self evolving sounds/animator circuits?????

JovianPyx · Posted: Sun Dec 16, 2007 8:42 am Post subject:

Not sure how much this applies...

I've done some math-to-wav experiments using a PC and a C compiler to emulate an additive synth (32 VCOs). The VCOs are all sine wave output. Each VCO operates such that VCO[0] is the fundamental, VCO[1] runs at VCO[0]*2, VCO[2] runs at VCO[0]*3. etc. as in a normal harmonic series. However, the amplitudes of the harmonic generating VCOs are controlled by heavily filtered noise so that each harmonic VCO is controlled by it's own random LFO. The sounds are interesting, and the "color" can be controlled by picking which harmonics contribute and which do not. Using all harmonics has a characteristic sound which is different from using all even harmonics which is different from using all odd harmonics. I messed with other selections as well. This makes some interesting drone sounds at least. As time goes by, you hear an evolving timbre which is ever changing and completely unpredictable. I believe that this could be done using an FPGA, but I'm not yet ready to commit to a design. If there is interest, I can create and post some sound files.

frijitz · Posted: Sun Dec 16, 2007 9:17 am Post subject:

JovianPyx · Posted: Sun Dec 16, 2007 9:34 am Post subject:

blue hell · Posted: Sun Dec 16, 2007 9:41 am Post subject:

JovianPyx · Posted: Sun Dec 16, 2007 10:33 am Post subject:

blue hell · Posted: Sun Dec 16, 2007 10:53 am Post subject:

JovianPyx · Posted: Sun Dec 16, 2007 11:08 am Post subject:

blue hell · Posted: Sun Dec 16, 2007 11:39 am Post subject:

JovianPyx · Posted: Sun Dec 16, 2007 12:02 pm Post subject:

Rykhaard · Joined: Sep 02, 2007 Posts: 1290 Location: Canada

Depending on the tweakabilities of the chip version, that could be very interesting! Smile

Outside of that - all 3 do, sound 'digital'. The odd harmonics one itself, also has that really nice 'woody character' to it.

Really liked all 3. Smile

I've loved Additive for years - even did 32 and 64 harmonic versions in Reason. Editing patches in that though, was beyond nightmarish. Shocked

lol

JovianPyx · Posted: Sun Dec 16, 2007 12:34 pm Post subject:

Here's one more sample, odd harmonics, lower in frequency with slower random evolution.

odd_lower_slower.wav

Description:

odd harmonics, lower frequency, slower evolution

Filesize:

3.81 MB

Duration:

0:45

Viewed:

1757 Time(s)

JovianPyx · Posted: Mon Dec 17, 2007 10:51 am Post subject:

Last night I tried using only the harmonics that are prime numbers and I got what I would call the most interesting evolving sound yet (this method). The only way I can describe it is "bell like". It might be interesting to use an ADSR to provide a strong attack with a slow decay.

I'm convinced that I need to do an FPGA version so that I can tweak it live as well as test to see how well it plays from a keyboard.

JovianPyx · Posted: Mon Dec 17, 2007 6:22 pm Post subject:

This is a sample of the last sound I mentioned, prime number harmonics. Now I gotta get busy...

prime_mid.wav

Description:

prime numbered harmonics

Filesize:

3.81 MB

Duration:

0:45

Viewed:

1739 Time(s)

Tim Servo

Okay, now THAT is neat (the "prime" sample). With the right controls, that could be a really nifty synth engine.

Tim (how many cylinders in that engine?) Servo

frijitz · Posted: Mon Dec 17, 2007 9:29 pm Post subject:

JovianPyx · Posted: Tue Dec 18, 2007 7:29 am Post subject:

Thanks frijitz & Tim, I've listened to it many times and I hear weird Martian melodies played on Jovian wind chimes... I made another one 4 times longer. I have no clue why the prime number scheme worked this way, but selection of harmonics is something to mess with. The reason I tried it was that IMO odd harmonics are more interesting than even or all harmonics. I wanted to try a set of harmonics that had no factors in common, so - prime. The way I want to put this together, each sine generator can be tuned with a fixed point binary value. This will allow individual generators to be detuned from the ideal harmonic series. Currently, I haven't done that with the C model

Here's what I have in mind for a design (includes MIDI):

32 sine generators all driven by one phase accumulator oscillator (with portamento).

Each sine generator has a multiplier for the current phase angle
The multiplier defines the harmonic.

Each sine generator's amplitude is controlled by a noise generator (large LFSR) connected to an IIR lowpass

Each IIR lowpass has a bandwidth value (controls "speed" of evolution) and amplitude compensation multiplier value (controls depth of modulation)

Each sine generator has a weight value (general strength of the harmonic)

Sum of all sine generators sent to a "vca" controlled by an ADSR
-------------------------------------------------------------------------
This is much more sophisticated than my C model and I think would provide a good deal of control.

fluxmonkey · Joined: Jun 24, 2005 Posts: 708 Location: cleve

perhaps next, try the Fibonacci series (1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144...)?

b

JovianPyx · Posted: Tue Dec 18, 2007 8:19 am Post subject:

v-un-v

jksuperstar · Posted: Tue Dec 18, 2007 10:02 pm Post subject:

JovianPyx · Posted: Wed Dec 19, 2007 5:55 am Post subject:

The equipment I have will handle up to 1 MHz sample rate (this is limited by the DAC). After doing some math and estimates regarding how many clock ticks each oscillator will likely need, I'm probably looking at no faster than 250 KHz for 32 oscillators. The XTAL oscillator on the board I use is 50 MHz, so 250 KHz sample rate gives me 200 clock ticks to play with. This leaves 6 ticks per oscillator to get things done. More features means more clock ticks (usually). Parallel operations can increase the throughput given that there's enough routing and logic resources available. I like to keep the sample rate high so that I don't need to add an antialiasing filter and to allow high resolution phase accumulation. Even 125 KHz would work (and give twice the potential oscillator count).

blue hell · Posted: Wed Dec 19, 2007 8:17 am Post subject:

JovianPyx · Posted: Wed Dec 26, 2007 11:43 am Post subject:

I've spent some time working on a digital version of this. I started by using the GateMan-III synth as a "chassis", the MIDI controller is already there as well as basic logic structures to do fine tuning (master tune for all NCOs), portamento, ADSRs, SVF filter, etc.

The project as it stands is still only 4 NCOs, but it is a digital proof of concept which works and sounds as I expected it to. I see that there are digital peculiarities, but nothing that's a deal breaker. Next step is to expand to 32 NCOs. Currently, the state machine requires 1 clock for initialization and 6 clocks per NCO. This is fast enough to support 32 NCOs at 250 KHz.

EDIT ADD (2007-12-27 11:47am):
The first cut of this works nicely, but is only 4 NCOs. Once I got it running I started playing with which harmonics it outputs. I have a patch that uses 1, 5, 9 and 13 (where 1 is the fundamental) that sounds very bell-like. This synth chassis already had an NCA and an ADSR driving it, so I applied a bell type envelope and it sounds very much like a bell. During sustain, subtle harmonic content changes are noticable, that too sounds bell like even though it is controlled by a pseudorandom number generator. I am happy enough with these results to move forward with a 32 NCO version.