Surround / quadraphonic sound techniques? Anyone?

[tomperson]

I'll check that.

Hey michael, you said you did some multichannel stuff in the past, could you tell us about your subjective experiences? What worked, what did not, where did you show the work, how about the public, what were your expectations, etc?

[Michael-SW]

I've only done two more serious multichannel attempts. On was my fake 4.1 piece (rear channels reversed front channels to create "diagonal tension" in the room, 1 sub bass channel). That was performed on 12 (10?)speakers for an audience of 50 or so people. Simultanesously, I did live "diffusion", ie I sat at the mixer table and controlled the levels on various speaker pairs (near/far, ambient/direct etc) to create more spatial movements. The piece worked very well, but how much was due to the diffusion and how much was the rear channels is impossible to say.

I had of listened to the piece in 4 channels earlier, and the reversed rear channels definitely gave an added spatial dimension. However, I didn't have any control over that of course.

Got to run now, but I'll be back and talk about my attempts to do 8 channel "real" spatialization later or tomorrow.

[tomperson]

Thanks Michael, look forward to that, sounds interesting, mate!

[Michael-SW]

For the 8-channel project (a project for my course in electroacoustic composition) I wanted to try my hand at an attempt to do real spatialization. I quickly came upon the concept of Ambisonics, and thought that that was a good solution. There are (free) VSTs for ambisonics encoding but after trying them in Live I realized that those VSTs really need a host that supports 4+ channel audio natively which I don't have. The VSTs I found were really written to be used in Nuendo.

Instead I thought to try to build everything in Max/MSP. Spatialization is more than just direction (which you get from Ambisonics). You also need distance/movement cues like volume, low pass filtering, doppler shift and direct/reflected(reverbed) audio ratio.

So I built a "Spatializer" object in Max/MSP that took a monophonic audio stream and room coordinates as input and produced an B-Format/ambisonics encoded audio stream as output, including all the types of cues mentioned above. I certainly didn't write everything from scratch but instead searched the net for ready made Max objects ("externals") to help me achieve this.

Just duplicate this object any number of times (CPU permitting), add the outputs together and hook up to an ambisonics decoding object and you can realize the spatialization of any number of audio sources on any number of speakers: 2, 4, 8 etc. The more speakers, the better spatialization of course.

In the end I did a realization where I had 8 soundsources moving in the room, driven by an algorithmic flocking behaviour ("Boids", Wikipedia has more info), moving very much like a computer driven bird flock through the room. Max was rewired into ProTools for recording and then onto an 8-channel speaker setup. I didn't use bird sounds of course - that would have been far too obvious!

The effect was cool and and the feeling of sounds moving in space was quite convincing. The flocking behaviour worked well - the movements were certainly not random, but not predictable either. However, it was by no means a finished piece and in that lies the problem.

If I want to make that work the into a composition I need to have conventional tools that will let me work in 8 channels in a more conventional DAW way. ProTools might be able to do it, but right now I have put the project on the shelf. It is a lot of work and I feel that conventional DAWs aren't really there yet.

[evernaut]

Ambisonics audio can either be recorded with a special mic or generated synthetically.

I'd put a shout in for Ambisonics too... Yes, ideally you should use a Soundfield mic ( multi-capsule), but you can fake it with 2 figure of 8's and an omni. The signals are then encoded into B-format.

Check out

http://www.york.ac.uk/inst/mustech/3d_a ... lcome.html

[John Sweet]

I'm looking at ways to convert stereo into as close an Ambisonic format as possible, then play it back in a 4-speaker environment.

I found this:

4.4 Converting Stereo to B-format :
Stereo signals may be considered as two mono signals and placed anywhere but the following equations allow the stereo image to be positioned to the front with a width (?) of up to 180 degrees:


W = 0.65? - 0.27j?(?/180) (24)
X = 0.98? + 0.4j?(?/180) (25)
Y = 0.75?(?/180) (26)

*EDIT-forum won't display certain characters. The paper's here:

http://www.icdvrat.reading.ac.uk/1996/p ... 996_24.pdf

I'm using a Lemur XY multiball for my panning control, which through expressions can do the trig to generate tons of different value scales to describe the relationship of a multiball point to anywhere in the rest of the XY space. You can assign CCs to all different kinds of vectors and have them all updating at the same time.

I want to put a rack in a stereo channel, add as many chains as are necessary to describe an X+Y+Z+W soundfield, do the necessary pan/shelf filter/phase cancellation/whatever in each channel using Live's Utility & EQ plugs and whatever else, and control those effects off the appropriate values generated by the Lemur's XY field. The channels of the rack could be broken out to as many dummy audio channels as are required, then the whole thing would then have to be brought down to a fixed 4-speaker rectangular setup.

It's something people would normally do in Max, but I think it could be approximated without it.

With the Lemur, a fader can add a height variable to the x-y calculations for proper placement in space. Then once all your calculations are correct, the one-channel setup can be upgraded to an 8 or 10 channel setup with no extra work (just a whole lot more CC mapping in Live). 8 balls in the XY field plus 8 sliders = 8 tracks moving over, behind, below, and in front of one another.

I'm dying to do this, but I can't get to the concept behind the equations, so I'm stuck about half the way there.

I'm gonna try playing a 4-speaker show on Thursday, but I don't think I'll have this off the ground by then. If I can find a panning plugin that's intel mac compatible, maybe I could break my stereo source out to 2 stereo channels & run duplicates of the plugin (one set to output 1&2, one set to output 3&4). I could map both plugs to the same CCs & hopefully get a good result.

Anybody have any ideas?

[John Sweet]

It's definitely possible to get a decent decode of Ambisonic audio using just Live:

This article describes how it was done in the old days with a mixer:

http://www.york.ac.uk/inst/mustech/3d_a ... rosdec.htm

Realistic XY panning while encoding seems pretty crazy though.

I can't get the CICM ambisonics externals to load in the current build of Pd. I've seen other people complaining about this, but no solution.

http://www.maxobjects.com/?v=objects&id ... lateforme=

There appears to be only one set of Max objects that's intel mac compatible at the moment, and while they're badass--16 mono channels with flipping and rotating the soundfield and more--the smallest speaker setup seems to be a 2D octagon.

http://www.grahamwakefield.net/soft/ambi%7E/index.htm

All the VSTs & AUs I've found so far are old and/or don't work with Live.

Still looking, though.

[southernsteppa]

Wow, there's some serious reading in there. I will have to set aside a day to try and come to terms with it all.

I've done some crude quadraphonic panning for theatre and performance pieces, simply by using a seperate bus for each corner speaker. I usually have another bus as a stereo pair for the back of the room, so the main mix isn't completely lost there.

Then I find that this little freebie called "Quadpan" does the trick with panning around: http://www.oli.adbe.org/freeware.shtml

Like I think most people have already touched on though - I find it's best to keep crazy panning to a minimum with this kind of setup.