Hi to all,
I would like to ask to set up a listening test in order to compare Wavpack hybrid modus to other lossy formats like mpc, ogg Vorbis, etc.
Until now I have ripped and encoded to mpc -level 6. But as capacities of drives get bigger and bigger, I consider the idea to switch over to wavpack hybrid modus. All I want to know is, wether wavpack hybrid modus (256/320/384 kbps) can compete with mpc -level 6?
Maybe some of you have alredy made some experiences with this matter and can give me some hints?
( if this is the wrong forum, please move it!)
Thanks
Martin
All I want to know is, wether wavpack hybrid modus (256/320/384 kbps) can compete with mpc -level 6?
Maybe some of you have alredy made some experiences with this matter and can give me some hints?
[a href="index.php?act=findpost&pid=244374"][{POST_SNAPBACK}][/a]
Hi Martin,
I was perfectly satisfied with MPC encoding whole albums, until noticing reproducable "glitches" (squeaky sounds, dunno if translated correctly) caused by either MPC or the playback routine (could not determine). This, for me, occured on severable albums...
Since then, for archiving, I mostly used flac for classical music and WavPack hybrid high quality @384 kpbs for about everything else.
For classical music, even 320 kbps may be overkill, for me sounding indistinguishable from the original and resulting in smaller .wvc than .wv. Flac ahoi!
If you ever plan to set up some large scale listening test, please count me in. I am not familiar with that statistics stuff and have to rely on others therefore.
First one has to understand that the hybrids are very different to the perceptual coders. Traditional psycoacoustics are not used so you need double the bitrates to reach transparency. There are no traditional artifacts other than white noise which may be abxed at lower bitrates such as 256k. Also quality increases with bitrate until lossless compression is reached so in other words past a certain bitrate the hybrids exceed all traditional coders in quality - they become fully transparent. Clipping , pre-echo and similar distortions don't occur, transcoding to other formats is transparent.
I've done some listening & transcoding tests. At 256k there is some noise present , 320k -high was transparent. >384k you will start to enter that no risks zone.
MPC Q6 200k will generaly beat wavpack at 256k. 320k -high is a contender and will be difficult for most to abx.
One major problem of such a test would be the selection of test samples. At your target bitrate/quality most stuff will be transparent anyway. The only way to get any positive results at all is using known problem samples. However these problematic cases are very different for the two types of codecs you plan to test ("hybrid" vs. perceptual). Interpreting the results of such a test will be most difficult.
I've done a few of these comparisons myself. If you compare Wavpack vs the others at identical bitrates, mpc etc will come out on top, ie they will hit transparency at lower bit rates than Wavpack lossy for most samples.
If you are talking about a lossy format to transcode from, it's a different story... Also Wavpack works well with many of the well known problem samples compared to Lame/Vorbis/mpc, but has some problem samples of its own.
As others have mentioned Wavpack lossy works quite differently from the other lossy codecs, so the sample selection and comparison is quite difficult.
Last but not least, if you are sticking to identical bitrates across the test, you will have difficulties finding test subjects who can ABX mpc, vorbis etc at say 320 kbit and above. (Truly only for those in the "golden ears" club.)
Good luck with it.
Den.