#define GAMMA 0.166
#define ALPHA 1.0
-static float hysteresis_limit = 3000.0 / 32768.0;
+static float hysteresis_upper_limit = 3000.0 / 32768.0;
+static float hysteresis_lower_limit = -3000.0 / 32768.0;
static bool do_calibrate = true;
static bool output_cycles_plot = false;
static bool do_crop = false;
static std::vector<float> train_snap_points;
static bool do_train = false;
+// The frequency to filter on (for do_auto_level), in Hertz.
+// Larger values makes the compressor react faster, but if it is too large,
+// you'll ruin the waveforms themselves.
+static float auto_level_freq = 200.0;
+
// The minimum estimated sound level (for do_auto_level) at any given point.
// If you decrease this, you'll be able to amplify really silent signals
// by more, but you'll also increase the level of silent (ie. noise-only) segments,
// possibly caused misdetected pulses in these segments.
static float min_level = 0.05f;
-// between [x,x+1]
-double find_zerocrossing(const std::vector<float> &pcm, int x)
+// search for the value <limit> between [x,x+1]
+double find_crossing(const std::vector<float> &pcm, int x, float limit)
{
- if (pcm[x] == 0) {
- return x;
- }
- if (pcm[x + 1] == 0) {
- return x + 1;
- }
-
- assert(pcm[x + 1] < 0);
- assert(pcm[x] > 0);
-
double upper = x;
double lower = x + 1;
while (lower - upper > 1e-3) {
double mid = 0.5f * (upper + lower);
- if (lanczos_interpolate(pcm, mid) > 0) {
+ if (lanczos_interpolate(pcm, mid) > limit) {
upper = mid;
} else {
lower = mid;
static struct option long_options[] = {
{"auto-level", 0, 0, 'a' },
+ {"auto-level-freq", required_argument, 0, 'b' },
{"output-leveled", 0, 0, 'A' },
+ {"min-level", required_argument, 0, 'm' },
{"no-calibrate", 0, 0, 's' },
{"plot-cycles", 0, 0, 'p' },
{"hysteresis-limit", required_argument, 0, 'l' },
fprintf(stderr, "decode [OPTIONS] AUDIO-FILE > TAP-FILE\n");
fprintf(stderr, "\n");
fprintf(stderr, " -a, --auto-level automatically adjust amplitude levels throughout the file\n");
+ fprintf(stderr, " -b, --auto-level-freq minimum frequency in Hertz of corrected level changes (default 200 Hz)\n");
fprintf(stderr, " -A, --output-leveled output leveled waveform to leveled.raw\n");
fprintf(stderr, " -m, --min-level minimum estimated sound level (0..32768) for --auto-level\n");
fprintf(stderr, " -s, --no-calibrate do not try to calibrate on sync pulse length\n");
{
for ( ;; ) {
int option_index = 0;
- int c = getopt_long(argc, argv, "aAm:spl:f:r:Fc:t:qh", long_options, &option_index);
+ int c = getopt_long(argc, argv, "ab:Am:spl:f:r:Fc:t:qh", long_options, &option_index);
if (c == -1)
break;
do_auto_level = true;
break;
+ case 'b':
+ auto_level_freq = atof(optarg);
+ break;
+
case 'A':
output_leveled = true;
break;
output_cycles_plot = true;
break;
- case 'l':
- hysteresis_limit = atof(optarg) / 32768.0;
+ case 'l': {
+ const char *hyststr = strtok(optarg, ": ");
+ hysteresis_upper_limit = atof(hyststr) / 32768.0;
+ hyststr = strtok(NULL, ": ");
+ if (hyststr == NULL) {
+ hysteresis_lower_limit = -hysteresis_upper_limit;
+ } else {
+ hysteresis_lower_limit = atof(hyststr) / 32768.0;
+ }
break;
+ }
case 'f': {
const char *coeffstr = strtok(optarg, ": ");
std::vector<float> do_rc_filter(const std::vector<float>& pcm, float freq, int sample_rate)
{
+ // This is only a 6 dB/oct filter, which seemingly works better
+ // than the Filter class, which is a standard biquad (12 dB/oct).
