Gedevaan
#1
21st April 2011
21st April 2011
#1
Gear interested

Spline Based Sound Engine

Hello!
My name is Sergey and I am present techno project GEDEVAAN.
So now I want to say not about Gedevaan and label Lead Square, but I as man who love sound and synthesizing, I tell you about interesting method of sound encoding. I haven't patents - I have just basic idea and basic digital knowledges for illustrate new method.

So let go!

This simple image tell us about possability of graphic spline based sound encoding. Sinus signal of 50 hz we may code with three points, actually we need two points, cause third point of fisrt wave phase is first point of second wave phase.

One point of wave have coordinates - time and amplitude. Also point have params - angle of tension and amount of tension. This information we may put into 64 bit forexample (its really narrow condition to illustrate thing). Next as on picture wrote - 5 minutes of 50 hz mono sound we may put into 0.235 Mb, nice yeah? So 50 hz its so simple. Discribe more difficult case ist 20000 hz. This hi freq signal we need multiple on 400 for 5 minutes of mono signal - about 92 Mb. So for stereo we need multiple on two. Its about 200 Mb also included file head information and other tags.

Next more detailed.

There we have basic 2 point splines from CorelDraw - as closed analog of sound spline. Fig b image that 2 point spline will able to describe sound wave form. Fig c its real sound spline as described spline based sound engine.

Go to the next.

Early I think that we can describe wave with 2 point spline. Fig d and e - case one and two - image how we make this. But sinusoidal wave is very simple. Actually we have more complex waves as square ans saw signals with short transients. Fig g - case three - describe 4 points spline. Its very versatile and powerful method.

So next.

This pic illustrate working (virtually) engine. One point have cords and params, more points its more sharp describing.

About realization of method. Since this spline sound is real analog sound as electric signal in wires we need to PCM this analog signal to wav file, or render in real time, digitize in real time to play in digital audio interfaces.

About pluses and positives. As analog wave there no need to dither, no loss of quality on convertation from one quantization freq to another, no loss on low amplitudes, no need to limiting and many others good things.

Minuses its storing of huge file im thing about 300-500 Mb and realization of this perspective method.

So as Im newbie there, may be we know about this method or similar codec or anything. I think its must to develop by Steinberg - commercial patience its very high.

#2
21st April 2011
21st April 2011
#2
Lives for gear

No immediate questions from me. But this is a fascinating idea you have.
#3
21st April 2011
21st April 2011
#3
Lives for gear

Nice!
Patent it! Quickly! Who knows, you might be famous in the future!

Cheers!
#4
22nd April 2011
22nd April 2011
#4
Gear maniac

Dude, patent that shit. seriously before someone else does!
restpause
#5
29th April 2011
29th April 2011
#5
restpause
Guest

I would do some research on interpolation filters because I think some digital audio engines already do this to some degree, but I'm not sure. But the render options in FL Studio and Reaper for example, sound like interpolation techniques and it seems like that's what you are describing if I understand you correctly.
#6
29th April 2011
29th April 2011
#6
Lives for gear

Hi,
Interesting, but I cannot undestand the usefulness of that concept, although it is interesting. I'm playing the devil's advocate, but don't take it as critizism, I encourage your enthusiasm,I'm testing you
Soooo.. questions:
Is this method intended for recording or just synthesis?

So basically you are making an analogy with Corel Draw (vector) and MS Paint (bitmap) applied to audio.
With vector drawing you can make beautiful and precise geometrical shapes (like splines) described mathematically and at the end you can render a bitmap at the resolution you want. The advantages are clear in the graphic world. But how do you represent photorealistic images with vectors? It is nearly impossible.

In digital audio, everything ends up at a D/A converter, and these work with PCM at very specific sample rates. It has no advantage to me working in vector-based audio when at the end you must convert to PCM anyway.
Furthermore, you must describe methods of processing vector-based audio. How do you perform a simple gain without affecting the wave shape?
How do you mix two signals? In PCM you multiply and sum (very simple)
How about a simple filter. In PCM you have fir-iir filters. What would be the equivalent? (Not to mention more complex things like dynamics or reverb..)
Are these splines supposed to be sine waves? I'm not sure of that...
Cheers! And keep on investigating!
#7
4th May 2011
4th May 2011
#7
Deal with it.

Very creative, but useless idea. Sorry.

PCM signals are already infinite resolution, and the angle/tension thing is the same as playing with harmonics and their phase. Of course this is used in synthesis, it's a pretty old trick (as old as synthesis itself) but there are far easier ways to do it. If you pay attention, you might notice that the huge file sizes you're getting are because the bits taken by the angle/tension information are the same bits that a PCM signal at a high enough sample rate to represent the highest harmonic in the spline would take. Or something like that. It's 5:30 in the morning and I didn't sleep (thank you GS!).

Actually unless I'm missing something, your "Fig g" is an exact illustration of how PCM works. Hey, looks like you just found a neat way to illustrate why PCM signals are lossless! Cool! (And "Fig d" would be more or less, if taken to a different context, an illustration of why inter-sample peaking occurs)

Replies
lettenmusic / Jobs Wanted
0
crowchaser / Jobs Wanted
0
Beaker02 / Live Sound
5
Sqye / Live Sound
1

Forum Jump

SEO by vBSEO ©2011, Crawlability, Inc.