Hi all

Eksample: 
I find a synth sound on Youtube or elsewhere and want to reverse-engineering
it, to figure out a recipe (in Csound) for this synth sound. 

I know that all sounds are made of a lot of sine frequences added, so I
choose to use additive synthesis as my methode. 

I use a Fourier transformation with pvsanal or pvstanal. I hope to find some
frequencies and envelopes to my additive synthesis recipe.

The problem is that these opcodes makes a f-signal file, and I can not see
which frequencies and envelopes it "hides inside". 



--
View this message in context: http://csound.1045644.n5.nabble.com/f-signal-hides-frequencies-envelopes-and-other-information-tp5719676.html
Sent from the Csound - General mailing list archive at Nabble.com.

It might be better to just load the sample into a spectral viewer to
see which frequencies are contained within the tone and go from there.
It might also help if you can figure out what synthesis technique was
used the begin with. Some of the classic synth sounds are pretty well
documented. Finally, there are some pretty good ears on this list for
identifying likely synth techniques for different sounds.

Rory.


> Hi all
>
> Eksample:
> I find a synth sound on Youtube or elsewhere and want to reverse-engineering
> it, to figure out a recipe (in Csound) for this synth sound.
>
> I know that all sounds are made of a lot of sine frequences added, so I
> choose to use additive synthesis as my methode.
>
> I use a Fourier transformation with pvsanal or pvstanal. I hope to find some
> frequencies and envelopes to my additive synthesis recipe.
>
> The problem is that these opcodes makes a f-signal file, and I can not see
> which frequencies and envelopes it "hides inside".
>
>
>
> --
> View this message in context: http://csound.1045644.n5.nabble.com/f-signal-hides-frequencies-envelopes-and-other-information-tp5719676.html
> Sent from the Csound - General mailing list archive at Nabble.com.
>
>
> Send bugs reports to the Sourceforge bug tracker
>             https://sourceforge.net/tracker/?group_id=81968&atid=564599
> Discussions of bugs and features can be posted here
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
>

"any sound that is not infinite in duration is made of an infine number of
sine frequencies, and conversely any sound that is finite in spectral
content is infinite in duration"

So, because i do not have an infinite life, i guess like anybody, i choose a
sound with a finite duration
but how many time i need to read all the sines? :)

ciao,
francesco.




--
View this message in context: http://csound.1045644.n5.nabble.com/f-signal-hides-frequencies-envelopes-and-other-information-tp5719676p5719683.html
Sent from the Csound - General mailing list archive at Nabble.com.

The format of f-signals is simple -- usually magnitude and frequency or
phase for each element, or real-imaginary pairs (all three variants
exist).  You can write a f-var to a table with  pvsftw  and then unpick
it.
Whether it is useful is another issue.

Or you could use the pvoc analysis utility and decode with pv_export --
again no promises that it will lead to what you want....


> Hi all
>
> Eksample:
> I find a synth sound on Youtube or elsewhere and want to
> reverse-engineering
> it, to figure out a recipe (in Csound) for this synth sound.
>
> I know that all sounds are made of a lot of sine frequences added, so I
> choose to use additive synthesis as my methode.
>
> I use a Fourier transformation with pvsanal or pvstanal. I hope to find
> some
> frequencies and envelopes to my additive synthesis recipe.
>
> The problem is that these opcodes makes a f-signal file, and I can not see
> which frequencies and envelopes it "hides inside".
>
>
>
> --
> View this message in context:
> http://csound.1045644.n5.nabble.com/f-signal-hides-frequencies-envelopes-and-other-information-tp5719676.html
> Sent from the Csound - General mailing list archive at Nabble.com.
>
>
> Send bugs reports to the Sourceforge bug tracker
>             https://sourceforge.net/tracker/?group_id=81968&atid=564599
> Discussions of bugs and features can be posted here
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe
> csound"
>
>
>
>

Yes,
You are right, i was joking.
Not having a scientific background, always i wonder how we can speak or feel
about infinite ...
But it a really old question, i guess.

never mind ...

ciao,
francesco.




