Skip to content
/ kmh_ls Public

Decoders for the Klangkupolen at KMH, Lilla Salen

5 stars 0 forks Branches Tags Activity
Star
Notifications You must be signed in to change notification settings

friskgit/kmh_ls

BranchesTags

Repository files navigation

General information

Scroll down to Klangkupolen, Lilla Salen for specifics

Ambisonics in KMH studios and Lilla salen

About ambisonics

The general idea behind using ambisonics for difusing sounds in a space is to, in the first stage, encode an audio signal to a representation of a sound field - a B-format signal. This B-format signal is idependent of whatever speaker array the signal will later be diffused on. In the second stage the B-format signal is decoded to the layout of a particular speaker array. In between these two stages a number of transformations may be applied, but the sound source specification may conveniently be specified in the encoder.

In principle, any kind of encoder may be combined with any kind of decoder provided that the same channel order is used. Among the encoders (VST) that work with the decoders presented here are Ambix and IEM Plugin suite. Using these makes it simple to automate panning in a DAW such as Reaper.

About the compiled decoders

All decoders use ACN channel order (see Component order) and SN3D normalization (see Normalisation) and are dual-band, max-rE/rV.

  • A guide to the file names

    For efficiency file names are composed of the following parts

    1. The space (for Lilla Salen this is KMHLS_AE)
    2. Speaker configuration (Full or Normal where the former in 108/118 and 114 means the center speaker is included, and in Lilla Salen it refers to the 45.4 setup)
    3. Order (e.g. 7H7P)
    4. Redundant speaker configuration (full, but only in certain cases for Lilla Salen)
    5. Finally, a number depicting the type of decoder according to the following codes:

    • (1) AllRad mixed order (4/3)

      Mixed order ambisonics decoder.

    • (2) Decoding with the pseudo-inverse, even-energy

      Pinv decoders in 1st, 3rd order with even-energy for inversion variants.

      The higher order decoders in pseudo-inverse do not work well, so do not try those above 3rd order.

    • (3) Decoding with the pseudo-inverse, mode-matching

      Pinv decoders in 1st, 3rd order, mode-matching for inversion variants. This has been removed from the repository.

    • (4) Same as (2) but with a blend coefficient for inversion variants (energy limited/mode matching 50%)

      1st and 3rd and 6th order.

      The higher order decoders in pseudo-inverse do not work well, so do not try those above 3rd order - though sixth order has provecd useful.

    • (5) Spherical slepian decoders

      SSF decoders in 1st, 3rd and 6th order (6th order is here only as VST, but I can compile the other formats if needed.)

    • (6) All-round decoders

      AllRAD decoders in 1st, 3rd, 5th and 7th order.

  • Ouput formats

    In the directory of each decoder there are four subdirectories:

    • ambix: contains the decoder matrix for the ambix VST-plugin for certain settings.
    • bin: contains all other decoders
    • doc: documentation, a bare minimal (but a more comprehensive documentation can be built from the source, check the individual Makefiles for each compilere)
    • src: the faust sources for the decoders

    The binaries are prepared in six different formats:

    1. a jack application (start from commandline: $ KMHLS_Dome_4h3v_normal_1)
    2. a double clickable CoreAudio application
    3. a 'max' directory with Max/MSP objects (as package with help files to be put in ~/Documents/Max 7/Packages)
    4. a 'pd' direcory (bang left most input to get parameter names)
    5. a 'sc' directory with SuperCollider classes (put all .sc and .scx in ~/Library/Application Support/SuperCollider/Extensions , the supernova files are of course optional)
    6. an 'ambix' directory with Ambix config file

    • Install on your system

      You may simply move the decoders you want to use to the right locations on your system. They are compiled and ready to be used in the repository. As a convenience one may run the corresponding install scripts in the Makefile for the decoders to install them in your default directory:

      1. Open a terminal (such as Terminal.app)
      2. At the prompt type $ cd and drag one of the directories, say 'KMHLS_AE_normal_6' of your downloaded repository onto the terminal and press enter.
      3. Enter $ make maxinstall at the prompt and press enter.

      The last step can be made for scinstall, vstinstall and pdinstall

      Note that to create a Max/MSP compatible package you may have to run maxinstall and have the reslevant dependencies resolved. Else, contact me!

