DSP Robotics Support

Advice needed if poss. I want to do some experiments using tuned noise as a polyphonic modulation source. (This is just to persue my very latest sonic hunch (i.e. obsession), if I'm honest )

Typically the kind of noises you get if you gently blow into a bottle or tube, or drag a pole along the ground. Wind in chimneys, rub screwdriver shafts together and so on. Things that are not properly oscillating but heading that way. Make sense?

So I dare say I could generate random numbers, and put them through hi-q filters to tune it. Hmm, would probably sound very underwhelming, and anyway for a polyphonic generator the CPU demand is going to be horrendous.

So I wondered about sampling, where the sounds would be stored as wavs and could be played back at keyboard pitch. BUT my question is .. how do you go about looping a random signal so that it can play continously and sound natural? Clearly there won't be any obvious loop-points, but I believe there are techniques using cross-fades, forward-backwards playing, and so on?

This is totally new territory to me so any help would be welcome. Anyone been here?

H

Hello Hugh,

xoxos has a couple of vsts that go into that direction. https://www.xoxos.net/
or https://www.xoxos.net/vst/
They used to have a couple of theoretical dsp papers on the old site. Can't find them on the new one.
Maybe if you asked ...?

Regards
Phil

Hello Hugh,

Here are a few Random Number Generators to start you off. Good luck! Sounds like an interesting project indeed.

Later then, BobF.....

HughBanton wrote:how do you go about looping a random signal so that it can play continously and sound natural? Clearly there won't be any obvious loop-points, but I believe there are techniques using cross-fades, forward-backwards playing, and so on?

The advantage of chaotic sound like noise is that you can loop them pretty much at any zero crossing. But natural sounding? Pretty much impossible with looped noise, as it loses its randomness and builds a pattern. You could maybe use two WAVs and blend them in varying amounts to restore some of the randomness. But that's just theory from my head, never tested it.

What if you use a continuous stream of random numbers and scale the shaping compomemts (those that give it color, like filters etc.)? I have found comb filters particularly useful for pan flute or blown bottle synthesis.

Hi Hugh

To get truly polyphonic noise (to avoid cancellation effects) I think you would maybe need a seed-able noise generator in poly and seed it with a random number for every channel opened (note played).

Then you would use a polyphonic filter. In this way every note would have its own unique noise source so would sound subtly different from any concurrent notes starting at the exact same moment.

Just a thought experiment really.

Cheers

Spogg

Thanks guys, it's all most interesting. Sorry, by the way, to have been quiet of late; I've mostly been (as we used to say in the 70s) making my new solo album, man. Which heavily features my FS HB3 Organ generator ... more on that when I've something to distribute but multi-tracked, to create a massive organ - well even I'm impressed HB3 has evolved considerably since I last posted a version.

I've long been interested in what it is that makes acoustic instruments sound unmistakably 'real'. I've been studying pipe organ sound for decades now, and although I'm undoubtedly getting ever closer there's still an easily identifiable difference. Much as I enjoy an electronic/digitally created sound I'm still struck by how there are subtle elements in the sound of the 'real thing' that still defy analysis. But I feel certain that randomness is the key, it's just figuring how to apply it. Simply adding noise always sounds, of course, like 'added noise'; seems to be the modulation method that's key. I'm already applying a number of techniques in HB3, but definitely more to discover yet.

(Sampled-sound organs are plentiful, but a) there are 101 drawbacks to the method, b) they mostly don't sound very good, and c) I want to know what's going on!)

On this subject ... the other day I was in the Albert Hall, and listening to the orchestra tuning up and practicing bits I was particularly struck by the sound of some very low Tuba notes - their fundamental will be down in the 35-45 Hz range. It can really only be described as .. 'scuse my French .. a fart! (Can I say that here?? OK OK .. it stands for Frequency Aligned Random Transients) or a rasberry. (Which stands for Random Additive Synthesis Binary Expone .. oh forget it )

But it's clear that what we hear from a real Tuba is not a regular cyclical waveform at all, more like rapid bursts (at 40-odd Hz) of pulsed noise, or chopped up noise if you like. The 'noise' in this instance must originate as the player's ffffffff.. sound that he or she is introducing at the mouthpiece.

I might have a go at generating that, see what I get! And if it turns out to be true at low frequencies, why not further up the scale?

Hello Hugh,

maybe this one helps:

https://www.researchgate.net/profile/Ve ... ion_detail

Although vintage, the paper is frequently cited in other publications on the topic of DSP and simulating wood-winds.

Regards
Phil

Mouth work is key to the sound. Here is some insight. It is a whole series describing all kinds of orchestra instruments, I just link to the bass trombone as an example, because he talks about his mouth work quite a lot. Just click on the channel to see other instruments.

https://www.youtube.com/watch?v=UJOa3t2OuPc

That's just great, thanks. Loved it. I think I'll give up organ and take up the trombone
H