2 #define _AUDIO_MIXER_H 1
4 // The audio mixer, dealing with extracting the right signals from
5 // each capture card, resampling signals so that they are in sync,
6 // processing them with effects (if desired), and then mixing them
7 // all together into one final audio signal.
9 // All operations on AudioMixer (except destruction) are thread-safe.
13 #include <zita-resampler/resampler.h>
24 #include "alsa_pool.h"
25 #include "correlation_measurer.h"
28 #include "ebu_r128_proc.h"
30 #include "input_mapping.h"
31 #include "resampling_queue.h"
32 #include "stereocompressor.h"
34 class DeviceSpecProto;
40 // Convert the given audio from {16,24,32}-bit M-channel to 32-bit N-channel PCM.
41 // Assumes little-endian and chunky, signed PCM throughout.
42 std::vector<int32_t> convert_audio_to_fixed32(const uint8_t *data, unsigned num_samples, bmusb::AudioFormat audio_format, unsigned num_destination_channels);
53 AudioMixer(unsigned num_capture_cards, unsigned num_ffmpeg_inputs);
54 void reset_resampler(DeviceSpec device_spec);
57 // Add audio (or silence) to the given device's queue. Can return false if
58 // the lock wasn't successfully taken; if so, you should simply try again.
59 // (This is to avoid a deadlock where a card hangs on the mutex in add_audio()
60 // while we are trying to shut it down from another thread that also holds
62 bool add_audio(DeviceSpec device_spec, const uint8_t *data, unsigned num_samples, bmusb::AudioFormat audio_format, std::chrono::steady_clock::time_point frame_time);
63 bool add_silence(DeviceSpec device_spec, unsigned samples_per_frame, unsigned num_frames);
65 // If a given device is offline for whatever reason and cannot deliver audio
66 // (by means of add_audio() or add_silence()), you can call put it in silence mode,
67 // where it will be taken to only output silence. Note that when taking it _out_
68 // of silence mode, the resampler will be reset, so that old audio will not
69 // affect it. Same true/false behavior as add_audio().
70 bool silence_card(DeviceSpec device_spec, bool silence);
72 std::vector<float> get_output(std::chrono::steady_clock::time_point ts, unsigned num_samples, ResamplingQueue::RateAdjustmentPolicy rate_adjustment_policy);
74 float get_fader_volume(unsigned bus_index) const { return fader_volume_db[bus_index]; }
75 void set_fader_volume(unsigned bus_index, float level_db) { fader_volume_db[bus_index] = level_db; }
77 bool get_mute(unsigned bus_index) const { return mute[bus_index]; }
78 void set_mute(unsigned bus_index, bool muted) { mute[bus_index] = muted; }
80 // Note: This operation holds all ALSA devices (see ALSAPool::get_devices()).
81 // You will need to call set_input_mapping() to get the hold state correctly,
82 // or every card will be held forever.
83 std::map<DeviceSpec, DeviceInfo> get_devices();
85 // See comments on ALSAPool::get_card_state().
86 ALSAPool::Device::State get_alsa_card_state(unsigned index)
88 return alsa_pool.get_card_state(index);
91 // See comments on ALSAPool::create_dead_card().
92 DeviceSpec create_dead_card(const std::string &name, const std::string &info, unsigned num_channels)
94 unsigned dead_card_index = alsa_pool.create_dead_card(name, info, num_channels);
95 return DeviceSpec{InputSourceType::ALSA_INPUT, dead_card_index};
98 void set_display_name(DeviceSpec device_spec, const std::string &name);
100 // Note: The card should be held (currently this isn't enforced, though).
101 void serialize_device(DeviceSpec device_spec, DeviceSpecProto *device_spec_proto);
103 enum class MappingMode {
104 // A single bus, only from a video card (no ALSA devices),
105 // only channel 1 and 2, locked to +0 dB. Note that this is
106 // only an UI abstraction around exactly the same audio code
107 // as MULTICHANNEL; it's just less flexible.
110 // Full, arbitrary mappings.
114 // Automatically sets mapping mode to MappingMode::SIMPLE.
115 void set_simple_input(unsigned card_index);
117 // If mapping mode is not representable as a MappingMode::SIMPLE type
118 // mapping, returns numeric_limits<unsigned>::max().
