[nageru] / nageru / audio_mixer.h

#ifndef _AUDIO_MIXER_H
#define _AUDIO_MIXER_H 1

// The audio mixer, dealing with extracting the right signals from
// each capture card, resampling signals so that they are in sync,
// processing them with effects (if desired), and then mixing them
// all together into one final audio signal.
//
// All operations on AudioMixer (except destruction) are thread-safe.

#include <assert.h>
#include <stdint.h>
#include <zita-resampler/resampler.h>
#include <atomic>
#include <chrono>
#include <functional>
#include <map>
#include <memory>
#include <mutex>
#include <set>
#include <string>
#include <vector>

#include "alsa_pool.h"
#include "correlation_measurer.h"
#include "decibel.h"
#include "defs.h"
#include "ebu_r128_proc.h"
#include "filter.h"
#include "input_mapping.h"
#include "resampling_queue.h"
#include "stereocompressor.h"

class DeviceSpecProto;

namespace bmusb {
struct AudioFormat;
}  // namespace bmusb

// Convert the given audio from {16,24,32}-bit M-channel to 32-bit N-channel PCM.
// Assumes little-endian and chunky, signed PCM throughout.
std::vector<int32_t> convert_audio_to_fixed32(const uint8_t *data, unsigned num_samples, bmusb::AudioFormat audio_format, unsigned num_destination_channels);

enum EQBand {
        EQ_BAND_BASS = 0,
        EQ_BAND_MID,
        EQ_BAND_TREBLE,
        NUM_EQ_BANDS
};

class AudioMixer {
public:
        AudioMixer(unsigned num_capture_cards, unsigned num_ffmpeg_inputs);
        void reset_resampler(DeviceSpec device_spec);
        void reset_meters();

        // Add audio (or silence) to the given device's queue. Can return false if
        // the lock wasn't successfully taken; if so, you should simply try again.
        // (This is to avoid a deadlock where a card hangs on the mutex in add_audio()
        // while we are trying to shut it down from another thread that also holds
        // the mutex.)
        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);
        bool add_silence(DeviceSpec device_spec, unsigned samples_per_frame, unsigned num_frames);

        // If a given device is offline for whatever reason and cannot deliver audio
        // (by means of add_audio() or add_silence()), you can call put it in silence mode,
        // where it will be taken to only output silence. Note that when taking it _out_
        // of silence mode, the resampler will be reset, so that old audio will not
        // affect it. Same true/false behavior as add_audio().
        bool silence_card(DeviceSpec device_spec, bool silence);

        std::vector<float> get_output(std::chrono::steady_clock::time_point ts, unsigned num_samples, ResamplingQueue::RateAdjustmentPolicy rate_adjustment_policy);

        float get_fader_volume(unsigned bus_index) const { return fader_volume_db[bus_index]; }
        void set_fader_volume(unsigned bus_index, float level_db) { fader_volume_db[bus_index] = level_db; }

        bool get_mute(unsigned bus_index) const { return mute[bus_index]; }
        void set_mute(unsigned bus_index, bool muted) { mute[bus_index] = muted; }

        // Note: This operation holds all ALSA devices (see ALSAPool::get_devices()).
        // You will need to call set_input_mapping() to get the hold state correctly,
        // or every card will be held forever.
        std::map<DeviceSpec, DeviceInfo> get_devices();

        // See comments on ALSAPool::get_card_state().
        ALSAPool::Device::State get_alsa_card_state(unsigned index)
        {
                return alsa_pool.get_card_state(index);
        }

        // See comments on ALSAPool::create_dead_card().
        DeviceSpec create_dead_card(const std::string &name, const std::string &info, unsigned num_channels)
        {
                unsigned dead_card_index = alsa_pool.create_dead_card(name, info, num_channels);
                return DeviceSpec{InputSourceType::ALSA_INPUT, dead_card_index};
        }

        void set_display_name(DeviceSpec device_spec, const std::string &name);

        // Note: The card should be held (currently this isn't enforced, though).
        void serialize_device(DeviceSpec device_spec, DeviceSpecProto *device_spec_proto);

        enum class MappingMode {
                // A single bus, only from a video card (no ALSA devices),
                // only channel 1 and 2, locked to +0 dB. Note that this is
                // only an UI abstraction around exactly the same audio code
                // as MULTICHANNEL; it's just less flexible.
                SIMPLE,

