Put random phases to each frequency of the spectrum
Do a single Inverse Fourier Transform of the whole spectrum.

I had taken a look at this and I think I got stuck doing the inverse
FFT.  I think I had some trouble figuring out how to do the iFFT from
complex numbers within Csound.  If anything, I assume it was a
deficiency in my understanding of FFT and DSP code (a summertime
project to brush up on).

The author provides a C++ class for free use, which wouldn't be very
difficult I think to create a csound opcode from, but I was interested
in convert it into csound ORC code.  As far as I understand, the
algorithm essentially generates a table for use with a table-based
oscillator. Also, the ZynAddSubFX site mentions exporting for use with
other instruments, which I would then imagine you could load into an
ftable and run in Csound (have not tried this as I was interested to
have this in Csound itself).

I'll probably resume on this sometime this summer as time permits,
though if someone else with better understanding of FFT's takes a look
and sees an easy path, that'd be great to hear. :)

On Wed, May 16, 2012 at 7:05 AM, Richard van Bemmelen
 wrote:
> Very interesting concepts in this article!
>
> Richard
>
>
> 2012/5/16 Stefan Thomas 
>>
>> Dear community,
>> I've read about the padsynth of zynaddsubfx, at
>> http://zynaddsubfx.sourceforge.net/doc/PADsynth/PADsynth.htm
>> I want to know how theses two things can be done within csound:
>>>
>>> Put random phases to each frequency of the spectrum
>>> Do a single Inverse Fourier Transform of the whole spectrum.
>>
>> Has someone tried something like this? What do You think about the
>> abilities of zynnaddsubfx?
>
>

I hope that this opcode is added to Csound

Sent from my iPad.

On May 16, 2012, at 7:59 AM, Steven Yi  wrote:

> I had taken a look at this and I think I got stuck doing the inverse
> FFT.  I think I had some trouble figuring out how to do the iFFT from
> complex numbers within Csound.  If anything, I assume it was a
> deficiency in my understanding of FFT and DSP code (a summertime
> project to brush up on).
> 
> The author provides a C++ class for free use, which wouldn't be very
> difficult I think to create a csound opcode from, but I was interested
> in convert it into csound ORC code.  As far as I understand, the
> algorithm essentially generates a table for use with a table-based
> oscillator. Also, the ZynAddSubFX site mentions exporting for use with
> other instruments, which I would then imagine you could load into an
> ftable and run in Csound (have not tried this as I was interested to
> have this in Csound itself).
> 
> I'll probably resume on this sometime this summer as time permits,
> though if someone else with better understanding of FFT's takes a look
> and sees an easy path, that'd be great to hear. :)
> 
> On Wed, May 16, 2012 at 7:05 AM, Richard van Bemmelen
>  wrote:
>> Very interesting concepts in this article!
>> 
>> Richard
>> 
>> 
>> 2012/5/16 Stefan Thomas 
>>> 
>>> Dear community,
>>> I've read about the padsynth of zynaddsubfx, at
>>> http://zynaddsubfx.sourceforge.net/doc/PADsynth/PADsynth.htm
>>> I want to know how theses two things can be done within csound:
>>>> 
>>>> Put random phases to each frequency of the spectrum
>>>> Do a single Inverse Fourier Transform of the whole spectrum.
>>> 
>>> Has someone tried something like this? What do You think about the
>>> abilities of zynnaddsubfx?
>> 
>> 
> 
> 
> 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"
> 

Theoretically, this could be accomplished by multiplying (in the frequency
domain) a signal which has all frequencies and a signal representing the
peaks.
Noise comes to mind as a source, but it has random fluctuations.
The peaks could be created by parallel band pass filters as a first
approximation.

Ideally, if it is possible to create such a 'wall of sound' opcode
(containing all frequencies) and an opcode that  can represent any frequency
spectrum content, this would open up a whole new way of sound creation.

