Skip to content

Commit

Permalink
Multibands: Major upgrade to DSP for crossover filters, flat response
Browse files Browse the repository at this point in the history 
Signed-off-by: Damien Zammit <damien@zamaudio.com>
  • Loading branch information
@zamaudio
zamaudio committed Aug 3, 2016
1 parent 404058e commit a0a21f0
Show file tree
Hide file tree
Showing 4 changed files with 145 additions and 194 deletions.
134 changes: 58 additions & 76 deletions plugins/ZaMultiComp/ZaMultiCompPlugin.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -701,10 +701,15 @@ void ZaMultiCompPlugin::activate()
old_ll=old_l1=0.f;

for (i = 0; i < MAX_FILT; i++) {
c1[i] = c2[i] = c3[i] = c4[i] = 0.f;
z1[i] = z2[i] = z3[i] = 0.f;
z4[i] = z5[i] = z6[i] = 0.f;
gl[i] = gh[i] = 0.f;
simper[0][i].k = 0.f;
simper[0][i].g = 0.f;
simper[0][i].s[0] = 0.f;
simper[0][i].s[1] = 0.f;

simper[1][i].k = 0.f;
simper[1][i].g = 0.f;
simper[1][i].s[0] = 0.f;
simper[1][i].s[1] = 0.f;
}
max = 0.f;
for (i = 0; i < MAX_SAMPLES; i++) {
Expand All @@ -720,76 +725,59 @@ void ZaMultiCompPlugin::activate()
oldxover2 = 0.f;
}

// LR4 filters adapted from zita-lrx
/*
* A. Simper's filters
* https://cytomic.com/files/dsp/SvfInputMixing.pdf
*/

void ZaMultiCompPlugin::linear_svf_set_xover(struct linear_svf *self, float sample_rate, float cutoff, float resonance)
{
double w;

self->k = 2. - 2. * resonance;
w = M_PI * cutoff / sample_rate;
self->g = tan(w);

}

void ZaMultiCompPlugin::linear_svf_reset(struct linear_svf *self)
{
self->s[0] = self->s[1] = 0.f;
}

float ZaMultiCompPlugin::run_linear_svf_xover(struct linear_svf *self, float in, float mixlow, float mixhigh)
{
double v[3];
double g = self->g;
double k = self->k;
double s0 = self->s[0];
double s1 = self->s[1];
double g2 = g*g;
double vhigh = in * mixhigh;
double vband = in * 0.5;
double vlow = in * mixlow;

v[0] = in;
v[1] = -1. / (1. + g2 + g*k) * (-s0 + g*s1 - g*k*s0 + g2*vband + g*vhigh - g*vlow - g2*k*vlow);
v[2] = -1. / (1. + g2 + g*k) * (-g*s0 - s1 - g*vband + g2*vhigh + g*k*vhigh - g2*vlow);
self->s[0] = 2. * v[1] - s0;
self->s[1] = 2. * v[2] - s1;

return (float)(vhigh + v[2]);
}

void ZaMultiCompPlugin::calc_lr4(float f, int i)
{
float a, b, d1, d2, q, p, s;

s = 0.f;

// Damping for first and second sections, from pole locations.
d1 = 2.f * cosf (2.f * M_PI / 8.f);
d2 = d1;
// Basic coefficients.
a = tanf (M_PI * f);
b = a * a;
// First section.
q = a / d1;
p = q * b;
s = p + q + b;
c1[i] = (4 * p + 2 * b) / s;
c2[i] = (4 * p) / s;
gl[i] = p / s;
gh[i] = q / s;
// Second section.
q = a / d2;
p = q * b;
s = p + q + b;
c3[i] = (4 * p + 2 * b) / s;
c4[i] = (4 * p) / s;
gl[i] *= p / s;
gh[i] *= q / s;
float srate = getSampleRate();

linear_svf_set_xover(&simper[0][i], srate, f, 0.5);
linear_svf_set_xover(&simper[1][i], srate, f, 0.5);
}

