replace generic xyzgain by diffusedecoder with detection of 2d/3d

HoerTech-gGmbH · Nov 15, 2024 · 54d87b0 · 54d87b0
1 parent c50ae39
commit 54d87b0
Show file tree

Hide file tree

Showing 22 changed files with 56 additions and 24 deletions.
diff --git a/libtascar/include/speakerarray.h b/libtascar/include/speakerarray.h
@@ -99,6 +99,7 @@ namespace TASCAR {
     spk_array_t(const spk_array_t&);
 
   public:
+    enum dffusedecoder_t { basic, maxre, inphase };
     double get_rmax() const { return rmax; };
     double get_rmin() const { return rmin; };
     class didx_t {
@@ -116,7 +117,7 @@ namespace TASCAR {
   private:
     double rmax;
     double rmin;
-    float xyzgain;
+    dffusedecoder_t diffusedecoder_enum = maxre;
     std::string onload;
     std::string onunload;
     std::vector<didx_t> didx;
@@ -212,8 +213,6 @@ namespace TASCAR {
     // lowpass filters for subwoofer (24 dB/Oct):
     std::vector<TASCAR::biquadf_t> flt_lowp1;
     std::vector<TASCAR::biquadf_t> flt_lowp2;
-    // allpass filters for transition phase matching, broad band speakers:
-    //std::vector<TASCAR::biquad_t> flt_allp;
     std::vector<std::vector<float>> subweight;
     std::vector<std::string>
         convolution_ir; //< file name of impulse response for convolution

diff --git a/libtascar/src/speakerarray.cc b/libtascar/src/speakerarray.cc
@@ -79,7 +79,7 @@ spk_array_cfg_t::~spk_array_cfg_t()
 spk_array_t::spk_array_t(tsccfg::node_t e, bool use_parent_xml,
                          const std::string& elementname_, bool allow_empty)
     : spk_array_cfg_t(e, use_parent_xml), elayout(e_layout), rmax(0), rmin(0),
-      xyzgain(1.0), elementname(elementname_), mean_rotation(0)
+      elementname(elementname_), mean_rotation(0)
 {
   clear();
   {
@@ -119,12 +119,23 @@ spk_array_t::spk_array_t(tsccfg::node_t e, bool use_parent_xml,
   }
   for(auto& sn : tsccfg::node_get_children(e_layout, elementname))
     emplace_back(sn);
-  elayout.GET_ATTRIBUTE(xyzgain, "", "XYZ-gain for FOA decoding");
   elayout.GET_ATTRIBUTE(name, "", "Name of layout, for documentation only");
   elayout.GET_ATTRIBUTE(onload, "",
                         "system command to be executed when layout is loaded");
   elayout.GET_ATTRIBUTE(
       onunload, "", "system command to be executed when layout is unloaded");
+  std::string diffusedecoder = "maxre";
+  elayout.GET_ATTRIBUTE(diffusedecoder, "basic|maxre|inphase",
+                        "Diffuse-decoder method");
+  if(diffusedecoder == "maxre")
+    diffusedecoder_enum = maxre;
+  else if(diffusedecoder == "basic")
+    diffusedecoder_enum = basic;
+  else if(diffusedecoder == "inphase")
+    diffusedecoder_enum = inphase;
+  else
+    throw TASCAR::ErrMsg("Invalid diffuse decoder type \"" + diffusedecoder +
+                         "\". Valid values are basic, maxre or inphase.");
   //
   if(empty() && (!allow_empty))
     throw TASCAR::ErrMsg("Invalid " + elementname_ + " array (no " +
@@ -141,13 +152,35 @@ spk_array_t::spk_array_t(tsccfg::node_t e, bool use_parent_xml,
         rmin = r;
     }
   }
+  bool isflat = true;
   std::complex<double> p_xy(0);
   for(uint32_t k = 0; k < size(); k++) {
+    if(fabsf(operator[](k).unitvector.z) > 1e-6f)
+      isflat = false;
     operator[](k).spkgain *= operator[](k).norm() / rmax;
     operator[](k).dr = rmax - operator[](k).norm();
     p_xy += std::exp(-(double)k * TASCAR_2PI * i / (double)(size())) *
             (operator[](k).unitvector.x + i * operator[](k).unitvector.y);
   }
+  // calculate xyzgain according to Daniel (2001), table 3.10:
+  double xyzgain = 1.0;
+  switch(diffusedecoder_enum) {
+  case basic:
+    xyzgain = 1.0;
+    break;
+  case maxre:
+    if(isflat)
+      xyzgain = sqrt(0.5);
+    else
+      xyzgain = 0.577;
+    break;
+  case inphase:
+    if(isflat)
+      xyzgain = 0.5;
+    else
+      xyzgain = 1.0/3.0;
+    break;
+  }
   mean_rotation = std::arg(p_xy);
   didx.resize(size());
   for(uint32_t k = 0; k < size(); ++k) {

diff --git a/manual/manual.tex b/manual/manual.tex
@@ -1181,7 +1181,7 @@ \subsection{The {\tt <receiver .../>} element}\label{sec:receiver}\index{receive
 first order Ambisonics signal of the $l$-th diffuse sound field; $L$ is the
 number of all diffuse sound fields, including diffuse reverberation inputs.
 
