Skip to content

BR_Loudness: make loudness calc work for Ambisonics and more surround formats #2018

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
kronihias wants to merge 1 commit into
base: master
Choose a base branch
from
Open

BR_Loudness: make loudness calc work for Ambisonics and more surround formats #2018

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
102 changes: 88 additions & 14 deletions Breeder/BR_Loudness.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1043,15 +1043,21 @@ unsigned WINAPI BR_LoudnessObject::AnalyzeData (void* loudnessObject)
int mode = integratedOnly ? EBUR128_MODE_I : EBUR128_MODE_M | EBUR128_MODE_S | EBUR128_MODE_I | EBUR128_MODE_LRA;
if (!integratedOnly && doTruePeak)
mode |= EBUR128_MODE_TRUE_PEAK;
loudnessState = ebur128_init((size_t)data.channels, (size_t)data.samplerate, mode);

// relevantChannels: how many channels to request from the audio accessor and pass to
// ebur128. For ambisonics only W (ch0) is measured, so we request 1 channel — this
// avoids ingesting all HOA channels (~16× speedup for TOA, ~9× for SOA, etc.).
// For all other modes we need every channel, so relevantChannels == data.channels.
int relevantChannels = data.channels;

// Ignore channels according to channel mode. Note: we can't partially request samples, i.e. channel mode is mono, but take is stereo...asking for
// 1 channel only won't work. We must always request the real channel count even though reaper interleaves active channels starting from 0
if (data.channelMode > 1)
{
loudnessState = ebur128_init((size_t)data.channels, (size_t)data.samplerate, mode);
// Mono channel modes
if (data.channelMode <= 66)
{
{
if (doDualMonoMode) {
// Actually there's a check in ebur128_set_channel() for dual mono mode:
// if ebur128_init() is called with channels > 1 than dual mono mode is disabled.
Expand Down Expand Up @@ -1081,18 +1087,86 @@ unsigned WINAPI BR_LoudnessObject::AnalyzeData (void* loudnessObject)
// Dual mono mode for mono takes
else if (doDualMonoMode && _this->m_take && data.channels == 1)
{
loudnessState = ebur128_init((size_t)data.channels, (size_t)data.samplerate, mode);
ebur128_set_channel(loudnessState, 0, EBUR128_DUAL_MONO);
for (int i = 1; i <= data.channels; ++i)
ebur128_set_channel(loudnessState, i, EBUR128_UNUSED);
}
// Normal channel mode
// Normal channel mode — auto-detect format from channel count
else
{
ebur128_set_channel(loudnessState, 0, EBUR128_LEFT);
ebur128_set_channel(loudnessState, 1, EBUR128_RIGHT);
ebur128_set_channel(loudnessState, 2, EBUR128_CENTER);
ebur128_set_channel(loudnessState, 3, EBUR128_LEFT_SURROUND);
ebur128_set_channel(loudnessState, 4, EBUR128_RIGHT_SURROUND);
// Ambisonic channel counts are perfect squares: (N+1)^2
// FOA=4, SOA=9, TOA=16, 4th=25, ... 10th=121 — all detected correctly.
// Standard surround formats (6, 8, 10, 12, 14) are not perfect squares — no collision.
// 16ch is TOA; 9.1.6 would also be 16ch but TOA takes priority here.
// Use lround() instead of (int) cast to avoid floating-point rounding on large values.
const int sqrtCh = (int)lround(sqrt((double)data.channels));
const bool isAmbisonics = (sqrtCh * sqrtCh == data.channels && data.channels >= 4);

