The Blog of Bruce

Each year, I get my 2nd year Spatial Audio students to produce a spatial audio production, which they can then attach 360 pictures to, and upload to YouTube (360 pictures are used, as I want the focus to be the audio).

How to do this was quite well documented soon after YouTube enabled the feature using a combination of FFMPEG and Google’s Spatial Media Tools on github.

YouTube’s spatial audio feature has had a rocky time over the years breaking for prolonged periods and then coming back, but although it’s currently working, I’ve found that the instructions and tools don’t work as they once did, so here’s my quick guide to what’s changed and how I currently get it to work (on an Apple Mac, but I’m assuming similar issues for Windows users, too – I’ll try and get round to testing on a Windows machine soon).

Assuming we have a video in .MOV format: 360PicVideo (it MUST be a .MOV)
Assuming we have a 4-channel 16 bit wave file in AmbiX format: AmbiX_1o.wav

I’ve found that using any channel mapping (i.e. -channel_layout 4.0) which is what I’ve always used in the past, now results in the channels being swapped in the resulting video file. The only way I can get the audio in the video to match the original audio (tested using both VLC and AVFoundation filters to read the video in Reaper) is to omit this command entirely – e.g.

ffmpeg -i AmbiX_1o.wav 
       -i 360PicVideo.mov 
       -map 0:a -map 1:v 
       -c:a copy -c:v copy 
       OutputVideoFinal.mov

Next, the video needs to be tagged using the Google Spatial Media Tools. However, I’ve found that the GUI.py version won’t tag the video as equirectangular anymore (using latest GITHUB version as of 29th June 2026). In order to get round this, we need to use the command line version instead – e.g.

python3 spatialmedia -i --spatial-audio OutputVideoFinal.mov OutputVideoFinal_injected.mov

I then carry out two checks to make sure this has worked.

1. Run python3 spatialmedia OutputVideoFinal_injected.mov to inspect the metadata
2. Load the video into Reaper and compare the audio with the wave file used as the input (this very quickly shows up channel swapping issues).

Output for 1. should look something like (it’s the true for Spherical and Stitched that’s missing when using the GUI version for me):

	Track 0
		Ambisonic Type: periphonic
		Contains Head-Locked Stereo: False
		Ambisonic Order: 1
		Ambisonic Channel Ordering: ACN
		Ambisonic Normalization: SN3D
		Number of Channels: 4
		Channel Map: [0, 1, 2, 3]
	Track 1
		Spherical = true
		Stitched = true
		StitchingSoftware = Spherical Metadata Tool
		ProjectionType = equirectangular

The correct, and incorrect, channel mappings are shown in Reaper below, compared to the original audio file. Note channel 4 in the audio and correct video has been incorrectly mapped to channel 3 in the incorrect video (using -channel_layout 4.0):

I still need to do tests on whether this is a problem for YouTube, or if it fixes the mapping – I only came across the issue when I was troubleshooting the spatial media metadata problem!

Once figuring this out, my students are back to successfully creating YouTube videos with Spatial Audio – This one is based in and around Derby Cathedral (thanks to the Derby Cathedral staff for letting us come in a measure Ambisonic Impulse Responses for this!):

How to install FFMPEG and Python3 on a Mac: https://www.brucewiggins.co.uk/?p=781

Last year, Bo-Erik Sandholm (https://www.ohti.xyz) asked me if it was possible to generate spaced stereo microphone outputs from 3rd order Ambisonic recordings.

It is possible, and the details of the start of this work can be found in:

Girijavallabhan, A.. and Wiggins, B., (2024). A Subjective Comparison Of Virtual Stereo Microphone Techniques For Rendering Ambisonics. Reproduced Sound 2024, Vol. 46 Pt. (3). Available at:
https://www.ioa.org.uk/catalogue/paper/subjective-comparison-virtual-stereo-microphone-techniques-rendering-ambisonics

I’ll add more accessible detail on the process soon (the paper is out today, and I present the work on Thursday at Reproduced Sound 2024 in Bristol), but below is a downloadable version of a Reaper project that incorporates the filters for a variety of standard space mic arrays!

Spaced Microphone from Ambisonics Reaper Project (37149 downloads )

As a continuation of my previous post (UHJ to B-Format/Quad Transcoder/Decoder) I’ve now made a plug-in that transcodes/encodes from 1st Order B-Format (AmbiX or FuMa) to UHJ. As before, it’s based on Gerzon’s original paper (http://decoy.iki.fi/dsound/ambisonic/motherlode/source/11730.pdf) and also includes what Gerzon calls a PQ meter which is a simple visualisation that can be used to look at matrix systems (as detailed in Pictures of 2-Channel Directional Reproduction Systems, also by Gerzon).

Here’s an animation showing it working with a simple panned sine wave (it gets much more interesting for real recordings/reverberant material!)

