Julien Bayle

I'm using the magic HOA Library for ambisonic spatialization sound field synthesis.

I'm encoding/decoding 3 or 4 sources at the same time, moving them, etc.

BUT, I have a misunderstanding problem (actually, a real one)

All my sources are normalized. Edited samples, no synthesis in that part.

I move them in my sound field in ambisonic harmonics domains and sometimes, in the end, on speakers, sound is a bit saturated because too high level.

I could use a limiter per speaker but it would break the ambisonic sound field. Obviously, the idea is only to avoid saturation so basically, I'll try to avoid to reach those limiters' threshold. But I don't like to alter dynamic in the end, like that.

I was thinking about compressing my sources before ambisonic encoding (before harmonic domains, I mean) but that wouldn't make sense because each source is encoded separately in order to be placed/located wherever I need in the sound field.

This is an installation. Should I alter gain in real time ? But I don't know the sound field in advance as I want the program to place the source randomly sometimes, on following path etc.

Hi there,
I'm using the magic HOA Library for ambisonic spatialization sound field synthesis.
I'm encoding/decoding 3 or 4 sources at the same time, moving them, etc.
All works ridiculously fine!

All my sources are normalized. Edited samples, no synthesis in that part.
I move them in my sound field in ambisonic harmonics domains and sometimes, in the end, on speakers, sound is a bit saturated because too high level.

Hi there,
I'm using the magic HOA Library for ambisonic spatialization sound field synthesis.
I'm encoding/decoding 3 or 4 sources at the sa

ambisonic-speaker-levels

I have to admit, I don't have any experience in ambisonics but in mixing and distributing sound sources to mono and multichannel environments. My experience does not go any further than the general aspects you find anywhere in the books or anyone's instinct.

From your post, it's not clear where the saturation is happening. 

Inside the computer before or after encoding, in the soundcard, in the amps or the speakers? And this is the most crucial part, you have to go through all the links in the chain, one after another.

Like you said, any dynamic processing like limiting or compressing will affect the "surround"-field. This is why I usually lower gain on all sources evenly to a constant value inside the computer before mixing (in your case before encoding) till I get a comfortable headroom on the mixed (decoded) result, at least 2 dB. 

I'm not afraid loosing dynamic range by lowering the overall gain of the sources. You can raise the gain of the speaker amps later.

Altering the gain in realtime? This is a compositional decision.

You may check also the output options on your soundcard. My lovely RME has 3 output options (-10, +4, hi), which makes a big difference.

I have to admit, I don't have any experience in ambisonics but in mixing and distributing sound sources to mono and multichannel environments. My experience does not go any further than the general aspects you find anywhere in the books or anyone's instinct. 

From your post, it's not clear where the saturation is happening. 
Inside the computer before or after encoding, in the soundcard, in the amps or the speakers? And this is the most crucial part,  you have to go through all the links in the chain, one after another. 

Like you said, any dynamic processing like limiting or compressing will affect the "surround"-field. This is why I usually lower gain on all sources evenly to a constant value inside the computer before mixing (in your case before encoding) till I get a comfortable headroom on the mixed (decoded) result, at least 2 dB. 

I'm not afraid loosing dynamic range by lowering the overall gain of the sources. You can raise the gain of the speaker amps later. 

Altering the gain in realtime? This is a compositional decision. 

You may check also the output options on your soundcard. My lovely RME has 3 output options (-10, +4, hi), which makes a big difference. 

Hi, I think that the normalisation of your sources is the problem. Ambisonic encoding and decoding adds and scales multiple signals. And you are doing this with 4 sources, all with a possible maximum of 0dB which can easily cause overloads. A quick fix would be to reduce the amplitude of your sources (before the decoder) by 6 or 10 dB or so (depends how many sources you are mixing)

and then increase the volume later. With the high bit resolution we are all working at these days you shouldn't notice any increase in noise. No need for compression, unless you like it!

Hi, I think that the normalisation of your sources is the problem. Ambisonic encoding and decoding adds and scales multiple signals. And you are doing this with 4 sources, all with a possible maximum of 0dB which can easily cause overloads. A quick fix would be to reduce the amplitude of your sources (before the decoder) by 6 or 10 dB or so (depends how many sources you are mixing)
and then increase the volume later. With the high bit resolution we are all working at these days you shouldn't notice any increase in noise. No need for compression, unless you like it!

And there is another reason, why normalization isn’t a good idea. It’s the lack of volume control (he?what?). Usually normalization looks for the highest peek in your audio file. Now just imagine, your file is very quiet but it has a very short peek somewhere (this might be a peek of a duration of 1 sample), then normalization wouldn’t increase the volume at all.

I agree with justin.
And there is another reason, why normalization isn’t a good idea. It’s the lack of volume control (he?what?). Usually normalization looks for the highest peek in your audio file. Now just imagine, your file is very quiet but it has a very short peek somewhere (this might be a peek of a duration of 1 sample), then normalization wouldn’t increase the volume at all.

Thanks for normalization bits, but this is ok on that part.

Basically, the main problem is not (only) volume control but a 

I discussed with friend at research lab here (

) and this is the question of two paradigms racing together.

Stereo (obviously) doesn't mean anything in a complex positionning source system and this is my need of reducing my problem to (sometimes) kind of 2 sides (~ stereo like) in my ambisonic global system.

It seems I found a way by using the most far speakers for making a kind of stereo, changing the contribution of each speakers in the system in real time (thanks to Max framework!)

Thanks for normalization bits, but this is ok on that part.
Basically, the main problem is not (only) volume control but a duality problem.
I discussed with friend at research lab here (LMA-CNRS) and this is the question of two paradigms racing together.
Stereo (obviously) doesn't mean anything in a complex positionning source system and this is my need of reducing my problem to (sometimes) kind of 2 sides (~ stereo like) in my ambisonic global system.
It seems I found a way by using the most far speakers for making a kind of stereo, changing the contribution of each speakers in the system in real time (thanks to Max framework!)


ambisonic & speaker levels