+ // The b/c calculations come from libnyquist (atone.c);
+ // I haven't checked, but I suppose they fall out of the bilinear
+ // transform of the transfer function H(s) = s/(s + w).
std::vector<float> filtered_pcm;
filtered_pcm.resize(pcm.size());
- Filter filter = Filter::hpf(M_PI * freq / sample_rate);
+ const float b = 2.0f - cos(2.0 * M_PI * freq / sample_rate);
+ const float c = b - sqrt(b * b - 1.0f);
+ float prev_in = 0.0f;
+ float prev_out = 0.0f;
for (unsigned i = 0; i < pcm.size(); ++i) {
- filtered_pcm[i] = filter.update(pcm[i]);
+ float in = pcm[i];
+ float out = c * (prev_out + in - prev_in);
+ filtered_pcm[i] = out;
+ prev_in = in;
+ prev_out = out;
}
if (output_filtered) {
std::vector<pulse> pulses;
// Find the flanks.
- int last_bit = -1;
+ enum State { START, ABOVE, BELOW } state = START;
double last_downflank = -1;
for (unsigned i = 0; i < pcm.size(); ++i) {
- int bit = (pcm[i] > 0) ? 1 : 0;
- if (bit == 0 && last_bit == 1) {
- // Check if we ever go up above <hysteresis_limit> before we dip down again.
- bool true_pulse = false;
- unsigned j;
- int min_level_after = 32767;
- for (j = i; j < pcm.size(); ++j) {
- min_level_after = std::min<int>(min_level_after, pcm[j]);
- if (pcm[j] > 0) break;
- if (pcm[j] < -hysteresis_limit) {
- true_pulse = true;
- break;
+ if (pcm[i] > hysteresis_upper_limit) {
+ state = ABOVE;
+ } else if (pcm[i] < hysteresis_lower_limit) {
+ if (state == ABOVE) {
+ // down-flank!
+ double t = find_crossing(pcm, i - 1, hysteresis_lower_limit) * (1.0 / sample_rate) + crop_start;
+ if (last_downflank > 0) {
+ pulse p;
+ p.time = t;
+ p.len = t - last_downflank;
+ pulses.push_back(p);
}
+ last_downflank = t;
}
-
- if (!true_pulse) {
-#if 0
- fprintf(stderr, "Ignored down-flank at %.6f seconds due to hysteresis (%d < %d).\n",
- double(i) / sample_rate, -min_level_after, hysteresis_limit);
-#endif
- i = j;
- continue;
- }
-
- // down-flank!
- double t = find_zerocrossing(pcm, i - 1) * (1.0 / sample_rate) + crop_start;
- if (last_downflank > 0) {
- pulse p;
- p.time = t;
- p.len = t - last_downflank;
- pulses.push_back(p);
- }
- last_downflank = t;
+ state = BELOW;
}
- last_bit = bit;
}
return pulses;
}
bool any_moved = false;
for (unsigned i = 0; i < initial_centers.size(); ++i) {
if (num[i] == 0) {
- printf("K-means broke down, can't output new reference training points\n");
+ fprintf(stderr, "K-means broke down, can't output new reference training points\n");
return;
}
float new_center = sums[i] / num[i];
break;
}
}
- printf("New reference training points:");
+ fprintf(stderr, "New reference training points:");
for (unsigned i = 0; i < last_centers.size(); ++i) {
- printf(" %.3f", last_centers[i]);
+ fprintf(stderr, " %.3f", last_centers[i]);
}
- printf("\n");
+ fprintf(stderr, "\n");
}
void spsa_train(const std::vector<float> &pcm, int sample_rate)
}
if (do_auto_level) {
- pcm = level_samples(pcm, min_level, sample_rate);
+ pcm = level_samples(pcm, min_level, auto_level_freq, sample_rate);
if (output_leveled) {
FILE *fp = fopen("leveled.raw", "wb");
fwrite(pcm.data(), pcm.size() * sizeof(pcm[0]), 1, fp);
projects / c64tapwav / blobdiff