-For all loudspeaker-based receiver types, $D$ is a first order Ambisonics decoder matrix, with optional loudspeaker density compensation and decorrelation filters. To get a $\max rE$-decoder, set the \indattr{xyzgain} attribute to 0.707 for regular horizontal loudspeaker layouts and to 0.577 for regular 3D loudspeaker layouts.  For Ambisonics based receiver types, $D$ is a diagonal matrix.  By default, the decoded output of the first order Ambisonics rendering is de-correlated using FIR all-pass filters to achieve diffuse sound fields and avoid coloration artifacts (see \indattr{decorr} and \indattr{decorr\_length} for details).
+For all loudspeaker-based receiver types, $D$ is a first order Ambisonics decoder matrix, with optional loudspeaker density compensation and decorrelation filters. By default, a $\max rE$-decoder is used. The order gain $g_{xyz}$ is set according to Table 3.10 of \citet{Daniel2001}. A loudspeaker layout is assumed to reproduce in 3D when at least one loudspeaker has non-zero elevation. For Ambisonics based receiver types, $D$ is a diagonal matrix.  By default, the decoded output of the first order Ambisonics rendering is de-correlated using FIR all-pass filters to achieve diffuse sound fields and avoid coloration artifacts (see \indattr{decorr} and \indattr{decorr\_length} for details).
 
 Figure \ref{fig:renderer} presents the typical connections in \tascar{}
 and may help to visualize the role of the receiver.

diff --git a/test/36ch.spk b/test/36ch.spk
@@ -1,5 +1,5 @@
 <?xml version="1.0" encoding="UTF-8"?>
-<layout xyzgain="0.70710678118654757274" >
+<layout>
   <speaker az="0"/>
   <speaker az="10"/>
   <speaker az="20"/>

diff --git a/test/4ch.spk b/test/4ch.spk
@@ -1,5 +1,5 @@
 <?xml version="1.0"?>
-<layout xyzgain="0.70710678118654757274" decorr="false" densitycorr="false">
+<layout decorr="false" densitycorr="false">
   <speaker az="0"/>
   <speaker az="90"/>
   <speaker az="180"/>

diff --git a/test/4ch_decorr.spk b/test/4ch_decorr.spk
@@ -1,5 +1,5 @@
 <?xml version="1.0"?>
-<layout xyzgain="0.70710678118654757274" diffusegain="6.45">
+<layout diffusegain="6.45">
   <speaker az="0"/>
   <speaker az="90"/>
   <speaker az="180"/>

diff --git a/test/4ch_nodecorr.spk b/test/4ch_nodecorr.spk
@@ -1,5 +1,5 @@
 <?xml version="1.0"?>
-<layout xyzgain="0.70710678118654757274" decorr="false" diffusegain="-2.89">
+<layout decorr="false" diffusegain="-2.89">
   <speaker az="0"/>
   <speaker az="90"/>
   <speaker az="180"/>

diff --git a/test/8ch.spk b/test/8ch.spk
@@ -1,5 +1,5 @@
 <?xml version="1.0" encoding="UTF-8"?>
-<layout xyzgain="0.70710678118654757274" >
+<layout>
   <speaker az="0" connect="system:playback_1"/>
   <speaker az="45" connect="system:playback_2"/>
   <speaker az="90" connect="system:playback_3"/>

diff --git a/test/8ch_calib.spk b/test/8ch_calib.spk
@@ -1,5 +1,5 @@
 <?xml version="1.0" encoding="UTF-8"?>
-<layout xyzgain="0.70710678118654757274" caliblevel="70">
+<layout caliblevel="70">
   <speaker az="0"/>
   <speaker az="45"/>
   <speaker az="90"/>

diff --git a/test/8ch_diffuse.spk b/test/8ch_diffuse.spk
@@ -1,5 +1,5 @@
 <?xml version="1.0" encoding="UTF-8"?>
-<layout xyzgain="0.70710678118654757274" decorr="false" densitycorr="false">
+<layout decorr="false" densitycorr="false">
   <speaker az="0"/>
   <speaker az="45"/>
   <speaker az="90"/>

diff --git a/test/8ch_diffuse_zero.spk b/test/8ch_diffuse_zero.spk
@@ -1,5 +1,5 @@
 <?xml version="1.0" encoding="UTF-8"?>
-<layout xyzgain="0.70710678118654757274" decorr="true" decorr_length="0" densitycorr="false">
+<layout decorr="true" decorr_length="0" densitycorr="false">
   <speaker az="0"/>
   <speaker az="45"/>
   <speaker az="90"/>