119 unsigned get_simple_input() const;
121 // Implicitly sets mapping mode to MappingMode::MULTICHANNEL.
122 void set_input_mapping(const InputMapping &input_mapping);
124 MappingMode get_mapping_mode() const;
125 InputMapping get_input_mapping() const;
127 unsigned num_buses() const;
129 void set_locut_cutoff(float cutoff_hz)
131 locut_cutoff_hz = cutoff_hz;
134 float get_locut_cutoff() const
136 return locut_cutoff_hz;
139 void set_locut_enabled(unsigned bus, bool enabled)
141 locut_enabled[bus] = enabled;
144 bool get_locut_enabled(unsigned bus)
146 return locut_enabled[bus];
149 bool is_mono(unsigned bus_index);
151 void set_stereo_width(unsigned bus_index, float width)
153 stereo_width[bus_index] = width;
156 float get_stereo_width(unsigned bus_index)
158 return stereo_width[bus_index];
161 void set_eq(unsigned bus_index, EQBand band, float db_gain)
163 assert(band >= 0 && band < NUM_EQ_BANDS);
164 eq_level_db[bus_index][band] = db_gain;
167 float get_eq(unsigned bus_index, EQBand band) const
169 assert(band >= 0 && band < NUM_EQ_BANDS);
170 return eq_level_db[bus_index][band];
173 float get_limiter_threshold_dbfs() const
175 return limiter_threshold_dbfs;
178 float get_compressor_threshold_dbfs(unsigned bus_index) const
180 return compressor_threshold_dbfs[bus_index];
183 void set_limiter_threshold_dbfs(float threshold_dbfs)
185 limiter_threshold_dbfs = threshold_dbfs;
188 void set_compressor_threshold_dbfs(unsigned bus_index, float threshold_dbfs)
190 compressor_threshold_dbfs[bus_index] = threshold_dbfs;
193 void set_limiter_enabled(bool enabled)
195 limiter_enabled = enabled;
198 bool get_limiter_enabled() const
200 return limiter_enabled;
203 void set_compressor_enabled(unsigned bus_index, bool enabled)
205 compressor_enabled[bus_index] = enabled;
208 bool get_compressor_enabled(unsigned bus_index) const
210 return compressor_enabled[bus_index];
213 void set_gain_staging_db(unsigned bus_index, float gain_db)
215 std::lock_guard<std::mutex> lock(compressor_mutex);
216 level_compressor_enabled[bus_index] = false;
217 gain_staging_db[bus_index] = gain_db;
220 float get_gain_staging_db(unsigned bus_index) const
222 std::lock_guard<std::mutex> lock(compressor_mutex);
223 return gain_staging_db[bus_index];
226 void set_gain_staging_auto(unsigned bus_index, bool enabled)
228 std::lock_guard<std::mutex> lock(compressor_mutex);
229 level_compressor_enabled[bus_index] = enabled;
232 bool get_gain_staging_auto(unsigned bus_index) const
234 std::lock_guard<std::mutex> lock(compressor_mutex);
235 return level_compressor_enabled[bus_index];
238 void set_final_makeup_gain_db(float gain_db)
240 std::lock_guard<std::mutex> lock(compressor_mutex);
241 final_makeup_gain_auto = false;
242 final_makeup_gain = from_db(gain_db);
245 float get_final_makeup_gain_db()
247 std::lock_guard<std::mutex> lock(compressor_mutex);
248 return to_db(final_makeup_gain);
251 void set_final_makeup_gain_auto(bool enabled)
253 std::lock_guard<std::mutex> lock(compressor_mutex);
254 final_makeup_gain_auto = enabled;
257 bool get_final_makeup_gain_auto() const
259 std::lock_guard<std::mutex> lock(compressor_mutex);
260 return final_makeup_gain_auto;
263 void reset_peak(unsigned bus_index);
266 float current_level_dbfs[2]; // Digital peak of last frame, left and right.
267 float peak_level_dbfs[2]; // Digital peak with hold, left and right.
268 float historic_peak_dbfs;
269 float gain_staging_db;
270 float compressor_attenuation_db; // A positive number; 0.0 for no attenuation.
