mirror of
https://github.com/ZDoom/qzdoom.git
synced 2024-11-10 14:51:51 +00:00
401 lines
11 KiB
C++
401 lines
11 KiB
C++
/*
|
|
reverb.c
|
|
|
|
Midi Wavetable Processing library
|
|
|
|
Copyright (C) Chris Ison 2001-2011
|
|
Copyright (C) Bret Curtis 2013-2014
|
|
|
|
This file is part of WildMIDI.
|
|
|
|
WildMIDI is free software: you can redistribute and/or modify the player
|
|
under the terms of the GNU General Public License and you can redistribute
|
|
and/or modify the library under the terms of the GNU Lesser General Public
|
|
License as published by the Free Software Foundation, either version 3 of
|
|
the licenses, or(at your option) any later version.
|
|
|
|
WildMIDI 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 and
|
|
the GNU Lesser General Public License for more details.
|
|
|
|
You should have received a copy of the GNU General Public License and the
|
|
GNU Lesser General Public License along with WildMIDI. If not, see
|
|
<http://www.gnu.org/licenses/>.
|
|
*/
|
|
|
|
//#include "config.h"
|
|
|
|
#include <math.h>
|
|
#include <stdlib.h>
|
|
|
|
#include "common.h"
|
|
#include "reverb.h"
|
|
|
|
namespace WildMidi
|
|
{
|
|
/*
|
|
reverb function
|
|
*/
|
|
void _WM_reset_reverb(struct _rvb *rvb) {
|
|
int i, j, k;
|
|
for (i = 0; i < rvb->l_buf_size; i++) {
|
|
rvb->l_buf[i] = 0;
|
|
}
|
|
for (i = 0; i < rvb->r_buf_size; i++) {
|
|
rvb->r_buf[i] = 0;
|
|
}
|
|
for (k = 0; k < 8; k++) {
|
|
for (i = 0; i < 6; i++) {
|
|
for (j = 0; j < 2; j++) {
|
|
rvb->l_buf_flt_in[k][i][j] = 0;
|
|
rvb->l_buf_flt_out[k][i][j] = 0;
|
|
rvb->r_buf_flt_in[k][i][j] = 0;
|
|
rvb->r_buf_flt_out[k][i][j] = 0;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
/*
|
|
_WM_init_reverb
|
|
|
|
=========================
|
|
Engine Description
|
|
|
|
8 reflective points around the room
|
|
2 speaker positions
|
|
1 listener position
|
|
|
|
Sounds come from the speakers to all points and to the listener.
|
|
Sound comes from the reflective points to the listener.
|
|
These sounds are combined, put through a filter that mimics surface absorbtion.
|
|
The combined sounds are also sent to the reflective points on the opposite side.
|
|
|
|
*/
|
|
struct _rvb *
|
|
_WM_init_reverb(int rate, float room_x, float room_y, float listen_x,
|
|
float listen_y) {
|
|
|
|
/* filters set at 125Hz, 250Hz, 500Hz, 1000Hz, 2000Hz, 4000Hz */
|
|
double Freq[] = {125.0, 250.0, 500.0, 1000.0, 2000.0, 4000.0};
|
|
|
|
/* numbers calculated from
|
|
* 101.325 kPa, 20 deg C, 50% relative humidity */
|
|
double dbAirAbs[] = {-0.00044, -0.00131, -0.002728, -0.004665, -0.009887, -0.029665};
|
|
|
|
/* modify these to adjust the absorption qualities of the surface.
|
|
* Remember that lower frequencies are less effected by surfaces
|
|
* Note: I am currently playing with the values and finding the ideal surfaces
|
|
* for nice default reverb.
