Carlo

Hi, I read about "SDIF-stft~" and "SDIF-istft~" modules from CNMAT in a paper. 

These objects are something I am desperately looking for a project I am doing now, but I'm not able to find them in any web site (not even in CNMAT's site).

Are they still under test? Or they will never be released?

Hi, I read about "SDIF-stft~" and "SDIF-istft~" modules from CNMAT in a paper. 
These objects are something I am desperately looking for a project I am doing now, but I'm not able to find them in any web site (not even in CNMAT's site).
Does someone know where to find them?
Are they still under test? Or they will never be released?

sdif-stft-and-sdif-istft

Quote: carlo-laurenzi wrote on Sat, 15 September 2007 11:13

----------------------------------------------------

> Hi, I read about "SDIF-stft~" and "SDIF-istft~" modules from CNMAT in a paper. 

> These objects are something I am desperately looking for a project I am doing now, but I'm not able to find them in any web site (not even in CNMAT's site).

> Are they still under test? Or they will never be released?

These are ancient objects that were never finished and aren't under development.

Quote: carlo-laurenzi wrote on Sat, 15 September 2007 11:13
----------------------------------------------------
> Hi, I read about "SDIF-stft~" and "SDIF-istft~" modules from CNMAT in a paper. 
> These objects are something I am desperately looking for a project I am doing now, but I'm not able to find them in any web site (not even in CNMAT's site).
> Does someone know where to find them?
> Are they still under test? Or they will never be released?
>
> Thank you very much for any info
>
> Carlo Laurenzi
>
>
----------------------------------------------------

Well, I thought that making spectral manipulations (time-variant) on sdif 

files (just numbers, not signals) created from complex audio material could 

be easier and could give better results (than using fft operations) since 

the enourmous potential of Additive Synthesis (with many oscs) is involved 

But it turned out that Sdif-based applications (spear included) aren't 

developed enough to handle very complex and heavy material (think of a 

recording of an entire piece of orchestral music, for example). And also, 

the dimension of a sdif file which can guarantee a decent result in 

resynthesis is so many many times high that a computer has big problems in 

I thought those problems were solved with those modules I mentioned... pity.

Thank you very much for your preciuos help.

Sent: Monday, September 17, 2007 11:45 PM

Subject: [maxmsp] Re: SDIF-stft~ and SDIF-istft~ ???

> These are ancient objects that were never finished and aren't under 

> || michael f. zbyszynski -- molecular gastronimist

Well, I thought that making spectral manipulations (time-variant) on sdif 
files (just numbers, not signals) created from complex audio material could 
be easier and could give better results (than using fft operations) since 
the enourmous potential of Additive Synthesis (with many oscs) is involved 
in the resynthesis.
But it turned out that Sdif-based applications (spear included) aren't 
developed enough to handle very complex and heavy material (think of a 
recording of an entire piece of orchestral music, for example). And also, 
the dimension of a sdif file which can guarantee a decent result in 
resynthesis is so many many times high that a computer has big problems in 
handling such amounts of data.
I thought those problems were solved with those modules I mentioned... pity.
Thank you very much for your preciuos help.

----- Original Message ----- 
From: "mzed" 
To: 
Sent: Monday, September 17, 2007 11:45 PM
Subject: [maxmsp] Re: SDIF-stft~ and SDIF-istft~ ???

>
> These are ancient objects that were never finished and aren't under 
> development.
>
> Maybe pfft~ does what you need?
>
> mz
>
> --
> || michael f. zbyszynski -- molecular gastronimist
> || http://www.cnmat.berkeley.edu/
> || http://www.mikezed.com/


> Well, I thought that making spectral manipulations (time-variant) on 

> sdif files (just numbers, not signals) created from complex audio 

> material could be easier and could give better results (than using fft 

> operations) since the enourmous potential of Additive Synthesis (with 

> many oscs) is involved in the resynthesis.

Spear is using fft, and the resulting SDIF file is just a representation 

of that data. It contains information which is reaching across fft frames...

> But it turned out that Sdif-based applications (spear included) aren't 

> developed enough to handle very complex and heavy material (think of a 

> recording of an entire piece of orchestral music, for example). And 

> also, the dimension of a sdif file which can guarantee a decent result 

> in resynthesis is so many many times high that a computer has big 

> problems in handling such amounts of data.

mp3 is also based on fft's, and a decent computer oldtimer can 

resynthesize even complex orchestra pieces...

But we still don't know what you're after. What do you want to do with 

the data? I can just tell you, that for simple oscillator based 

resynthesis of complex sounds. You can get pretty far with much less 

Carlo Laurenzi schrieb:
> Well, I thought that making spectral manipulations (time-variant) on 
> sdif files (just numbers, not signals) created from complex audio 
> material could be easier and could give better results (than using fft 
> operations) since the enourmous potential of Additive Synthesis (with 
> many oscs) is involved in the resynthesis.