273 typedef std::function<void(float level_lufs, float peak_db,
274 std::vector<BusLevel> bus_levels,
275 float global_level_lufs, float range_low_lufs, float range_high_lufs,
276 float final_makeup_gain_db,
277 float correlation)> audio_level_callback_t;
278 void set_audio_level_callback(audio_level_callback_t callback)
280 audio_level_callback = callback;
283 typedef std::function<void()> state_changed_callback_t;
284 void set_state_changed_callback(state_changed_callback_t callback)
286 state_changed_callback = callback;
289 state_changed_callback_t get_state_changed_callback() const
291 return state_changed_callback;
294 void trigger_state_changed_callback()
296 if (state_changed_callback != nullptr) {
297 state_changed_callback();
301 // A combination of all settings for a bus. Useful if you want to get
302 // or store them as a whole without bothering to call all of the get_*
303 // or set_* functions for that bus.
305 float fader_volume_db;
309 float eq_level_db[NUM_EQ_BANDS];
310 float gain_staging_db;
311 bool level_compressor_enabled;
312 float compressor_threshold_dbfs;
313 bool compressor_enabled;
315 static BusSettings get_default_bus_settings();
316 BusSettings get_bus_settings(unsigned bus_index) const;
317 void set_bus_settings(unsigned bus_index, const BusSettings &settings);
321 std::unique_ptr<ResamplingQueue> resampling_queue;
322 std::string display_name;
323 unsigned capture_frequency = OUTPUT_FREQUENCY;
324 // Which channels we consider interesting (ie., are part of some input_mapping).
325 std::set<unsigned> interesting_channels;
326 bool silenced = false;
329 const AudioDevice *find_audio_device(DeviceSpec device_spec) const
331 return const_cast<AudioMixer *>(this)->find_audio_device(device_spec);
334 AudioDevice *find_audio_device(DeviceSpec device_spec);
336 void find_sample_src_from_device(const std::map<DeviceSpec, std::vector<float>> &samples_card, DeviceSpec device_spec, int source_channel, const float **srcptr, unsigned *stride);
337 void fill_audio_bus(const std::map<DeviceSpec, std::vector<float>> &samples_card, const InputMapping::Bus &bus, unsigned num_samples, float stereo_width, float *output);
338 void reset_resampler_mutex_held(DeviceSpec device_spec);
339 void apply_eq(unsigned bus_index, std::vector<float> *samples_bus);
340 void update_meters(const std::vector<float> &samples);
341 void add_bus_to_master(unsigned bus_index, const std::vector<float> &samples_bus, std::vector<float> *samples_out);
342 void measure_bus_levels(unsigned bus_index, const std::vector<float> &left, const std::vector<float> &right);
343 void send_audio_level_callback();
344 std::vector<DeviceSpec> get_active_devices() const;
345 void set_input_mapping_lock_held(const InputMapping &input_mapping);
347 unsigned num_capture_cards, num_ffmpeg_inputs;
349 mutable std::timed_mutex audio_mutex;
352 AudioDevice video_cards[MAX_VIDEO_CARDS]; // Under audio_mutex.
353 AudioDevice alsa_inputs[MAX_ALSA_CARDS]; // Under audio_mutex.
354 std::unique_ptr<AudioDevice[]> ffmpeg_inputs; // Under audio_mutex.
356 std::atomic<float> locut_cutoff_hz{120};
357 StereoFilter locut[MAX_BUSES]; // Default cutoff 120 Hz, 24 dB/oct.
358 std::atomic<bool> locut_enabled[MAX_BUSES];
359 StereoFilter eq[MAX_BUSES][NUM_EQ_BANDS]; // The one for EQBand::MID isn't actually used (see comments in apply_eq()).
361 // First compressor; takes us up to about -12 dBFS.
362 mutable std::mutex compressor_mutex;
363 std::unique_ptr<StereoCompressor> level_compressor[MAX_BUSES]; // Under compressor_mutex. Used to set/override gain_staging_db if <level_compressor_enabled>.
364 float gain_staging_db[MAX_BUSES]; // Under compressor_mutex.
365 float last_gain_staging_db[MAX_BUSES]; // Under compressor_mutex.