Richard 

--
View this message in context: http://csound.1045644.n5.nabble.com/padsynth-like-zynaddsubfx-tp5710678p5711287.html
Sent from the Csound - General mailing list archive at Nabble.com.

I already did this late last year, but due to a busy job I didn't finish 
the last bits. I actually only wanted to add optional formants to 
produce vowels. Today is a holiday, but I have to go somewhere soon, 
maybe I can post a csd tonight without the formants.

It's actually been on my mind often to share, because it is a nice 
instrument.

Mark

Op 16/05/2012 12:57, Stefan Thomas schreef:
> Dear community,
> I've read about the padsynth of zynaddsubfx, at 
> http://zynaddsubfx.sourceforge.net/doc/PADsynth/PADsynth.htm
> I want to know how theses two things can be done within csound:
>
>     Put random phases to each frequency of the spectrum
>     Do a single Inverse Fourier Transform of the whole spectrum.
>
> Has someone tried something like this? What do You think about the 
> abilities of zynnaddsubfx?

Ok, here it is: a csd file and a blue file with my CSound implementation 
of this instrument. It includes 7 different shapes for the peaks, 
scaling of bandwidth, detuning, random phase start and stereo phase 
difference, like described in the article. I also added the option to 
use only peaks at odds harmonics.

I remember I experimented quite a bit more with it, like making the 
phases less random by making them a random number between 0 and 0.2 
instead of 0 and 1, or adding a constant small value to every amplitude, 
but that made it less interesting.

Enjoy!

Mark

Op 17/05/2012 11:19, Mark Van Peteghem schreef:
> I already did this late last year, but due to a busy job I didn't 
> finish the last bits. I actually only wanted to add optional formants 
> to produce vowels. Today is a holiday, but I have to go somewhere 
> soon, maybe I can post a csd tonight without the formants.
>
> It's actually been on my mind often to share, because it is a nice 
> instrument.
>
> Mark
>
> Op 16/05/2012 12:57, Stefan Thomas schreef:
>> Dear community,
>> I've read about the padsynth of zynaddsubfx, at 
>> http://zynaddsubfx.sourceforge.net/doc/PADsynth/PADsynth.htm
>> I want to know how theses two things can be done within csound:
>>
>>     Put random phases to each frequency of the spectrum
>>     Do a single Inverse Fourier Transform of the whole spectrum.
>>
>> Has someone tried something like this? What do You think about the 
>> abilities of zynnaddsubfx?
>
>
>
> 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"
>
>

-- 
   Mark
   _________________________________________
   When you get lemons, you make lemonade.
   When you get hardware, you make software.

Hi Mark,

Thanks for sharing this!  I see now that you are using gen33; I hadn't
thought to look through the gen routines as I so rarely use them, and
was looking at only opcode solutions.  This is great. :)

When I have some time, I'll probably rework what I had and incorporate
what you've done, and modify it to work with realtime changes
(automatable, but reading the values at note start with i()).  I have
a few more instruments I need to finish tweaking and am planning to
release to blueShare within the next month as well.