// LR4 filters adapted from zita-lrx
void ZaMultiCompPlugin::run_lr4(int i, float in, float *outlo, float *outhi)
{
float x, y, zz1, zz2, zz3, zz4, zz5, zz6;
float glo = 1.f;
float ghi = 1.f;

// Get filter state.
zz1 = z1[i];
zz2 = z2[i];
zz3 = z3[i];
zz4 = z4[i];
zz5 = z5[i];
zz6 = z6[i];
// Get gain factors.
glo *= gl[i];
ghi *= gh[i];
// Run filter for 1 sample
x = in;
x -= c1[i] * zz1 + c2[i] * zz2 + EPS;
y = x + 4 * (zz1 + zz2);
zz2 += zz1;
zz1 += x;
x -= c3[i] * zz3 + c4[i] * zz4 - EPS;
*outhi = ghi * x;
zz4 += zz3;
zz3 += x;
y -= c3[i] * zz5 + c4[i] * zz6 - EPS;
x = y + 4 * (zz5 + zz6);
*outlo = glo * x;
zz6 += zz5;
zz5 += y;
// Save filter state.
z1[i] = zz1;
z2[i] = zz2;
z3[i] = zz3;
z4[i] = zz4;
z5[i] = zz5;
z6[i] = zz6;
*outlo = run_linear_svf_xover(&simper[0][i], in, 1., 0.);
*outhi = run_linear_svf_xover(&simper[1][i], in, 0., 1.);
}

void ZaMultiCompPlugin::run_comp(int k, float in, float *out)
Expand Down Expand Up @@ -877,19 +865,13 @@ void ZaMultiCompPlugin::run(const float** inputs, float** outputs, uint32_t fram
int listen3 = (listen[2] > 0.5f) ? 1 : 0;

if (oldxover1 != xover1) {
// clear filter state
z1[0] = z2[0] = z3[0] = 0.f;
z4[0] = z5[0] = z6[0] = 0.f;
// recalculate coeffs
calc_lr4(xover1/srate, 0);
calc_lr4(xover1, 0);
oldxover1 = xover1;
}
if (oldxover2 != xover2) {
// clear filter state
z1[1] = z2[1] = z3[1] = 0.f;
z4[1] = z5[1] = z6[1] = 0.f;
// recalculate coeffs
calc_lr4(xover2/srate, 1);
calc_lr4(xover2, 1);
oldxover2 = xover2;
}

Expand Down
28 changes: 13 additions & 15 deletions plugins/ZaMultiComp/ZaMultiCompPlugin.hpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -168,6 +168,18 @@ class ZaMultiCompPlugin : public Plugin
void activate() override;
void run(const float** inputs, float** outputs, uint32_t frames) override;

struct linear_svf {
double k;
double g;

double s[2];
};

struct linear_svf simper[2][MAX_FILT];
void linear_svf_set_xover(struct linear_svf *self, float sample_rate, float cutoff, float resonance);
void linear_svf_reset(struct linear_svf *self);
float run_linear_svf_xover(struct linear_svf *self, float in, float mixlow, float mixhigh);

void pushsample(float samples[], float sample, int k);
float averageabs(float samples[]);
// -------------------------------------------------------------------
Expand All @@ -182,21 +194,7 @@ class ZaMultiCompPlugin : public Plugin
float outlevelold[3][MAX_SAMPLES];
float oldxover1, oldxover2;
bool reset;
// Crossover filter coefficients
float c1[MAX_FILT];
float c2[MAX_FILT];
float c3[MAX_FILT];
float c4[MAX_FILT];
float gl[MAX_FILT];
float gh[MAX_FILT];

//Crossover filter states
float z1[MAX_FILT];
float z2[MAX_FILT];
float z3[MAX_FILT];
float z4[MAX_FILT];
float z5[MAX_FILT];
float z6[MAX_FILT];

};