Both the encoder and the decoder can be downloaded via the link below:

Reaper JSFX UHJ Encoder/Decoder/Transcoder (65796 downloads )

Back in 1983, Michael Gerzon presented a 2, 3 or 4 channel encoding system called UHJ where the 1st two channels are stereo compatible Left and Right signals. This was, in the 4-channel incarnation, a lossless transcoding of the W, X, Y and Z channels of Ambisonics. If only the first two channels are kept, then horizontal Ambisonic surround is still able to be extracted, albeit with less isolation between the resulting channels (much like Dolby Stereo used in cinemas from 1976).

I won’t bore you with the equations and maths of the system, but Gerzon’s original paper can be found at http://decoy.iki.fi/dsound/ambisonic/motherlode/source/11730.pdf

To cut a long story short, I’ve made a UHJ decoder plug-in based on Gerzon’s paper above, as there isn’t one currently available that works in the same way:

Although UHJ encoded material works best (a discography of UHJ releases can be found at http://www.surrounddiscography.com/uhjdisc/uhjhtm.htm), any two-channel recording can be fed into the decoder and either a square/rectangle decoding can be produced, or the B-Format for decoding using another plug-in. Do note that the ‘shelf-filters’ recommended for UHJ are different from the ones used for ‘standard’ Ambisonic B-Format decoding.

There are other UHJ implementations available, but this plug-in has a few features not available elsewhere:

• Allpass filter based phase shift networks. This is similar to the techniques used in the original hardware, and can sound more natural that FIR filter alternatives.
• Correct shelf filtering based on Gerzon’s recommendations.
• Speaker distance compensation (dial in, and compensate for, the distance the speakers are from the sweet spot.

The decoded polar patterns (coming from a 2-channel UHJ encoded source) can be seen below. UHJ does well considering it’s only coming from UHJ stereo!

The plugin, which is a JSFX for Reaper, can be downloaded below. To install, unzip the files into the Effects folder in the Reaper Resource Path. A guide for doing this can be found at: https://reaperblog.net/2015/06/quick-tip-how-to-install-js-plugins/

Reaper JSFX UHJ Encoder/Decoder/Transcoder (65796 downloads )

The restrictions imposed by the pandemic thwarted the continuation of ‘in-person’ listening tests into Ambisonic Order and transparency over head-tracked headphones in 2020/21, which is an ongoing project using Very High Order Ambisonics (up to 35th) and hardware head-tracking. It raised the question, “How do we maintain our essential test features using remote systems?”. Many people had access to webcams, laptops and headphones due to remote working, so we sought to leverage this, with the results being WHAM!

The WHAM (Webcam Head-tracked Ambisonics) website (www.brucewiggins.co.uk/WHAM) takes the approach of using a webcam to measure head rotation via the browser to provide the dynamic cues necessary for a convincing room auralisation using Binaural Room Impulse Response (BRIR) data. Visitors to the site can experience up to 17th Order Ambisonics over headphones which incorporate asymmetric binaural filtering of captured room responses that reacts to the rotation of the head.

This approach is only able to model a single source position, but with the associated room response captured to a much higher order than is currently possible using Ambisonic microphones (which max out at 5th order) and allowing for accurate dynamic head movement cues to be processed in real-time in the browser.

The project has added extra functions to the JS Ambisonics library, to enable asymmetrical filtering (left/right symmetry is a common method for increasing the efficiency of Ambisonics to Binaural processing, but isn’t valid if room responses are to be used). The forked JSAmbisonics can be found at: https://github.com/DrWig/JSAmbisonics

If you have a webcam and headphones, do give it a try!

• Dring, M. and Wiggins, B., 2021. WHAM: TO ASYMMETRY AND BEYOND!. Reproduced Sound 2021, Vol. 43 Pt. (2). Available at: <https://www.ioa.org.uk/catalogue/paper/wham-asymmetry-and-beyond>.
• Dring, M., Wiggins, B. (2020) WHAM: Webcam Head-tracked AMbisonics.  Reproduced Sound 2020 – Proceedings of the Institute of Acoustics, Vol. 42. Pt. 3 2020, UK
• Dring, M., Wiggins, B. (2019)  The Transparency of Binaural Auralisation Using Very High Order Circular Harmonics.  Reproduced Sound 2019 – Proceedings of the Institute of Acoustics, p165-173, Vol. 41. Pt. 3 2019, Bristol, UK 

After some time, I decided to both update my Ambisonic tools to support the AmbiX standard (now widely used for immersive audio, 360 videos and VR) and rework my speaker array decoders using JSFX to make them a little more powerful and with a better workflow for me to quickly create new ones! My plug-ins implement near-field compensation, distance filtering and a few other things that other Ambisonic tools don’t, hence why I decided to bite the bullet and update them all. My 3D Ambisonic Reverb (AmbiXFreeverb) is also available in the bundle.