diff --git a/test/8ch_nodecorr.spk b/test/8ch_nodecorr.spk
@@ -1,5 +1,5 @@
 <?xml version="1.0" encoding="UTF-8"?>
-<layout xyzgain="1" decorr="false">
+<layout diffusedecoder="basic" decorr="false">
   <speaker az="0" connect="system:playback_1"/>
   <speaker az="45" connect="system:playback_2"/>
   <speaker az="90" connect="system:playback_3"/>

diff --git a/test/8ch_sub.spk b/test/8ch_sub.spk
@@ -1,5 +1,5 @@
 <?xml version="1.0" encoding="UTF-8"?>
-<layout xyzgain="0.70710678118654757274" fcsub="80">
+<layout fcsub="80">
   <speaker az="0" connect="system:playback_1"/>
   <speaker az="45" connect="system:playback_2"/>
   <speaker az="90" connect="system:playback_3"/>

diff --git a/test/expected_snd_diffuse_maxre.wav b/test/expected_snd_diffuse_maxre.wav
diff --git a/test/irreg.spk b/test/irreg.spk
@@ -1,5 +1,5 @@
 <?xml version="1.0" encoding="UTF-8"?>
-<layout xyzgain="0.70710678118654757274" >
+<layout >
   <speaker az="0" connect="system:playback_1"/>
   <speaker az="37" connect="system:playback_2"/>
   <speaker az="90" connect="system:playback_3"/>

diff --git a/test/lidhan_3d.spk b/test/lidhan_3d.spk
@@ -1,5 +1,5 @@
 <?xml version="1.0" encoding="UTF-8" standalone="no"?>
-<layout name="glab" xyzgain="0.577" decorr="false">
+<layout name="glab" decorr="false">
 	<speaker az="11.25"/>
 	<speaker az="33.75"/>
 	<speaker az="56.25"/>

diff --git a/test/lidhan_3d45.spk b/test/lidhan_3d45.spk
@@ -1,5 +1,5 @@
 <?xml version="1.0" encoding="UTF-8" standalone="no"?>
-<layout xyzgain="0.577" decorr="false">
+<layout decorr="false">
   <speaker az="11.25"  el="0.0" label="B.1"/>
   <speaker az="33.75"  el="0.0" label="B.2"/>
   <speaker az="56.25"  el="0.0" label="B.3"/>

diff --git a/test/lidhan_3d45_sub.spk b/test/lidhan_3d45_sub.spk
@@ -1,5 +1,5 @@
 <?xml version="1.0" encoding="UTF-8" standalone="no"?>
-<layout xyzgain="0.577" decorr="false">
+<layout decorr="false">
   <speaker az="11.25"  el="0.0" label="B.1"/>
   <speaker az="33.75"  el="0.0" label="B.2"/>
   <speaker az="56.25"  el="0.0" label="B.3"/>

diff --git a/test/test_snd_diffuse_basic.tsc b/test/test_snd_diffuse_basic.tsc
@@ -10,7 +10,7 @@
     </diffuse>
     <receiver type="nsp">
       <!-- basic decoding is achieved with g_m = 1.0 (Daniel, 2001) -->
-      <layout xyzgain="1.0" decorr="false">
+      <layout diffusedecoder="basic" decorr="false">
         <speaker az="0" r="1"/>
       </layout>
     </receiver>

diff --git a/test/test_snd_diffuse_inphase.tsc b/test/test_snd_diffuse_inphase.tsc
@@ -10,7 +10,7 @@
     </diffuse>
     <receiver type="nsp">
       <!-- inphase decoding is achieved with g_m = 0.5 (Daniel, 2001) -->
-      <layout xyzgain="0.5" decorr="false">
+      <layout diffusedecoder="inphase" decorr="false">
         <speaker az="0" r="1"/>
       </layout>
     </receiver>

diff --git a/test/test_snd_diffuse_maxre.tsc b/test/test_snd_diffuse_maxre.tsc
@@ -10,7 +10,7 @@
     </diffuse>
     <receiver type="nsp">
       <!-- maxre decoding is achieved with g_m = 0.7071 (Daniel, 2001) -->
-      <layout xyzgain="0.7071" decorr="false">
+      <layout diffusedecoder="maxre" decorr="false">
         <speaker az="0" r="1"/>
       </layout>
     </receiver>

diff --git a/test/vbap6.spk b/test/vbap6.spk
@@ -1,5 +1,5 @@
 <?xml version="1.0"?>
-<layout>
+<layout diffusedecoder="basic">
   <speaker az="0"/>
   <speaker az="90"/>
   <speaker az="-90"/>