Thanks again!
steven


On Thu, May 17, 2012 at 3:25 PM, Mark Van Peteghem
 wrote:
> Ok, here it is: a csd file and a blue file with my CSound implementation of
> this instrument. It includes 7 different shapes for the peaks, scaling of
> bandwidth, detuning, random phase start and stereo phase difference, like
> described in the article. I also added the option to use only peaks at odds
> harmonics.
>
> I remember I experimented quite a bit more with it, like making the phases
> less random by making them a random number between 0 and 0.2 instead of 0
> and 1, or adding a constant small value to every amplitude, but that made it
> less interesting.
>
> Enjoy!
>
> Mark
>
> Op 17/05/2012 11:19, Mark Van Peteghem schreef:
>
>> I already did this late last year, but due to a busy job I didn't finish
>> the last bits. I actually only wanted to add optional formants to produce
>> vowels. Today is a holiday, but I have to go somewhere soon, maybe I can
>> post a csd tonight without the formants.
>>
>> It's actually been on my mind often to share, because it is a nice
>> instrument.
>>
>> Mark
>>
>> Op 16/05/2012 12:57, Stefan Thomas schreef:
>>>
>>> Dear community,
>>> I've read about the padsynth of zynaddsubfx, at
>>> http://zynaddsubfx.sourceforge.net/doc/PADsynth/PADsynth.htm
>>> I want to know how theses two things can be done within csound:
>>>
>>>    Put random phases to each frequency of the spectrum
>>>    Do a single Inverse Fourier Transform of the whole spectrum.
>>>
>>> Has someone tried something like this? What do You think about the
>>> abilities of zynnaddsubfx?
>>
>>
>>
>>
>> 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"
>>
>>
>
> --
>  Mark
>  _________________________________________
>  When you get lemons, you make lemonade.
>  When you get hardware, you make software.
>
>
>
> 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"
>

Hi Stefan,

I'm glad you like it. Yes it's complicated, probably because I added 
features one at a time, then optimized some things, so if you see the 
whole at once, it looks complicated. The most important to understand is 
the double loop: a loop over the harmonics, and in that a loop that 
calculates the amplitudes for a range of frequencies around that 
harmonic. For some peak shapes I first calculate at what frequencies to 
start and end; for gauss shapes this is only an approximation.

If you have more questions, ask me.

Mark

Op 17/05/2012 21:56, Stefan Thomas schreef:
> Dear Mark,
> thank You for sharing Your csd-file!
> For me it looks a bit complitated (for my understanding), but I will 
> try to understand it!
>
>
> 2012/5/17 Mark Van Peteghem  >
>
>     Ok, here it is: a csd file and a blue file with my CSound
>     implementation of this instrument. It includes 7 different shapes
>     for the peaks, scaling of bandwidth, detuning, random phase start
>     and stereo phase difference, like described in the article. I also
>     added the option to use only peaks at odds harmonics.
>
>     I remember I experimented quite a bit more with it, like making
>     the phases less random by making them a random number between 0
>     and 0.2 instead of 0 and 1, or adding a constant small value to
>     every amplitude, but that made it less interesting.
>
>     Enjoy!
>
>     Mark
>
>     Op 17/05/2012 11:19, Mark Van Peteghem schreef:
>
>         I already did this late last year, but due to a busy job I
>         didn't finish the last bits. I actually only wanted to add
>         optional formants to produce vowels. Today is a holiday, but I
>         have to go somewhere soon, maybe I can post a csd tonight
>         without the formants.
>
>         It's actually been on my mind often to share, because it is a
>         nice instrument.
>
>         Mark
>
>         Op 16/05/2012 12:57, Stefan Thomas schreef:
>
>             Dear community,
>             I've read about the padsynth of zynaddsubfx, at
>             http://zynaddsubfx.sourceforge.net/doc/PADsynth/PADsynth.htm
>             I want to know how theses two things can be done within
>             csound:
>
>                Put random phases to each frequency of the spectrum
>                Do a single Inverse Fourier Transform of the whole
>             spectrum.
>
>             Has someone tried something like this? What do You think
>             about the abilities of zynnaddsubfx?
>
>
>
>
>         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"
>
>
>
>     -- 
>      Mark
>      _________________________________________
>      When you get lemons, you make lemonade.
>      When you get hardware, you make software.
>
>
>
>     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"
>
>

-- 
   Mark
   _________________________________________
   When you get lemons, you make lemonade.
   When you get hardware, you make software.

Hi Steven,

I did not allow for the table to be adjusted by realtime changes because 
I thought generating the table would not be fast enough. I measured it 
with rtclock yesterday, it seems that it takes between 80 and 120 ms. 
Reducing the table size from 2^18 to 2^16 would probably make it more 
than 4 times faster (since time needed for FFT is proportional to n*log 
n) and not reduce quality too much. I'm not sure  if that would make it 
fast enough for real time changes.