366 bool level_compressor_enabled[MAX_BUSES]; // Under compressor_mutex.
368 static constexpr float ref_level_dbfs = -14.0f; // Chosen so that we end up around 0 LU in practice.
369 static constexpr float ref_level_lufs = -23.0f; // 0 LU, more or less by definition.
371 StereoCompressor limiter;
372 std::atomic<float> limiter_threshold_dbfs{ref_level_dbfs + 4.0f}; // 4 dB.
373 std::atomic<bool> limiter_enabled{true};
374 std::unique_ptr<StereoCompressor> compressor[MAX_BUSES];
375 std::atomic<float> compressor_threshold_dbfs[MAX_BUSES];
376 std::atomic<bool> compressor_enabled[MAX_BUSES];
378 // Note: The values here are not in dB.
380 float current_level = 0.0f; // Peak of the last frame.
381 float historic_peak = 0.0f; // Highest peak since last reset; no falloff.
382 float current_peak = 0.0f; // Current peak of the peak meter.
383 float last_peak = 0.0f;
384 float age_seconds = 0.0f; // Time since "last_peak" was set.
386 PeakHistory peak_history[MAX_BUSES][2]; // Separate for each channel. Under audio_mutex.
388 double final_makeup_gain = 1.0; // Under compressor_mutex. Read/write by the user. Note: Not in dB, we want the numeric precision so that we can change it slowly.
389 bool final_makeup_gain_auto = true; // Under compressor_mutex.
391 MappingMode current_mapping_mode; // Under audio_mutex.
392 InputMapping input_mapping; // Under audio_mutex.
393 std::atomic<float> fader_volume_db[MAX_BUSES] {{ 0.0f }};
394 std::atomic<bool> mute[MAX_BUSES] {{ false }};
395 float last_fader_volume_db[MAX_BUSES] { 0.0f }; // Under audio_mutex.
396 std::atomic<float> stereo_width[MAX_BUSES] {{ 0.0f }}; // Default 1.0f (is set in constructor).
397 std::atomic<float> eq_level_db[MAX_BUSES][NUM_EQ_BANDS] {{{ 0.0f }}};
398 float last_eq_level_db[MAX_BUSES][NUM_EQ_BANDS] {{ 0.0f }};
400 audio_level_callback_t audio_level_callback = nullptr;
401 state_changed_callback_t state_changed_callback = nullptr;
402 mutable std::mutex audio_measure_mutex;
403 Ebu_r128_proc r128; // Under audio_measure_mutex.
404 CorrelationMeasurer correlation; // Under audio_measure_mutex.
405 Resampler peak_resampler; // Under audio_measure_mutex.
406 std::atomic<float> peak{0.0f};
409 std::atomic<double> metric_audio_loudness_short_lufs{0.0 / 0.0};
410 std::atomic<double> metric_audio_loudness_integrated_lufs{0.0 / 0.0};
411 std::atomic<double> metric_audio_loudness_range_low_lufs{0.0 / 0.0};
412 std::atomic<double> metric_audio_loudness_range_high_lufs{0.0 / 0.0};
413 std::atomic<double> metric_audio_peak_dbfs{0.0 / 0.0};
414 std::atomic<double> metric_audio_final_makeup_gain_db{0.0};
415 std::atomic<double> metric_audio_correlation{0.0};
417 // These are all gauges corresponding to the elements of BusLevel.
418 // In a sense, they'd probably do better as histograms, but that's an
419 // awful lot of time series when you have many buses.
421 std::vector<std::pair<std::string, std::string>> labels;
422 std::atomic<double> current_level_dbfs[2]{{0.0/0.0},{0.0/0.0}};
423 std::atomic<double> peak_level_dbfs[2]{{0.0/0.0},{0.0/0.0}};
424 std::atomic<double> historic_peak_dbfs{0.0/0.0};
425 std::atomic<double> gain_staging_db{0.0/0.0};
426 std::atomic<double> compressor_attenuation_db{0.0/0.0};
428 std::unique_ptr<BusMetrics[]> bus_metrics; // One for each bus in <input_mapping>.
431 extern AudioMixer *global_audio_mixer;
433 #endif // !defined(_AUDIO_MIXER_H)