if (isAmbisonics)
{
// HOA / ambisonics: measure only the W channel (0th-order, omnidirectional).
// Per Peters & Epain (AES 154th 2023), applying BS.1770 directly to W is as
// accurate as full loudspeaker rendering, with no order-specific correction needed.
// Request only 1 channel from the accessor — no need to ingest all HOA channels.
relevantChannels = 1;
loudnessState = ebur128_init(1, (size_t)data.samplerate, mode);
ebur128_set_channel(loudnessState, 0, EBUR128_LEFT); // W channel, G=1.00
}
else if (data.channels == 14)
{
// 9.1.4: L, R, C, LFE, Lss, Rss, Ls, Rs, Lb, Rb, Ltf, Rtf, Ltb, Rtb
//
// Dolby/SMPTE ordering for extended theatrical bed + 4 height channels.
// Lss/Rss = wide side at ~±60°, Ls/Rs = surround at ~±110° — both ear-level in
// the 60°–120° azimuth range → G=1.41 per BS.1770-5 Annex 3 Table 4.
// Lb/Rb = rear back at ~±150° (|θ|>120°, ear-level) → G=1.00.
// All height channels (elevation ≥30°) → G=1.00 regardless of azimuth.
loudnessState = ebur128_init((size_t)data.channels, (size_t)data.samplerate, mode);
ebur128_set_channel(loudnessState, 0, EBUR128_LEFT); // L
ebur128_set_channel(loudnessState, 1, EBUR128_RIGHT); // R
ebur128_set_channel(loudnessState, 2, EBUR128_CENTER); // C
ebur128_set_channel(loudnessState, 3, EBUR128_UNUSED); // LFE — excluded
ebur128_set_channel(loudnessState, 4, EBUR128_LEFT_SURROUND); // Lss ~±60°, G=1.41
ebur128_set_channel(loudnessState, 5, EBUR128_RIGHT_SURROUND); // Rss ~±60°, G=1.41
ebur128_set_channel(loudnessState, 6, EBUR128_LEFT_SURROUND); // Ls ~±110°, G=1.41
ebur128_set_channel(loudnessState, 7, EBUR128_RIGHT_SURROUND); // Rs ~±110°, G=1.41
ebur128_set_channel(loudnessState, 8, EBUR128_LEFT); // Lb ~±150°, G=1.00
ebur128_set_channel(loudnessState, 9, EBUR128_RIGHT); // Rb ~±150°, G=1.00
ebur128_set_channel(loudnessState, 10, EBUR128_LEFT); // Ltf height, G=1.00
ebur128_set_channel(loudnessState, 11, EBUR128_RIGHT); // Rtf height, G=1.00
ebur128_set_channel(loudnessState, 12, EBUR128_LEFT); // Ltb height, G=1.00
ebur128_set_channel(loudnessState, 13, EBUR128_RIGHT); // Rtb height, G=1.00
}
else
{
// Dolby/SMPTE channel ordering up to 7.1.4 (12ch):
// L, R, C, LFE, Ls, Rs, Lb, Rb, Ltf, Rtf, Ltb, Rtb
//
// Per BS.1770-5 Annex 3 Table 4 (position-dependent weightings):
// G=1.41 : ear-level side surrounds at azimuth 60°-120° (Ls, Rs at ~±110°)
// G=1.00 : all other channels — front (L,R,C), rear-backs (Lb,Rb at ~±135°),
// and ALL height channels (any elevation >= 30°)
//
// 10ch is treated as 7.1.2 (L,R,C,LFE,Ls,Rs,Lb,Rb,Ltf,Rtf); 9.1 has the same
// channel count and cannot be distinguished without file metadata.
// libebur128 only has LEFT_SURROUND/RIGHT_SURROUND for G=1.41;
// for G=1.00 channels we reuse LEFT/RIGHT/CENTER as appropriate.
loudnessState = ebur128_init((size_t)data.channels, (size_t)data.samplerate, mode);
if (data.channels >= 1) ebur128_set_channel(loudnessState, 0, EBUR128_LEFT);
if (data.channels >= 2) ebur128_set_channel(loudnessState, 1, EBUR128_RIGHT);
if (data.channels >= 3) ebur128_set_channel(loudnessState, 2, EBUR128_CENTER);
if (data.channels >= 4) ebur128_set_channel(loudnessState, 3, EBUR128_UNUSED); // LFE — excluded
if (data.channels >= 5) ebur128_set_channel(loudnessState, 4, EBUR128_LEFT_SURROUND); // Ls ~±110°, G=1.41
if (data.channels >= 6) ebur128_set_channel(loudnessState, 5, EBUR128_RIGHT_SURROUND);// Rs ~±110°, G=1.41
if (data.channels >= 7) ebur128_set_channel(loudnessState, 6, EBUR128_LEFT); // Lb ~±135°, G=1.00
if (data.channels >= 8) ebur128_set_channel(loudnessState, 7, EBUR128_RIGHT); // Rb ~±135°, G=1.00
if (data.channels >= 9) ebur128_set_channel(loudnessState, 8, EBUR128_LEFT); // Ltf height, G=1.00
if (data.channels >= 10) ebur128_set_channel(loudnessState, 9, EBUR128_RIGHT); // Rtf height, G=1.00
if (data.channels >= 11) ebur128_set_channel(loudnessState, 10, EBUR128_LEFT); // Ltb height, G=1.00
if (data.channels >= 12) ebur128_set_channel(loudnessState, 11, EBUR128_RIGHT); // Rtb height, G=1.00
// Channels beyond 12 (other than 14ch/9.1.4) stay at libebur128 defaults (UNUSED)
}
}

