WavPack, MPC, transcoding and so on 2003-04-11 12:20:05 First of all, I don't want to start any kind of flame war. The second, this is no promotion or nothing alike. I have no reason for something like this. The only reason is to get the best possible quality in lossy compression and to provide the possibility for transcoding along (which is not necessarily an evil, as many here use to say...). I'd like to start a serious "think-tank" exchange, because I've been thinking about this idea for some time...As for MPC, it's without doubt the best lossy codec, very near to transparency (or better to say almost transparent) on very effective bitrates...But this is the only one reason why I use it along with WavPack (in lossy mode). I can't describe my "problems" with MPC properly and even can't ABX my reservations (compared with wav). One evening I can, on the other I can't... It is in the sound, in the very nature of the sound, it has a lot to do with spatial resolution, how the space is defined and so. I think it must be very closely linked with the lower bitrate on which MPC usually operates. For me is MPC and WavPack (~ 352-448 kbps) definitely the best solution for archiving music, I do not need any lossless (except for classical music, but in this field is WavPack very effective). Most of the time it's just wasting a space. And tagging was never an issue for me.Why using WavPack?WavPack's lossy mode does not currently use any psychoacoustic model because those models are based on the idea that there are decisions to be made about how to divide the available bitrate to the various frequency bands (to minimize the audibility of the added noise). Because WavPack does not divide the audio into subbands, there is no decision to be made there. The only artifact is quantization noise added evenly across all frequencies. In this aspect it is more similar to ADPCM, but because WavPack is not limited to a fixed number of bits per sample it is able to accommodate transients much more accurately, and of course the predictor that WavPack uses has been far more tuned than any for ADPCM.The default mode is true stereo (L/R). For joint stereo to be really useful it must be able to detect those situations where it makes things worse and switch back to true stereo, and of course then it can run into the situation where the encoder might switch back and forth so often that that becomes audible! If you have sufficient bitrate then true stereo is the safest (although for material with little stereo separation you can specify -J1 to force joint stereo).The explanation for WavPack's lossy mode sounding more "solid" and with better spatial resolution than other codecs could be because there is absolutely no "time smearing" or "pre-echo" in WavPack because there are no digital filters in the signal path. Most of the experts here at HA would argue that these effects are below the threshold of direct audibility and are therefore irrelevant, but I'm not sure I buy this. I'm definitely not the first person to mention this perception nor the first to be afraid of mentioning it on HA (as far as I know). But I must risk it on my account...I keep in mind that WavPack is far less of a standard than the others (although being native in FooBar2000 doesn't hurt) and obviously has poorer seeking and tagging capabilities. AFAIK bryant is working on an entirely new format for WavPack 4.0 that will address the seeking and tagging issues and provide better lossy quality, but this will certainly be months in the future and it will be fully compatible with existing files.I personally use only WavPack for decoding back to wavs. Once the quality exceeds what can be measured in ABX tests, I suspect that most of the people at HA would say that all codecs are the same, but I don't believe that. There is no doubt that WavPack's lossy mode provides better quality than other lossy codecs above some bitrate because it provides a straight line to lossless encoding which the others do not.Just my 2 cents.P.S. First time I realized the spatial "problems" with MPC was when I compared wavs/mpcs/wavpacks from "Men In Black 2" soundtrack (to be exact, it was track 4). Sometimes it was quite apparent even without abxing... For MPC (1.15r) I use --quality 7 --xlevel.