It's imperfections in the [fft~] objects which come from float precision and error propagation. There are so many multiplications and additions in the fft butterflies that it's almost impossible not to get any junk, even with very precise inputs and twiddle factors. Also, each bin leaks a little error into each other bin because of the way the fft is calculated. You'll notice that if you increase the block size, the junk is a little more significant.

Do a "find text" for "error per bin" here, and then read the few posts following:
https://www.dsprelated.com/showthread/comp.dsp/94142-1.php

On Mon, Apr 25, 2016 at 7:42 AM, Christof Ressi <christof.ressi@gmx.at> wrote:
Sounds great!

Two more things I noticed in the test version:

1)
Pd sporadically crashes really hard when opening help patches... Can't reproduce it. I'm on Win 7.

2)
There's a object called [framp~] (just discovered that it's been there earlier, but I wasn't aware of it).
In the 'list of objects' aka help-intro-pd it is described as "output a ramp for each block", but actually it does something completely different ;-). I'm still figuring out how to read/interprete the analysis data.
Anyway, the help patch says:
"A sinusoidal component should appear in four components (or three in the special case of a sinusoid exactly tuned to a bin."
That's not what I get... For feeding fractional bin frequencies I sometimes get 2, 3 or even 5 components. And for non-fractional bin frequencies I hardly ever get 3 components, on the contrary, the output is quite messy, with lots of tiny values (rounding errors because of Pd's 6 digit float precision? imperfection of [osc~]?).
Just something I noticed.




> Gesendet: Sonntag, 24. April 2016 um 22:08 Uhr
> Von: "Miller Puckette" <msp@ucsd.edu>
> An: "Christof Ressi" <christof.ressi@gmx.at>
> Cc: pd-list@iem.at, "Miller Puckette" <mpuckett@imusic1.ucsd.edu>
> Betreff: Re: [PD] [PD-announce] pd 0.47-0 test 1 released
>
> I want to do this but have 3 or more conflicting ideas right now how to
> do it... so can't decide :)
>
> M
>
> On Sun, Apr 24, 2016 at 01:49:49PM +0200, Christof Ressi wrote:
> > Thanks a lot! [clone] is a really awesome feature!
> >
> > I also like the new 'delete' method for [text]. Very nice!
> >
> > Just one question: Are there any plans to get rid of some deep copying in the code for [list]? Some month ago we had a discussion about iterating through lists in Pd and how it's not possible in linear time in pure vanilla style (at least not in a trivial way, think of the [list-drip] abstraction). You kind of made a hint that you might revise the code in a future release. I'm just curious if this is something we could expect. I checked and the 'naive' style with two [list split] objects still behaves kind of exponentially.
> > Generally I find it a bit curious that in Pd one needs to split a list twice (with all the copying involved) just to retrieve a single item. Have you thought of implementing something like an 'iterate' method (I think I even saw this in one of your to-do-lists) or even getter/setter methods, like in [text] and [array]?
> >
> > Thanks a again for the release!
> >
> >
> > > Gesendet: Samstag, 23. April 2016 um 20:27 Uhr
> > > Von: "Miller Puckette" <msp@ucsd.edu>
> > > An: pd-announce@iem.at
> > > Betreff: [PD] [PD-announce] pd 0.47-0 test 1 released
> > >
> > > To Pd-announce:
> > >
> > > Pd version 0.47-0 test 1 is available on http://msp.ucsd.edu/software.htm
> > > or via git from sourceforge:
> > >   git clone git://git.code.sf.net/p/pure-data/pure-data
> > >
> > > I'm still fixing bugs but the major changes are in place: a "clone" object
> > > for making voice banks etc., and a "zoom" feature for people whose
> > > high-resolution displays make Pd's windows microscopically small.  (Also,
> > > setting the Pd window font size now extends to menus and dialogs).
> > >
> > > cheers
> > > Miller
> > >
> > > _______________________________________________
> > > Pd-announce mailing list
> > > Pd-announce@lists.iem.at
> > > https://lists.puredata.info/listinfo/pd-announce
> > > _______________________________________________
> > > Pd-list@lists.iem.at mailing list
> > > UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list
> > >
> >
> > _______________________________________________
> > Pd-list@lists.iem.at mailing list
> > UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list
>

_______________________________________________
Pd-list@lists.iem.at mailing list
UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list