Ambi.decode~ (Ambisonic playback)help

Hi

You're absolutely right, you should be able to play back w x y z
channels into an [ambi.decode~ 3d 1st] for ambsionic playback, but
there's a few possible gotchas: Ambisonics is a theory in which a
variety of conventions have been used and no particular set is
standard (though there has been quite a lot of activity on the
Sursound list recently to tackle this issue, see [1]; note that my
externals and the ICST ones are predate and do not currently follow
this standard).

Besides the 2d/3d and 1st, 2nd etc. order (or in the proposed new
terminology, degree), there are multiple 'flavors' or coefficients
that may or may not be assumed to apply to the ambisonic domain
signals, various different sequences in which the domain signals are
arranged in the domain (not necessarily intuitive, again see [1]
which suggests W Y Z X for example) The channel order is easy to
remedy, if it doesn't match, by patch cabling (The inlet/outlet
assist messages (mouse over -> status bar) shows the channel names in
all the ambi. objects). The importance channel coefficients however
is often overlooked.

Another standards gotcha that applies to both my externals and the
ICST ones also is the use of left-hand rule for axes (i.e. azimuth
90' is to the right, whereas much ambisonic software uses right-hand
rule). If your spatial image is flipped in the Y axis, it can be
remedied by reversing the sign of your speaker configuration azimuth
coordinates (or at 1st order, flipping the sign of the Y signal).

I don't know what conventions were used in this test file, it is not
documented. Could you describe the 'unexpected' results?

Graham

[1] http://ambisonics.ch/standards/

On Oct 17, 2008, at 7:49 AM, Mike Andrews wrote:

>
> Hi,
> I am currently attempting .amb file playback using Graham
> Wakefield's Ambi.decode external. I am using this test file as my
> source http://www.ambisonia.com/Members/ajh/ambisonicfile.
> 2007-05-29.7251031563
>
> I am opening the test file in audacity which splits the file into 4
> mono tracks (which I assume are wxyz). I am then feeding these
> individual tracks into Ambi.decode~ for playback. Ambi decode is
> set with the arguments 3d 1st Default 4 as the .amb file is in 1st
> order 3D. However on playback the sounds don't appear in the
> correct positions. I am currently playing back on a quad setup.
> When using ambiencode with decode a mono input everything is
> positioned as it should be. Am I right in thinking I should be able
> to feed w x y z into ambi decode and get correct ambisonic
> playback? If so does anyone know what could be causing my problem?
>
> Thanks for any help in advance.
>
> Mike. mandrews04@qub.ac.uk
>
>

Be seeing you

grrr waaa
www.grahamwakefield.net

Hi Graham I appreciate the reply,
The results seem to be pretty much random with the only constant being the centre front position. Unfortunately I will not have access to the system until tomorrow so I can not check in more detail until then. I have also tried patch cable re-ordering and have tried all possible combinations but with no luck. I have also tried this file http://www.ambisonia.com/Members/michael/ambisonicfile.2008-07-05.2994956247 and am having similar problems. When you say "If your spatial image is flipped in the Y axis, it can be
remedied by reversing the sign of your speaker configuration azimuth
coordinates (or at 1st order, flipping the sign of the Y signal)." How would I do this? Do you mean by reversing the polarity of the Y axis? The 8 position .amb file I am trying to use was created using a perl script (http://mchapman.com/amb/soft/positions/) with ambisuite so I don't know if that gives any clue on the conventions used?
Thanks again, Mike.

Quote: lists@grahamwakefield wrote on Fri, 17 October 2008 12:50
----------------------------------------------------
> Hi
>
> You're absolutely right, you should be able to play back w x y z
> channels into an [ambi.decode~ 3d 1st] for ambsionic playback, but
> there's a few possible gotchas: Ambisonics is a theory in which a
> variety of conventions have been used and no particular set is
> standard (though there has been quite a lot of activity on the
> Sursound list recently to tackle this issue, see [1];
...
> [1] http://ambisonics.ch/standards/

No, the situation on standards is not nearly as confused
as you suggest. All the files on Ambisonia.com are in
the ".amb" standard. This is fairly old and is very
stable. The details of the standard can be found at:
http://www.ambisonia.com/Members/mleese/file-format-for-b-format/

This document contains links to pictures and stuff which
should help.

