What exactly is Granular Synthesis?

What is it? I know it's obviously not just sample manipulation, but something much more than that... what is it? In order to use it properly, I need to understand how it works, so I don't blindly tweak parameters. I'm sure I can create some pretty awesome sounds with it, but I want to really know what it's about... (like I know the theory of subtractive synthesis in order to craft sounds the way I like to.) I can "hear a sound" and craft it with substractive synthesis methods, but how do I use granular synthesis to make sounds I "hear" in my head?

Thanks...

 

Subtractive synthesis is good in the way that you start with a sound, and then you can hear how it changes as you remove pieces from it (hence, subtractive synthesis). Granular synthesis isn't so direct of a method. Basically, granular synthesis chops up either incoming audio, or some form of a sample/table, into very small groups of samples called grains. Then, each grain can be processed individually without affecting the timing of other grains. For example. For example, you can take a bass line which is in the key of A minor, and via granular resynthesis, you can pitch-shift it without affecting how quickly the sample plays back (traversal time). On the other end of the spectrum, you can change traversal time (how quickly the grains are moved through) while keeping the pitch the same, just by simply speeding up the order of how quickly each of the grains play back. There are other ways you can modify grains, but this should give you a basic understanding.

For an example of a Reaktor device which utalizes granular resynthesis that is also geared toward live performance/improvisation, check out my Red2 ensemble: http://homepage.mac.com/springscansing/red2.jpg

If you'd like a copy of it, be sure to email me. Or I suppose I could put it up in the user library.

In fact, yes, I will put it there later today.

 

Thanks for sort of clearing things up, but the examples you've given me (other than the idea that there is a manipulation of the "grains" by the modulators provided in a granular synthesis ensemble)are similar to time stretch/sans pitch affecting, or pitch affecting/sans time stretch that is inherent in wave editors like Pro Tools or Cool Edit Pro where there is either a removal of the samples themselves to shorten a waveform, hence time shortening (or stretching if desired.)

What I can deduce from what you said is that the "grains" may be sample sets which are affected by the modulators withing the Reaktor ensemble, and can be "synthesized" like any other oscillated waveform through filters/envelopes/blah blah blah...

Please tell me how on/or off I am on understanding this premise...

Thanks again.

 

You're dead on pretty much. There's a resynthesis module in Reaktor which lets you do just that, which is, use a sample as a waveform. You can then route it through filters, and combine granular synthesis techniques with subtractive filtering techniques, and even use it to modulate the sum of sines via additive synthesis with FM, to create a very nice hybrid synthesizer.

Once we have unit delays in R4 and can get some nice linear-phase filters in order, we can do some hi-quality crossovers and convolution technicques and do some very intense processing. Couple that with the internal patch randomization features of R4, and the future looks very bright. Looks like Csound might not be seeing me quite as much anymore, hehe.

 

I'd like a copy but what's your email address?

 

It will be in the user library. I'll post here once its up.

I meant to get it up today but I have a bad stomach bug and I've been sleeping. Expect it tomorrow, along with a demo song made entirely in Red2.