123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268 |
- /* -*- linux-c -*-
- Copyright (C) 2004 Tom Szilagyi
-
- 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; if not, write to the Free Software
- Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
- $Id: tap_utils.h,v 1.5 2004/02/21 17:33:36 tszilagyi Exp $
- */
- #ifndef _ISOC99_SOURCE
- #define _ISOC99_SOURCE
- #endif
- #include <stdint.h>
- #ifndef M_PI
- #define M_PI 3.14159265358979323846264338327
- #endif
- /* push a sample into a ringbuffer and return the sample falling out */
- static inline
- LADSPA_Data
- push_buffer(LADSPA_Data insample, LADSPA_Data * buffer,
- unsigned long buflen, unsigned long * pos) {
- LADSPA_Data outsample;
- outsample = buffer[*pos];
- buffer[(*pos)++] = insample;
- if (*pos >= buflen)
- *pos = 0;
- return outsample;
- }
- /* read a value from a ringbuffer.
- * n == 0 returns the oldest sample from the buffer.
- * n == buflen-1 returns the sample written to the buffer
- * at the last push_buffer call.
- * n must not exceed buflen-1, or your computer will explode.
- */
- static inline
- LADSPA_Data
- read_buffer(LADSPA_Data * buffer, unsigned long buflen,
- unsigned long pos, unsigned long n) {
- while (n + pos >= buflen)
- n -= buflen;
- return buffer[n + pos];
- }
- /* overwrites a value in a ringbuffer, but pos stays the same.
- * n == 0 overwrites the oldest sample pushed in the buffer.
- * n == buflen-1 overwrites the sample written to the buffer
- * at the last push_buffer call.
- * n must not exceed buflen-1, or your computer... you know.
- */
- static inline
- void
- write_buffer(LADSPA_Data insample, LADSPA_Data * buffer, unsigned long buflen,
- unsigned long pos, unsigned long n) {
- while (n + pos >= buflen)
- n -= buflen;
- buffer[n + pos] = insample;
- }
- /* Please note that the majority of the definitions and helper
- functions below have been derived from the source code of Steve
- Harris's SWH plugins (particularly from the "biquad.h" file). While I
- give him credit for his excellent work, I reserve myself to be blamed
- for any bugs or malfunction. */
- #define db2lin(x) ((x) > -90.0f ? powf(10.0f, (x) * 0.05f) : 0.0f)
- #define ABS(x) (x)>0.0f?(x):-1.0f*(x)
- #define LN_2_2 0.34657359f
- #define LIMIT(v,l,u) ((v)<(l)?(l):((v)>(u)?(u):(v)))
- #define BIQUAD_TYPE float
- typedef BIQUAD_TYPE bq_t;
- /* Biquad filter (adapted from lisp code by Eli Brandt,
- http://www.cs.cmu.edu/~eli/) */
- /* The prev. comment has been preserved from Steve Harris's biquad.h */
- typedef struct {
- bq_t a1;
- bq_t a2;
- bq_t b0;
- bq_t b1;
- bq_t b2;
- bq_t x1;
- bq_t x2;
- bq_t y1;
- bq_t y2;
- } biquad;
- static inline void biquad_init(biquad *f) {
- f->x1 = 0.0f;
- f->x2 = 0.0f;
- f->y1 = 0.0f;
- f->y2 = 0.0f;
- }
- static inline
- void
- eq_set_params(biquad *f, bq_t fc, bq_t gain, bq_t bw, bq_t fs) {
- bq_t w = 2.0f * M_PI * LIMIT(fc, 1.0f, fs/2.0f) / fs;
- bq_t cw = cosf(w);
- bq_t sw = sinf(w);
- bq_t J = pow(10.0f, gain * 0.025f);
