[LAU] ambisonics: fons ambdec and muse.demon.co.uk ambidec

[Jörn Nettingsmeier]

On 01/07/2011 10:55 AM, sh0099 at gmx.de wrote:
>
> Jörn Nettingsmeier schrieb:

>> why would you want to do that? ambdec comes with a number of example
>> configurations, and iirc, they cover all the speaker layouts that are
>> covered on richard's page, plus a few more.
>
> my problem was that i realised that there are not so many ambisonics
> order 2 (9channel) presets with ambdec.
> and my files are 9channel files.
> do i remember right that i can not decode a 9channel file in a 1st order
> etc. decoder?

i guess you are confusing things here. there are a number of 2nd order 
decoders in ambdec. disregarding your source files for the moment, what 
is your speaker layout? is it capable of reproducing 2nd order?
for that, it would have to be at least a 5.0, better yet a hexagon.
for both rigs, ambdec ships example configurations.

as for the input channel count: you won't need 9 channels for horizontal 
only, just WXYUV. if you want to do with-height, you will need at least 
10 speakers (12 is more practical, because then you can use a regular 
dodecahedron).

>> or do you want to create a custom layout and have amb*i*dec compute a
>> configuration for you?
>
> actually i was thinking about it.

if you do that, check out bruce wiggins' paper on optimizing for 
irregular layouts. but it's nowhere near a recipe, you'll have to do 
some hefty number crunching to arrive at anything useful.

Wiggins, Bruce: "The Generation of Panning Laws for Irregular Speaker 
Layouts using Heuristic Methods", AES 31st International Conference, 
London 2007

>> if so, conversion between polar and cartesian is not too hard... if
>> your listening position is at the origin of both coordinate systems,
>> then (off the top of my head), you should get something like:
>>
>> azimuth = arctan (y/x)
>> elevation = arctan (z/(sqrt(x^2 + y^2))
>> r = sqrt (x^2 + y^2 + z^2)
> thanks i will dubble check with fons answer :-)

to my great relief, it looks like they are congruent :)
of course, when you implement it, you have to do something about the 
singularities at 90 and -90 degrees elevation.

>> i'm pretty sure that the custom layout matrix you will get out of
>> amb*i*dec will yield worse results than ambdec, because it does not do
>> proper dual-band decoding.
>>
>> out of curiosity, what layout are you looking for?
>
> i will start tomorrow to set up a ambisonic setup for a presentation and
> i have two days to do so. because i have to mix different speakertyps i
> have to look a little bit around to know what will fit.

ouch. that is tricky. watch out: in the LF band, you have to have 
matching phase responses, otherwise the speaker will work against one 
another. if at all possible, get the horizontal ring right, with all 
matching speakers. for height, you can get away with using a different 
model. try the following: create a stereo pair from your two different 
speaker models (i hope it's only two, otherwise it's a pretty lost 
case). match the levels carefully, and then listen to some stereo 
material you know well, with lots of spatial information in the LF band 
(an orchestra recording would do). play with the polarity and (if you 
can) with delay, until you get the best LF localisation result. use the 
same polarity and delay for the rig. that way, you have hopefully 
aligned the LF phase responses.
or easier but more expensive: get a dual-fft measurement system such as 
smaart and plot the phase responses.

or screw perfectionism and just try it as-is. don't be disappointed, though.

the reason i'm so nit-pickish about all this is that every single audio 
professional i've met who has had prior exposure to ambisonics has heard 
at least one distinctly unconvincing demo, which makes live harder for 
all of us.
i'm proud to say that of all the HOA demos i've rigged so far (about 12 
in total), only one has been sub-optimal, and only because i pushed 
delay compensation too far. i like to believe the reason for this is a) 
that i totally don't believe in ambisonics as a silver bullet, and b) i 
work very precisely, with the same speakers directivities, phase and 
amplitude responses, placed within +/- 2° and +/- 2cms, if at all possible.