[Bug 14717] resampled sound is horrible

[wine-bugs at winehq.org]

http://bugs.winehq.org/show_bug.cgi?id=14717

--- Comment #19 from Krzysztof Nikiel <zzdz2 at yahoo.pl> 2010-11-24 13:00:50 CST ---
Created an attachment (id=32114)
 --> (http://bugs.winehq.org/attachment.cgi?id=32114)
Transfer function of the resample filter

Hopefully the attached image will get correct mime type from bugzilla.
It's created using the same table as in the patch.
The approach is rather conservative, transition is well below Nyquist
frequency. Usually(in commercial/hardware apps) the transition band seems
rather centered around Nyquist and that's probably the best setting.

As to the comparison to the other resampler engines, my goal is to create
ultimate resampler. It may be the other way around, our resampler will be
better than them.

>cutoff = 1.0 - 1.0/filter_size;
>if (cutoff < 0.80)
>    cutoff = 0.80;

What does it mean, is it passband width, start of the stopband?

The whole band may be divided into parts: 
passband: signal passes unattenuated;
transition band: signal becomes more and more attenuated;
and finally the stopband: signal is attenuated as much as possible.

In my code the passband if from 0 to (0.96-0.15)*Nyquist and transition from
(0.96-0.15) to 0.96.
Stopband ripples are more than 80dB below the main signal.
That gives just top quality output.

I don't remember now where the "0.96" came from. It actually was quite old code
of mine, initially based on the original "resample" source, but then rewritten
a few times. Now rewritten again.

-- 
Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email
Do not reply to this email, post in Bugzilla using the
above URL to reply.
------- You are receiving this mail because: -------
You are watching all bug changes.