--
View this message in context: http://csound.1045644.n5.nabble.com/f-signal-hides-frequencies-envelopes-and-other-information-tp5719676p5719688.html
Sent from the Csound - General mailing list archive at Nabble.com.

On 29/01/2013 19:03, Justin Smith wrote:
> and pedantically, any sound that is not infinite in duration is made of
> an infine number of sine frequencies, and conversely any sound that is
> finite in spectral content is infinite in duration
>
>

I don't think things are quite as dire as all that. This description 
really applies more to a filter - the more vertical the brickwall, the 
longer the impulse response, until for a (strictly theoretical) 
literally discontinuous step frequency response, the impulse response is 
infinite - it rings forever.

In practice, for the sort of running analysis sought, the "short-time" 
FFT is used with overlapping windows tastefully squished at the ends to 
minimise spectral leakage (and hopefully present all genuine components 
in clear relief). As sound analysis goes, it wroks pretty well so long 
as the limitations are understood. So, an fsig (or a PVOCEX file) is 
really just a first stage in the much more involved process of detecting 
and tracking peaks frame by frame.

The problem with using Csound for this sort of task is that it is really 
predicated on streaming and transformation, not "scientific" offline 
analysis. There are opcodes that will extract time-varying peaks, which 
can be rendered via oscillator bank etc, and likewise utilities that 
will create various forms of partial tracking output files (ats, 
partials, even good old hetro). So processing and resynthesisng partial 
tracks is relatively easy.

Studying them is more of a problem, as displaying (and analysing) the 
resulting data is quite a task. There can be a lot of data involved, 
perhaps 1000s of individual tracks. Reverse engineering how a sound was 
made will be a complex and decidedly non-trivial exercise, except in 
reasonably obvious (trivial?) cases. As suggested, the Mark-1 human ears 
plus the stuff between them may be more successful. Cecilia was one of 
the few graphic front ends which offered a means to visually inspect 
hetro data as editable harmonic-partial breakpoint files. For plain 
monophonic pitched tones, without too much variation of frequency or 
noise, hetro could work surprisingly well!


Richard Dobson

Thank you all

My idea was that I could pick, say, the 20 most significant partials and
then use them to make an additive
synth. It should not be a perfect resynthesis, only an approximation. If the
sound source is another synthesizer, then the partials should be quite
regular. The opposite schould be the case if I use a natural sound. e.g. a
singing bird. If I could spot a special relation between some of the
partials, then I perhaps could use a saw osc or puls modulation etc.

I have a little question:
What is the difference between a fsig or a PVOCEX file ?



--
View this message in context: http://csound.1045644.n5.nabble.com/f-signal-hides-frequencies-envelopes-and-other-information-tp5719676p5719716.html
Sent from the Csound - General mailing list archive at Nabble.com.

On 30/01/2013 17:50, Bjoern Houdorf wrote:
> Thank you all
>
> My idea was that I could pick, say, the 20 most significant partials and
> then use them to make an additive
> synth. It should not be a perfect resynthesis, only an approximation. If the
> sound source is another synthesizer, then the partials should be quite
> regular. The opposite schould be the case if I use a natural sound. e.g. a
> singing bird. If I could spot a special relation between some of the
> partials, then I perhaps could use a saw osc or puls modulation etc.
>
> I have a little question:
> What is the difference between a fsig or a PVOCEX file ?
>


The frequency domain counterpart to the difference between an asig and a 
soundfile. The one is the (real-time) streaming version of the other. An 
  fsig can be read from a PVOCEX file, and can similarly be written to one.

Richard Dobson

hi bjoern -

i did not follow the discussion, but reading

 > pick, say, the 20 most significant partials and
 > then use them to make an additive synth

reminds me what i tried to implement in this way. if you use csoundqt, 
you can give a try to
Examples > Synths > Imitative Additive

