Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: WavPack 3.97a hybrid mode ABX (Read 5221 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

WavPack 3.97a hybrid mode ABX

I decided to test out WavPack again and see what I think because it has been a couple of versions since and it is now open source as well.  I decided to see if I could ABX the hybrid 320 mode.  I used the WavPack frontend along with the 3.97a compile that john33 put up on his website.  I am using my Altec Lansing computer speakers which have ok sound.  I encoded "Earshot - Headstrong.wav" ( a song I have often used for testing because of familiarity and the complexity of the music ) using the high quality setting.  I then made a copy of the resulting file, naming it "Copy of Earshot - 01 - Headstrong.wv" to make sure the decode was just the lossy part, and not the lossy + correction file.  So, I decoded it with the frontend, and then compared file sizes, which were equal, though the contents were of course different, as I did a binary comparison.  I have a head cold and earache, so I am not testing with the best conditions, but that should only ADD to the significance of my results.  I was able to ABX in two separate tests of the original WAV and the decoded lossy WavPack 6 of 9 and 11 of 15.  I think what I am hearing is not necessarily a sound quality difference, but possibly a difference in loudness or a difference in the output of the left and right channels.  I have a pretty good idea of the difference and what I am looking for, and might be able to get even better ABX results now, but I wanted to post my original results first.  Could this have to do with Joint Stereo as compared to Normal Stereo?  Is Joint Stereo on as default?  If so, have others tested the Stereo modes side by side and has anyone else tested WavPack and obtained similar results?  Possibly I am not using the best settings for WavPack, as I am not very experienced with the program as of yet, but ABX does not lie that I am hearing SOMETHING different.  I will listen some more while I wait for a response to see if I can better describe what I hear...

Edit:  in a related question:  What work has been made to make better the seeking capabilities of this format and/or the Winamp plugin?
WARNING:  Changing of advanced parameters might degrade sound quality.  Modify them only if you are expirienced in audio compression!

WavPack 3.97a hybrid mode ABX

Reply #1
Actually, you may be the first person to do any sort of formal listening tests on the WavPack hybrid mode (or at least the first to admit it), so thanks!

There aren't a lot of settings for the hybrid mode, so if you go with the defaults you should get just about the best quality possible. The exception is the "high quality" mode which should be used for critical listening. I have a couple samples in which I can hear problems at 320 kbps that disappear with the "high" mode (this is particularly audible on sine sweeps). The downside is slower encoding, slower decoding, and slower seeking in winamp.

The first version of the hybrid mode used joint stereo by default because it generally gives measurably better noise performance. However, I found that when listening in headphones to music with exaggerated stereo I could sometimes hear the artifact of sound in one channel making noise in the other channel. I am not sure how MP3 (and other codecs with joint stereo) avoids this problem, but I decided to make the default true stereo.

As for the difference you are hearing, I'm not sure how there could be any level or channel balance problem. Because of the very simple nature of WavPack's compression, the only artifact is the addition of modulated quantization noise, which is either pink or white depending on whether noise shaping is enabled. This noise is perceived directly as noise when it is high enough in level, or as a "grainy" character when lower in level. It sometimes can make the sound simply "brighter", even though WavPack cannot actually alter the bandwidth because the signal never leaves the time domain. It is possible that you are hearing the noise as a level change because it is essentially being added to the signal, but I have never perceived it that way.

There are a couple things you can try to narrow it down (assuming that "high" mode doesn't fix it). If you try the test again and encode at 265 kbps, the effect should be exaggerated and easier to hear. Also, switching to 384 kbps should probably clear it up (I have never even thought I heard a difference with any music at that bitrate). If you can ABX a sample at 384 kbps high mode then I would definitely appreciate getting a short sample for my own testing.

As for the seeking issue, the unfortunate thing is that the problem is not with the winamp plugin, but is a fundamental limitation of the format (i.e. no blocks) and any sort of fix would be an incredible hack. I have begun work on an all new compression format that will eliminate that (and several other) WavPack limitations, but in the meantime this will have to be. I assume that you have noticed that once you have played a track you can instantly seek anywhere in that track (at least until you play another track). And, of course, on faster computers the seeking speed increases proportionally. 

Thanks again for giving the hybrid mode a critical test. A lot of people I hear from use it casually and claim that 320 gives them transparent operation (and this is what I have found over dozens of CDs I have encoded and listen to regularly), but this is not the same as rigorous ABX tests. Some might feel that the purpose of the mode is basically lossless storage and that it's silly to complain about the quality of the lossy file because it's sort of a "freebie" anyway (and it obviously can't compete with a perceptual lossy codec on bitrate alone). But I am still very interested in the actual bitrate that critical listeners find transparent and am often working on ways I might be able to improve the lossy quality without sacrificing the lossless performance.