Non-randomized pseudo-random generators (for noise)

[FarleyCZ]

Hi all!

Lately I've been experimenting with synthetising my own kicks. Zebra is amazing for that (as well), mainly becouse those sweat MSEG's.

I found a little problem. In contemporary music for few years back, everyone is using samples which usually doesn't change at all while played over and over and over and over again..... That means, that if I don't want to sample the synth and use it actually for in-session generating, I need to meake it behave absolutely the same every note.

That is pretty easy, just reseting phases ... until you use noise for something. Especially when beefing kick's attack with it, it always change the sound a bit.

That made me think: Is it possible, to make noise behave the same every note, but still being noise? Then I remembered my lame highschool programing. It should be!

In computer, random values (used for noise generating I suppose) are made by pseudorandom generators. It's some equations that you feed with some "classic" ascending or descending values and it gets you back different number every time, but if you take the same sequence you feed it with and run it through the same equation, you get the same sequence of random values again. (In Delphi, you needed to hit "randomize" comand to make it's random generators generate different sequence everytime you run your app. )

...now, do you think it's possible to implement that into some next version of Zebra? May be with choise of reseting sequence start every note, or on every DAW playback start? Or may be make Zabra behave that way even now?

Discuss we may!

[Bronto Scorpio]

I'd definitely like to see something like that too
Would also be great to fake D50-ish attack samples

For now you could use an FMO in FM Self (+) or FM Self2 (+) mode with a very high FM value. It should sound the same every time. It can sound rather digital/combfilter like though (doesn't have to be a bad thing!).
You can also modulate several FMOs by each other to get different noise flavours (would probably be a bit to much for the task though).

Cheers
Dennis

[FarleyCZ]

Exactly. One can do loads of wierd modulations that comes close to noise, but there is always some characteristic that's "non-noisy".

This solution shouldn't be even that hard to implement, as every language has those generators somewhere inside. ...and as far as I'm awere of, it'd be first synth that would do that.

[xh3rv]

How long of a noise are you interested in? I'm just thinking of snappy impulses, which can be done with just about anything when the amp envelope is shorter than a single wave cycle, so any nice harmonics get a bit thrashed. I don't think this is really different from a short 'sampled' burst of noise, which will have some frequency content as well over this fixed length of time.

[e] I haven't delved into the number theory of much past the co-prime mod sort of deal, not sure where things go after that ... this sort of RNG in infinite output I think would tend to have some frequency content as well but it's very easy to engineer it out of perceptible range (Sorry, total tangent)

[pdxindy]

I use Geist!

It is so easy to sample with. Zebra, Diva or other, just sample it when you want the same sound repeated... or when you want to control the randomness sample multiple hits and round robin them or put them on different pads etc. Easy to make every fourth kick slightly different with the others exactly the same for example.

I genrally prefer synths over samples, but in this case it is easy and fun enough in Geist that I will sample synth sounds per project and leave the synth disabled on its own track if I want to resample. The flexibility of the synth and the easy repeatability and low cpu use of samples.

Anyway, I also enjoy figuring out how to do things so carry on the conversation!

[Sendy]

Perhaps something with linear bit shift registers (IIRC?) might be up to the task. I'm a bit sketchy on the details, but retro soundchips use them. The longer and more complicated (more bits?) the RNG is, the less 'loopy' it is, and the less sense of pitch it has. For example, the Nintendo NES has two types of noise, one is for general noise duties, the other is a small noise loop for gritty "sproingy" sounds.

And they're deterministic, so starting from the same initialization (set of numbers, or seed) will always produce the same noise pattern.

[FarleyCZ]

Yeah, thx for tips. Well I'm tallking about 30 - 100 ms. That pretty much shows lots of frequency content. Those sampled noise bursts arn't bad idea though, yes. If there's still that little sample ROM for Zebra in plan, it could do the trick quite quickly.