On Sat, 13 May 2023 06:49:31 +0200
ichthyo <p...@ichthyostega.de> wrote:
>Am 12.05.23 um 23:56 schrieb Will Godfrey:
>> Aaand we come screeching to a halt :( I ran the testsuite and i failed really
>> badly on the SVfilter tests.
>
>Unfortunately, I didn't see your latest Mail message in time.
>And additionally I made the mistake first to do the packaging work
>and only after the packages were pushed out, I run the testsuite and
>saw the failed cases.
>
>
>Anyway, the good news is that by a quick check there is no obvious *audible*
>difference. From looking closely at the waveforms, I get the impression that
>the filter also behaves similar, and we've just gotten a minute phase shift,
>which is most notably at the lower frequencies.
>
>Filters can be quite nasty beasts and amplify even the tiniest changes.
>Incidentally, also the AnalogFilter test cases show differences, albeit very
>small and close to rounding errors.
>
>
>Thus I'd say, there is no reason to raise a big glaring alarm, rather we
>should take our time to do a closer analysis. If it's really a "number dust"
>error leaking into some frequency parameter, this would be something we just
>have to admit (and which has no practical consequences). I had similar changes
>with my "Padthread" rework (where a very minute numerical error in the
>frequency
>domain caused a significant deviatation after FFT, but also in practice
>the change was not audible).
>
>-- Hermann
I also couldn't hear any difference, but them I don't trust my hearing much
these days :(
However I just did a test using the x42-spectra spectrum display module.
I set up both Yoshimi V2.2.3 and my latest build (mostly doc updates from rc3)
each with one AddSynth instrument of the power waveform and StVar filter. These
were linked to the same USB keyboard, and separate instance of the spectrum
analyser. I placed these one above the other and stretched out so that the pair
filled the screen. The result was very interesting and suggests it is a phase
issue.
Holding a C maj chord starting from middle C produced a nice rich spectrum in
both displays, with the classic tail off at around 10kHz, and lots of visible
beating. Overall the harmonic lines seemed identical. However the *pattern* of
amplitude variations from the beats was quite different.
--
Will J Godfrey {apparently now an 'elderly'}
https://willgodfrey.bandcamp.com/
http://yoshimi.github.io
Say you have a poem and I have a tune.
Exchange them and we can both have a poem, a tune, and a song.
_______________________________________________
Yoshimi-devel mailing list
Yoshimi-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/yoshimi-devel
