Christian takes us through BT Phobos and the concept of polyconvolution using signal diagrams, plasticine and BT Phobos itself. By A-B-ing two duplicate sets of sounds working as dry signals and as convolvers, he hopes to demonstrate the power of this new, revolutionary form of sample-synthesis.

When attacking the question of Polyconvolution, I think it best to approach it in three stages:

  • What is convolution?
  • How does Phobos work?
  • What does Phobos do that puts the Poly in Polyconvolution?

So without further ado...

What is convolution?

Well, in simple mathematical terms, it is the combination of two things to create a third. In our musical / sound world, it is usually utilised in reverb plugins - an impulse response (IR) is captured by recording and processing a loud transient sound in a space (for example, by firing a starter pistol into a hall or room), and we use convolution to put a dry sound source ‘into’ that space. So it’s not really mixing two sounds together, it’s like wrapping one around the other. Take two balls of plasticine as representing two sounds - you can control having more of one ball than another, and you can move them closer together or further apart. This is mixing. However, if you wrap one ball around the other - like in the convolution process, with sound - you don’t end up with the two balls, you end up with the a single ball that has one ball in the inside and another on the outside. 

Like with plasticine, either ball can be wrapped around the other, and this is the fundamental idea behind BT Phobos; that there is no denomination between sounds, and IRs; they’re interchangeable. So, an IR can be something tonal, a hit or indeed a loop. BT Phobos takes your dry sound source and filters it through another. Imagine forcing a drum kit through the key bed of a synth and seeing what comes out the other end - it would still be drums, but there’d be all sorts of innards from the synth wrapped around it too.

Check out this signal diagram. 

There are two key bits of sound creation in BT Phobos: the sound source and the IR. There are also modulation stages that you would get in any synth or modular system. The source sound can pass through filters, gates and amps to make them louder or pan them left to right. When you hit a key, it triggers a sound source which goes through all of those modular stages before passing to the ‘dry’ knob - the amount you tweak this will govern how much of this signal, i.e. the sound source with all the modules added, is output:

This signal is also split into the ‘wet’ knob, which controls the amount of signal that goes to the convolver triangle. The convolver triangle is simply a device that distributes the amount of send to the three different convolution stages. We use a puck to determine how much or how little of the signal is distributed to each convolver bus. We then give these modulated source sounds a few more modulation stages to prepare the sound for the convolution.

So, we’ve triggered the source sound with MIDI, but this MIDI signal also splits so that it triggers the IRs too. Again, this is just another sound, but one we’re using as an IR. This has a bunch of modulators before passing into the convolution stage. We finally have a few more modulation stages before passing it down to the output, at which point it is simply mixed with dry sound source. Back to the plasticine:

So is this Polyconvolution?

No, this is just us taking a sound and passing it through convolvers. It’s unusual that there are three, yes, and unusual that there is a triangle shaped distributor, and that we’re using synth sounds and loops as IRs, but this isn’t polyconvolution.

Now, when you hit more than one note simultaneously on the keyboard it will trigger two of each sound source (usually at different pitches) and create duplicate modulation stages for each of those notes. Is this polyconvolution? Well no, this is polyphony. But the same principal has been applied to our convolvers, so that when you hit more than one note on the keyboard, two of each sound source are created AND two IRs are triggered (say at different pitches), and two convolution processes and all the modulation stages are created for each convolver stage (w,x,y). THIS IS POLYCONVOLUTION.

And the real bit of wizardry here is that everything is rendered in realtime - the IRs are prepared just at the point that they’re needed (following a touch of the keyboard).

In my video, I demonstrate the difference between how different sounds affect the original ones when used as convolvers.

I hope this has shed light on our latest and maddest of ideas!