Larger Scale Ambisonics
Post categories: Production,Ìýproject updates
| 15:00 UK time, Tuesday, 18 May 2010
A few weeks ago I had a phone call asking if we could put on a demonstration of Ambisonics for a TV studio full of people. As you may have noticed, I'm always keen to talk about my work and I agreed to figure out how to put on this demonstration. In R&D we've a modest amount of experience setting up small rigs, Chris Baume set up one rig for some listening tests and another for some demonstrations for our internal Audio & Music festival. I set up our Manchester Ambisonics listening room, and rigs in Newcastle for the Maker Faire and for an internal demonstration we did at a month ago (ok, I admit it, I've been geeky enough to set up a four speaker array at home too.) All of these rigs were set up in contollable acoustic environments and were optimised to be suitable for 1 to 6 listeners. The TV studio was going to have an audience of about 150. After a chat with Chris and Andrew Mason we set about trying to identify some of the problems we were lightly to face.Ìý
Once the speakers were in position and the cabling was all rigged back to the rack of amplifiers we were able to measure the speaker locations relative to the listener. To decode Ambisonics you have to calculate the correct decoding matrix. This decoding matrix can be arrived at using either heuristic search methods or matrix inversion, both methods have their drawbacks. Heuristic search methods can lead to horribly complex algorithms, and matrix inversion can lead to singularity (which is also pretty horrible).
We had a number of problems to solve, not least of which was we only had 2 days to organise, set up and test everything. The demonstration was due to take place in studio 3 in . For those of you who have not been in a TV studio, it's a large, flexible space which is designed to have lots of equipment hung up from the ceiling, our major constraint was having to work around the set, the cameras and crew, and the lighting rig.
Although Ambisonics can be played back on almost any speaker array the ideal array is a sphere of speakers. In all but laboratory conditions a spherical array is probably unachievable. A compromise is to roughly equally space speakers on a hemisphere above the audience. ÌýAn added complication in the studio 3 was the raked seating, which resulted in our hemisphere being tilted forwards at approximately the same angle as the seating rake. We settled for 16 speakers, 12 flown from the lighting rig of the studio, two on stands at the front left and right of the audience and two were on the floor at the front left and right of the stage. This gave is fairly equal spacing, without getting in the way of the lighting rig, the crew or the audience.
We invited Richard Furse, a colleague who has lent his expertise to the ambisonics work we've done in the past, to help with the decoding. Richard runs a company called which has an aim of bringing Ambisonics to the masses.
Richard and me working out the best compromise for speaker locations
Once the speakers were in position and the cabling was all rigged back to the rack of amplifiers we were able to measure the speaker locations relative to the listener. To decode Ambisonics you have to calculate the correct decoding matrix. This decoding matrix can be arrived at using either heuristic search methods or matrix inversion, both methods have their drawbacks. Heuristic search methods can lead to horribly complex algorithms, and matrix inversion can lead to singularity (which is also pretty horrible).
Once we had a working decoder we tested the system and it gave a reasonably good experience over the whole of the seating area. One of the noticeable advantages of Ambisonics over discreet channel systems that we benefited from here is the sound image can work over a wider 'sweet spot' before the image seems to fold into the closest speaker (if you get your decoder right).
Using the speaker locations in studio 3, Television Centre to calculate the best decoder
It was good to prove to ourselves that Ambisonics is scaleable and can work for events such as this, because when I received the phone call asking for the demonstration I didn't know for sure that it could.Ìý
Thanks go to Richard Furse, Chris Pike and Max Leonard, without whom, Chris and I would have been unable to put on this demonstration.
Comments Post your comment