  • Component order

    The component order for the input to the decoder follows the ACN (Ambisonics Channel Number) standard according to the table below (for third order):

          0      
        1 2 3    
      4 5 6 7 8  
    9 10 11 12 13 13 15

    Or: W Y Z X V T R S U Q O M K L N P

  • Normalisation

    The decoder is using SN3D normalisation. With SN3D normalisation no component will exceed the peak value of the 0th order component.

  • Compiling the decoders

    This is only if you wish to tweak the settings in any way. All scripts needed for the compilation is in a speparate GitHub repository that may be found here:

    https://github.com/friskgit/kmh_decoders

    In order to compile the decoders there are a number of dependencies that need to be resolved:

    • ADT (Ambisoncs Decoder Toolkit)

    • Faust

    • VST SDK

    • Pure Data

    • Max MSP SDK

    • Links to Supercollider classes

    • Running Makefile scripts for ADT

      Run the following for each function that needs to be compiled for lilla salen (108_all and 114_all for the studios):

      $ make ls_all function=6

      Note that for fuction 2 & 4 only orders up to 3 are usable.

      The decoding matrices end up in a directory above your current directory named 'decoders'. Move into that directory with the following command:

      $ cd ../decoders

      With the decoding matrices calculated the binary decoders for all orders can be compiled by

      $ make -k target=all all

      The target can be either of [all,sc,max,pd,vst]. Following this step the binaries can be installed (to install only one target, use that as the argument for 'target'.:

      $ make -k target=install all

Klangkupolen, Lilla Salen

Specific info for Klangkupolen

Output

In Klangkupolen the channel order from the mixer is channel 1-29 for the regular dome. The lower ring starts at channel 33-48 and the subs are at 49-52. This is the reason that in the decoder the names for the sixteen floor speakers start at 33.

The layout is ordered clockwise starting at speaker one.

Speaker positions

All distances and heigths are measured from 1.25m height from floor.

ndx azimuth elevation distance radius
1 -34.689614 12.910417 4.61 4.5789148
2 -13.383763 12.910417 4.61 4.5577825
3 10.440725 12.910417 4.61 4.5635592
4 32.117788 12.910417 4.61 4.5386017
5 55.741675 12.910417 4.61 4.4464181
6 78.207673 12.910417 4.61 4.4723909
7 101.49442 12.910417 4.61 4.4461719
8 124.85167 12.910417 4.61 4.5148007
9 147.91193 12.910417 4.61 4.5536496
10 169.17789 12.910417 4.61 4.6069350
11 -167.82013 12.910417 4.61 4.7018379
12 -145.63454 12.910417 4.61 4.6822723
13 -123.78400 12.910417 4.61 4.7368687
14 -102.64182 12.910417 4.61 4.6743181
15 -79.887731 12.910417 4.61 4.7101696
16 -57.926139 12.910417 4.61 4.6119481
17 -22.349553 34.696822 4.882 4.0761992
18 22.843958 34.696822 4.882 3.7761827
19 69.013292 34.696822 4.882 4.0485676
20 115.56544 34.696822 4.882 4.1015729
21 158.89992 34.696822 4.882 3.9444645
22 -158.89763 34.696822 4.882 3.8051807
23 -114.65354 34.696822 4.882 3.6199724
24 -68.170128 34.696822 4.882 3.7380476
25 -45. 69.185799 5.317 1.7225121
26 45. 69.185799 5.317 1.7225121
27 135. 69.185799 5.317 1.7225121
28 -135. 69.185799 5.317 1.7225121
29 0. 90. 5.576 0
33 -24.036689 -3.089583 4.61 4.5789148
34 -1.471519 -3.089583 4.61 4.5577825
35 21.279257 -3.089583 4.61 4.5635592
36 43.929732 -3.089583 4.61 4.5386017
37 66.974674 -3.089583 4.61 4.4464181
38 89.851047 -3.089583 4.61 4.4723909
39 113.17305 -3.089583 4.61 4.4461719
40 136.3818 -3.089583 4.61 4.5148007
41 158.56486 -3.089583 4.61 4.5536496
42 179.81087 -3.089583 4.61 4.6069350
43 -156.31914 -3.089583 4.61 4.7018379
44 -134.54175 -3.089583 4.61 4.6822723
45 -112.85873 -3.089583 4.61 4.7368687
46 -92.07073 -3.089583 4.61 4.6743181
47 -68.510687 -3.089583 4.61 4.7101696
48 -46.945343 -3.089583 4.61 4.6119481

Spat

  • 29 speakers

    A setup file for the current speaker arrangments in Lilla Salen (29 speakers) according to Speaker positions for use with the Ircam spat package.