|
|
*/
|
|
double dbAttn[8][6] = {
|
|
{-1.839, -6.205, -8.891, -12.059, -15.935, -20.942},
|
|
{-0.131, -6.205, -12.059, -20.933, -20.933, -15.944},
|
|
{-0.131, -6.205, -12.059, -20.933, -20.933, -15.944},
|
|
{-1.839, -6.205, -8.891, -12.059, -15.935, -20.942},
|
|
{-1.839, -6.205, -8.891, -12.059, -15.935, -20.942},
|
|
{-0.131, -6.205, -12.059, -20.933, -20.933, -15.944},
|
|
{-0.131, -6.205, -12.059, -20.933, -20.933, -15.944},
|
|
{-1.839, -6.205, -8.891, -12.059, -15.935, -20.942}
|
|
};
|
|
/*
|
|
double dbAttn[6] = {
|
|
// concrete covered in carpet
|
|
// -0.175, -0.537, -1.412, -4.437, -7.959, -7.959
|
|
// pleated drapes
|
|
-0.630, -3.223, -5.849, -12.041, -10.458, -7.959
|
|
};
|
|
*/
|
|
|
|
/* distance */
|
|
double SPL_DST[8] = {0.0};
|
|
double SPR_DST[8] = {0.0};
|
|
double RFN_DST[8] = {0.0};
|
|
|
|
double MAXL_DST = 0.0;
|
|
double MAXR_DST = 0.0;
|
|
|
|
double SPL_LSN_XOFS = 0.0;
|
|
double SPL_LSN_YOFS = 0.0;
|
|
double SPL_LSN_DST = 0.0;
|
|
|
|
double SPR_LSN_XOFS = 0.0;
|
|
double SPR_LSN_YOFS = 0.0;
|
|
double SPR_LSN_DST = 0.0;
|
|
|
|
|
|
struct _rvb *rtn_rvb = (struct _rvb*)malloc(sizeof(struct _rvb));
|
|
int j = 0;
|
|
int i = 0;
|
|
|
|
struct _coord {
|
|
double x;
|
|
double y;
|
|
};
|
|
|
|
#if 0
|
|
struct _coord SPL = {2.5, 5.0}; /* Left Speaker Position */
|
|
struct _coord SPR = {7.5, 5.0}; /* Right Speaker Position */
|
|
/* position of the reflective points */
|
|
struct _coord RFN[] = {
|
|
{ 5.0, 0.0},
|
|
{ 0.0, 6.66666},
|
|
{ 0.0, 13.3333},
|
|
{ 5.0, 20.0},
|
|
{ 10.0, 20.0},
|
|
{ 15.0, 13.3333},
|
|
{ 15.0, 6.66666},
|
|
{ 10.0, 0.0}
|
|
};
|
|
#else
|
|
struct _coord SPL; /* Left Speaker Position */
|
|
struct _coord SPR; /* Right Speaker Position */
|
|
/* position of the reflective points */
|
|
struct _coord RFN[8];
|
|
|
|
SPL.x = room_x / 4.0;
|
|
SPR.x = room_x / 4.0 * 3.0;
|
|
SPL.y = room_y / 10.0;
|
|
SPR.y = room_y / 10.0;
|
|
|
|
RFN[0].x = room_x / 3.0;
|
|
RFN[0].y = 0.0;
|
|
RFN[1].x = 0.0;
|
|
RFN[1].y = room_y / 3.0;
|
|
RFN[2].x = 0.0;
|
|
RFN[2].y = room_y / 3.0 * 2.0;
|
|
RFN[3].x = room_x / 3.0;
|
|
RFN[3].y = room_y;
|
|
RFN[4].x = room_x / 3.0 * 2.0;
|
|
RFN[4].y = room_y;
|
|
RFN[5].x = room_x;
|
|
RFN[5].y = room_y / 3.0 * 2.0;
|
|
RFN[6].x = room_x;
|
|
RFN[6].y = room_y / 3.0;
|
|
RFN[7].x = room_x / 3.0 * 2.0;
|
|
RFN[7].y = 0.0;