Spear is using fft, and the resulting SDIF file is just a representation 
of that data. It contains information which is reaching across fft frames...

> But it turned out that Sdif-based applications (spear included) aren't 
> developed enough to handle very complex and heavy material (think of a 
> recording of an entire piece of orchestral music, for example). And 
> also, the dimension of a sdif file which can guarantee a decent result 
> in resynthesis is so many many times high that a computer has big 
> problems in handling such amounts of data.

mp3 is also based on fft's, and a decent computer oldtimer can 
resynthesize even complex orchestra pieces...

But we still don't know what you're after. What do you want to do with 
the data? I can just tell you, that for simple oscillator based 
resynthesis of complex sounds. You can get pretty far with much less 
than 1024 oscillators...

-- 
Stefan Tiedje------------x-------
--_____-----------|--------------
--(_|_ ----|-----|-----()-------
-- _|_)----|-----()--------------
----------()--------www.ccmix.com


>> But it turned out that Sdif-based applications (spear included) aren't 

>> developed enough to handle very complex and heavy material (think of a 

>> recording of an entire piece of orchestral music, for example). And also, 

>> the dimension of a sdif file which can guarantee a decent result in 

>> resynthesis is so many many times high that a computer has big problems 

> mp3 is also based on fft's, and a decent computer oldtimer can 

> resynthesize even complex orchestra pieces...

> But we still don't know what you're after. What do you want to do with the 

> data? I can just tell you, that for simple oscillator based resynthesis of 

> complex sounds. You can get pretty far with much less than 1024 

Well, the idea is to have a piece of music (a complete one, edited mixed and 

mastered) changing every specific amount of time through automatic 

Resinthesizing is not an issue when you're dealing with a simple 

analysis-resinthesis operation. But it becoms an issue if you do some heavy 

Artifacts and poor sound quality are some of the problems one can get into 

Sdif format, especially when the time between frames is very little, can 

give some results, but there is always a problem of too many data being 

Additive Synthesis is very powerful, but requires a lot of resources.

Too much power is needed to use sdif for those kinds of complex 

Maybe discrete wavelet transform might be a solution, but I need to know 

>> But it turned out that Sdif-based applications (spear included) aren't 
>> developed enough to handle very complex and heavy material (think of a 
>> recording of an entire piece of orchestral music, for example). And also, 
>> the dimension of a sdif file which can guarantee a decent result in 
>> resynthesis is so many many times high that a computer has big problems 
>> in handling such amounts of data.
>
> mp3 is also based on fft's, and a decent computer oldtimer can 
> resynthesize even complex orchestra pieces...
>
> But we still don't know what you're after. What do you want to do with the 
> data? I can just tell you, that for simple oscillator based resynthesis of 
> complex sounds. You can get pretty far with much less than 1024 
> oscillators...
>
> Stefan
>
> -- 
> Stefan Tiedje------------x-------
> --_____-----------|--------------
> --(_|_ ----|-----|-----()-------
> -- _|_)----|-----()--------------
> ----------()--------www.ccmix.com

Well, the idea is to have a piece of music (a complete one, edited mixed and 
mastered) changing every specific amount of time through automatic 
manipulations of its spectrum.
Resinthesizing is not an issue when you're dealing with a simple 
analysis-resinthesis operation. But it becoms an issue if you do some heavy 
manipulations before resynthesis.
Artifacts and poor sound quality are some of the problems one can get into 
at that point.
Sdif format, especially when the time between frames is very little, can 
give some results, but there is always a problem of too many data being 
handled by the computer.
Additive Synthesis is very powerful, but requires a lot of resources.
Too much power is needed to use sdif for those kinds of complex 
manipulations on complex spectra.
Maybe discrete wavelet transform might be a solution, but I need to know 
more about that.
Bye,

> Resinthesizing is not an issue when you're dealing with a simple 

> analysis-resinthesis operation. But it becoms an issue if you do some 

> heavy manipulations before resynthesis.

> Artifacts and poor sound quality are some of the problems one can get 

But if the resynthesizing is fine, the technology is fine. You have to 

enhance your manipulation if you get artefacts with the manipulation... 

That requires more knowledge about psychoacoustics than about DSP 

> Maybe discrete wavelet transform might be a solution, but I need to know 

At least it delivers a desription of the sound which is closer to the 

psychoacoustic listening model. But you can easily thin out the Spear 

data according to wavelets and see how far you get with that...

Its basically less resolution for higher frequency ranges...

I had good results by splitting a plain fft into tonal and noise parts, 

and just don't touch the noise part, only manipulate the tonal part...

