FLAC on Mac OS X

[ziggyzack]

I'll tone it down, by all means. It seemed to me there's a lot of wooly debate going on, and I think that kind of thing can get out of hand, but fair dos, I'll can it. I'm just a punter, not a pro, but I've picked up enough to know where I'm generally on the ball, and where I'm fuzzy - I'll always say when I'm speculating.

 

Right - I think the misunderstanding here is that I'm talking from the point of view of ethernet dacs, like the Transporter, the Linn DS range, and the PS Audio Perfectwave dac (once PS Audio finally release their ethernet interface). I was attempting to refer to these generically as a 'dac box', distinct from the dac chip itself - apologies if that wasn't clear. Of course, when receiving an audio stream via ethernet, in flac, alac, wav, mp3, etc., these boxes have to process the stream before feeding the DAC chip itself.

 

 

Can you share with everyone what is the source of your "received wisdom?"

 

Like I know penicillin is an antibiotic, I know flac is a lossless codec. I haven't verified it for myself! And you can only establish losslessness from the design, and from file comparisons, not from listening tests, which is why I thought it odd you asked if I'd performed listening tests to see whether flac was lossless.

 

 

If I have this right, your "received wisdom" tells you something is technically doubtful because of the people who believe in it?

 

Where subjective, Yes! Otherwise you ignore psychology.

 

On the penultimate couple of paragraphs, sorry but I think you're off-base. Some people using ethernet dacs do express a preference for uncompressed files formats over losslessly compressed formats. This, to me and many others, seems far-fetched, especially as there's expectation bias in favour of uncompressed (because uncompressed is 'closer to the truth'). I was putting forward the possible explanations I have heard - I'm glad they sound unconvincing to you, because they sound unconvincing to me! It's really clutching at straws. E.g. by 'an inward stream being half as busy with FLAC?', I mean that the ethernet port and everything up to the decoder has to deal with half the data rate for flac than for uncompressed, as fewer data packets need to be sent for the compressed format. That's a fact - but does it explain perceived differences in sound quality between uncompressed and flac?

 

The point is that the only reasons anyone can give for uncompressed sounding better or worse than lossless, in the context of ethernet dacs at least, are pretty tenuous. Note I never said it was impossible: 'I am NOT saying that there is no benefit in transcoding flac to an uncompressed format before it's streamed to a dac box, though I think it's highly doubtful - I certainly wouldn't take anyone's word for it.'

 

ZZ

 

[Lance]

I got that 16KHz upper limit from years of research in audiology and linguistics. Some sources show 17KHz, others less. It's an exponential change in that there are only 12 keys of a piano between 10KHz and 20KHz (mathematically speaking, not on any real piano). Human hearing begins to fall off at the high end starting at age 8! By age 20, that 20KHz limit is a myth. Many casual statements are made about that 20Hz-20KHz range because one source heard it from another. They can't all have actually tested human hearing and found that the average is exactly 20,000Hz, not 19,000, not 21,000.

From http://hypertextbook.com/facts/2003/ChrisDAmbrose.shtml

"The highest frequency that a normal middle-aged adult can hear is only 12-14 kilohertz."

 

Before I entered academics, I was in electronics and can tell you that your grandfather may hear a whine from a CRT set. That is typically caused by a leaky capacitor and may produce a sound anywhere in the range of 8KHz to 16KHz.

 

Even though the FLAC developers have found a way to make it lossless (in theory), we have a range of compressions available for it as well as VBR. If all of these compression levels and VBR result in a lossless file, why not use the max compression and VBR? What would be the reason for creating a larger file than necessary?

 

I argue that FLAC is pointless is because it doesn't fit the definition of the purest, uncompromised form of computer sound file format. That would be WAV or AIF since they are uncompressed and required for nearly every professional application. For an archive of a precious, unreplaceable CD, I am choosing AIF. I didn't say FLAC is lossy or faulty. I said it isn't needed. Those who want something that is the paragon of uncompromized computer sound file have WAV and AIF. Those who want compression, but not a loss in audible quality have AAC 256K. We don't need sixteen different formats or even six for that matter. The biggest limitation of FLAC is that it isn't compatible with a very wide range of hardware. Sure, it works on a dozen portable players that most of us have never heard of, but it doesn't work on the most popular one without altering the firmware. I agree that an iPod is not an audiophile's dream, but who doesn't like to hear music on the go? If we have a format that has "perfect" sound (by that I mean you can't hear any fault in it), and that format makes a file that is 25% of the uncompressed AIF, and we can't hear any difference, and it plays on virtually any portable MP3 device, well, isn't that as good as it needs to get for all but archiving? We are chasing the FLAC, trying our best to justify clutter up our computers with first one utility and then another to enable use of this thing. We just heard it through the grapevine that it is our panacea.

 

[Dan Gravell]

Even though the FLAC developers have found a way to make it lossless (in theory), we have a range of compressions available for it as well as VBR. If all of these compression levels and VBR result in a lossless file, why not use the max compression and VBR? What would be the reason for creating a larger file than necessary?

 

Surely if it's lossless, it's lossless. It's just a compressed form of the same thing. In algebreic terms, it's like difference between:

a * a * a * a * a

And

5a

The semantics are the same, the manifested artifact is different.

 

That seems rather formal to me, the very opposite of 'whooly'.

 

Compression and fidelity are two entirely separate things.

 

[Audio_ELF]

take the sequence ...

 

a a a b b b a a b b b a a b b b

 

This could be compressed as

 

3a 3b 2a 3b 2a 3b

 

Or as

 

a 3(aaabbb)

 

Or even

 

a 3(2a 2b)

 

All can be extracted back to the original, but the final takes up less space - but took me longer to think about.