|
|
#endif
|
|
|
|
SPL_LSN_XOFS = SPL.x - listen_x;
|
|
SPL_LSN_YOFS = SPL.y - listen_y;
|
|
SPL_LSN_DST = sqrt((SPL_LSN_XOFS * SPL_LSN_XOFS) + (SPL_LSN_YOFS * SPL_LSN_YOFS));
|
|
|
|
if (SPL_LSN_DST > MAXL_DST)
|
|
MAXL_DST = SPL_LSN_DST;
|
|
|
|
SPR_LSN_XOFS = SPR.x - listen_x;
|
|
SPR_LSN_YOFS = SPR.y - listen_y;
|
|
SPR_LSN_DST = sqrt((SPR_LSN_XOFS * SPR_LSN_XOFS) + (SPR_LSN_YOFS * SPR_LSN_YOFS));
|
|
|
|
if (SPR_LSN_DST > MAXR_DST)
|
|
MAXR_DST = SPR_LSN_DST;
|
|
|
|
if (rtn_rvb == NULL) {
|
|
return NULL;
|
|
}
|
|
|
|
for (j = 0; j < 8; j++) {
|
|
double SPL_RFL_XOFS = 0;
|
|
double SPL_RFL_YOFS = 0;
|
|
double SPR_RFL_XOFS = 0;
|
|
double SPR_RFL_YOFS = 0;
|
|
double RFN_XOFS = listen_x - RFN[j].x;
|
|
double RFN_YOFS = listen_y - RFN[j].y;
|
|
RFN_DST[j] = sqrt((RFN_XOFS * RFN_XOFS) + (RFN_YOFS * RFN_YOFS));
|
|
|
|
SPL_RFL_XOFS = SPL.x - RFN[i].x;
|
|
SPL_RFL_YOFS = SPL.y - RFN[i].y;
|
|
SPR_RFL_XOFS = SPR.x - RFN[i].x;
|
|
SPR_RFL_YOFS = SPR.y - RFN[i].y;
|
|
SPL_DST[i] = sqrt(
|
|
(SPL_RFL_XOFS * SPL_RFL_XOFS) + (SPL_RFL_YOFS * SPL_RFL_YOFS));
|
|
SPR_DST[i] = sqrt(
|
|
(SPR_RFL_XOFS * SPR_RFL_XOFS) + (SPR_RFL_YOFS * SPR_RFL_YOFS));
|
|
/*
|
|
add the 2 distances together and remove the speaker to listener distance
|
|
so we dont have to delay the initial output
|
|
*/
|
|
SPL_DST[i] += RFN_DST[i];
|
|
|
|
/* so i dont have to delay speaker output */
|
|
SPL_DST[i] -= SPL_LSN_DST;
|
|
|
|
if (i < 4) {
|
|
if (SPL_DST[i] > MAXL_DST)
|
|
MAXL_DST = SPL_DST[i];
|
|
} else {
|
|
if (SPL_DST[i] > MAXR_DST)
|
|
MAXR_DST = SPL_DST[i];
|
|
}
|
|
|
|
SPR_DST[i] += RFN_DST[i];
|
|
|
|
/* so i dont have to delay speaker output */
|
|
SPR_DST[i] -= SPR_LSN_DST;
|
|
|
|
if (i < 4) {
|
|
if (SPR_DST[i] > MAXL_DST)
|
|
MAXL_DST = SPR_DST[i];
|
|
} else {
|
|
if (SPR_DST[i] > MAXR_DST)
|
|
MAXR_DST = SPR_DST[i];
|
|
}
|
|
|
|
RFN_DST[j] *= 2.0;
|
|
|
|
if (j < 4) {
|
|
if (RFN_DST[j] > MAXL_DST)
|
|
MAXL_DST = RFN_DST[j];
|
|
} else {
|
|
if (RFN_DST[j] > MAXR_DST)
|
|
MAXR_DST = RFN_DST[j];
|
|
}
|
|
|
|
for (i = 0; i < 6; i++) {
|
|
double srate = (double) rate;
|
|
double bandwidth = 2.0;
|
|
double omega = 2.0 * M_PI * Freq[i] / srate;
|
|
double sn = sin(omega);
|
|
double cs = cos(omega);
|
|
double alpha = sn * sinh(M_LN2 / 2 * bandwidth * omega / sn);
|
|
double A = pow(10.0, ((/*dbAttn[i]*/dbAttn[j][i] +