That research isn't finished yet, I can't deliver the results in this 

Carlo Laurenzi schrieb:
> Resinthesizing is not an issue when you're dealing with a simple 
> analysis-resinthesis operation. But it becoms an issue if you do some 
> heavy manipulations before resynthesis.
> Artifacts and poor sound quality are some of the problems one can get 
> into at that point.

But if the resynthesizing is fine, the technology is fine. You have to 
enhance your manipulation if you get artefacts with the manipulation... 
That requires more knowledge about psychoacoustics than about DSP 
technology btw...

> Maybe discrete wavelet transform might be a solution, but I need to know 
> more about that.

At least it delivers a desription of the sound which is closer to the 
psychoacoustic listening model. But you can easily thin out the Spear 
data according to wavelets and see how far you get with that...
Its basically less resolution for higher frequency ranges...

I had good results by splitting a plain fft into tonal and noise parts, 
and just don't touch the noise part, only manipulate the tonal part...
That research isn't finished yet, I can't deliver the results in this 
stage, but stay tuned...

-- 
Stefan Tiedje------------x-------
--_____-----------|--------------
--(_|_ ----|-----|-----()-------
-- _|_)----|-----()--------------
----------()--------www.ccmix.com

I know I need to improve my manipulation algorithms to get good results out 

of sdif representation of long and complex music materials.

I should also buy a powerful workhorse computer to handle all those quite 

heavy sdif files (almost 50 times the size in bits of the original wav 

I am into discrete wavelet transform right now, yet still interested in 

using a bunch of time-variant oscillators to do complex manipulations on 

I'm very interested in what you're doing. Maybe you can bring some light in 

Subject: Re: [maxmsp] Re: SDIF-stft~ and SDIF-istft~ ???

>> Resinthesizing is not an issue when you're dealing with a simple 

>> analysis-resinthesis operation. But it becoms an issue if you do some 

>> heavy manipulations before resynthesis.

>> Artifacts and poor sound quality are some of the problems one can get 

> But if the resynthesizing is fine, the technology is fine. You have to 

> enhance your manipulation if you get artefacts with the manipulation... 

> That requires more knowledge about psychoacoustics than about DSP 

>> Maybe discrete wavelet transform might be a solution, but I need to know 

> At least it delivers a desription of the sound which is closer to the 

> psychoacoustic listening model. But you can easily thin out the Spear data 

> according to wavelets and see how far you get with that...

> Its basically less resolution for higher frequency ranges...

> I had good results by splitting a plain fft into tonal and noise parts, 

> and just don't touch the noise part, only manipulate the tonal part...

> That research isn't finished yet, I can't deliver the results in this 

Hi Stefan,
I know I need to improve my manipulation algorithms to get good results out 
of sdif representation of long and complex music materials.
I should also buy a powerful workhorse computer to handle all those quite 
heavy sdif files (almost 50 times the size in bits of the original wav 
data).. !!
I am into discrete wavelet transform right now, yet still interested in 
using a bunch of time-variant oscillators to do complex manipulations on 
signals spectra..
I'm very interested in what you're doing. Maybe you can bring some light in 
my world!
I'll wait for your results.
Best,

----- Original Message ----- 
From: "Stefan Tiedje" 
To: 
Sent: Sunday, September 23, 2007 7:26 PM
Subject: Re: [maxmsp] Re: SDIF-stft~ and SDIF-istft~ ???

> Carlo Laurenzi schrieb:
>> Resinthesizing is not an issue when you're dealing with a simple 
>> analysis-resinthesis operation. But it becoms an issue if you do some 
>> heavy manipulations before resynthesis.
>> Artifacts and poor sound quality are some of the problems one can get 
>> into at that point.
>
> But if the resynthesizing is fine, the technology is fine. You have to 
> enhance your manipulation if you get artefacts with the manipulation... 
> That requires more knowledge about psychoacoustics than about DSP 
> technology btw...
>
>> Maybe discrete wavelet transform might be a solution, but I need to know 
>> more about that.
>
> At least it delivers a desription of the sound which is closer to the 
> psychoacoustic listening model. But you can easily thin out the Spear data 
> according to wavelets and see how far you get with that...
> Its basically less resolution for higher frequency ranges...
>
> I had good results by splitting a plain fft into tonal and noise parts, 
> and just don't touch the noise part, only manipulate the tonal part...
> That research isn't finished yet, I can't deliver the results in this 
> stage, but stay tuned...
>
> Stefan
>
> -- 
> Stefan Tiedje------------x-------
> --_____-----------|--------------
> --(_|_ ----|-----|-----()-------
> -- _|_)----|-----()--------------
> ----------()--------www.ccmix.com
