1
Fork 0
mirror of https://github.com/RGBCube/serenity synced 2025-05-31 16:18:12 +00:00
serenity/Userland/Libraries/LibAudio/Buffer.h
David Isaksson fa4255bcf1 LibAudio: Refactor out linear_to_log function and add its inverse
The conversion from a linear scale (how we think about audio) to a
logarithmic scale (how audio actually works) will be useful for other
operations, so let's extract it to its own utility function. Its inverse
will also allow reversible operations to be written more easily.
2021-11-08 16:29:25 -08:00

241 lines
6.9 KiB
C++

/*
* Copyright (c) 2018-2020, Andreas Kling <kling@serenityos.org>
* Copyright (c) 2021, kleines Filmröllchen <malu.bertsch@gmail.com>
*
* SPDX-License-Identifier: BSD-2-Clause
*/
#pragma once
#include <AK/ByteBuffer.h>
#include <AK/Math.h>
#include <AK/MemoryStream.h>
#include <AK/String.h>
#include <AK/Types.h>
#include <AK/Vector.h>
#include <LibCore/AnonymousBuffer.h>
#include <string.h>
namespace Audio {
using namespace AK::Exponentials;
// Constants for logarithmic volume. See Frame::operator*
// Corresponds to 60dB
constexpr double DYNAMIC_RANGE = 1000;
constexpr double VOLUME_A = 1 / DYNAMIC_RANGE;
double const VOLUME_B = log(DYNAMIC_RANGE);
// A single sample in an audio buffer.
// Values are floating point, and should range from -1.0 to +1.0
struct Frame {
constexpr Frame() = default;
// For mono
constexpr Frame(double left)
: left(left)
, right(left)
{
}
// For stereo
constexpr Frame(double left, double right)
: left(left)
, right(right)
{
}
void clip()
{
if (left > 1)
left = 1;
else if (left < -1)
left = -1;
if (right > 1)
right = 1;
else if (right < -1)
right = -1;
}
// Logarithmic scaling, as audio should ALWAYS do.
// Reference: https://www.dr-lex.be/info-stuff/volumecontrols.html
// We use the curve `factor = a * exp(b * change)`,
// where change is the input fraction we want to change by,
// a = 1/1000, b = ln(1000) = 6.908 and factor is the multiplier used.
// The value 1000 represents the dynamic range in sound pressure, which corresponds to 60 dB(A).
// This is a good dynamic range because it can represent all loudness values from
// 30 dB(A) (barely hearable with background noise)
// to 90 dB(A) (almost too loud to hear and about the reasonable limit of actual sound equipment).
//
// Format ranges:
// - Linear: 0.0 to 1.0
// - Logarithmic: 0.0 to 1.0
ALWAYS_INLINE double linear_to_log(double const change)
{
// TODO: Add linear slope around 0
return VOLUME_A * exp(VOLUME_B * change);
}
ALWAYS_INLINE double log_to_linear(double const val)
{
// TODO: Add linear slope around 0
return log(val / VOLUME_A) / VOLUME_B;
}
ALWAYS_INLINE Frame& log_multiply(double const change)
{
double factor = linear_to_log(change);
left *= factor;
right *= factor;
return *this;
}
ALWAYS_INLINE Frame log_multiplied(double const volume_change) const
{
Frame new_frame { left, right };
new_frame.log_multiply(volume_change);
return new_frame;
}
ALWAYS_INLINE Frame& log_pan(double const pan)
{
left *= linear_to_log(min(pan * -1 + 1.0, 1.0));
right *= linear_to_log(min(pan + 1.0, 1.0));
return *this;
}
ALWAYS_INLINE Frame log_pan(double const pan) const
{
Frame new_frame { left, right };
new_frame.log_pan(pan);
return new_frame;
}
constexpr Frame& operator*=(double const mult)
{
left *= mult;
right *= mult;
return *this;
}
constexpr Frame operator*(double const mult)
{
return { left * mult, right * mult };
}
constexpr Frame& operator+=(Frame const& other)
{
left += other.left;
right += other.right;
return *this;
}
constexpr Frame& operator+=(double other)
{
left += other;
right += other;
return *this;
}
constexpr Frame operator+(Frame const& other)
{
return { left + other.left, right + other.right };
}
double left { 0 };
double right { 0 };
};
// Supported PCM sample formats.
enum PcmSampleFormat : u8 {
Uint8,
Int16,
Int24,
Int32,
Float32,
Float64,
};
// Most of the read code only cares about how many bits to read or write
u16 pcm_bits_per_sample(PcmSampleFormat format);
String sample_format_name(PcmSampleFormat format);
// Small helper to resample from one playback rate to another
// This isn't really "smart", in that we just insert (or drop) samples.
// Should do better...
template<typename SampleType>
class ResampleHelper {
public:
ResampleHelper(u32 source, u32 target);
// To be used as follows:
// while the resampler doesn't need a new sample, read_sample(current) and store the resulting samples.
// as long as the resampler needs a new sample, process_sample(current)
// Stores a new sample
void process_sample(SampleType sample_l, SampleType sample_r);
// Assigns the given sample to its correct value and returns false if there is a new sample required
bool read_sample(SampleType& next_l, SampleType& next_r);
Vector<SampleType> resample(Vector<SampleType> to_resample);
void reset();
u32 source() const { return m_source; }
u32 target() const { return m_target; }
private:
const u32 m_source;
const u32 m_target;
u32 m_current_ratio { 0 };
SampleType m_last_sample_l;
SampleType m_last_sample_r;
};
// A buffer of audio samples.
class Buffer : public RefCounted<Buffer> {
public:
static RefPtr<Buffer> from_pcm_data(ReadonlyBytes data, int num_channels, PcmSampleFormat sample_format);
static RefPtr<Buffer> from_pcm_stream(InputMemoryStream& stream, int num_channels, PcmSampleFormat sample_format, int num_samples);
static NonnullRefPtr<Buffer> create_with_samples(Vector<Frame>&& samples)
{
return adopt_ref(*new Buffer(move(samples)));
}
static NonnullRefPtr<Buffer> create_with_anonymous_buffer(Core::AnonymousBuffer buffer, i32 buffer_id, int sample_count)
{
return adopt_ref(*new Buffer(move(buffer), buffer_id, sample_count));
}
const Frame* samples() const { return (const Frame*)data(); }
int sample_count() const { return m_sample_count; }
const void* data() const { return m_buffer.data<void>(); }
int size_in_bytes() const { return m_sample_count * (int)sizeof(Frame); }
int id() const { return m_id; }
const Core::AnonymousBuffer& anonymous_buffer() const { return m_buffer; }
private:
explicit Buffer(const Vector<Frame> samples)
: m_buffer(Core::AnonymousBuffer::create_with_size(samples.size() * sizeof(Frame)).release_value())
, m_id(allocate_id())
, m_sample_count(samples.size())
{
memcpy(m_buffer.data<void>(), samples.data(), samples.size() * sizeof(Frame));
}
explicit Buffer(Core::AnonymousBuffer buffer, i32 buffer_id, int sample_count)
: m_buffer(move(buffer))
, m_id(buffer_id)
, m_sample_count(sample_count)
{
}
static i32 allocate_id();
Core::AnonymousBuffer m_buffer;
const i32 m_id;
const int m_sample_count;
};
// This only works for double resamplers, and therefore cannot be part of the class
NonnullRefPtr<Buffer> resample_buffer(ResampleHelper<double>& resampler, Buffer const& to_resample);
}