PKZip and RAR are lossless, even Microsoft admits that WMA Lossless is "mathematically" lossless (heh..... see figure 1 in the original article, the screenshot even says it). Yes, I'm aware that "mathematically lossless" is close enough for 99% of the music and people out there... but be aware that there is some music out there that does not pass an inversion comparison test with WMA Lossless, and therefore it's not truely lossless in all cases.
Do you have some evidence to back this up? Lossless audio compression isn't a new concept and compression rates of 40-60% are quite usual (for listenable music). I highly doubt that MS would screw this up.
Maybe you compared the .wav files on a binary level? Two files can contain the same audio content but differ in their meta information.
Lossless means lossless. "Mathematicially" isn't the same as "Statistically." You can google for Windows Media Audio lossless and see several articles outside of MS's site that confirm it is a lossless compression technology, just like PKZip.
Do you have some evidence to back this up? Lossless audio compression isn't a new concept and compression rates of 40-60% are quite usual (for listenable music). I highly doubt that MS would screw this up.
Maybe you compared the .wav files on a binary level? Two files can contain the same audio content but differ in their meta information.
No..... Inversion Mixing test: Rip to PCM WAV, make a copy of that WAV as WMA Lossless, convert the WMA Lossless back to PCM WAV. Use DBPowerAmp or Cool Edit 2 or the like to read in one of the WAVs, invert it, mix in the other WAV.... the result is not always silence (flat-line).. e.g. Not always lossless. I do find it interesting that you say:
Quote:
Originally Posted by Hylas
compression rates of 40-60% are quite usual (for listenable music)
The way I'd normally read that is that if you get compression ratios of 40-60% that the music is listenable.... implying lossy? or am I reading that wrong?
Lossless means lossless. "Mathematicially" isn't the same as "Statistically."
Heh, and we all believed Bill Gates when he said that "... Microsoft software is inherently bug free...".
Quote:
Originally Posted by Ed Hansberry
You can google for Windows Media Audio lossless and see several articles outside of MS's site that confirm it is a lossless compression technology, just like PKZip.
When people ask me for the definitive site for PPC information, I send em to Pocket PC Thoughts.... for the definitive Digital Audio site, I suggest Hydrogen Audio There you'll also find articles and tests that show WMA Lossless is in fact lossless, and others that show it isn't. Fact is that it is what Microsoft says it is: "Mathmatically Lossless". It can be shown to a mathmatical certanty that WMA lossless is in fact lossless. Note that there is also a mathmatical proof that .999999... (.9 repeating) is equal to 1. In principle, both are technically true; in practice, there are examples where it is not true.
Look.... I'm not MS bashing... if WMA Lossless works for you, by all means use it. I use another lossless format that works for me (not quite as good as WMA Lossless for compression, but more efficient CPU utilization, broader software support, and actual hardware support too).
No..... Inversion Mixing test: Rip to PCM WAV, make a copy of that WAV as WMA Lossless, convert the WMA Lossless back to PCM WAV. Use DBPowerAmp or Cool Edit 2 or the like to read in one of the WAVs, invert it, mix in the other WAV.... the result is not always silence (flat-line).. e.g. Not always lossless.
I find this really hard to believe. A lossless audio codec should by definition never generate a difference in this test (the only exception to this rule, that I know of, is the codec used on DVD-Audio, that can fallback to a lossy-mode in extreme bitrate situations. But this is only for the sake of DVD-player drive speed limitations.)
WMA-Lossless would be quite different then from APE, FLAC, RAR, ...
Quote:
I do find it interesting that you say:
Quote:
Originally Posted by Hylas
compression rates of 40-60% are quite usual (for listenable music)
The way I'd normally read that is that if you get compression ratios of 40-60% that the music is listenable.... implying lossy? or am I reading that wrong?
As it is standard for every information encoding, it is impossible to reduce the size of every input. It is possible to generate special test cases that cannot be compressed by e.g. ZIP.
However, for musical information (things recorded by a microphone, synth., ...) the stated numbers usually apply. Classical music, and low volume music in general, is particulary suited for strong compression.
Nothing of this has anything to do with perceptual entropy encoding (lossy).
I find this really hard to believe. A lossless audio codec should by definition never generate a difference in this test (the only exception to this rule, that I know of, is the codec used on DVD-Audio, that can fallback to a lossy-mode in extreme bitrate situations. But this is only for the sake of DVD-player drive speed limitations.)
WMA-Lossless would be quite different then from APE, FLAC, RAR, ...
But again, this is my point..... WMA Lossless is "Mathmatically Lossless" which is mostly, but not completely, lossless in all cases. Honestly the same can be said for Mpeg3 and 4, Jpeg, OGG-Vorbis, WMA Lossy or any other Lossy compression at suitably high bitrates... although likely higher bitrates than would be acceptable for typical use. At some point, the lossy-ness falls off to the point where it's Mathmatically Lossless. At this point, most data (music or otherwise) will decompress to waveform-identical or bit-identical compared to the original.... Most, but not all. APE, FLAC, RAR, Huffman-based (ZIP/RAR/etc) are all lossless, and WMA Lossless (by Microsoft's own admission) is Mathmatically lossless.
Now..... the real question is what does "Most, but not all" mean and is it really worth arguing. Honestly?? It don't mean much and it's not worth arguing. What..... we're talking fractions of a percent of artifacts in the data?? 2 or 3 bits out of place across hundreds (WMA9) or thousands (WMA10) of songs? Again.... nit picking (trust me, I'm a master nit-picker :lol: )
If it works for you, use it. I choose something different, and not because it's mostly lossless or not.... but for other reasons that work for me.
Quote:
Originally Posted by Hylas
As it is standard for every information encoding, it is impossible to reduce the size of every input. It is possible to generate special test cases that cannot be compressed by e.g. ZIP.
However, for musical information (things recorded by a microphone, synth., ...) the stated numbers usually apply. Classical music, and low volume music in general, is particulary suited for strong compression.
Nothing of this has anything to do with perceptual entropy encoding (lossy).
That's a longer version of what I thought you said. The first time I read the part of the phrase "... (for listenable music)" it implies to me that you were starting to discuss perceptual transparency, which is usually reserved for personal evaluation of lossy compression. I didn't think that what you were saying, and thanks for confirming that.
I don't get it. Do you claim, that there is a difference between "lossless" and MS' "mathematical lossless"?
Can you guide me to a thread at HA, where this is discussed? From what I read in the thread Which is the best lossless codec?, this fundamental flaw is not widely known.
(About lossy codecs: No, they never reach lossless quality, even at extreme high bitrates. This is by design, as the used fourier transformations are numerical operations and are thus not perfectly invertible).
You can google for Windows Media Audio lossless and see several articles outside of MS's site that confirm it is a lossless compression technology, just like PKZip.
Linear Mode
LinkBacks (?)
