Do Apple Lossless files really sound the same as AIFF?

[sandyk]

sone

 

I expect that the same will happen with .wav and .flac. Results of comparisons were reported in The Absolute Sound 220, and 221.Furthermore, I have uploaded comparison. wav files with identical check sums that sound different to Martin Colloms in HiFi Critic Forum.(Martin is also a member of this forum.) The results of these comparisons were published in HiFi Critic Vol.6 No.1. For the sceptics I have made the differences even more obvious by converting the worse sounding .wav file to .flac and back again to .wav.

Almost 2 years after I provided the original comparison .wav files, I recently supplied a further improved version of the previous best .wav file, (From Dire Straits-Love Over Gold 1985 CD.) but still having identical check sums to the original. The results are reported at this link.

Problem when hocus-pocus works? - Page 5 - General HIFI Discussion - HIFICRITIC FORUM - HIFICRITIC FORUM : hi fi audio systems forum

 

As with analogue equipment, the quality of the PSU area is also equally as important with Digital, and perhaps even more so.

This applies also to even Asynchronous USB DACs and their USB cables,, despite what their manufacturers may claim.

I have further RECENT confirmation of my claims, that unfortunately I am unable to post in open forum.

I would be happy to give either Chris or Eloise, (as an old sparring partner) a rundown on this if requested.

Regards

Alex

 

 

 

 

 

 

 

 

Signature

--------------------------------------------------------------------------------

"If you can't hear the difference between an original CD and a copy of your CD, you might as well give up your career as a tester. The difference between a reconstituted FLAC and full size WAV is much less than that, but it does exist."-Cookie Marenco. cookiemarenco.com/

[Paul R]

I am afraid the issue *is* settled. Some people can hear differences, and other people do not. Issue definitively settled.

 

-Paul

 

 

ALAC is compressed, but lossless. It's the exact same bits being sent to the DAC as the uncompressed file.

 

If anyone doubts this, or astonishingly believes that a computer with Gigabytes of memory cannot reliably decode 10s of Megabytes of audio content, perform this simple test: redirect your output audio stream to Soundflower, then using mplayer or Audacity record the ALAC and WAV and/or AIFF outputs to two files, then do a bytewise comparison of these files. They will be the same.

 

If anyone can identify a numeric difference in the output stream between ALAC and WAV/AIFF, please post it and settle the issue.

 

Otherwise, there simply is neither evidence nor any reason why lossless compressed should differ from uncompressed. The burden of the proof is on anyone claiming that a difference exists.

[Jud]

If anyone can identify a numeric difference in the output stream between ALAC and WAV/AIFF, please post it and settle the issue.

 

Otherwise, there simply is neither evidence nor any reason why lossless compressed should differ from uncompressed. The burden of the proof is on anyone claiming that a difference exists.

 

I feel the burden of proof is on anyone claiming this discussion differs from all those had before on this topic. ;-)

[Teresa]

Back when I was stuck using a 1.5GHz G4 Mac mini w/1GB RAM and Leopard, it was very easy to tell the difference between AIFF rips and Apple Lossless, but only if you listened to one version then the other. Soundstage was flatter, female voices did not have the same air, emotion, or reality, and my friend used to comment how all the ALAC files in our library had a "sameness" to their presentation. We even demonstrated this to a few select people in our (Hovland Co.) room at CES '08. I know I did the comparison--using a female vocal--for John Atkinson, Robert Harley, and if I recall correctly, CA's own Chris Connaker.

 

I always had a nagging suspicion (and others here suggested as much) that the differences heard were due to the real-time decompression processing taxing the G4.

 

So last year, when I finally got a new Mac mini (2010 2.4GHz Core 2 Duo w/8GB RAM, slimmed Mountain Lion, A+, etc.), I did the comparison again. Well, darn, as much as I hate to admit it (I went into it biased to hear a difference) I now can NOT tell any appreciable difference between an AIFF track and the same one converted to ALAC.

 

-ALEX

 

Alex I can confirm that as I also went from a 1 GB RAM Mac Mini with Leopard to a 8 GB RAM Mac Mini with Mountain Lion. On my old 2007 1 GB RAM Mac Mini AIFF music files were clearly superior to Apple Lossless. With my new Mac Mini's 8 GB of RAM the differences are lessened to a great degree and I find WAV, AIFF, FLAC and Apple Lossless to sound almost the same. Since both FLAC and Apple Lossless are bit perfect when unpacked, I feel that the reason that the sonic differences were larger on the old Mac Mini had to do with the real time unpacking while the music file was being played and now that I have 8 times the operating memory that the unpacking process leaves hardly any sonic footprint on the sound.