Mark

Op 18/05/2012 3:22, Steven Yi schreef:
> Hi Mark,
>
> Thanks for sharing this!  I see now that you are using gen33; I hadn't
> thought to look through the gen routines as I so rarely use them, and
> was looking at only opcode solutions.  This is great. :)
>
> When I have some time, I'll probably rework what I had and incorporate
> what you've done, and modify it to work with realtime changes
> (automatable, but reading the values at note start with i()).  I have
> a few more instruments I need to finish tweaking and am planning to
> release to blueShare within the next month as well.
>
> Thanks again!
> steven
>
>
> On Thu, May 17, 2012 at 3:25 PM, Mark Van Peteghem
>   wrote:
>> Ok, here it is: a csd file and a blue file with my CSound implementation of
>> this instrument. It includes 7 different shapes for the peaks, scaling of
>> bandwidth, detuning, random phase start and stereo phase difference, like
>> described in the article. I also added the option to use only peaks at odds
>> harmonics.
>>
>> I remember I experimented quite a bit more with it, like making the phases
>> less random by making them a random number between 0 and 0.2 instead of 0
>> and 1, or adding a constant small value to every amplitude, but that made it
>> less interesting.
>>
>> Enjoy!
>>
>> Mark
>>
>> Op 17/05/2012 11:19, Mark Van Peteghem schreef:
>>
>>> I already did this late last year, but due to a busy job I didn't finish
>>> the last bits. I actually only wanted to add optional formants to produce
>>> vowels. Today is a holiday, but I have to go somewhere soon, maybe I can
>>> post a csd tonight without the formants.
>>>
>>> It's actually been on my mind often to share, because it is a nice
>>> instrument.
>>>
>>> Mark
>>>
>>> Op 16/05/2012 12:57, Stefan Thomas schreef:
>>>> Dear community,
>>>> I've read about the padsynth of zynaddsubfx, at
>>>> http://zynaddsubfx.sourceforge.net/doc/PADsynth/PADsynth.htm
>>>> I want to know how theses two things can be done within csound:
>>>>
>>>>     Put random phases to each frequency of the spectrum
>>>>     Do a single Inverse Fourier Transform of the whole spectrum.
>>>>
>>>> Has someone tried something like this? What do You think about the
>>>> abilities of zynnaddsubfx?
>>>
>>>
>>>
>>> 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"
>>>
>>>
>> --
>>   Mark
>>   _________________________________________
>>   When you get lemons, you make lemonade.
>>   When you get hardware, you make software.
>>
>>
>>
>> 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"
>>
>
> 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"
>
>

-- 
   Mark
   _________________________________________
   When you get lemons, you make lemonade.
   When you get hardware, you make software.

Hi Mark,

I think in this case the time required wouldn't be an issue, as it'd
be part of the musical workflow when one would be experimenting with
sounds, rather than during a live performance.  Already one needs to
take the hit with the current design for the first note.  The proposed
design would do some caching, checking if there's a change.  If not,
reuse the existing table, if so, regenerate.

The scenario for this design I am picturing is starting up blueLive
and playing MIDI keyboard or using blueLive's live object system with
repeat on.  I think it'd speed up the process of exploring patch
creation a good bit, even if there's a bit of wait and breakup while
the new table is generated; it'd still be faster than restarting a new
Csound render, IMO.

In general, I've been designing my blue instruments this way, where
parameters that need to be fixed at init-time use i() with k-rate vars
set by blueSynthBuilder knobs/sliders/etc.  It ends up being that some
parameters can be modified while the instrument is running, while
others change only when new notes are initiated.  On the whole, I
think it's a fair-enough tradeoff between some possible breakup in
audio versus improvement in workflow.