more explanation are here:
http://lac.linuxaudio.org/2011/papers/20.pdf

best -

	joachim


Am 30.01.2013 18:50, schrieb Bjoern Houdorf:
> Thank you all
>
> My idea was that I could pick, say, the 20 most significant partials and
> then use them to make an additive
> synth. It should not be a perfect resynthesis, only an approximation. If the
> sound source is another synthesizer, then the partials should be quite
> regular. The opposite schould be the case if I use a natural sound. e.g. a
> singing bird. If I could spot a special relation between some of the
> partials, then I perhaps could use a saw osc or puls modulation etc.
>
> I have a little question:
> What is the difference between a fsig or a PVOCEX file ?
>
>
>
> --
> View this message in context: http://csound.1045644.n5.nabble.com/f-signal-hides-frequencies-envelopes-and-other-information-tp5719676p5719716.html
> Sent from the Csound - General mailing list archive at Nabble.com.
>
>
> Send bugs reports to the Sourceforge bug tracker
>              https://sourceforge.net/tracker/?group_id=81968&atid=564599
> Discussions of bugs and features can be posted here
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
>
>

Oh Lord, forgive my sines.

only if you confess all your cosines.

Am 30.01.2013 20:33, schrieb Oeyvind Brandtsegg:
> Oh Lord, forgive my sines.
>
> Den 29. jan. 2013 20:41 skrev "francesco"  > følgende:
>
>
>     "any sound that is not infinite in duration is made of an infine
>     number of
>     sine frequencies, and conversely any sound that is finite in spectral
>     content is infinite in duration"
>
>     So, because i do not have an infinite life, i guess like anybody, i
>     choose a
>     sound with a finite duration
>     but how many time i need to read all the sines? :)
>
>     ciao,
>     francesco.
>
>
>
>
>     --
>     View this message in context:
>     http://csound.1045644.n5.nabble.com/f-signal-hides-frequencies-envelopes-and-other-information-tp5719676p5719683.html
>     Sent from the Csound - General mailing list archive at Nabble.com.
>
>
>     Send bugs reports to the Sourceforge bug tracker
>     https://sourceforge.net/tracker/?group_id=81968&atid=564599
>     Discussions of bugs and features can be posted here
>     To unsubscribe, send email sympa@lists.bath.ac.uk
>      with body "unsubscribe csound"
>

Ego Te Absolvo

ciao,
pater francesco



--
View this message in context: http://csound.1045644.n5.nabble.com/f-signal-hides-frequencies-envelopes-and-other-information-tp5719676p5719723.html
Sent from the Csound - General mailing list archive at Nabble.com.

Richard Dobson wrote
> The frequency domain counterpart to the difference between an asig and a 
> soundfile. The one is the (real-time) streaming version of the other. An 
>   fsig can be read from a PVOCEX file, and can similarly be written to
> one.

Nice to know!



joachim-3 wrote
> if you use csoundqt, 
> you can give a try to
> Examples > Synths > Imitative Additive
> 
> more explanation are here:
> http://lac.linuxaudio.org/2011/papers/20.pdf

Interesting paper!


/Bjørn





--
View this message in context: http://csound.1045644.n5.nabble.com/f-signal-hides-frequencies-envelopes-and-other-information-tp5719676p5719730.html
Sent from the Csound - General mailing list archive at Nabble.com.

Waveforms FTW!

On 1/31/13, Bjoern Houdorf  wrote:
>
>
> Richard Dobson wrote
>> The frequency domain counterpart to the difference between an asig and a
>> soundfile. The one is the (real-time) streaming version of the other. An
>>   fsig can be read from a PVOCEX file, and can similarly be written to
>> one.
>
> Nice to know!
>
>
>
> joachim-3 wrote
>> if you use csoundqt,
>> you can give a try to
>> Examples > Synths > Imitative Additive
>>
>> more explanation are here:
>> http://lac.linuxaudio.org/2011/papers/20.pdf
>
> Interesting paper!
>
>
> /Bjørn
>
>
>
>
>
> --
> View this message in context:
> http://csound.1045644.n5.nabble.com/f-signal-hides-frequencies-envelopes-and-other-information-tp5719676p5719730.html
> Sent from the Csound - General mailing list archive at Nabble.com.
>
>
> Send bugs reports to the Sourceforge bug tracker
>             https://sourceforge.net/tracker/?group_id=81968&atid=564599
> Discussions of bugs and features can be posted here
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe
> csound"
>
>