Multi-listener approach

[colonynofi]

Hi there,

I’m looking at using SPAT for a fairly simple (conceptually at least) application, but just a little unsure as to where to start regarding researching different approaches / evaluating possible solutions.

I have a med sized (32x8m) space covered by around 20 speakers hung from the ceiling (pointed down around 3 m high) in a matrix (two lines of 10 speakers)

The space will have numerous people at various times. Think of it as the “museum” model - a significant number of speakers covering a significant number of people.

My use for spat would be as simple as a multi-channel panner where the speaker positions are defined, and objects are panned but NOT localised for a particular user position…

I’m digging down into spat now - but I keep running into (fun!) deep dives about other use types (which are all exciting) but I cannot find any literature / writing on this. I’m kind of assuming I’ll just be able to turn off different parts of the processing and rely on one of the panning methods only to do this (and perhaps using some of the room processing… though for this particular project it may not be necessary!)

Does this make sense? I apologise if I come to this from a slightly ignorant space. I come to this from using systems such as IOSONO, various matrix systems, or even a custom built system at one time… or hacking together various things to make large channel count panners work… (some people would die at the simple methods employed yet perfectly acceptable for the type of experience they’ve been used for!)

(The final project needs to run for quite a long time on a PC - but I’m developing everything on a mac! So I can’t use outside max packages that are not windows compatible )

Cheers all - Brendan.

[colonynofi]

For reference, The best I’ve come up with so far is from the flux pages :
https://desk.fluxhome.com/portal/kb/articles/panning-types
In which case I’m most looking into KNN and DBAP. I’d love to read more about the approaches that SPAT takes with these… as I’m not sure I’m going to have tonnes of experimentation time once I’m in the space.

[nadirb]

Hi,

You may check these panning types in the spat5 overview Pan tab.

Otherwise you may use a simple envelop spatialisation triggered with a random or urn object Cf. image caps attached (and if you consider planning a determined spatialisation sketch you may use a qlist (or antescofo in play mode i.e. as a sequencer…) to write your it

Best

N.

[tcarpent]

Hi Brendan,

For such kind of “museum model” (matrix of speakers), I would definitely recommend KNN or DBAP.
(they’re based on the same approach – distance-based panning – but KNN has a few more tweaking options)
Although I admit it’s a bit difficult to anticipate what the result will sound like without experimenting in the space.
Using an additional spat5.spat~ it should be possible to somehow simulate (over headphones) what listeners would hear in various spots (or when moving into the space).

Cheers,
T.

[colonynofi]

Ah many thanks for this both.

I’ll keep experimenting with KNN - and great idea for simulating over headphones using a second spat.

Quick question… and again, it comes from not having documentation (that I’ve found at least) on KNN. Is it ok to use when speakers are pointing directly down mounted from the ceiling? I guess I’m just surprised there is (that I’ve found yet) no requirement to show the direction of the speaker when setting their positions. I’m sure it just comes from not using this system before.

Do any of the panning types require positioning (this is just interest now)

Is there anywhere I can read more about the different panning types - a paper perhaps? (I’ve been thru the help patches, but there’s no real descriptions there I can find)… I’m definitely interested in learning more about SPAT (especially since revolution is also around now, which makes a lot of sense for other experiential projects for the types of things I do) but I feel like I’m short of a bit of knowledge.

Cheers!

[colonynofi]

And by extra options (for KNN) are you referring to the ability to “spread” the sound / the neighbours function as found in the help tab?
Are there others I’m not seeing?
Thanks.

[tcarpent]

Hi,

Spat does not take speaker orientation into account.
For listener-centric panning (vbap, ambisonic, etc.), it is assumed that all speakers are pointing towards the sweet spot.
For KNN/DBAP, a similar rule of thumb applies, and speakers should point “inside the field of speakers” (i.e. inside their convex envelope).

The only documentation I can think of is:
T. Lossius, P. Balthazar, and T. de la Hogue.
DBAP - Distance-Based Amplitude Panning.
In Proc. of the International Computer Music Conference, Montreal, Canada, 2009.

And yes, I was referring to the “neighbors” option that KNN offers.

Best,
T.

[colonynofi]

Ah I think I’ve got it now - having done a bunch more exploration / trial and error and much more reading.

Tomorrow I get to try listen to my work in thru the 20 speaker array in the space. In this case, I’m going to just try panning first with no room / reverb - and then start playing with those options once I have the movement of the direct signal working to my liking.

Loving the depth of SPAT - it certainly goes SO much further than my use case - but it also brings up some really interesting ideas for other creative uses I might want to try one day.

Many thanks for the paper reference!