                // Full, arbitrary mappings.
                MULTICHANNEL
        };

        // Automatically sets mapping mode to MappingMode::SIMPLE.
        void set_simple_input(unsigned card_index);

        // If mapping mode is not representable as a MappingMode::SIMPLE type
        // mapping, returns numeric_limits<unsigned>::max().
        unsigned get_simple_input() const;

        // Implicitly sets mapping mode to MappingMode::MULTICHANNEL.
        void set_input_mapping(const InputMapping &input_mapping);

        MappingMode get_mapping_mode() const;
        InputMapping get_input_mapping() const;

        unsigned num_buses() const;

        void set_locut_cutoff(float cutoff_hz)
        {
                locut_cutoff_hz = cutoff_hz;
        }

        float get_locut_cutoff() const
        {
                return locut_cutoff_hz;
        }

        void set_locut_enabled(unsigned bus, bool enabled)
        {
                locut_enabled[bus] = enabled;
        }

        bool get_locut_enabled(unsigned bus)
        {
                return locut_enabled[bus];
        }

        bool is_mono(unsigned bus_index);

        void set_stereo_width(unsigned bus_index, float width)
        {
                stereo_width[bus_index] = width;
        }

        float get_stereo_width(unsigned bus_index)
        {
                return stereo_width[bus_index];
        }

        void set_eq(unsigned bus_index, EQBand band, float db_gain)
        {
                assert(band >= 0 && band < NUM_EQ_BANDS);
                eq_level_db[bus_index][band] = db_gain;
        }

        float get_eq(unsigned bus_index, EQBand band) const
        {
                assert(band >= 0 && band < NUM_EQ_BANDS);
                return eq_level_db[bus_index][band];
        }

        float get_limiter_threshold_dbfs() const
        {
                return limiter_threshold_dbfs;
        }

        float get_compressor_threshold_dbfs(unsigned bus_index) const
        {
                return compressor_threshold_dbfs[bus_index];
        }

        void set_limiter_threshold_dbfs(float threshold_dbfs)
        {
                limiter_threshold_dbfs = threshold_dbfs;
        }

        void set_compressor_threshold_dbfs(unsigned bus_index, float threshold_dbfs)
        {
                compressor_threshold_dbfs[bus_index] = threshold_dbfs;
        }

        void set_limiter_enabled(bool enabled)
        {
                limiter_enabled = enabled;
        }

        bool get_limiter_enabled() const
        {
                return limiter_enabled;
        }

        void set_compressor_enabled(unsigned bus_index, bool enabled)
        {
                compressor_enabled[bus_index] = enabled;
        }

        bool get_compressor_enabled(unsigned bus_index) const
        {
                return compressor_enabled[bus_index];
        }

        void set_gain_staging_db(unsigned bus_index, float gain_db)
        {
                std::lock_guard<std::mutex> lock(compressor_mutex);
                level_compressor_enabled[bus_index] = false;
                gain_staging_db[bus_index] = gain_db;
        }

        float get_gain_staging_db(unsigned bus_index) const
        {
                std::lock_guard<std::mutex> lock(compressor_mutex);
                return gain_staging_db[bus_index];
        }

        void set_gain_staging_auto(unsigned bus_index, bool enabled)
        {
                std::lock_guard<std::mutex> lock(compressor_mutex);
                level_compressor_enabled[bus_index] = enabled;
        }

        bool get_gain_staging_auto(unsigned bus_index) const
        {
                std::lock_guard<std::mutex> lock(compressor_mutex);
                return level_compressor_enabled[bus_index];
        }

        void set_final_makeup_gain_db(float gain_db)
        {
                std::lock_guard<std::mutex> lock(compressor_mutex);
                final_makeup_gain_auto = false;
                final_makeup_gain = from_db(gain_db);
        }

        float get_final_makeup_gain_db()
        {
                std::lock_guard<std::mutex> lock(compressor_mutex);
                return to_db(final_makeup_gain);
        }

        void set_final_makeup_gain_auto(bool enabled)
        {
                std::lock_guard<std::mutex> lock(compressor_mutex);
                final_makeup_gain_auto = enabled;
        }