    The following file may be used for a spat.viewer @numspeakers 29 @showlistener 1 @viewpoint top @width 1200 @height 600:

    speakers aed
0 12.910417 4.61 -26.34 12.910417 4.61 -58.7 12.910417 4.61 -106 12.910417 4.61 -129.35 12.910417 4.61 129.35 12.910417 4.61 106 12.910417 4.61 58.7 12.910417 4.61 45 12.910417 4.61 -45 12.910417 4.61 -135 12.910417 4.61 135 12.910417 4.61 0 12.910417 4.61 hline 12.910417 4.61

  • 45 speakers

    A setup file for the current speaker arrangments in Lilla Salen (45 speakers) according to Speaker positions for use with the Ircam spat package.

    The following file may be used for a spat.viewer @numspeakers 29 @showlistener 1 @viewpoint top @width 1200 @height 600:

    speakers aed
0 12.910417 4.61 -26.34 12.910417 4.61 -58.7 12.910417 4.61 -106 12.910417 4.61 -129.35 12.910417 4.61 129.35 12.910417 4.61 106 12.910417 4.61 58.7 12.910417 4.61 45 12.910417 4.61 -45 12.910417 4.61 -135 12.910417 4.61 135 12.910417 4.61 0 12.910417 4.61 hline 12.910417 4.61

Panoramix

  • 29 speakers

    A setup file for the current speaker arrangments in Lilla Salen (29 speakers) according to Speaker positions for use with the Ircam panoramix package.

    img

    The following file may be loaded in a HOA bus track (See fig:hoabus) in panormaix by selecting the text below, clicking the spekers... button and click paste in the upper left corner of the newly opened window.

    /speaker/number 13
/speaker/correction/delay off
/speaker/correction/gain off
/speakers/aed 24.6 0 3.26 -26.34 0 3.3 -58.7 0 3.35 -106 0 3.27 -129.35 0 3.364 129.35 0 3.376 106 0 3.262 58.7 0 3.296 45 18 3.02 -45 18 3.02 -135 18 3.05 135 18 3.02 0 90 1.6
/speaker/1/delay 0
/speaker/2/delay 0
/speaker/3/delay 0
/speaker/4/delay 0
/speaker/5/delay 0
/speaker/6/delay 0
/speaker/7/delay 0
/speaker/8/delay 0
/speaker/9/delay 0
/speaker/10/delay 0
/speaker/11/delay 0
/speaker/12/delay 0
/speaker/13/delay 0

Zirkonium

Two setup files for Zirkonium are available in the corresponding directory. There is one for normal setup (29 speakers) and one for full setup (45 speakers).

Convolution

It is possible to do work in Lilla Salen at a distance with only a stereosystem and a pair of headpohnes using a BRIR (Binaural Room, Impulse Response) of lilla salen. The recording of the BRIR was made in 2018 and before we moved the speakers a few degrees. I will make a new one shortly that reflects the current speaker setup. The process is to record a sine sweep from each speaker with a dummy head in the sweet spot that generates 29 binaural signals. At the stage of convolution each input is convoluted with each of the 29 signals and summed up producing one binaural stereo signal.

To get this working there are a couple of options of which I now describe sing a VST host (such as Reaper):

  1. Get Matthias Kronlacher's mcfx suite of plugins and install
  2. Install the preset files in ~/Library/Application Support/mcfx (if the directory doesn't exist, create it). The preset files is to be found in the binaural directory above.
  3. Now you should be able to load the plugin in Reaper (see fig:convol below). You need to make sure that the track you load the plugin in has at least 29 channels: It will take 29 channels input and 2 channels out.
  4. Once the plugin is loaded and you have set up the track you can load the preset (see fig:mcfx_window)

img

img

At this point any sound you send to the track will be convolved with each of the convolution signals and, if you send a sound to channel 1, it will appear to come from the speaker number one in lilla salen and so forth. The plugin here should be inserted at the very end of your mix where you have 29 discrete channels ready to be sent to Klangkupolen.

There is an example file in the binaural directory that demonstrates the effect. It is a simple snare drum circling the lower 16 speakers. It should be listened to in headphones.

About

Decoders for the Klangkupolen at KMH, Lilla Salen

Resources

Readme
Activity

Stars

5 stars

Watchers

2 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

No packages published

Languages