posted 17 July 2012 08:12            

I am working on a new dance piece in the studio with the dancers and we have an 8 Meyer loudspeaker array setup in a circle - this is being driven by both kyma and Ambisonic reproduction of field recordings.

The choreographer asked me today to make a mechinic spinning sound - spinning around the speaker array which sitts outside the audience and performer - they are in the image.

I have 2 questions:
1. when placing a sound in the timeline and then using the panner to move it around the array, there was signal accross 4 speakers at any time - in an 8 channel array that means the sonic object is quite wide - I would like to narrow that so that the image moved perhaps between 2 loudspeakers at a time, but when directed at a single loudspeaker ie. 0, -45, -90, -135, 180, 135, 90, 45, the sound would come from that loudspeaker only.

I played with the distance setting to try to achieve this, and radius but did not come up with anything satisfactory so am seeking your collective wisdom on the matter. I think a width variable would be terrific in the spatialization toolkit in Kyma.

2. I am thinking some doppler or pitch variation is necessary to achieve the result the choreographer is thinking of, however the sound is constantly rotating - faster and slower - and so am not sure how to achieve such a sound - imagine being in the centre of a room when a large mechine element was spinning around the circumference of the room

Thanks in advance for your suggestions

posted 17 July 2012 17:03            

A lot of engineers went through stages when stereo was first born. They eventually realized the all the vocals on the left and all the instruments on the right wasn't really giving the best stereo results. They then panned things around, vocals in the middle and various instruments dotted around but very little at the extreme.

The goal was then to be able to make it sound like you could actually walk into the sound field and it wasn't coming from just two speakers.

One of the main tricks was to use a delay on one of the two speakers as this stopped it sounding like the sound was sitting in the cone of the speakers and pulled the apparent sound source out in to mid air.

Different signal types had different delays as percussive sound with a long delay would wreck the mix, where as strings could have hugh delays and would add distance to the instrument/s. Also by putting the delay on the left or the right it would give an apparent pan that was more perceptually positioned than simple left/right level differences.

At that time delays and reverbs cost an arm and a leg and it was not uncommon for a 24 track recording to have only 12 different tracks with every other track being a delayed version of the previous track. This way one delay unit could be used multiple times. This may sound wasteful, but the effect gave such an expensive sounding mix that it was worth it.

Now days reverbs are ten a penny and tend to be used more often to give space. Also it's a bit of a chore to route the signal up two channels and insert a delay in one of them. That would mean having to think about two faders instead of one, but I think sound mixes have suffered a bit IMHO.

With this in mind maybe you could use the technique in your surround application. Think of three speakers next to each other, say 4,5,6. Quiet, Loud, Quiet. long delay , short/no delay, long delay. As you slowly move the sound 4 and 5 would go quieter while 6 goes louder and 7 starts to be introduced. As the sound goes louder so the delay becomes shorted and vice versa. This way you could move the sound around as if it were really moving (not just being passed from speaker to speaker). It should also give the impression that the sound is coming from is outside the speaker field as if the speaker are not really doing anything. You can see the speaker but feel that they really don't need to be there.

I hope this makes sense

posted 17 July 2012 18:02            

How do you make such a beast?

Module called QuadOscilator.

Get two of them and make sure that the frequency field has identical controls. This will keep them in phase. Feed one with constant value 1 on it's left and constant value zero on it's right. Feed the other with constant value zero in it's left and constant value 1 on it's right.

Now you have two modules that are generating a sine and a matching cosine waveform. you can control the rate and direction with the frequency fields.

Speaker 1 is a mix one sine only
Speaker 2 is a mix of 0.707*sine + 0.707*cos
Speaker 3 is a mix of cos only
Speaker 4 is a mix of (-0.707*sine) + 0.707*cos
Speaker 5 is a mix one (-sine) only
Speaker 6 is a mix of (-0.707*sine) +(- 0.707*cos)
Speaker 7 is a mix of (-cos) only
Speaker 8 is a mix of 0.707*sine +(- 0.707*cos)

Now you have 8 sine waves all phase shifted by 45 deg from each other.

Now you only want the tops of these of these sine waves so using 8 scale and offset modules. Offset them down by -1 plus some extra negative value and amplify them by the same but positive extra value so that the bottoms are clipped off at -1 but the tops just touch zero. Then offset them again by +1 so that the flat, clipped bottom is now sitting at zero and the tops just touch +1.

By varying you extra value in the first scale and offsets you can vary the overlap between speakers. Put these wave forms into 8 VCAs to control the speaker levels. You can use the same wave forms to control the delay amount in 8 delay modules with some multiplies and offsets (additions in cappy talk) to adjust the amount of delay effect.

I hope this makes sense.

Pete

posted 17 July 2012 21:38            

Thanks for your amazingly generous sharing :-)

I am trying to build the patch as described - we have 4 signals from 2 QuadOscilators. I have used Channeller to separate out the Sine and Cosine and a SoundToGlobalCOntroller to do the math on those 2 signals.

So it seems to me that we only need one of the QuadOscilators and then multiply by -1 to get the negative offsets in the math? I am assuming that you are using them to get oposite phase signals and then I am assuming that I use one for channels 1-4 and the other for 5-8?

re the width - yes I understand the field idea - in fact as a bed in parts of this show I am using Ambisonic A-Format recordings decoded to the 8 speaker array in order to produce a contiguous and inhabitable space - the audience should feel like they walked into the theatre and are then sitting in long grass with the wind blowing across the grass heads around them - the machine I am trying to create them slowly spins up and sucks up the grass field which leads to the audience being abducted (aliens...) - the sound will then evolve towards an ambisonic recording I am making with some dancers at the Integratron (http://www.integratron.com/) where I will create a contiguous white noise space, place the Soundfield Mic in the middel and then hopefully record the noise shadow of the dancers moving between the source (noise) and the microphone - this recreated in the performance space I am hoping will generate the sensation of unsees bodies moving around and through the audience.

its a fun piece :-) GP

posted 18 July 2012 16:45            

Oh if one Quad Oscillator gives you both sin and cos thats great. I looked at the drawing on the icon and thought it only gave out a mono signal. I would be worried about using SoundToGlobalCOntroller to carry the controlling signal, as they are at cappytalk rate not sample rate so you may get 1 kHz ripper noise when you start moving things fast around the room.

If the quad oscillator has sine on the left and cos on the right then you don't need to split the legs as you can use 8 channel crossers with the formula in the fields.

Pete