 

So adding my experience to yours and Boris with enough operating memory the sonic differences between lossless and uncompressed music files seem to disappear.

[sone]

I am afraid the issue *is* settled. Some people can hear differences, and other people do not. Issue definitively settled.

 

Saying that you hear a difference isn't an achievement—it's saying that you hear a difference at a rate that's better than chance. That's why science journals demand sufficiently small significance or p values of before they will publish a discovery.

 

There is no evidence presented in this thread or elsewhere that anyone can detect a difference between compressed and uncompressed formats. Anyone who says they can hear a difference would be able to either show us the numerical difference in the audio stream, or show us the p value of their detection. Any claims without such evidence can and should be dismissed out-of-hand.

[Jud]

Saying that you hear a difference isn't an achievement—it's saying that you hear a difference at a rate that's better than chance. That's why science journals demand sufficiently small significance or p values of before they will publish a discovery.

 

There is no evidence presented in this thread or elsewhere that anyone can detect a difference between compressed and uncompressed formats. Anyone who says they can hear a difference would be able to either show us the numerical difference in the audio stream, or show us the p value of their detection. Any claims without such evidence can and should be dismissed out-of-hand.

 

Dismissed out of hand if that is your inclination and you aren't curious whether there's some mechanism that might explain the difference. If you are curious about such things, then you can think more about them.

 

Couple of general points:

 

- It does not logically follow from the fact that losslessly compressed files can be recovered into bit-identical format that the compressed file sounds identical to the uncompressed one, any more than the fact that ice's chemical composition doesn't change when it melts means you can walk on it in its liquid state just the same as when it was frozen. Is there some reasonable mechanism that might account for reports of being able to hear differences between losslessly compressed and uncompressed formats? At least one (computer processing necessary to expand the file) has been proposed, and some people even report the degree of sonic difference between formats appears to depend on the computer resources available, such as RAM. You can certainly choose to doubt the reports and the proposed mechanism, but I don't actually see a good scientific reason why anyone else should simply dismiss it and not at least wonder whether it's possible.

 

- For people who are curious and also good, careful engineers, there are intuitively absurd propositions that actually turn out to be amenable to scientific/engineering proof. The example I always return to is when Keith Johnson of Spectral managed to show the mechanism behind the apparently absurd claim that green marker changes the sound of a CD (it turns out to result from the laser having difficulty tracking precisely due to the reflection from the marker). So be very careful about what you decide (as a non-expert? unless you're an audio engineer) can be dismissed out of hand.

[sone]

Dismissed out of hand if that is your inclination and you aren't curious whether there's some mechanism that might explain the difference.

 

Don't make a common-sense call for evidence about me. I obviously am interested in this subject, as evidenced by my posts and personal detailed, quantified tests on the differences between various formats.

 

ice's chemical composition doesn't change when it melts means you can walk on it in its liquid state just the same as when it was frozen. Is there some reasonable mechanism that might account for reports of being able to hear differences between losslessly compressed and uncompressed formats? At least one (computer processing necessary to expand the file) has been proposed, and some people even report the degree of sonic difference between formats appears to depend on the computer resources available, such as RAM. You can certainly choose to doubt the reports and the proposed mechanism, but I don't actually see a good scientific reason why anyone else should simply dismiss it and not at least wonder whether it's possible.

 

We're talking about computers here. We can look at the exact byte stream being sent to the computer's DAC/DOUT. The DAC doesn't care whether those bytes came originally from AIFF or ALAC. If there are any codec or RAM issues, then we can detect these by looking directly at DOUT. No one has posted any differences between these two outputs. If there is such a difference -- possible, but not plausible -- that indicates a software failing, as the file sizes and sample rates are minuscule relative to modern computer capabilities. But again, no evidence has been presented that such differences exist, even though it would be very easy to do so on a computer.

[Jud]

Don't make a common-sense call for evidence about me. I obviously am interested in this subject, as evidenced by my posts and personal detailed, quantified tests on the differences between various formats.

 

 

I did not see a p value quoted for detection/non-detection in your post, which if I remember correctly, at least some folks think is a necessity before we can take any such personal report seriously.

[Jud]

We're talking about computers here.

 

Strangely, some folks on these forums who know a little bit about computers (developers who code audio software) disagree.

[wgscott]

If there is an audible difference, how do we know it is intrinsic to the file container differences rather than a bug in the audio kernel extension code or whatever it is?