        bool get_final_makeup_gain_auto() const
        {
                std::lock_guard<std::mutex> lock(compressor_mutex);
                return final_makeup_gain_auto;
        }

        void reset_peak(unsigned bus_index);

        struct BusLevel {
                float current_level_dbfs[2];  // Digital peak of last frame, left and right.
                float peak_level_dbfs[2];  // Digital peak with hold, left and right.
                float historic_peak_dbfs;
                float gain_staging_db;
                float compressor_attenuation_db;  // A positive number; 0.0 for no attenuation.
        };

        typedef std::function<void(float level_lufs, float peak_db,
                                   std::vector<BusLevel> bus_levels,
                                   float global_level_lufs, float range_low_lufs, float range_high_lufs,
                                   float final_makeup_gain_db,
                                   float correlation)> audio_level_callback_t;
        void set_audio_level_callback(audio_level_callback_t callback)
        {
                audio_level_callback = callback;
        }

        typedef std::function<void()> state_changed_callback_t;
        void set_state_changed_callback(state_changed_callback_t callback)
        {
                state_changed_callback = callback;
        }

        state_changed_callback_t get_state_changed_callback() const
        {
                return state_changed_callback;
        }

        void trigger_state_changed_callback()
        {
                if (state_changed_callback != nullptr) {
                        state_changed_callback();
                }
        }

        // A combination of all settings for a bus. Useful if you want to get
        // or store them as a whole without bothering to call all of the get_*
        // or set_* functions for that bus.
        struct BusSettings {
                float fader_volume_db;
                bool muted;
                bool locut_enabled;
                float stereo_width;
                float eq_level_db[NUM_EQ_BANDS];
                float gain_staging_db;
                bool level_compressor_enabled;
                float compressor_threshold_dbfs;
                bool compressor_enabled;
        };
        static BusSettings get_default_bus_settings();
        BusSettings get_bus_settings(unsigned bus_index) const;
        void set_bus_settings(unsigned bus_index, const BusSettings &settings);

private:
        struct AudioDevice {
                std::unique_ptr<ResamplingQueue> resampling_queue;
                std::string display_name;
                unsigned capture_frequency = OUTPUT_FREQUENCY;
                // Which channels we consider interesting (ie., are part of some input_mapping).
                std::set<unsigned> interesting_channels;
                bool silenced = false;
        };

        const AudioDevice *find_audio_device(DeviceSpec device_spec) const
        {
                return const_cast<AudioMixer *>(this)->find_audio_device(device_spec);
        }

        AudioDevice *find_audio_device(DeviceSpec device_spec);

        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);
        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);
        void reset_resampler_mutex_held(DeviceSpec device_spec);
        void apply_eq(unsigned bus_index, std::vector<float> *samples_bus);
        void update_meters(const std::vector<float> &samples);
        void add_bus_to_master(unsigned bus_index, const std::vector<float> &samples_bus, std::vector<float> *samples_out);
        void measure_bus_levels(unsigned bus_index, const std::vector<float> &left, const std::vector<float> &right);
        void send_audio_level_callback();
        std::vector<DeviceSpec> get_active_devices() const;
        void set_input_mapping_lock_held(const InputMapping &input_mapping);

        unsigned num_capture_cards, num_ffmpeg_inputs;

        mutable std::timed_mutex audio_mutex;

        ALSAPool alsa_pool;
        AudioDevice video_cards[MAX_VIDEO_CARDS];  // Under audio_mutex.
        AudioDevice alsa_inputs[MAX_ALSA_CARDS];  // Under audio_mutex.
        std::unique_ptr<AudioDevice[]> ffmpeg_inputs;  // Under audio_mutex.

        std::atomic<float> locut_cutoff_hz{120};
        StereoFilter locut[MAX_BUSES];  // Default cutoff 120 Hz, 24 dB/oct.
        std::atomic<bool> locut_enabled[MAX_BUSES];
        StereoFilter eq[MAX_BUSES][NUM_EQ_BANDS];  // The one for EQBand::MID isn't actually used (see comments in apply_eq()).