// -----------------------------------------------------------------------
Expand Down
144 changes: 59 additions & 85 deletions plugins/ZaMultiCompX2/ZaMultiCompX2Plugin.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -722,12 +722,12 @@ void ZaMultiCompX2Plugin::activate()
for (j = 0; j < 2; j++)
old_ll[j]=old_l1[j]=0.f;

for (i = 0; i < MAX_FILT; i++) {
for (j = 0; j < 2; j++) {
c1[j][i] = c2[j][i] = c3[j][i] = c4[j][i] = 0.f;
z1[j][i] = z2[j][i] = z3[j][i] = 0.f;
z4[j][i] = z5[j][i] = z6[j][i] = 0.f;
gl[j][i] = gh[j][i] = 0.f;
for (j = 0; j < 2; j++) {
for (i = 0; i < MAX_FILT; i++) {
simper[j][i].k = 0.f;
simper[j][i].g = 0.f;
simper[j][i].s[0] = 0.f;
simper[j][i].s[1] = 0.f;
}
}
maxL = maxR = 0.f;
Expand All @@ -744,76 +744,58 @@ void ZaMultiCompX2Plugin::activate()
oldxover2 = 0.f;
}

// LR4 filters adapted from zita-lrx
void ZaMultiCompX2Plugin::calc_lr4(float f, int i, int ch)
/*
* A. Simper's filters
* https://cytomic.com/files/dsp/SvfInputMixing.pdf
*/

void ZaMultiCompX2Plugin::linear_svf_set_xover(struct linear_svf *self, float sample_rate, float cutoff, float resonance)
{
float a, b, d1, d2, q, p, s;
double w;

s = 0.f;
self->k = 2. - 2. * resonance;
w = M_PI * cutoff / sample_rate;
self->g = tan(w);
}

// Damping for first and second sections, from pole locations.
d1 = 2.f * cosf (2.f * M_PI / 8.f);
d2 = d1;
// Basic coefficients.
a = tanf (M_PI * f);
b = a * a;
// First section.
q = a / d1;
p = q * b;
s = p + q + b;
c1[ch][i] = (4 * p + 2 * b) / s;
c2[ch][i] = (4 * p) / s;
gl[ch][i] = p / s;
gh[ch][i] = q / s;
// Second section.
q = a / d2;
p = q * b;
s = p + q + b;
c3[ch][i] = (4 * p + 2 * b) / s;
c4[ch][i] = (4 * p) / s;
gl[ch][i] *= p / s;
gh[ch][i] *= q / s;
void ZaMultiCompX2Plugin::linear_svf_reset(struct linear_svf *self)
{
self->s[0] = self->s[1] = 0.f;
}

// LR4 filters adapted from zita-lrx
void ZaMultiCompX2Plugin::run_lr4(int i, int ch, float in, float *outlo, float *outhi)
float ZaMultiCompX2Plugin::run_linear_svf_xover(struct linear_svf *self, float in, float mixlow, float mixhigh)
{
float x, y, zz1, zz2, zz3, zz4, zz5, zz6;
float glo = 1.f;
float ghi = 1.f;
double v[3];
double g = self->g;
double k = self->k;
double s0 = self->s[0];
double s1 = self->s[1];
double g2 = g*g;
double vhigh = in * mixhigh;
double vband = in * 0.5;
double vlow = in * mixlow;

v[0] = in;
v[1] = -1. / (1. + g2 + g*k) * (-s0 + g*s1 - g*k*s0 + g2*vband + g*vhigh - g*vlow - g2*k*vlow);
v[2] = -1. / (1. + g2 + g*k) * (-g*s0 - s1 - g*vband + g2*vhigh + g*k*vhigh - g2*vlow);
self->s[0] = 2. * v[1] - s0;
self->s[1] = 2. * v[2] - s1;

