Twice

I'm currently working on a dissertation project and keep hitting walls. Any advice/pointers/guiding would be much appreciated. My apologies if it seems like i'm being lazy, but i've spent so long changing ideas/approaches that i've started to lose sight/mind of what I actually need to do and how.

I want to record an environment ambisonically using the Soundfield microphone; this records the W, X, Y and Z axis and I am left with four separate sound files (one for each axis). My next intention is also to record video in 360º to then be played back on a Virtual Reality device. When the user is wearing the VR device and turns their head, the video will rotate and look around the environment (this is achieved by using a mobile device and its internal gyroscope). I want this to also happen with the Ambisonic recordings. However, my peers have told me that it is explicitly hard to rotate ambisonic recordings unless they are virtual also (individual recordings of sounds)

To rotate the recordings, my plan was to use Arduino with a gyroscope (mounted on the VR headset), the gyroscope data would then control the rotation of the recordings, allowing for playback through a loudspeaker array.

So I guess what i'm asking is, is it possible to rotate recordings from one microphone rather than having to use individual virtual recordings? If so, what is the best approach?

Literally ANY pointers, guiding, error correcting, obvious mistakes... are all much appreciated

I'm currently working on a dissertation project and keep hitting walls. Any advice/pointers/guiding would be much appreciated. My apologies if it seems like i'm being lazy, but i've spent so long changing ideas/approaches that i've started to lose sight/mind of what I actually need to do and how. 

To rotate the recordings, my plan was to use Arduino with a gyroscope (mounted on the VR headset), the gyroscope data would then control the rotation of the recordings, allowing for playback through a loudspeaker array.
So I guess what i'm asking is, is it possible to rotate recordings from one microphone rather than having to use individual virtual recordings? If so, what is the best approach?

rotating-ambisonic-recordings-from-a-soundfield-microphone

The W, X, Y, Z channels are the B-format (1st order Ambisonics) and fortunately, the rotation of the sound field isn't very complicated in this case ! You can have a look at this topic 

 (the patches that perform the rotation are at the end of the discussion).

Ps : perhaps the numbering and the normalization of your channels differ.

Hi Tyson,
The W, X, Y, Z channels are the B-format (1st order Ambisonics) and fortunately, the rotation of the sound field isn't very complicated in this case  ! You can have a look at this topic HoaLibrary (the patches that perform the rotation are at the end of the discussion).
Ps : perhaps the numbering and the normalization of your channels differ. 

Thankyou for the pointer Pierre! What a great library...and with a tutorial folder explaining it all as well, fantastic!

Just to confirm; So I am correct in presuming that the W, X, Y, Z recordings from the Soundfield microphone can be rotated the same as if I was to record sounds individually and map them to the same axis?

Thankyou for the pointer Pierre! What a great library...and with a tutorial folder explaining it all as well, fantastic!
Just to confirm; So I am correct in presuming that the W, X, Y, Z recordings from the Soundfield microphone can be rotated the same as if I was to record sounds individually and map them to the same axis?


Yes, in theory if you encode a set of sound and you rotate the sound field, the result is exactly the same as if you record a sound field with this sounds and you rotate it. But in practice there are few things you got to take into account :

- With a soundfield recording, you'll have the room effects (reverberations) and you're limited to 1st order.

- With individual recordings, you can avoid the room effects (or not, depending on the distance with the source, etc.) but you can encode to high orders.

- High order rotation becomes complicated in 3D.

- The synthesis of room effects sounds less natural.

- The simulation of the distance sounds less natural with encoding than with a soundfield recording.

In short, soundfield recording should sound more natural.

Yes, in theory if you encode a set of sound and you rotate the sound field, the result is exactly the same as if you record a sound field with this sounds and you rotate it. But in practice there are few things you got to take into account :
- With a soundfield recording, you'll have the room effects (reverberations) and you're limited to 1st order.
- With individual recordings, you can avoid the room effects (or not, depending on the distance with the source, etc.)  but you can encode to high orders.
- High order rotation becomes complicated in 3D.
- The synthesis of room effects sounds less natural.
- The simulation of the distance sounds less natural with encoding than with a soundfield recording.
In short, soundfield recording should sound more natural.


Thanks for your pointers. Just another question; if the user 'looks up' in the virtual reality, the sound will need to reflect the change. For this to be achieved, will I need to use a 3D system/3D array of loudspeakers, as I presume elevation differences can't be achieved on a 2D/horizontal system?

Thanks for your pointers. Just another question; if the user 'looks up' in the virtual reality, the sound will need to reflect the change. For  this to be achieved, will I need to use a 3D system/3D array of loudspeakers, as I presume elevation differences can't be achieved on a 2D/horizontal system?


Even in a 2D system this will have some effect. For example: Imagine that there is a sound coming from directly above (vertically) and with no reflections of any kind from any other direction. Depending on the decoder, this would either end up not being present in the decoded signals at all, or it would appear as a omnipresent signal (same signals in all speakers). When you move your head up or down, this cause a tumbling of the B-format sound field. Relative to your hear this sound source will start being located more to the front (looking up) or to the back (looking down). This will be reflected in how it gets decoded to the various speakers, and hence have an audible effect.

As for pointers to further readings, me and Joseph Anderson wrote a paper on Ambisonic Toolkit for Reaper in 2014. It can be found here:

This paper has references to several other papers on ambisonics as well. For more references, a bit of searching will lead you to several publications on e.g. HoaLibrary, and the ICST ambisonics externals. The sursound list is where ambisonics nerds hang out, and is a tremendously useful resource:

https://mail.music.vt.edu/mailman/listinfo/sursound

The wikipedia article on ambisonics in English is the joint efforts of many of the leading experts in the field, and apart from being informative, also have references to many of the core articles and papers on the theory of ambisonics:

Even in a 2D system this will have some effect. For example: Imagine that there is a sound coming from directly above (vertically) and with no reflections of any kind from any other direction. Depending on the decoder, this would either end up not being present in the decoded signals at all, or it would appear as a omnipresent signal (same signals in all speakers). When you move your head up or down, this cause a tumbling of the B-format sound field. Relative to your hear this sound source will start being located more to the front  (looking up) or to the back (looking down). This will be reflected in how it gets decoded to the various speakers, and hence have an audible effect.

https://en.wikipedia.org/wiki/Ambisonics


Thank you Trond! I'll have a search through those links, I did have a brief search to find publications on HOA but i'll search deeper!

Rotating Ambisonic recordings from a soundfield microphone.