- bq_t g = sw * sinhf(LN_2_2 * LIMIT(bw, 0.0001f, 4.0f) * w / sw);
- bq_t a0r = 1.0f / (1.0f + (g / J));
- f->b0 = (1.0f + (g * J)) * a0r;
- f->b1 = (-2.0f * cw) * a0r;
- f->b2 = (1.0f - (g * J)) * a0r;
- f->a1 = -(f->b1);
- f->a2 = ((g / J) - 1.0f) * a0r;
- }
- static inline void lp_set_params(biquad *f, bq_t fc, bq_t bw, bq_t fs) {
- bq_t omega = 2.0 * M_PI * fc/fs;
- bq_t sn = sin(omega);
- bq_t cs = cos(omega);
- bq_t alpha = sn * sinh(M_LN2 / 2.0 * bw * omega / sn);
- const float a0r = 1.0 / (1.0 + alpha);
- #if 0
- b0 = (1 - cs) /2;
- b1 = 1 - cs;
- b2 = (1 - cs) /2;
- a0 = 1 + alpha;
- a1 = -2 * cs;
- a2 = 1 - alpha;
- #endif
- f->b0 = a0r * (1.0 - cs) * 0.5;
- f->b1 = a0r * (1.0 - cs);
- f->b2 = a0r * (1.0 - cs) * 0.5;
- f->a1 = a0r * (2.0 * cs);
- f->a2 = a0r * (alpha - 1.0);
- }
- static inline
- void
- hp_set_params(biquad *f, bq_t fc, bq_t bw, bq_t fs)
- {
- bq_t omega = 2.0 * M_PI * fc/fs;
- bq_t sn = sin(omega);
- bq_t cs = cos(omega);
- bq_t alpha = sn * sinh(M_LN2 / 2.0 * bw * omega / sn);
- const float a0r = 1.0 / (1.0 + alpha);
- #if 0
- b0 = (1 + cs) /2;
- b1 = -(1 + cs);
- b2 = (1 + cs) /2;
- a0 = 1 + alpha;
- a1 = -2 * cs;
- a2 = 1 - alpha;
- #endif
- f->b0 = a0r * (1.0 + cs) * 0.5;
- f->b1 = a0r * -(1.0 + cs);
- f->b2 = a0r * (1.0 + cs) * 0.5;
- f->a1 = a0r * (2.0 * cs);
- f->a2 = a0r * (alpha - 1.0);
- }
- static inline
- void
- ls_set_params(biquad *f, bq_t fc, bq_t gain, bq_t slope, bq_t fs)
- {
- bq_t w = 2.0f * M_PI * LIMIT(fc, 1.0, fs/2.0) / fs;
- bq_t cw = cosf(w);
- bq_t sw = sinf(w);
- bq_t A = powf(10.0f, gain * 0.025f);
- bq_t b = sqrt(((1.0f + A * A) / LIMIT(slope, 0.0001f, 1.0f)) - ((A -
- 1.0f) * (A - 1.0)));
- bq_t apc = cw * (A + 1.0f);
- bq_t amc = cw * (A - 1.0f);
- bq_t bs = b * sw;
- bq_t a0r = 1.0f / (A + 1.0f + amc + bs);
- f->b0 = a0r * A * (A + 1.0f - amc + bs);
- f->b1 = a0r * 2.0f * A * (A - 1.0f - apc);
- f->b2 = a0r * A * (A + 1.0f - amc - bs);
- f->a1 = a0r * 2.0f * (A - 1.0f + apc);
- f->a2 = a0r * (-A - 1.0f - amc + bs);
- }
- static inline
- void
- hs_set_params(biquad *f, bq_t fc, bq_t gain, bq_t slope, bq_t fs) {
- bq_t w = 2.0f * M_PI * LIMIT(fc, 1.0, fs/2.0) / fs;
- bq_t cw = cosf(w);
- bq_t sw = sinf(w);
- bq_t A = powf(10.0f, gain * 0.025f);
- bq_t b = sqrt(((1.0f + A * A) / LIMIT(slope, 0.0001f, 1.0f)) - ((A -
- 1.0f) * (A - 1.0f)));
- bq_t apc = cw * (A + 1.0f);
- bq_t amc = cw * (A - 1.0f);
- bq_t bs = b * sw;
- bq_t a0r = 1.0f / (A + 1.0f - amc + bs);
- f->b0 = a0r * A * (A + 1.0f + amc + bs);
- f->b1 = a0r * -2.0f * A * (A - 1.0f + apc);
- f->b2 = a0r * A * (A + 1.0f + amc - bs);
- f->a1 = a0r * -2.0f * (A - 1.0f - apc);
- f->a2 = a0r * (-A - 1.0f + amc + bs);
- }
- static inline
- bq_t
- biquad_run(biquad *f, bq_t x) {
- union {
- bq_t y;
- uint32_t y_int;
- } u;
-
- u.y = f->b0 * x + f->b1 * f->x1 + f->b2 * f->x2
- + f->a1 * f->y1 + f->a2 * f->y2;
- if ((u.y_int & 0x7f800000) == 0)
- u.y = 0.0f;
- f->x2 = f->x1;
- f->x1 = x;
- f->y2 = f->y1;
- f->y1 = u.y;
- return u.y;
- }
|