1 /*
2 * Direct Stream Digital (DSD) decoder
3 * based on BSD licensed dsd2pcm by Sebastian Gesemann
4 * Copyright (c) 2009, 2011 Sebastian Gesemann. All rights reserved.
5 * Copyright (c) 2014 Peter Ross
6 *
7 * This file is part of FFmpeg.
8 *
9 * FFmpeg is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public
11 * License as published by the Free Software Foundation; either
12 * version 2.1 of the License, or (at your option) any later version.
13 *
14 * FFmpeg is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
18 *
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with FFmpeg; if not, write to the Free Software
21 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 */
23
24 #include <string.h>
25 #include "libavutil/attributes.h"
26 #include "libavutil/reverse.h"
27 #include "libavutil/thread.h"
28 #include "dsd.h"
29
30 #define CTABLES ((HTAPS + 7) / 8) /** number of "8 MACs" lookup tables */
31
32 /*
33 * Properties of this 96-tap lowpass filter when applied on a signal
34 * with sampling rate of 44100*64 Hz:
35 *
36 * () has a delay of 17 microseconds.
37 *
38 * () flat response up to 48 kHz
39 *
40 * () if you downsample afterwards by a factor of 8, the
41 * spectrum below 70 kHz is practically alias-free.
42 *
43 * () stopband rejection is about 160 dB
44 *
45 * The coefficient tables ("ctables") take only 6 Kibi Bytes and
46 * should fit into a modern processor's fast cache.
47 */
48
49 /**
50 * The 2nd half (48 coeffs) of a 96-tap symmetric lowpass filter
51 */
52 static const double htaps[HTAPS] = {
53 0.09950731974056658, 0.09562845727714668, 0.08819647126516944,
54 0.07782552527068175, 0.06534876523171299, 0.05172629311427257,
55 0.0379429484910187, 0.02490921351762261, 0.0133774746265897,
56 0.003883043418804416, -0.003284703416210726, -0.008080250212687497,
57 -0.01067241812471033, -0.01139427235000863, -0.0106813877974587,
58 -0.009007905078766049, -0.006828859761015335, -0.004535184322001496,
59 -0.002425035959059578, -0.0006922187080790708, 0.0005700762133516592,
60 0.001353838005269448, 0.001713709169690937, 0.001742046839472948,
61 0.001545601648013235, 0.001226696225277855, 0.0008704322683580222,
62 0.0005381636200535649, 0.000266446345425276, 7.002968738383528e-05,
63 -5.279407053811266e-05, -0.0001140625650874684, -0.0001304796361231895,
64 -0.0001189970287491285, -9.396247155265073e-05, -6.577634378272832e-05,
65 -4.07492895872535e-05, -2.17407957554587e-05, -9.163058931391722e-06,
66 -2.017460145032201e-06, 1.249721855219005e-06, 2.166655190537392e-06,
67 1.930520892991082e-06, 1.319400334374195e-06, 7.410039764949091e-07,
68 3.423230509967409e-07, 1.244182214744588e-07, 3.130441005359396e-08
69 };
70
71 static float ctables[CTABLES][256];
72
dsd_ctables_tableinit(void)73 static av_cold void dsd_ctables_tableinit(void)
74 {
75 int t, e, m, sign;
76 double acc[CTABLES];
77 for (e = 0; e < 256; ++e) {
78 memset(acc, 0, sizeof(acc));
79 for (m = 0; m < 8; ++m) {
80 sign = (((e >> (7 - m)) & 1) * 2 - 1);
81 for (t = 0; t < CTABLES; ++t)
82 acc[t] += sign * htaps[t * 8 + m];
83 }
84 for (t = 0; t < CTABLES; ++t)
85 ctables[CTABLES - 1 - t][e] = acc[t];
86 }
87 }
88
ff_init_dsd_data(void)89 av_cold void ff_init_dsd_data(void)
90 {
91 static AVOnce init_static_once = AV_ONCE_INIT;
92 ff_thread_once(&init_static_once, dsd_ctables_tableinit);
93 }
94
ff_dsd2pcm_translate(DSDContext * s,size_t samples,int lsbf,const uint8_t * src,ptrdiff_t src_stride,float * dst,ptrdiff_t dst_stride)95 void ff_dsd2pcm_translate(DSDContext* s, size_t samples, int lsbf,
96 const uint8_t *src, ptrdiff_t src_stride,
97 float *dst, ptrdiff_t dst_stride)
98 {
99 uint8_t buf[FIFOSIZE];
100 unsigned pos, i;
101 uint8_t* p;
102 double sum;
103
104 pos = s->pos;
105
106 memcpy(buf, s->buf, sizeof(buf));
107
108 while (samples-- > 0) {
109 buf[pos] = lsbf ? ff_reverse[*src] : *src;
110 src += src_stride;
111
112 p = buf + ((pos - CTABLES) & FIFOMASK);
113 *p = ff_reverse[*p];
114
115 sum = 0.0;
116 for (i = 0; i < CTABLES; i++) {
117 uint8_t a = buf[(pos - i) & FIFOMASK];
118 uint8_t b = buf[(pos - (CTABLES*2 - 1) + i) & FIFOMASK];
119 sum += ctables[i][a] + ctables[i][b];
120 }
121
122 *dst = (float)sum;
123 dst += dst_stride;
124
125 pos = (pos + 1) & FIFOMASK;
126 }
127
128 s->pos = pos;
129 memcpy(s->buf, buf, sizeof(buf));
130 }
131