An example of one of the decoders is shown below. Dual-band decoding is available, and you’re able to tweak the overall gain of the low frequency vs high frequency decodes too (as energy optimised ones tend to be a little quieter). They’ll default to all sensible options, so if you don’t know what some of the controls do, just don’t change them 😉

Both the platform-dependent VSTs (for Mac and PC) and the platform-independent decoders are available for download on the WigWare page, or below:

• Mac VST AmbiX Ambisonic Panner/Reverb VST Plugins (62634 downloads )
• PC VST AmbiX Ambisonic Panner/Reverb VSTs (61352 downloads )
• Reaper JSFX AmbiX Decoders (61461 downloads )

The VST plugins can be placed in the usual folders for Mac and PC, and a guide to installing JSFX plugins can be found at: https://reaperblog.net/2015/06/quick-tip-how-to-install-js-plugins/

The folder JSFX plugins go in is (Mac)
~/Library/Application Support/REAPER/Effects/WigWare/Amb Dec AmbiX/
or (PC)
%USERPROFILE%\AppData\Roaming\REAPER\Effects\WigWare\Amb Dec AmbiX\

You’ll need to create the WigWare folders. You can get to this quickly from Reaper by selecting Options->Show Resource Path in Explorer/Finder…

I’ve been using Angelo Farina‘s excellent Aurora tools with my students for a number of years now to help with Log Sine Sweep measurements, but they often struggle to get the modules working with Audacity. So, here are the instructions I gave them, this year, to help with that. I’ve shared as it’ll likely be useful for anyone that needs to install them!

Update Jan 2024: Audacity 2.4.1 doesn’t work past Catalina OS on the Mac. Angelo Farina recommends using PlayOnMac to run the Windows version, but that crashed for me. PortingKit seems to work out of the box (except I had to use the latest WINE when asked, which was WS11WineCX21.2.0….the default didn’t seem to work for Aurora), using the default options for everything else. If you have a Mac, install https://www.portingkit.com/ and then follow the Windows instructions below!

1. Download the latest version of the Aurora Plugins for Audacity, noting the version number of Audacity they are for:

http://www.angelofarina.it/Public/Aurora-for-Audacity/

The highest version for Mac is 2.4.1 and for Windows is 2.4.1, too (as of Feb 2021)

Here’s a direct link to the latest version (as of Feb 2021): http://www.angelofarina.it/Public/Aurora-for-Audacity/2.4.1/

When you’ve downloaded, extract the files.  If you’re using a Mac, the download is the whole Audacity package with Aurora included.  Just unzip and drag it into applications – jump to point 5 to get them enabled!

Let’s look at the PC version – your unzipped folder should look something like the figure below:

2. Now, install the corresponding version of Audacity (older versions than the latest can be found at https://www.fosshub.com/Audacity-old.html ).  I’m going to put it in a non-default folder to help keep track of versions!
   

3. Now, we need to overwrite files in the Audacity install folder with the Aurora + ASIO ones we downloaded and unzipped earlier:

4. You should get something like this warning that Windows needs to overwrite files!

5. Now run Audacity as usual.  We need to enable the Aurora modules/plugins so we can use them.  Find the Generate Menu and select Add/Remove Plugins:

6. On a PC, you should see an Aurora plugin at the top (there are others, as well!).  Click ‘Select All’ and then ‘Enable’ to turn them all on. If you’re using a Mac, you’ll probably see all your VST plugins in here too.  I’ll leave it up to you whether you enable them all…..order by Path (click it!) to have them grouped more sensibly allowing you to just enable the things you want!

Above is the view on a PC.  Below, is the view on a Mac after I’ve clicked ‘Path’ and found the Aurora modules:

7. Now, if you go back to the Generate menu, you’ll see the top entry is Aurora Sine Sweep Generator – this means it’s all working 🙂

My work and that of a colleague (Dr Adam Hill) is being featured as an impact case study for the upcoming Research Excellence Framework 2021 (REF2021). My work was previously featured as a case study in the last REF, too, as it happens (in REF2014). Here’s a video of a presentation we made to the rest of the University to let them know what we get up to. Further details can also be found at https://www.derby.ac.uk/research/showcase/audio-engineering-research/

New server……I’m alive!

This year’s Sounds in Space was our most successful yet!  We ran a two day event, with some really excellent demonstrations and talks.  The videos that were streamed live are now available on YouTube, with links to each talk and the corresponding slides available on the Sounds in Space 2018 page .  Attached to this post are also some photos from the day!

Read more for the pics, below!

Continue reading “Sounds in Space 2018 Videos, PowerPoints and Pics!”

A couple of weeks ago, I presented at the 4th International Conference on Spatial Audio in Graz, Austria (and it was a great event, too!)

The paper I presented was:  Analysis of Binaural Cue Matching Using Ambisonics to Binaural Decoding Techniques.

In the presentation, I demonstrated the performance of the algorithm by looking at inter-aural time and level differences between the ears of a centrally seated listener at different order (1st to 35th order), and a number of people asked if they could get access to the animations I presented at the event.  You can find them linked on a simple webpage below.  If the GIFs don’t sync, just go back and forward to the page again.  Once the GIFs have all been cached by your browser (1st time you load it), opening the page again should start them all at the same time!

Note, these probably won’t play as well on mobile devices (too many GIFs!)

Wiggins ICSA 2017 Animations

Wiggins ICSA 2017 Animations – half speed