 

The latter seems at least as likely an explanation.

[4est]

If this were the case, then it would not be cross platform. I have heard the differences here on Mac, Windows and Linux (whatever variant Auraliti PK90 uses).

 

For the record, I (and others here) noticed a distinct difference between wav and aiff using 4 gb of RAM, memory play and such when I used a 2009 mini. The differences between alac and aiff were not tested on platforms other than this, but were pretty distinct as well. Straightaway I converted all of my files to aiff upon testing and never looked back.

 

If there is an audible difference, how do we know it is intrinsic to the file container differences rather than a bug in the audio kernel extension code or whatever it is?

 

The latter seems at least as likely an explanation.

[sone]

I did not see a p value quoted for detection/non-detection in your post, which if I remember correctly, at least some folks think is a necessity before we can take any such personal report seriously.

 

p levels are necessary to support positive claims. I'm making a negative one for myself and a few others in that post.

[Paul R]

Once again, my poor attempt at humor fails. (*sigh*)

 

No, there is plenty of evidence, you simply choose to not accept it - as is your right. And speaking of rights, the plain fact here is that nobody here is being paid to provide you or anyone else with proof of anything, nor is anyone here trying to convince you of anything. It would be civilized to return that kind of respect and attitude. This is a hobby after all, at least for most of us.

 

As far as I can tell, which I bashfully admit is quite well indeed, nobody is claiming the data is different, only that the same data sounds different on different machines, playback software, and audio chains.

 

That is hardly a controversial claim by any reasonable standards.

 

-Paul

 

 

Saying that you hear a difference isn't an achievement—it's saying that you hear a difference at a rate that's better than chance. That's why science journals demand sufficiently small significance or p values of before they will publish a discovery.

 

There is no evidence presented in this thread or elsewhere that anyone can detect a difference between compressed and uncompressed formats. Anyone who says they can hear a difference would be able to either show us the numerical difference in the audio stream, or show us the p value of their detection. Any claims without such evidence can and should be dismissed out-of-hand.

[Paul R]

We're talking about computers here. We can look at the exact byte stream being sent to the computer's DAC/DOUT. The DAC doesn't care whether those bytes came originally from AIFF or ALAC. If there are any codec or RAM issues, then we can detect these by looking directly at DOUT. No one has posted any differences between these two outputs. If there is such a difference -- possible, but not plausible -- that indicates a software failing, as the file sizes and sample rates are minuscule relative to modern computer capabilities. But again, no evidence has been presented that such differences exist, even though it would be very easy to do so on a computer.

 

Which again nobody disputes.

 

But we are *not* talking computers here, except peripherally. We are talking about what people hear. A very different and much much more difficult to measure issue.

[ringenesherre]

As far as I can tell, which I bashfully admit is quite well indeed, nobody is claiming the data is different, only that the same data sounds different on different machines, playback software, and audio chains.

 

Except for sandyk who keeps claiming that bit-identical files played through the same output chain sound different ;-)

 

Ok, not going there

 

Cheers,

Peter

[elcorso]

Alex I can confirm that as I also went from a 1 GB RAM Mac Mini with Leopard to a 8 GB RAM Mac Mini with Mountain Lion. On my old 2007 1 GB RAM Mac Mini AIFF music files were clearly superior to Apple Lossless. With my new Mac Mini's 8 GB of RAM the differences are lessened to a great degree and I find WAV, AIFF, FLAC and Apple Lossless to sound almost the same. Since both FLAC and Apple Lossless are bit perfect when unpacked, I feel that the reason that the sonic differences were larger on the old Mac Mini had to do with the real time unpacking while the music file was being played and now that I have 8 times the operating memory that the unpacking process leaves hardly any sonic footprint on the sound.

 

So adding my experience to yours and Boris with enough operating memory the sonic differences between lossless and uncompressed music files seem to disappear.

 

Hi Teresa,

 

You can read from my signature I have plenty of RAM, a very fast CPU, super fast RAM and SSHD (very low latency), the best and transparent MAC music player, A+ (to my ears) and the evidence of the difference continue (to my ears again).

 

What you and others call "sound almost the same", this "almost" means to me light years in digital music evolution, from year 2008 until today, thanks to the endeavor of wise people developing and perfecting music players and DACs each day.

 

The other day was in my home a musician friend who carried a USB stick with some music he listen when jogging, asked me to play it (MP3 at high sample rate & some ALAC) and he was fascinated with the SQ (A+ & Playback Design DAC). I had some of his music in my music library (under AIFF & WAV), when I played it he told me "my God, there is a huge difference...!".