 

Eloise

 

[ziggyzack]

Even though the FLAC developers have found a way to make it lossless (in theory)

 

I argue that FLAC is pointless is because it doesn't fit the definition of the purest, uncompromised form of computer sound file format.

 

In theory? FLAC doesn't fit the definition of uncompromised? You're not sounding too credible now Lance!

 

[Audio_ELF]

Even though the FLAC developers have found a way to make it lossless (in theory)

There is no "in theory" here. FLAC (also Apple Lossless) is lossless in all accepted definitions: that is that the file is compressed for storage, but exactly the same data used to create the compressed file can be extracted by the algorithm to de-compress it. This has been checked, argued, rechecked and agreed on and shouldn't be in doubt. This IS a scientific certainty (unless the file is corrupted or errors occur).

 

I argue that FLAC is pointless is because it doesn't fit the definition of the purest, compromised form of computer sound file format.

Actually this statement can be correct. It is compromised because when playing back the data needs to be decoded (in real time) which is another process in the playback chain. You can argue if this extra step affects sound quality (people with revealing systems do think it does), but the fact is that the extra step is a potential compromise.

 

Eloise

 

[ziggyzack]

Eloise, in my earlier comment (titled 'Nope (and you knew that!)'), I made some suggestions about why playback of uncompressed and lossless files may, just, perhaps, on-the-edge-of-plausibility, sound different.

 

That was for an ethernet dac, where all of the DAC componentry (including sundries like control logic and I/O, as well as the DAC chip itself) is completely isolated from the source of the stream. That's the optimum system design imho. Music is streamed from the hard drives via ethernet. The ethernet dac can (and should) be in a different room, on a different power supply. This separates computer components from hi-fi components, which is a good, logical and probably necessary separation of tasks.

 

In such a system, you can for example store and tag the files in flac and have the computer transcode them to wav on the fly when they're streamed. Thus all the dac sees is a bit-correct wav stream. It has no notion that it was once stored in flac. With this approach, your (generous!) interpretation of Lance's second point is rendered moot.

 

What kind of systems are people using where they can identify a difference between uncompressed and lossless? If there's a good thread on this, can you point me to it? Fair enough, if the difference is there, it may be masked in some systems, but it most likely points to a flaw in system design imho. There really shouldn't be a difference.

 

[ziggyzack]

Chris, I looked at this one again, and I don't think there was anything wrong with it. Lizard King didn't seem to mind!

 

http://www.computeraudiophile.com/content/How-my-new-hard-drive-makes-my-sound-better

 

My guru told me...

 

I hope my posting helps others out there who share my passion yet maybe are not as savvy as me with this.

 

C'mon!

 

[Audio_ELF]

ziggyzack asked... "What kind of systems are people using where they can identify a difference between uncompressed and lossless? If there's a good thread on this, can you point me to it? Fair enough, if the difference is there, it may be masked in some systems, but it most likely points to a flaw in system design imho. There really shouldn't be a difference."

 

If I'm honest ziggyzack: I am unsure if I'm able to hear any difference between lossless (ALAC format) and uncompressed audio - I can tell a difference between AAC at 256k and lossless - on my system.

 

I do know that it was reported after the Computer Audiophile symposium that people could tell the difference between lossless and uncompressed and between systems with HDD, SSD and accessing files via NAS - all of which (I agree) shouldn't affect sound quality. Given the number of people who do report differences in their systems I have come to the conclusion that there must be differences in some systems.

 

IIRC Chris' system that he reports differences being noticable is Mac Pro with Lynx AES16e feeding a Berkerley DAC and onto MacIntosh amps and Focal speakers (I may be totally wrong on amps and speakers).

 

Eloise

 

[wgscott]

If you convert say FLAC to Apple Lossless and then back, do you get back to the original file?

 

The best way is to check with the MD5sum.

 

[Audio_ELF]

"If you convert say FLAC to Apple Lossless and then back, do you get back to the original file?

The best way is to check with the MD5sum."

 

Yes convert FLAC to Lossless then back will give you the same file so long as the same compression settings are used for FLAC. The actual MD5 checksum (at OS level) may vary with differences in the header, tag information, etc. IIRC Foobar 2000 on the PC has a "verify files" facility to check that internally the files are identical.

 

What is more valid would be start with WAV, compress to FLAC (or Apple Lossless) and back to WAV. The two WAV files should be identical even with MD5 checksum at an OS level.

 

Eloise

 

[kthu]

Hi I just found this site and was browsing around when I stumbled upon this thread. I thought I could help by conducting the experiment outlined by Eloise:

 

 

~> flac -s original.wav -o original.flac

~> flac -d -s original.flac -o copy.wav

~> md5sum *.wav

eec4543432b0bad0e61ceb9aa0ffa83f copy.wav

eec4543432b0bad0e61ceb9aa0ffa83f original.wav

~> ls -s1h *.wav *.flac

48M copy.wav

30M original.flac

48M original.wav

 

 

That should lay that one dead. If you can hear any difference between a wav and a flac, you either have a buggy decoder or a vivid imagination

 

[wgscott]

Yeah, I did that too, but never got around to posting it. (I realized afterward that md5sum is something the user will have to install. I forget we don't all have fink...).

 

For those wondering, the md5sum is essentially a coded fingerprint based on the byte structure of the file, so you can see if it has been tampered with. The two files must be identical to give the same md5sum value. The experiment has to be done with a file type (like wav) that doesn't encode stuff like conversion history in the metadata header.