The standard in your ref [1] is a new standard under
development, so is not yet stable. Unlike ".amb", the
new one will be able to go above third-order (16 channels).
But it is not finished yet, so ignore it.

Unfortunately, I have never used plug-in decoders, so
can't really help the original poster other than to
point them at the ".amb" standard document.

Regards,
Martin

On Oct 17, 2008, at 9:48 PM, Martin Leese wrote:

> No, the situation on standards is not nearly as confused
> as you suggest. All the files on Ambisonia.com are in
> the ".amb" standard. This is fairly old and is very
> stable. The details of the standard can be found at:
> http://www.ambisonia.com/Members/mleese/file-format-for-b-format/

Martin:

Sorry, I didn't want to suggest that there was a lack of convention
in the .amb format, which as you say is clear and well established.
I was talking about the literature and implementations of ambisonics
in the wider community over the past decades, and the recent Sursound
discussions.

Alex:

Double-checking the .amb definition, the channel order and channel
weights for .amb B-format matches [ambi.decode~ 3d 1st] with no
modifications needed. That narrows the problem down to either axis
conventions (or speaker configuration mismatch in the decode~ vs. the
actual speaker layout used maybe?)

Flipping the sign of the Y (or any axis) at 1st order is a case of
placing a [*~ -1] in the signal path between your playback and your
ambi.decode~ inlet for that channel.

Unfortunately I don't have a 3D speaker array to hand to test. I also
don't have a torrent client on this machine to dl the .amb file for a
visual test. But here's a patch - could you try this? I haven't done
all 26 positions, I figured 14 was enough to catch whatever the
problem is. In fact, the first 6 should be enough to get that (L, R,
F, B, U, D): if L/R are backwards, flip the X channel, etc. If this
is working fine but yours is still sounding strange, maybe the
speaker locations are not well defined in your patch/system?

Hi Graham,
Thanks for taking the time to create that patch. I loaded the sound file and played it while observing the meter bars. The meter bars do not correspond to the directions for example left- back comes through each output equally. Front however on its own seems to be correct. I have also tried inverting the connections but I am still not sucessful. I modified that patch in an attempt to playback correctly on my 4 speaker set up. The speaker definitions I have used are as below. Is this correct for a 4 speaker horizontal set up?

;
decoder speaker 0 -45 0;
decoder speaker 1 45 0;
decoder speaker 2 135 0;
decoder speaker 3 -135 0;
decoder dump;

Quote: Mikelib wrote on Mon, 20 October 2008 09:05
----------------------------------------------------
> Hi Graham,
> Thanks for taking the time to create that patch. I loaded the sound file and played it while observing the meter bars. The meter bars do not correspond to the directions for example left- back comes through each output equally. Front however on its own seems to be correct.
...
----------------------------------------------------

In Ambisonics, all speakers cooperate to localise a sound.
So, for example, while the speakers on the left push,
those on the right pull. Given this, looking at meters
is of limited use.

Hi Martin,
I understand this concept but surely a sound coming from 45 left for example with show a slightly higher reading on the meter of the corresponding speaker?

Martin is quite correct that all speakers contribute to the
construction of the sound field, and meters are thus not a very
helpful way to examine the output. In the patch below you can try out
a few cardinal directions and see what the meters show. Notice that
all channels are active (almost) all of the time, and that although
L, R, U, D, B, F all give helpful results for all the example
positions, the more complex speaker positions such as ULF are often
counter-intuitive. (This, amongst many other reasons, is why an
irregular speaker layout can totally corrupt the spatial image!) Then
change the decoding flavors (default, in-phase, etc.) and see how the
signals change again...

Mike, if your speaker layout is horizontal only, then decoding a 3D
encoded signal stream simply cannot reproduce positions that
incorporate vertical components!

BTW, a [ambi.decode~ 2d 1st 4] object will choose exactly those
speaker locations as a default.

On Oct 20, 2008, at 8:05 AM, Mike Andrews wrote:

> I modified that patch in an attempt to playback correctly on my 4
> speaker set up. The speaker definitions I have used are as below.
> Is this correct for a 4 speaker horizontal set up?
>
> ;
> decoder speaker 0 -45 0;
> decoder speaker 1 45 0;
> decoder speaker 2 135 0;
> decoder speaker 3 -135 0;
> decoder dump;