[LAU] Ambisonics for newbies - ambdec

[Fons Adriaensen]

On Tue, Mar 24, 2020 at 05:27:16PM +0100, Michael Jarosch wrote:

> I would have made my own configuration, but I couldn't find the button
> "new" in the configuration dialog, which was described in the manual of
> the 0.4.2 version I've found on Fons' page. Is that handy tool given up
> in later versions or am I too dump to handle it right?

The manual is indeed out of date. The new option was dropped because
manually entering a preset wasn't very useful in the first place.
Ambdec preset are just text files, so all you need is a text editor,
but normally they would be machine-generated.

> By the way, @Fons, if he is watching that thread: I can't understand,
> why counting in an audio application starts at "0". And I can't
> understand why it seems better not keeping the ambisonics' nomenclature
> for an ambisonics application… There may be reasons for this - maybe 3rd
> order ambisonics complicates things, here? Nevertheless: Is it possible
> to name the inputs at least by a number AND the corresponding ambisonics
> channel name? That would be great!

All current (that is, since AMB was picked up by the surround video world)
AMB software uses the 'Ambix' standard. The main reason for this is that
the original conventions (knows as Furse-Malham) don't scale well to higher
orders. In the Ambix standard, channels names are based on the Ambisonic
Channel Number (ACN) which is defined mathematically, and this is turn
means that starting from zero is the 'natural' thing to do.

> I made sure, that the outputs of ambdec are correctly connected, which
> is easy to test with the internal test signal. But there are a few
> things that let me think, I'm wrong with the inputs, here:
> 
> * If the "mono panner" is in a neutral position (Azimuth: 0, Elevation:
> 0), I would expect to get a signal from exactly in front of me. Instead
> I get an output signal of LF and LB which is 20dB higher than RF and RB.

That looks normal. Ambisonics doesn't work by using only the nearest
speakers - even if you pan a sound in the actual direction of a speaker
the others will not be silent. This is very apparent in low order AMB,
less so for higher orders.

> I can reach the signals position that corresponds to the one I expected
> if I set Elevation to 90. But then, changing azimuth does nothing
> (instead of  - my expectation - circling around me while changing it).

Elevation +90 means 'above your head'. Azimuth doesn't matter here -
in the same way as longitude is undefined when you are at the North 
Pole. 
 
> * Elevation: The Ardour page says, if there's no vertical dimension, I
> shouldn't connect "z" - if I do that, ambdec changes it's reaction to
> the "mono panners" values - but still it's not very intuitive to work
> with. Or to be more precise: It's even getting worse…

If the decoder expects Z (ACN 2), then it should be connected. 
The square decoder doesn't expect or use Z, it shouldn't matter
if it is connected or not (assuming the other connections are
correct).

> Conclusion: I'm sure, ambisonics is no rocket science, but the tools I'm
> depending on getting ambisonics to work don't make it that easy…

There is a lot of info on AMB on the web which is just plain wrong.
And the FuMa to Ambix conversion of course adds to the confusion. 

I should really retire the original AMB plugins set, and replace
it by and Ambix compliant one. Now that I'm more less in 'house
arrest' that may happen soon (I've been using a 3rd order Ambix
set for years, it just was never released). 

Ciao,

-- 
FA