What is Ambisonics and how can I get some??

IX 1999

by Joseph Lloyd Anderson

With one definition in the Merriam-Webster's Collegiate Dictionary the prefix ambi means around. From this root comes the word ambient for which two definitions are given: existing or present on all sides; an encompassing atmosphere. Of or involving sound is of course sonic. Simply put, Ambisonic is a cool way of saying surround sound. I'd like to pay particular attention on the second notion of ambient, an encompassing atmosphere, for this is a very important concept for the Ambisonic technique which hopefully will become more clear in this brief discussion.

Questions please…

Well just what is Ambisonics??

In its most complete form, Ambisonics is a periphonic (another fancy word, peri meaning surrounding) sound recording, synthesis and playback system. This means that it is capable of accurately recording, synthesizing and playing back sounds from left and right (as traditional two channel stereo), front and back (as currently realized commercial surround sound methods attempt), and more surprisingly up and down. So Ambisonics is truly 3D sound, not the 2D variety which is marketed as such by the entertainment industry. Ambisonics is a special case of wavefield synthesis, which seeks to resynthesize a recorded soundfield in a new location.

For recording, a special microphone (or microphone technique - tips on DIY later) is used to sample the directional information of the soundfield in question. For synthetic soundfields, this information can be generated for any recorded (mono or stereo) or synthesized sound by using appropriate encoding equations.

Ok, fine. It's multichannel sound right?? So how many channels is it??

I'm having a knee-jerk reaction here, having to restrain myself… actually, this is a very very good question. Ambisonics is very interesting in that it doesn't involve channels the way we're used to thinking about them. Soundfield recording (or synthesis) and playback are two separate activities. For playback we have many many choices. This is one of the reasons I like Ambisonics sooo much. We can playback in simple two channel stereo, for compatibility with everyone. We can do horizontal only over four, five, six, seven, eight, nine, ten, eleven, twelve, . . . actually as many loudspeakers as we like; and the placement of speakers of these playback channels is very flexible. And for big thrills we can playback complete 3D periphonic sound over eight, ten, twelve, fourteen, sixteen, . . . again with flexibility of layout. As Ambisonics is capable of N-channel playback, the "how many channels" question as related to loudspeakers can be determined as performance conditions require.

When it comes to recording and synthesizing soundfields, Ambisonics is encoded in a four channel form called B-format. The first channel, W, is the nondirectional pressure signal. The second channel, X, is the directional pressure gradient containing the front and back information. The third channel, Y, contains the left and right, and the Z the up and down. These channels define a first order sample of the complete soundfield at a point. When it comes to microphone response types, W is an omnidirectional mic, where X, Y, and Z are figure-of-eight mics.

This is something else I really like about Ambisonics, I can compose for N-channel loudspeaker playback, but I only need to deal with four channel files for complete 3D sound. Wow!! (Have I told you I'm working on a 2048-channel piece?? :-) )

So Ambisonics is really flexible on playback and very efficient on storage, but what really makes it different from other multichannel systems??

With Ambisonics the playback loudspeakers are not treated as simple monophonic "voices" in the playback space. Much early (and even contemporary) electroacoustic music has adopted the notion of the loudspeaker as a virtual performer or voice. This is a very limiting view of the loudspeaker which has been espoused by many musicians, most likely because the notion of "voices" is an important concept in music. The speakers also are not treated as "windows" from behind which sound travels, until finally reaching the listener. This is the most common approach to multichannel "surround" sound. Here sound is limited to the perimeter of the loudspeakers, with localization towards the sides and the rear tending to be very unstable. Ambisonics was created as the result of researchers addressing these localization problems.

Ambisonics uses all the available loudspeakers to recreate the sampled (or synthesized) soundfield within the array of loudspeakers. Surprisingly, for sources from the front of the soundfield, the loudspeakers in the rear of the array will be cranking away (especially if your sound is a crank!). While some loudspeakers are pushing, others are pulling. An interesting side effect of this cooperation is that loudspeakers which sound just OK running in stereo can sound amazingly good on Ambisonic playback.

