apple lossless is only 10.3% of original audio? - Page 2

He doesn't seem to understand the science of psychoacoustics. I do think that mp3 is a product of a different era, where storage was expensive and bandwidth was limited. I'd be happy to see lossless completely replace it. BUT that doesn't make it as evil as he thinks it is.

 

He seems misinformed about a lot of things. I recall him saying that 24/96 files were still missing 50% of the information.... As if ALL recordings are done in native 24/192 (HINT: they aren't). 

Either way long live FLAC.

Even Neil Young's Archives box isn't in native 24/192[it's upconverted from 24/96 audio]. 

I see no one has picked up a copy of this month's The Absolute Sound where they devote some 8-10 pages in their 4 part series on optimizing computer audio to testing of FLAC and their conclusions are saying that any compressed format, lossless or not (FLACs, AACs etc)  are compromised compared to raw, unprocessed WAV files.  Cue the rage in 3... 2  ...

 

Pre-empting all the math ragers who will proclaim that "lossless is lossless" and "there is no difference", fine.  Don't sell it to me, I didn't write the article.  Set up your own tests using their conditions and systems and prove your point.

 

But they do discuss multiple types of file degradation,  including converting from FLAC to WAV, which in theory should be identical, but their results show otherwise.

 

I am a skeptic as well, so I've used their upsampling set up from last month's issue: ripping raw WAVs, running them through Izotope RX2 and playing back through both JRiver MC (with the JPlay plug-in for 24-192) for PC and through PureMusic (also at 24-192) on an IMac and then comparing them to FLACs of the same files (at 0 compression ratio).  There is a difference. And it's noticeable, especially on my 2 channel system with Maggie 3.7s.  Whether it's due to the dithering from the upsampling, the file format itself or how the PC handles the files, there is a noticeable difference, most dramatically in lower/higher frequency extremes.  I've had several non-audiophile friends come to the same conclusions listening to the same files.

 

The trade off of course is that file sizes are enormous (100-400 Mbyte per song).  I won't rip every album using the upsampled WAVs, but ones I have, I've been listening to them a lot lately. 

 

But Neil Young has an advantage none of us has: he knows the master tapes of his music.  Also, he knows what it should sound like in his head.  Any kind of copy is going to sound inferior to his ears.

Is it too much to ask that you have blind test results?

 

I'd also like to know more about how they tested the files and determined there are differences.

Then I would suggest you read the articles, as testing and evaluation bias and methodology are covered extensively in the first installment, including this

 

All listening tests were conducted under strict single-blind conditions.
Upon occasion, double-blind testing was also employed where
practical. When appropriate, listening levels were matched to +/-
0.1dB. For most tests, this was unnecessary and redundant since
common digital files yielded identical playback volumes. We wish
to make one final comment about how we arranged our listening
tests when only the two of us were evaluating differences. In
conducting single-blind trials, one person would do the switching
while the primary listener would sit without moving, with eyes
closed, and with no talking allowed to eliminate any potentially
distracting sensory stimuli which could disturb our concentration
and long-term memory (musical excerpts lasted anywhere from
30 seconds to 5 minutes depending on the content of the test
track). This was essential when trying to make extremely subtle
quantitative judgments. The A/B comparison was then repeated
with our roles reversed. Only after we agreed that we had come
to a conclusion did we discuss between us what we had heard.

I should add that as soon as I post the words "The Absolute Sound", there will be rolling of the eyes and "haw haw"s from many.  Fine.  Changes in tech and formats make any kind of evaluation welcome in my books.  The authors  (Clawson and Zeilig) are at least putting some kind of information out there.  "Use it to make your own decisions" is my point.

And where are the articles? Are they online?

 

I don't like the fact that they discussed at all. They should have let the results speak for themselves. Did they record their choices first? Minds can be changed.

But -

 

even assuming, for the moment, that the tests in TAS show the differences claimed -

 

why would Young have said that Apple Lossless are "10.3 per cent" of the original?

 

 

Seems he's referring to some very specific information.

 

But the article doesn't mention what it is.

 

 

 

(Edit: formatting)

Edited by zazex - 2/1/12 at 9:07pm

 

The article is not online, *anywhere*, or perhaps my google-fu isn't strong enough, but it's controversial enough to have several threads dedicated to it on multiple forums.

They did claim an audible difference (in single blind) with 2 files ripped at different speeds, both verified to be bit accurate, the mind boggles .
 

 

 

85.7% of people have been known to make misleading statements using mythical statistics which may have have a .90 confidence of being within 95.5% range of factual, though idealized values.  Sorry, that was a bit sarcastic.

 

If you really want to dig into it, I suppose you could look at Apple lossless AAC rates in at a bit rate of between 700 (mono) - 1000+ (stereo) kb/s, depending upon the recording.  Call it 850 kB/s, the mean.  Take an uncompressed 24-176.4k WAV file, the bit rate is 8,467 kB/s (the 24-192k bit rate is 9216kB/s).  Here AAC is 10.03% of the full WAV.  Not quite Neil's number but. close enough for government work and  very much within the ball park. and (imo) well within reason for a burner like Neil to talk about off the top of his head

 

Yeah, I wouldn't expect to see it on line just yet. Gotta buy the issues!  They need to drive revenues!!  I'm sure RH and HP at TAS have been drooling waiting to publish this.

 

#1, they're still publishing the results (I just picked up part 3 at the newsstand last Saturday),  #2, it's surely a hugely controversial issue with us audio nerds, so any chance for a troll  can get more sales. #3, Despite John Atkinson's comments poo-pooing the results of this study (he noted on another forum that he turned down a study by the same authors back in the 1980's ), this is a huge scoop on Stereophile and a good dose for subscriptions.  (I've seen plenty of criticism of JA's measurement methods by the armchair elite as well, so take any and all criticisms with a grain of salt.  Or two.)

 

Still, listen for yourself and decide what you need.  If you can hear a difference, go for it.  If not, keep dancing with the partner that you brought.

 

I don't need to use their systems, conditions, or methods to disprove their conclusion.

I checked the MD5 sum of a WAV file, converted the file to ALAC, converted to ALAC file back to WAV, and checked the new WAV file's MD5 sum.

Result: Checksums matched exactly.

Conclusion: the information in the original file was in no way compromised by the conversion to ALAC.

PM with email address if you would like a copy of the original file to attempt to reproduce results exactly.

Edit to add:

I repeated the test using XLD to convert from WAV to FLAC (and back to WAV) instead of ALAC, and got the same result (checksums matched).
Edited by Jaywalk3r - 2/1/12 at 11:30pm

 

 

 Apparently they also claimed that FLAC > WAV > FLAC isn't transparent and that they can while repeating the process over 10 passes, the degradation over each pass is audible (blind), all of that without verifying the checksums).
The mind boggles even more, both at the lack of scientific rigor and the mystifying results
 

 

 

Ha, of course they didn't verify the checksums, who needs logic in hifi? Did they have a WAV player to plug? Or extol the virtues of CD players because computers aren't 'reliable'? I'm surprised nobody has started selling audiophile hard drives (sounds better!), motherboards (sound better!), or processors (sound better!!) yet, which of course would just be rebranded copies of existing components, but that wouldn't matter would it, as they'd clearly sound better. The woo in digital audio is the strongest yet - 'proprietary quantum process', vibration dampening for solid state gear... it's no wonder my engineer/electronics friends think audiophiles are a joke and I have to convince them not to tar me with the same brush.