        // First compressor; takes us up to about -12 dBFS.
        mutable std::mutex compressor_mutex;
        std::unique_ptr<StereoCompressor> level_compressor[MAX_BUSES];  // Under compressor_mutex. Used to set/override gain_staging_db if <level_compressor_enabled>.
        float gain_staging_db[MAX_BUSES];  // Under compressor_mutex.
        float last_gain_staging_db[MAX_BUSES];  // Under compressor_mutex.
        bool level_compressor_enabled[MAX_BUSES];  // Under compressor_mutex.

        static constexpr float ref_level_dbfs = -14.0f;  // Chosen so that we end up around 0 LU in practice.
        static constexpr float ref_level_lufs = -23.0f;  // 0 LU, more or less by definition.

        StereoCompressor limiter;
        std::atomic<float> limiter_threshold_dbfs{ref_level_dbfs + 4.0f};   // 4 dB.
        std::atomic<bool> limiter_enabled{true};
        std::unique_ptr<StereoCompressor> compressor[MAX_BUSES];
        std::atomic<float> compressor_threshold_dbfs[MAX_BUSES];
        std::atomic<bool> compressor_enabled[MAX_BUSES];

        // Note: The values here are not in dB.
        struct PeakHistory {
                float current_level = 0.0f;  // Peak of the last frame.
                float historic_peak = 0.0f;  // Highest peak since last reset; no falloff.
                float current_peak = 0.0f;  // Current peak of the peak meter.
                float last_peak = 0.0f;
                float age_seconds = 0.0f;   // Time since "last_peak" was set.
        };
        PeakHistory peak_history[MAX_BUSES][2];  // Separate for each channel. Under audio_mutex.

        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.
        bool final_makeup_gain_auto = true;  // Under compressor_mutex.

        MappingMode current_mapping_mode;  // Under audio_mutex.
        InputMapping input_mapping;  // Under audio_mutex.
        std::atomic<float> fader_volume_db[MAX_BUSES] {{ 0.0f }};
        std::atomic<bool> mute[MAX_BUSES] {{ false }};
        float last_fader_volume_db[MAX_BUSES] { 0.0f };  // Under audio_mutex.
        std::atomic<float> stereo_width[MAX_BUSES] {{ 0.0f }};  // Default 1.0f (is set in constructor).
        std::atomic<float> eq_level_db[MAX_BUSES][NUM_EQ_BANDS] {{{ 0.0f }}};
        float last_eq_level_db[MAX_BUSES][NUM_EQ_BANDS] {{ 0.0f }};

        audio_level_callback_t audio_level_callback = nullptr;
        state_changed_callback_t state_changed_callback = nullptr;
        mutable std::mutex audio_measure_mutex;
        Ebu_r128_proc r128;  // Under audio_measure_mutex.
        CorrelationMeasurer correlation;  // Under audio_measure_mutex.
        Resampler peak_resampler;  // Under audio_measure_mutex.
        std::atomic<float> peak{0.0f};

        // Metrics.
        std::atomic<double> metric_audio_loudness_short_lufs{0.0 / 0.0};
        std::atomic<double> metric_audio_loudness_integrated_lufs{0.0 / 0.0};
        std::atomic<double> metric_audio_loudness_range_low_lufs{0.0 / 0.0};
        std::atomic<double> metric_audio_loudness_range_high_lufs{0.0 / 0.0};
        std::atomic<double> metric_audio_peak_dbfs{0.0 / 0.0};
        std::atomic<double> metric_audio_final_makeup_gain_db{0.0};
        std::atomic<double> metric_audio_correlation{0.0};

        // These are all gauges corresponding to the elements of BusLevel.
        // In a sense, they'd probably do better as histograms, but that's an
        // awful lot of time series when you have many buses.
        struct BusMetrics {
                std::vector<std::pair<std::string, std::string>> labels;
                std::atomic<double> current_level_dbfs[2]{{0.0/0.0},{0.0/0.0}};
                std::atomic<double> peak_level_dbfs[2]{{0.0/0.0},{0.0/0.0}};
                std::atomic<double> historic_peak_dbfs{0.0/0.0};
                std::atomic<double> gain_staging_db{0.0/0.0};
                std::atomic<double> compressor_attenuation_db{0.0/0.0};
        };
        std::unique_ptr<BusMetrics[]> bus_metrics;  // One for each bus in <input_mapping>.
};

extern AudioMixer *global_audio_mixer;

#endif  // !defined(_AUDIO_MIXER_H)