
|
{
|
|
public:
|
|
stereoEnhancer( float _wide_coeff ) :
|
|
m_wideCoeff( _wide_coeff )
|
|
{
|
|
}
|
|
|
|
// Lou's Hack
|
|
void setWideCoeff( float _wideCoeff )
|
|
{
|
|
m_wideCoeff = _wideCoeff;
|
|
}
|
|
|
|
float getWideCoeff()
|
|
{
|
|
return m_wideCoeff;
|
|
}
|
|
// -----------
|
|
|
|
void nextSample( sample_t & _in_left, sample_t & _in_right )
|
|
{
|
|
/*
|
|
const float delta = ( _in_left -
|
|
( _in_left+_in_right ) / 2.0f ) * m_wideCoeff;
|
|
_in_left += delta;
|
|
_in_right -= delta;
|
|
*/
|
|
|
|
|
|
// Lou's Hack
|
|
// I really can't tell you why this math works, but it sounds good
|
|
float toRad = 3.141592 / 180;
|
|
_in_left += _in_right * sinf( m_wideCoeff * .5 * toRad);
|
|
_in_right -= _in_left * sinf( m_wideCoeff * .5 * toRad);
|
|
|
|
}
|
|
|
|
private:
|
|
// Lou's Hack
|
|
float m_wideCoeff;
|
|
//-----------
|
|
} ;
|
|
|
|
} ;
|
|
|
|
|
|
#endif
|