bool doShortTerm = true;
Expand All @@ -1115,7 +1189,7 @@ unsigned WINAPI BR_LoudnessObject::AnalyzeData (void* loudnessObject)
const double audioLength = data.audioEnd - data.audioStart;

int sampleCount = data.samplerate / refreshRateInHz;
int bufSz = sampleCount * data.channels;
int bufSz = sampleCount * relevantChannels;
double currentTime = data.audioStart;
bool momentaryFilled = true;
int processedSamples = 0;
Expand All @@ -1129,14 +1203,14 @@ unsigned WINAPI BR_LoudnessObject::AnalyzeData (void* loudnessObject)
if (remainingTime < bufferTime + numeric_limits<double>::epsilon())
{
sampleCount = static_cast<int>(data.samplerate * remainingTime);
bufSz = sampleCount * data.channels;
bufSz = sampleCount * relevantChannels;
skipIntervals = true;
}

// Get new 200 ms (or 10 ms in high precision mode) of samples
// GetAudioAccessorSamples() stops writing to the buffer once it reaches the item's end, everything from that point to sampleCount is garbage
std::vector<double> samples(bufSz);
GetAudioAccessorSamples(data.audio, data.samplerate, data.channels, currentTime, sampleCount, &samples[0]);
GetAudioAccessorSamples(data.audio, data.samplerate, relevantChannels, currentTime, sampleCount, &samples[0]);

// Correct for volume and pan/volume envelopes
int currentChannel = 1;
Expand Down Expand Up @@ -1166,10 +1240,10 @@ unsigned WINAPI BR_LoudnessObject::AnalyzeData (void* loudnessObject)

sample *= adjust;

if (++currentChannel > data.channels)
if (++currentChannel > relevantChannels)
currentChannel = 1;

if (currentChannel + 1 > data.channels)
if (currentChannel + 1 > relevantChannels)
sampleTime = sampleTime + sampleTimeLen;
}

Expand Down Expand Up @@ -1233,7 +1307,7 @@ unsigned WINAPI BR_LoudnessObject::AnalyzeData (void* loudnessObject)
ebur128_loudness_range(loudnessState, &range);
if (doTruePeak)
{
for (int i = 0; i < data.channels; ++i)
for (int i = 0; i < relevantChannels; ++i)
{
double channelTruePeak, channelTruePeakPos;
ebur128_true_peak(loudnessState, i, &channelTruePeak, &channelTruePeakPos);
Expand Down