How is this sound generated? / reverse engineer sound

AI Thread Summary
The discussion focuses on how to reverse engineer a specific sound generated by a digital soundscape generator. Participants suggest using software like Audacity to analyze the sound's waveform and envelope characteristics, emphasizing that the sound is likely electronically generated. Techniques for recreating the sound include mixing sawtooth and PWM waves, applying filters, and adjusting the amplitude envelope. The conversation highlights that various electronic components and adjustments can produce similar sounds, making precise replication a matter of experimentation. Ultimately, the consensus is that the sound is a product of software rather than traditional instruments.
btb4198
Messages
570
Reaction score
10
Engineering news on Phys.org
Sounds like bells. Say tubular bells struck by rubber hammers?
Of course - technically - the sound is generated by a bunch of electronics responding to digital input.

What do you mean by "reverse engineer"?
There are a lot of things that can make the same or very similar sounds, so working out what did it is a matter of guesswork based on context. In the context of the website - someone fiddled with some sliders on the UI perhaps?
 
There are several components to this. To recreate it try the following:

Make a repeating sequence of a couple of dozen notes playing simple arpeggios at about 180 notes a minute.

1) Generate the basic tone by mixing a sawtooth wave with a PWM wave where the pulse width is set to vary slowly (about 2 LFO cycles per minute) between 25% and 50% duty cycle. Send this signal through a low pass filter with a cutoff of about 500 Hz and very little resonance. Set the amplitude envelope to have a slowish attack and release, with high sustain and a decay of abut the same as the attack (but the sustain is high so this doesn't make much of a difference). Set the envelope rate to scale with the keyboard to make the envelope faster for higher pitched sounds and slower for lower pitched sounds to give high frequency "tinkles" and lower frequency "whoops". Modulate the filter cutoff a little with the envelope.
2) Add a very little (very little) amount of very slow phase shift.
3) Add some reverb and maybe a little delay of about 350 ms.
4) Pass this through a 10 band EQ.

Tinker with it until it sounds like the sound on that page.
 
Simon Bridge said:
Sounds like bells. Say tubular bells struck by rubber hammers?
Of course - technically - the sound is generated by a bunch of electronics responding to digital input.

What do you mean by "reverse engineer"?
There are a lot of things that can make the same or very similar sounds, so working out what did it is a matter of guesswork based on context. In the context of the website - someone fiddled with some sliders on the UI perhaps?

Reverse engineering is when you go backward. In this case from the sound.
 
btb4198 said:
Reverse engineering is when you go backward. In this case from the sound.
Pretty sure Simon knows the definition. We are asking what you mean by it in this context.

The music is almost certainly (i.e. approaching 1.0) electronically generated, so there's nothing to reverse engineer it to.

What you'd do it analyze the waveform and look at the envelope (attack, sustain, decay, release). Reproducing that would get you the sound.

http://en.wikipedia.org/wiki/Synthesizer#ADSR_envelope
 
Last edited:
  • Like
Likes dlgoff
DaveC426913 said:
Pretty sure Simon knows the definition. We are asking what you mean by it in this context.

The music is almost certainly (i.e. approaching 1.0) electronically generated, so there's nothing to reverse engineer it to.

What you'd do it analyze the waveform and look at the envelope (attack, sustain, decay, release). Reproducing that would get you the sound.

http://en.wikipedia.org/wiki/Synthesizer#ADSR_envelope
what is (i.e. approaching 1.0) electronically generated?
 
Blibbler said:
There are several components to this. To recreate it try the following:

Make a repeating sequence of a couple of dozen notes playing simple arpeggios at about 180 notes a minute.

1) Generate the basic tone by mixing a sawtooth wave with a PWM wave where the pulse width is set to vary slowly (about 2 LFO cycles per minute) between 25% and 50% duty cycle. Send this signal through a low pass filter with a cutoff of about 500 Hz and very little resonance. Set the amplitude envelope to have a slowish attack and release, with high sustain and a decay of abut the same as the attack (but the sustain is high so this doesn't make much of a difference). Set the envelope rate to scale with the keyboard to make the envelope faster for higher pitched sounds and slower for lower pitched sounds to give high frequency "tinkles" and lower frequency "whoops". Modulate the filter cutoff a little with the envelope.
2) Add a very little (very little) amount of very slow phase shift.
3) Add some reverb and maybe a little delay of about 350 ms.
4) Pass this through a 10 band EQ.

Tinker with it until it sounds like the sound on that page.
ok a few questions:
about your number 1:
1) what do you mean buy PWM wave where the pulse width is set to vary slowly (about 2 LFO cycles per minute) between 25% and 50% duty cycle?
can you please explain?
2) what frequency should the sawtooth wave be?
also do you know a good program that can do all of this ?
 
btb4198 said:
what is (i.e. approaching 1.0) electronically generated?
Sorry. That was cryptic.

I was simply saying the probability that the sound was electronically-generated is pretty much certain.

You asking if if can be reverse-engineered. I'm thinking you want to know what instruments made it? No physical instrument made the sound, it was made by a piece of software. So I'm not sure what you expect to get from it.

What you can do is analyze the waveform. Envelope will tell you about the sound of each note (this is on the order of seconds). Timbre (tome) is a property of wave shape - each cycle of an individual cycle (this is on the order of milliseconds).
 
Back
Top