|
|
(dbAirAbs[i] * RFN_DST[j])) / 40.0) );
|
|
/*
|
|
Peaking band EQ filter
|
|
*/
|
|
double b0 = 1 + (alpha * A);
|
|
double b1 = -2 * cs;
|
|
double b2 = 1 - (alpha * A);
|
|
double a0 = 1 + (alpha / A);
|
|
double a1 = -2 * cs;
|
|
double a2 = 1 - (alpha / A);
|
|
|
|
rtn_rvb->coeff[j][i][0] = (signed int) ((b0 / a0) * 1024.0);
|
|
rtn_rvb->coeff[j][i][1] = (signed int) ((b1 / a0) * 1024.0);
|
|
rtn_rvb->coeff[j][i][2] = (signed int) ((b2 / a0) * 1024.0);
|
|
rtn_rvb->coeff[j][i][3] = (signed int) ((a1 / a0) * 1024.0);
|
|
rtn_rvb->coeff[j][i][4] = (signed int) ((a2 / a0) * 1024.0);
|
|
}
|
|
}
|
|
|
|
/* init the reverb buffers */
|
|
rtn_rvb->l_buf_size = (int) ((float) rate * (MAXL_DST / 340.29));
|
|
rtn_rvb->l_buf = (int*)malloc(
|
|
sizeof(signed int) * (rtn_rvb->l_buf_size + 1));
|
|
rtn_rvb->l_out = 0;
|
|
|
|
rtn_rvb->r_buf_size = (int) ((float) rate * (MAXR_DST / 340.29));
|
|
rtn_rvb->r_buf = (int*)malloc(
|
|
sizeof(signed int) * (rtn_rvb->r_buf_size + 1));
|
|
rtn_rvb->r_out = 0;
|
|
|
|
for (i = 0; i < 4; i++) {
|
|
rtn_rvb->l_sp_in[i] = (int) ((float) rate * (SPL_DST[i] / 340.29));
|
|
rtn_rvb->l_sp_in[i + 4] = (int) ((float) rate
|
|
* (SPL_DST[i + 4] / 340.29));
|
|
rtn_rvb->r_sp_in[i] = (int) ((float) rate * (SPR_DST[i] / 340.29));
|
|
rtn_rvb->r_sp_in[i + 4] = (int) ((float) rate
|
|
* (SPR_DST[i + 4] / 340.29));
|
|
rtn_rvb->l_in[i] = (int) ((float) rate * (RFN_DST[i] / 340.29));
|
|
rtn_rvb->r_in[i] = (int) ((float) rate * (RFN_DST[i + 4] / 340.29));
|
|
}
|
|
|
|
rtn_rvb->gain = 4;
|
|
|
|
_WM_reset_reverb(rtn_rvb);
|
|
return rtn_rvb;
|
|
}
|
|
|
|
/* _WM_free_reverb - free up memory used for reverb */
|
|
void _WM_free_reverb(struct _rvb *rvb) {
|
|
if (!rvb) return;
|
|
free(rvb->l_buf);
|
|
free(rvb->r_buf);
|
|
free(rvb);
|
|
}
|
|
|
|
void _WM_do_reverb(struct _rvb *rvb, signed int *buffer, int size) {
|
|
int i, j, k;
|
|
signed int l_buf_flt = 0;
|
|
signed int r_buf_flt = 0;
|
|
signed int l_rfl = 0;
|
|
signed int r_rfl = 0;
|
|
int vol_div = 64;
|
|
|
|
for (i = 0; i < size; i += 2) {
|
|
signed int tmp_l_val = 0;
|
|
signed int tmp_r_val = 0;
|
|
/*
|
|
add the initial reflections
|
|
from each speaker, 4 to go the left, 4 go to the right buffers
|
|
*/
|
|
tmp_l_val = buffer[i] / vol_div;
|
|
tmp_r_val = buffer[i + 1] / vol_div;
|
|
for (j = 0; j < 4; j++) {
|
|