 

I should add that ALAC (never MP3) isn't total unlistenable, but to my taste I lose the emotion, detail, sweetness, soundstage, etc., etc. Then I repeat: Why to have a piece of the full cow if I have enough room on hard disk storage, and as today prices could be a no sense?

 

I like to repeat again, I don't need to proof anything, I'm very happy my way as @sone is happy on his one.

 

Happy listening,

 

Roch

[Jud]

p levels are necessary to support positive claims. I'm making a negative one for myself and a few others in that post.

 

Right, back to parsing burdens of proof and such. If you would like to make this about a scientifically valid degree of proof I have doubts we'll get there soon, unless you know of someone who has grant money for a real experiment, rather than amateurs working with limited consumer grade equipment, limited consumer grade minds (except for the 4 or 5 qualified audio engineers who know better than to get involved for the umpteenth time in one of these threads - I joked about that a couple of posts back, and should have had enough sense to take my own advice), and individual personal anecdotes rather than statistically valid data.

 

So barring actual valid lab experimentation, we're all just here chatting, OK? I don't know if you do the positive and negative claim, burden of proof, bla, bla thing when you're chatting with friends or acquaintances, but I personally try to avoid that, as it reminds me far too much of my day job in law.

[Paul R]

(grin) Hi Peter - Actually, I think Alex is saying the same files played from different recorded media can sound different. Controversial, but not without some merit.

-Paul

 

 

Except for sandyk who keeps claiming that bit-identical files played through the same output chain sound different ;-)

 

Ok, not going there

 

Cheers,

Peter

[Audio_ELF]

About as relevant...

 

Deep Throat: Mister Mulder, why are those like yourself, who believe in the existence of extraterrestrial life on this Earth, not dissuaded by all the evidence to the contrary?

Mulder: Because, all the evidence to the contrary is not entirely dissuasive.

Deep Throat: Precisely.

Mulder: They're here, aren't they?

Deep Throat: Mister Mulder, they've been here for a long, long time.

[sone]

Jesse Pinkman: So you do have a plan! Yeah, Mr White! Yeah Science!!

 

Richard Feynman: The first principle is that you must not fool yourself, and you are the easiest person to fool.

 

Raising the issue of the cost of experiments of missing the point. The tests we are discussing are simple, can be done for free in little time, and do have an impact on an audiophile's budget and time. First, storage costs about $100/TB. Second, and just speaking for myself, I have thousands and thousands of dollars in nice equipment that I want to sound nice and want to enjoy listening to. Like everyone else, I must make subjective judgements on audio quality, but draw the line when subjectivity is easily trumped by objectivity, as it often can be on a computer.

 

To achieve a significance of less than 5%, you only need to listen independently 5 or 11 times or so. If you listen five times, you must be correct all five times to be reasonably certain you can actually distinguish a difference. If you listen 11 times, you're allowed to be wrong twice.

 

This doesn't take long, and if you keep yourself honest, or have someone test you, then you can be reasonably certain that you're doing no better than chance and shouldn't waste your time and effort worrying about the difference.

 

The binomial significance tests for these home experiments:

 

Wolfram Alpha:

 

BetaRegularized[1-1/2, 5-0, 0+1] - Wolfram|Alpha

BetaRegularized[1-1/2, 11-2, 2+1] - Wolfram|Alpha

 

Python:

 

>>> scipy.stats.binom.sf(5-1,5,0.5)

0.03125

>>> scipy.stats.binom.sf(11-2-1,11,0.5)

0.032714843749999986

[4est]

I think the point was: "Why do I need to convince you?"

 

Raising the issue of the cost of experiments of missing the point. The tests we are discussing are simple, can be done for free in little time, and do have an impact on an audiophile's budget and time. First, storage costs about $100/TB. Second, and just speaking for myself, I have thousands and thousands of dollars in nice equipment that I want to sound nice and want to enjoy listening to. Like everyone else, I must make subjective judgements on audio quality, but draw the line when subjectivity is easily trumped by objectivity, as it often can be on a computer.

 

To achieve a significance of less than 5%, you only need to listen independently 5 or 11 times or so. If you listen five times, you must be correct all five times to be reasonably certain you can actually distinguish a difference. If you listen 11 times, you're allowed to be wrong twice.

 

This doesn't take long, and if you keep yourself honest, or have someone test you, then you can be reasonably certain that you're doing no better than chance and shouldn't waste your time and effort worrying about the difference.