// Get filter state.
zz1 = z1[ch][i];
zz2 = z2[ch][i];
zz3 = z3[ch][i];
zz4 = z4[ch][i];
zz5 = z5[ch][i];
zz6 = z6[ch][i];
// Get gain factors.
glo *= gl[ch][i];
ghi *= gh[ch][i];
// Run filter for 1 sample
x = in;
x -= c1[ch][i] * zz1 + c2[ch][i] * zz2 + EPS;
y = x + 4 * (zz1 + zz2);
zz2 += zz1;
zz1 += x;
x -= c3[ch][i] * zz3 + c4[ch][i] * zz4 - EPS;
*outhi = ghi * x;
zz4 += zz3;
zz3 += x;
y -= c3[ch][i] * zz5 + c4[ch][i] * zz6 - EPS;
x = y + 4 * (zz5 + zz6);
*outlo = glo * x;
zz6 += zz5;
zz5 += y;
// Save filter state.
z1[ch][i] = zz1;
z2[ch][i] = zz2;
z3[ch][i] = zz3;
z4[ch][i] = zz4;
z5[ch][i] = zz5;
z6[ch][i] = zz6;
return (float)(vhigh + v[2]);
}

void ZaMultiCompX2Plugin::calc_lr4(float f, int i)
{
float srate = getSampleRate();

linear_svf_set_xover(&simper[0][i], srate, f, 0.5);
linear_svf_set_xover(&simper[1][i], srate, f, 0.5);
}

void ZaMultiCompX2Plugin::run_lr4(int i, float in, float *outlo, float *outhi)
{
*outlo = run_linear_svf_xover(&simper[0][i], in, 1., 0.);
*outhi = run_linear_svf_xover(&simper[1][i], in, 0., 1.);
}

void ZaMultiCompX2Plugin::run_comp(int k, float inL, float inR, float *outL, float *outR)
Expand Down Expand Up @@ -941,22 +923,14 @@ void ZaMultiCompX2Plugin::run(const float** inputs, float** outputs, uint32_t fr
int listen3 = (listen[2] > 0.5f) ? 1 : 0;

if (oldxover1 != xover1) {
for (i = 0; i < MAX_FILT; i++) {
z1[0][i] = z2[0][i] = z3[0][i] = 0.f;
z4[0][i] = z5[0][i] = z6[0][i] = 0.f;
}
calc_lr4(xover1/srate, 0, 0);
calc_lr4(xover1/srate, 0, 1);
calc_lr4(xover1, 0);
calc_lr4(xover1, 1);
oldxover1 = xover1;
}

if (oldxover2 != xover2) {
for (i = 0; i < MAX_FILT; i++) {
z1[1][i] = z2[1][i] = z3[1][i] = 0.f;
z4[1][i] = z5[1][i] = z6[1][i] = 0.f;
}
calc_lr4(xover2/srate, 1, 0);
calc_lr4(xover2/srate, 1, 1);
calc_lr4(xover2, 2);
calc_lr4(xover2, 3);
oldxover2 = xover2;
}

Expand All @@ -973,10 +947,10 @@ void ZaMultiCompX2Plugin::run(const float** inputs, float** outputs, uint32_t fr
int listenmode = 0;

// Interleaved channel processing
run_lr4(0, 0, inl, &fil1[0], &fil2[0]);
run_lr4(0, 1, inr, &fil1[1], &fil2[1]);
run_lr4(1, 0, fil2[0], &fil3[0], &fil4[0]);
run_lr4(1, 1, fil2[1], &fil3[1], &fil4[1]);
run_lr4(0, inl, &fil1[0], &fil2[0]);
run_lr4(1, inr, &fil1[1], &fil2[1]);
run_lr4(2, fil2[0], &fil3[0], &fil4[0]);
run_lr4(3, fil2[1], &fil3[1], &fil4[1]);

pushsample(outlevelold[0], std::max(fil1[0], fil1[1]), 0);
outlevel[0] = averageabs(outlevelold[0]);
Expand Down
Loading

0 comments on commit a0a21f0

Please sign in to comment.