rvb->l_buf[rvb->l_sp_in[j]] += tmp_l_val;
|
|
rvb->l_sp_in[j] = (rvb->l_sp_in[j] + 1) % rvb->l_buf_size;
|
|
rvb->l_buf[rvb->r_sp_in[j]] += tmp_r_val;
|
|
rvb->r_sp_in[j] = (rvb->r_sp_in[j] + 1) % rvb->l_buf_size;
|
|
|
|
rvb->r_buf[rvb->l_sp_in[j + 4]] += tmp_l_val;
|
|
rvb->l_sp_in[j + 4] = (rvb->l_sp_in[j + 4] + 1) % rvb->r_buf_size;
|
|
rvb->r_buf[rvb->r_sp_in[j + 4]] += tmp_r_val;
|
|
rvb->r_sp_in[j + 4] = (rvb->r_sp_in[j + 4] + 1) % rvb->r_buf_size;
|
|
}
|
|
|
|
/*
|
|
filter the reverb output and add to buffer
|
|
*/
|
|
l_rfl = rvb->l_buf[rvb->l_out];
|
|
rvb->l_buf[rvb->l_out] = 0;
|
|
rvb->l_out = (rvb->l_out + 1) % rvb->l_buf_size;
|
|
|
|
r_rfl = rvb->r_buf[rvb->r_out];
|
|
rvb->r_buf[rvb->r_out] = 0;
|
|
rvb->r_out = (rvb->r_out + 1) % rvb->r_buf_size;
|
|
|
|
for (k = 0; k < 8; k++) {
|
|
for (j = 0; j < 6; j++) {
|
|
l_buf_flt = ((l_rfl * rvb->coeff[k][j][0])
|
|
+ (rvb->l_buf_flt_in[k][j][0] * rvb->coeff[k][j][1])
|
|
+ (rvb->l_buf_flt_in[k][j][1] * rvb->coeff[k][j][2])
|
|
- (rvb->l_buf_flt_out[k][j][0] * rvb->coeff[k][j][3])
|
|
- (rvb->l_buf_flt_out[k][j][1] * rvb->coeff[k][j][4]))
|
|
/ 1024;
|
|
rvb->l_buf_flt_in[k][j][1] = rvb->l_buf_flt_in[k][j][0];
|
|
rvb->l_buf_flt_in[k][j][0] = l_rfl;
|
|
rvb->l_buf_flt_out[k][j][1] = rvb->l_buf_flt_out[k][j][0];
|
|
rvb->l_buf_flt_out[k][j][0] = l_buf_flt;
|
|
buffer[i] += l_buf_flt / 8;
|
|
|
|
r_buf_flt = ((r_rfl * rvb->coeff[k][j][0])
|
|
+ (rvb->r_buf_flt_in[k][j][0] * rvb->coeff[k][j][1])
|
|
+ (rvb->r_buf_flt_in[k][j][1] * rvb->coeff[k][j][2])
|
|
- (rvb->r_buf_flt_out[k][j][0] * rvb->coeff[k][j][3])
|
|
- (rvb->r_buf_flt_out[k][j][1] * rvb->coeff[k][j][4]))
|
|
/ 1024;
|
|
rvb->r_buf_flt_in[k][j][1] = rvb->r_buf_flt_in[k][j][0];
|
|
rvb->r_buf_flt_in[k][j][0] = r_rfl;
|
|
rvb->r_buf_flt_out[k][j][1] = rvb->r_buf_flt_out[k][j][0];
|
|
rvb->r_buf_flt_out[k][j][0] = r_buf_flt;
|
|
buffer[i + 1] += r_buf_flt / 8;
|
|
}
|
|
}
|
|
|
|
/*
|
|
add filtered result back into the buffers but on the opposite side
|
|
*/
|
|
tmp_l_val = buffer[i + 1] / vol_div;
|
|
tmp_r_val = buffer[i] / vol_div;
|
|
for (j = 0; j < 4; j++) {
|
|
rvb->l_buf[rvb->l_in[j]] += tmp_l_val;
|
|
rvb->l_in[j] = (rvb->l_in[j] + 1) % rvb->l_buf_size;
|
|
|
|
rvb->r_buf[rvb->r_in[j]] += tmp_r_val;
|
|
rvb->r_in[j] = (rvb->r_in[j] + 1) % rvb->r_buf_size;
|
|
}
|
|
}
|
|
}
|
|
|
|
}
|