On Sun, 2014-08-17 at 13:51 +0600, Alexander E. Patrakov wrote: > 17.08.2014 12:38, Tanu Kaskinen wrote: > > Patch review status updated: > > http://www.freedesktop.org/wiki/Software/PulseAudio/PatchStatus/ > > ... > > > Resampler quality testing > > > > From: poljar (Damir Jeli?) > > Submission date: 2013-08-26(?) > > In a github branch: https://github.com/poljar/pulseaudio/commits/resampler_quality > > Review: does not answer the question "is this distortion audible?" and > provides no way to evaluate quality of non-PulseAudio resamplers. > > So I have decided to redo all of this using a different approach. The > new work is based on a psychoacoustical model in order to correct the > first objection, and is supposed to judge any resampler according to its > output as a wav file (including Windows output recorded by KVM) when > given a linear sine sweep. I have already implemented the model, can > answer the "is this distortion audible" question given the spectrum of > the signal and the distortion, but have not used this yet to analyze the > resampler output. Damir also provided useful contributions. So this is > definitely no longer "Waiting for Review". > > As this successor is now a side project that does not share code with > PulseAudio and does not even use PulseAudio, I don't expect it to > produce a PulseAudio patch. If Damir agrees, let's remove it from the page. Ok, I'm waiting for "yes" or "no" from Damir. > I think I will have some announceable results in about a week and full > results in about a month, and would like to talk about this (maybe > unofficially) at Audio mini conference 2014. > > > Resampler implementations > > > > From: poljar (Damir Jeli?) > > Submission date: 2013-09-06(?) > > Reviewed: 2013-11-29 > > In a github branch: https://github.com/poljar/pulseaudio/commits/resampler_implementations_v2 > > Status update: http://article.gmane.org/gmane.comp.audio.pulseaudio.general/18991 > > Performance tests (mono): http://poljar.blogspot.com/2013/08/vol-2-resampling-methods.html > > Performance tests (stereo & 5.1): TBD > > Bandwidth & Aliasing tests: TBD, with explanatory material at http://poljar.blogspot.com/2013/10/epilogue-fourier-analysis-and-testing.html > > Features (such as variable-rate support, input & output formats, optimizations for particular combinations of sample rates, compatibility with rewinds): TBD > > Duplicate of this one, and all TBDs still apply to the new submission > (please copy): Notes copied from this old entry to the new. > > resamplers > > > > 3 patches > > From: Damir Jeli? / Peter Meerwald > > Submission date: 2014-08-04 > > On the mailing list: http://thread.gmane.org/gmane.comp.audio.pulseaudio.general/20739 > > These have been reviewed by Arun and Alexander, and it's a bit unclear what will happen next. The first patch appears to be (mostly?) ready to be merged, the two other patches probably need at least some additional justification. > > Yes, the first patch should be merged, with or without taking my > speex_is_fixed_point objection into account. I would have to apply the > other two patches locally in order to do the quality evaluation, but > don't want them to be upstreamed yet, until we have tools to judge them. > Hopefully this clears the "what's next" question. I applied the first patch (changed the author to Damir) without taking your objection into account, because that was the easier option for me. I didn't look in detail what the objection was about, so I'm not saying that I don't agree with your objection. If you or someone else wants to change the code, please send a patch. -- Tanu