Continuing with this important point, since the loudspeakers are treated as simply parts of a larger wavefield synthesis engine, they tend to disappear into the darkness. The speakers fade away, the perimeter of the speakers disappears, and as listeners our ears and brains are transported to the encompassing recorded or synthesized Ambisonic soundscape. Lovely.

That's cool, but as a composer do you find any specific advantages of Ambisonics??

I'm glad I asked. Aside from the above recording and playback channel advantages, the answer continues to be yes, yes, yes. And maybe another yes just to be sure. There are a number of techniques to re-image B-format recorded or synthesized sounds. We can rotate across any axis (x, y, z) the complete 3D image. We can "zoom-in" on a region of the image using a technique called "dominance". (Put your silly leather toys away, now.) This dominance trick is very cool as it can be used to crunch a sound down to mono in a single direction, say front, then slowly expand so it takes over the complete soundfield, engulfing the listener. We can control the "width" of the sound across any of the three axes by scaling the X, Y, or Z signals. We can very easily do some funky filtering where we split up the spectrum so different frequencies are spread in different directions.

Yeah, so there are some very powerful imaging techniques that make Ambisonics a joy for the composer.

Yet another thing I like about Ambisonics as a composer, is the previously mentioned microphone technique. So moving from stereo to Ambisonics means just using a different microphone (technique). I can continue to bang on pots, scrape ceramic bowls, wander at the seaside. But the recordings I come away with are actual 3D soundfield samples. The difference between stereo and Ambisonic recording is as substantial as the move from mono to stereo. In comparison to Ambisonics, stereo sounds quite flat.

Now I'm all fired up and ready to Ambisonisize, how can I get some??

A very good clearing house with a number of general articles and links is producer/recording engineer Richard Elen's Ambisonic.Net. Much of the information on his pages deals with more high level user issues, but his is a very good resource with links to hardware manufactures and the Surround Sound discussion list. To get up and running with Ambisonics, David Malham's tutorial is excellent, listing the simple equations needed for synthesizing and playing back B-format Ambisonic soundfields. Malham has two pages (1, 2) of very useful references to further develop your Ambisonic comprehension. Equations for generating loudspeaker feeds for more than four horizontal speakers can be found in Jeff Bamford's research. Richard Furse has kindly calculated tables of coefficients for multiple arrays of loudspeakers. Also, here's a link to the patent detailing the latest research on Ambisonic playback for 5.1 sound systems. Oh, joy!!

And here's the little mic trick mentioned earlier you can try. This won't be perfect soundfield sampling, but the results can be very useful with a little experimentation. Grab two small figure-of-8 mics and one small omni. (Nimbus records all their releases with the Ambisonic technique and uses two Schoeps figure-of-8's plus a B&K 4006 omni.) We're trying with three mics first for horizontal only recording. For full periphonic (four channel) B-format, there is another way to go about it with four separate cardioid microphones. (See Malham's references above.) We're going to stack these microphones up, as close as we can cram them. It works best if they're oriented on their sides instead of up and down. Starting at the bottom, place one figure-of-eight facing sideways to the left. Place the omni facing forward on top of this one. At the very top goes the final figure-of-eight facing forward. The middle omni is W, the top figure-of-eight is X, and the very bottom mic is Y. On the mic preamp attenuate the W signal by -3dB. Ideally all the capsules should be coincident and this isn't exactly the case with our little setup. (Nimbus uses some extra electronics to deal with these issues.) The Soundfield microphone is really the easiest way to go, and Josephson Engineering's C700S will work for horizontal only surround. But, our experiment gives reasonable coincidence for the horizontal plane and can give very nice results in the studio and the concert hall. Decode the resulting W, X, Y signals for four playback channels with the equations found in Malham's tutorial. Even easier, Richard Furse has a decoder for Windows 98. Or, if you have ProTools TDM, use these sessions to mix and monitor your Ambisonic delight. Have fun!