Thanks!
steven

On Sun, May 20, 2012 at 3:52 AM, Mark Van Peteghem
 wrote:
> Hi Steven,
>
> I did not allow for the table to be adjusted by realtime changes because I
> thought generating the table would not be fast enough. I measured it with
> rtclock yesterday, it seems that it takes between 80 and 120 ms. Reducing
> the table size from 2^18 to 2^16 would probably make it more than 4 times
> faster (since time needed for FFT is proportional to n*log n) and not reduce
> quality too much. I'm not sure  if that would make it fast enough for real
> time changes.
>
> Mark
>
> Op 18/05/2012 3:22, Steven Yi schreef:
>
>> Hi Mark,
>>
>> Thanks for sharing this!  I see now that you are using gen33; I hadn't
>> thought to look through the gen routines as I so rarely use them, and
>> was looking at only opcode solutions.  This is great. :)
>>
>> When I have some time, I'll probably rework what I had and incorporate
>> what you've done, and modify it to work with realtime changes
>> (automatable, but reading the values at note start with i()).  I have
>> a few more instruments I need to finish tweaking and am planning to
>> release to blueShare within the next month as well.
>>
>> Thanks again!
>> steven
>>
>>
>> On Thu, May 17, 2012 at 3:25 PM, Mark Van Peteghem
>>   wrote:
>>>
>>> Ok, here it is: a csd file and a blue file with my CSound implementation
>>> of
>>> this instrument. It includes 7 different shapes for the peaks, scaling of
>>> bandwidth, detuning, random phase start and stereo phase difference, like
>>> described in the article. I also added the option to use only peaks at
>>> odds
>>> harmonics.
>>>
>>> I remember I experimented quite a bit more with it, like making the
>>> phases
>>> less random by making them a random number between 0 and 0.2 instead of 0
>>> and 1, or adding a constant small value to every amplitude, but that made
>>> it
>>> less interesting.
>>>
>>> Enjoy!
>>>
>>> Mark
>>>
>>> Op 17/05/2012 11:19, Mark Van Peteghem schreef:
>>>
>>>> I already did this late last year, but due to a busy job I didn't finish
>>>> the last bits. I actually only wanted to add optional formants to
>>>> produce
>>>> vowels. Today is a holiday, but I have to go somewhere soon, maybe I can
>>>> post a csd tonight without the formants.
>>>>
>>>> It's actually been on my mind often to share, because it is a nice
>>>> instrument.
>>>>
>>>> Mark
>>>>
>>>> Op 16/05/2012 12:57, Stefan Thomas schreef:
>>>>>
>>>>> Dear community,
>>>>> I've read about the padsynth of zynaddsubfx, at
>>>>> http://zynaddsubfx.sourceforge.net/doc/PADsynth/PADsynth.htm
>>>>> I want to know how theses two things can be done within csound:
>>>>>
>>>>>    Put random phases to each frequency of the spectrum
>>>>>    Do a single Inverse Fourier Transform of the whole spectrum.
>>>>>
>>>>> Has someone tried something like this? What do You think about the
>>>>> abilities of zynnaddsubfx?
>>>>
>>>>
>>>>
>>>>
>>>> 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"
>>>>
>>>>
>>> --
>>>  Mark
>>>  _________________________________________
>>>  When you get lemons, you make lemonade.
>>>  When you get hardware, you make software.
>>>
>>>
>>>
>>> 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"
>>>
>>
>> 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"
>>
>>
>
> --
>  Mark
>  _________________________________________
>  When you get lemons, you make lemonade.
>  When you get hardware, you make software.
>
>
>
> 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"
>

Nice sounds. How would I make this midi playable?
I tried several midi options in Blue, but to no effect.
On the other hand, I see no line of code in the csd relating to midi. 
Is Blue adding these 'behind the scenes' or not?


Richard 

--
View this message in context: http://csound.1045644.n5.nabble.com/padsynth-like-zynaddsubfx-tp5710678p5713128.html
Sent from the Csound - General mailing list archive at Nabble.com.