[PD] oscillators (osc~ / cycle~) not working well in FM?

Wed Nov 25 00:01:22 CET 2015

Okay, here is a compromise.  What may be a fix, or a new feature.  It changes the oscillator to double precision, but costs a few more CPU cycles in the process.  This seems to keep things more accurate. Attached is the source and a compiled version for OS 10.10+ and this is the git repository if anyone is interested. 
  But I still say let’s keep osc~ as is and just add the double precision as an alternative.  

https://bitbucket.org/resler/osc2/overview <https://bitbucket.org/resler/osc2/overview>

Best,
Rob

> On Nov 24, 2015, at 10:39 AM, Jonathan Wilkes <jancsika at yahoo.com> wrote:
> 
> Does anyone have an example of a working patch that depends on the 
> current behavior?
> 
> -Jonathan
> 
> 
> 
> On Tuesday, November 24, 2015 9:52 AM, Alexandre Torres Porres <porres at gmail.com> wrote:
> 
> 
> > It's not that hard to add microscopic DC
> > if that's what you want
> 
> I almost raised that up :) and it's easier than trying to fix it every time.
> 
> I've lived with osc~ for over a decade, but it's like I finally found its dirty secret bug. Had I not found it I'd still be ok with it, but had I found this it years ago I'd have brought it up and asked for a fix.
> 
> I think it's pushing it to compare legacy analog hardware with their signature sounds (which would mostly come down to filters, not sine waves) to an imprecise and easily fix digital code - it's not like it's the secret formula for digitally emulating the moog filter or something - it's just bad.
> 
> Don't be afraid of changes and fixes, not sure what children's book teach us that lesson, but there might or ought to be one.
> 
> cheers 
> 
> 2015-11-24 3:18 GMT-02:00 Matt Barber <brbrofsvl at gmail.com <mailto:brbrofsvl at gmail.com>>:
> Right, there are good reasons to want that behavior, but should it be the default in a program that aspires to be "deterministic"? It's not that hard to add microscopic DC if that's what you want, or add a tiny, tiny bit of low-pass-filtered noise to you oscillator to make it act more like acoustic gear.
> 
> The other thing is, Pd isn't only an audio application. The quality of an oscillator is context dependent, and "how does it sound" isn't always the most important consideration. "Can I predict how this will behave?" is the more important question much of the time.
> 
> On Mon, Nov 23, 2015 at 11:31 PM, Robert Esler <robert at urbanstew.org <mailto:robert at urbanstew.org>> wrote:
> I think you just found one of the nuances I’m referencing.  Think of analog gear, none of the sinusoids are anywhere near perfect, yet we still like how they sound.  
> We’ve known about these issues of microscopic DC, phasing, etc of unit generators for a long time.   I recall an old pd thread explaining how [osc~] is working: http://music.columbia.edu/pipermail/music-dsp/2004-november/061991.html <http://music.columbia.edu/pipermail/music-dsp/2004-november/061991.html>
> Moreover, we’ve all lived with [osc~] for what, about 15-20 years?  It’s legacy code.    
>      I’m probably being so adamant because I’ve been reading the "Ugly Duckling" to my daughter, but aren’t children’s stories also lessons in computer synthesis too?
> 
> 
>> On Nov 23, 2015, at 9:05 PM, Alexandre Torres Porres <porres at gmail.com <mailto:porres at gmail.com>> wrote:
>> 
>> moreover, I really doubt there's any particular nuance that comes out of this... or that a fix would break it. All I know is that it's preventing FM patches from achieving stable waveforms as they should.
>> 
>> 2015-11-24 0:31 GMT-02:00 Matt Barber <brbrofsvl at gmail.com <mailto:brbrofsvl at gmail.com>>:
>> I usually agree in cases like these, but a sinusoid oscillator with built-in DC is not the expected behavior in most any synthesis environment. Notice how everyone in this thread was genuinely surprised by this behavior.
>> 
>> On Mon, Nov 23, 2015 at 9:20 PM, Robert Esler <robert at urbanstew.org <mailto:robert at urbanstew.org>> wrote:
>> I would call this more of a feature than a bug that needs fixing.  I would hope that [osc~] and [cos~] don't change, simply because many of us like these little nuances.
>>   If it is really bothersome then perhaps create a new version, but let’s not change a legacy object.  A simple “fix” might break someone else’s patch.
>> 
>> Just my opinion,
>> -Rob
>> 
>>> Did you make it work in a patch? if so, can you share it? :)
>>> 
>>> Maybe someone could work on a "fix" on the source and send it to miller,
>>> perhaps this could be updated for the next version release (0.47).
>>> 
>>> cheers
>>> 
>>> 2015-11-22 19:32 GMT-02:00 Matt Barber <brbrofsvl at gmail.com <mailto:brbrofsvl at gmail.com>>:
>>> 
>>>> Yeah, so all that really needs to be done is to force symmetry by copying
>>>> the 0-pi phase inverted to the pi-2pi phase + guard points for [tabosc4~].
>>>> I did that and it's been stable for 3.5 hours. It wouldn't be too hard to
>>>> fix this in the Pd source; it would be a marked improvement to [osc~] even
>>>> with the 512-pt table and linear interpolation.
>> 
>> 
>> _______________________________________________
>> Pd-list at lists.iem.at <mailto:Pd-list at lists.iem.at> mailing list
>> UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list <http://lists.puredata.info/listinfo/pd-list>
>> 
>> 
>> 
> 
> 
> 
> 
> _______________________________________________
> Pd-list at lists.iem.at <mailto:Pd-list at lists.iem.at> mailing list
> UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list <http://lists.puredata.info/listinfo/pd-list>
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puredata.info/pipermail/pd-list/attachments/20151124/fc8f15d2/attachment-0002.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: osc2~.zip
Type: application/zip
Size: 8072 bytes
Desc: not available
URL: <http://lists.puredata.info/pipermail/pd-list/attachments/20151124/fc8f15d2/attachment-0001.zip>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puredata.info/pipermail/pd-list/attachments/20151124/fc8f15d2/attachment-0003.html>