 

The binomial significance tests for these home experiments:

 

Wolfram Alpha:

 

BetaRegularized[1-1/2, 5-0, 0+1] - Wolfram|Alpha

BetaRegularized[1-1/2, 11-2, 2+1] - Wolfram|Alpha

 

Python:

 

>>> scipy.stats.binom.sf(5-1,5,0.5)

0.03125

>>> scipy.stats.binom.sf(11-2-1,11,0.5)

0.032714843749999986

[sone]

I think the point was: "Why do I need to convince you?"

 

This answers the OP's question quantitatively—you'd have to reliably hear a difference 5/5 times, or at least 9/11 times to positively assert an audible difference in a set of fair trials. This is basic math and has nothing to do with me.

[4est]

Fair enough, he did ask for blind testing.

 

This answers the OP's question quantitatively—you'd have to reliably hear a difference 5/5 times, or at least 9/11 times to positively assert an audible difference in a set of fair trials. This is basic math and has nothing to do with me.

[Teresa]

Hi Teresa,

 

You can read from my signature I have plenty of RAM, a very fast CPU, super fast RAM and SSHD (very low latency), the best and transparent MAC music player, A+ (to my ears) and the evidence of the difference continue (to my ears again).

 

What you and others call "sound almost the same", this "almost" means to me light years in digital music evolution, from year 2008 until today, thanks to the endeavor of wise people developing and perfecting music players and DACs each day.

 

The other day was in my home a musician friend who carried a USB stick with some music he listen when jogging, asked me to play it (MP3 at high sample rate & some ALAC) and he was fascinated with the SQ (A+ & Playback Design DAC). I had some of his music in my music library (under AIFF & WAV), when I played it he told me "my God, there is a huge difference...!".

 

I should add that ALAC (never MP3) isn't total unlistenable, but to my taste I lose the emotion, detail, sweetness, soundstage, etc., etc. Then I repeat: Why to have a piece of the full cow if I have enough room on hard disk storage, and as today prices could be a no sense?

 

I like to repeat again, I don't need to proof anything, I'm very happy my way as @sone is happy on his one.

 

Happy listening,

 

Roch

 

Hi Roch,

 

Good catch, yes the differences to my ears are considerably smaller with 8 GB RAM versus 1 GB RAM but they are still there. I see you have 16 GB of RAM and superb hardware (Mojo Audio Mac Mini and Playback Designs DAC.) I am using Mojo Audio's Lucent Copper Ribbon Interconnects from DAC to Preamp and from Preamp to Power Amp.

 

In the comparisons I performed with the new 8 GB Mac Mini in every single case the original music file format I downloaded sounded the best so I wonder if in addition to differences heard being due to unpacking lossless files if there are any unaccountable changes in format conversions even though they are bit perfect? I use XLD for format conversions, Pure Music software and Teac UD-501 USB DAC.

 

With a FLAC download I found I preferred in this order: FLAC, WAV, AIFF, ALAC, however with a WAV download I preferred WAV, AIFF, FLAC, ALAC. I was surprised I actually preferred the original lossless FLAC music file over a conversion to either uncompressed music format. On my older 1 GB Mac Mini I always preferred WAV, AIFF, ALAC no matter the file format of the original file, I couldn't play FLAC on the older computer. Of course I realize in the case of lossless formats there was a file conversion by the record company or reseller as masters are either WAV or AIFF.

 

Also since I now have "Memory" playback with Pure Music, could that be a third reason the differences are smaller?

 

Thus that is now three possible reasons for sonic differences:

 

1) unpacking lossless files in real time while the music is playing

2) format conversions

3) lack of memory playback

 

However these differences are so small now I have to really concentrate to hear them whereas on my older Mac Mini they were considerably more pronounced. Thus WAV, AIFF, FLAC and ALAC all sound superb as long as the original master is superb so I am leaving music in the format I download it in unless the original download is WAV in which case I convert it to AIFF for the album cover.

[Cycleman]

Three questions, but only for those here who prefer AIFF files to ALAC.

 

1. If I took my music files, ripped as ALAC files, and converted them to AIFF, would they be the same AIFF files you prefer?
   
2. And if they would be the same, does it matter which software program I used initially to rip those CDs into ALAC files?
   
3. And would it matter which software program I use now to convert the ALAC to AIFF?
   

Just curious about your perspective, really, since I have no specific religion in this argument (and yes, that's what it sounds like to this observer, an argument).

 

Dave, who purposely used the term "software program" above rather than "apps" just to be old-fashioned