Cut Me Some FLAC!

Do FLAC files sound inferior to WAV files? Or more generally, do lossless compressed file formats (FLAC, ALAC, APE...) sound worse than their uncompressed twins? I've seen the argument that due to the extra processing imposed on your computer by FLAC, for example, during playback, this method is more prone to computer-introduced timing errors. Stepping back from this scenario for a moment, everyone seems to agree that the less your computer does during music playback the better it sounds so it would make sense to apply this same logic to the actual playback process. No?

If you Google this topic "FLAC v WAV" you'll find all kinds of arguments from and for both sides; the rational - there can be no difference between FLAC and WAV or if there is it's not audible, to the other side that says there's clearly an audible difference. Bruce A. Brown of Puget Sound Studios, who does work for HD Tracks, had this to say on the What's Best Forum:

If I put a wav file on one track and a FLAC file on the other track in Pyramix, I can't tell them apart, sighted or blind. Pyramix also has their own proprietary format as well called .pmf and .pmi which is similar to FLAC, because it can contain metadata, but it's not compressed.

Doing transfers for HDtracks, we spent months testing formats and sample rates.

Let's compare this to what Cookie Marenco of Blue Coast Records had to say on the Computer Audiophile forum:
About FLAC... We did a lot of early testing for FLAC because it is also less expensive for us to send when a file is smaller, but, when listening back to uncompressed FLAC against the full WAV file when both are sent through email... there is a difference and the WAV has consistently sounded better. We realize that much has been published that FLAC is bit to bit accurate, but with a listening test, this is not the case. I have suggested to many of our customers to do the experiment themselves and all have reported the same results as we have found.
And more recently, we have Todd Garfinkle of MA Recordings:
Downloads are for the most part FLAC and therefore not at the same sonic level as untouched WAV files, no matter what some people say.
Let's be clear—no one is talking about whether or not FLAC compression messes with the bits. The question is does playing back a FLAC file negatively affect the sound quality. On one hand given the ever decreasing cost (and size) of storage combined with increasing bandwidth, the entire compression issue is nearly moot (if that sounds sensible to you, feel free to stop reading). Yet this remains an important topic since the decision needs to be made by anyone ripping, downloading and storing their music—what format should I choose? One very commonsense solution I've seen offered goes like this - if the issue with FLAC is introduced because of the 'unpacking' process during playback, just convert your FLAC files to AIFF or WAV for playback.

I'd go one step further and suggest why not just do a batch conversion of all your FLAC files and keep your FLAC copies as backup which is something that Chris Conniker of the Computer Audiophile recommends in his excellent and thorough CD Ripping Strategy and Methodology and call it a day? Mac users can convert to AIFF and PC users can go WAV but realize there may be issues dealing with metadata or a lack thereof down the road. More on that metadata subject some other day. For the conversion process, there are a number of applications out there including Max (free) for Mac users and dBpoweramp for the PC ($38).

My feeling on the subject is there's no need to deal with compressed files due to the price of storage so I convert all of my downloaded FLAC files as well as rip all of my music to AIFF and WAV files to accommodate my Mac and PC systems. But for the cause and to see if I can hear a difference, I went through the tedious process, at least it feels tedious to me, of creating multiple copies of a few of the same tracks in ALAC (Apple Lossless), AIFF (Apple Uncompressed), and/or WAV formats. The tracks up for dissection:

1. Track 1 & 2 from J.S. Bach Das Wohltemperierte Clavier as performed by Davitt Moroney (harpsichord) [HMC 901285.88]
2. "Worm Tamer" from Grinderman 2 [Anti 87110-1]
3. "Bemsha Swing" from Don Cherry's Art Deco [A&M CD 5258]
4. "String Quartet No. 13 in B Flat Op. 130" from Beethoven: The Late String Quartets as performed by the Quartetto Italiano [PHILIPS 426 050-2]
5. "Cuando Silva el Viento" from Sera Una Noche's La Segunda [M062A-HR]
I figured it would be good to have a wide range of sounds more and less difficult and I loaded the above tracks into a playlist titled "Am I Really Doing This?" and listened to a few seconds of each and moved to the next and back 'n forth (I did this via my iPad using the Remote app and the tracks all had the same title so I did not know which was which). And there's no way I could accurately tell the difference between which version of the track was playing unless I got up, walked over and peeked at my MacBook Pro which I had set to show file "Kind". When looking and when I knew what was playing, I'd think I could hear a difference, maybe a bit more air (audiophile hint—when you think you hear a change and that change is limited to "air" the change you're hearing more than likely is not there and if it is, its more than likely not worth fretting over. Take a deep breath instead).

But wait, what about FLAC? So I loaded up 1 track, Bonnie "Prince" Billy's "Then The Letting Go" which is from the album The Letting Go into Decibel (Decibel plays FLAC natively on a Mac) in two versions, FLAC and WAV. I let these play over and over all the way through throughout the day* and night. Thankfully I really like this song and the rest of the album is wonderful too (as are Dawn McCarthy's vocals). After a while, over time I began to hear differences. I'm not saying the FLAC version sounded bad or even hugely different but what I did notice was the FLAC version sounded less controlled. Less harmonically right like the microphone(s) were ever so slightly overloaded at times and lost, momentarily mind you, the music's snap, grip and flow.

There was also something going on with the way the different parts relate to each other; Bonnie "Prince" Billy is very much upfront in the mix with Dawn McCarthy off somewhere in the distance. Like a ghost or a conscience. With the FLAC version this sense of separation was somewhat stunted so you could almost picture them taking turns at the same mic whereas with the WAV version Dawn is clearly far away from and behind Bonnie (or Billy). And this distance conveys a sense of loss and some other weird stuff making the WAV version more emotionally moving. It tells a deeper story**.

How large was this difference? It was large enough that I had to really listen completely and intently most of the day and into the night to feel comfortable talking about it. So the large difference appears rather small in the grand scheme of things. But so does the proposition of converting FLAC files to your uncompressed format of choice. Besides storage is cheap so why not just hedge your bets and listen uncompressed? I can't think of a single reason against this seemingly sensible proposition unless you just don't care to.

I'd take away a few things from this little experiment; I would absolutely recommend playing back your music files in an uncompressed format and I'd also suggest that when trying to hear subtle differences in the presentation of music, you have to listen to songs not pieces of sounds and you have to listen over time. Listening to a few snippets of a track and switching to another and then to another and back again is an exercise in proving that this method of listening proves nothing more than this method has no bearing whatsoever on how we listen to and more importantly how we enjoy music.



* Thanks to John DeVore for recommending the 1 track over and over test because I was ready to give up after the whac-a-mole approach which gave very different results. That reminds me of another thing I'd take away from this—people who make music or who make things that make music (or both) and listen for a living are worth listening to.

Andrea del Sarto (Andrea d'Agnolo) (Italian, Florentine, 1486–1530)
The Holy Family with the Infant St. John, 1530, oil on panel
The Metropolitan Museum of Art, New York

** For those so inclined, I'd liken this sense of space to the space represented in some paintings (like this one by Andrea Del Sarto) where the depiction of physical separation is pregnant with meaning. But you really need to go see the original in the Met to get it—this .jpg is compressed so you lose a lot in that translation.

Share | |
COMMENTS
dextro's picture

Michael, I do not know what is your scientific background, but IMHO in your little experiment there is a number of factors that can influence the results. To name a few: computer used, software used to reproduce, software used to convert, initial files...

Any of those (and others), can in itself be considered the cause of any difference. I, by no means, infer you could find a difference, what I mean is perhaps it comes from somewhere else. 

FLAC is bit perfect as is WAV, by definition there is no difference, but one thing is a definition, and the other is an implementation of it. 

Michael Lavorgna's picture

But that is not the question I'm addressing - the question is do compressed formats like FLAC sound different during playback. As with any experiment (and I use that word with tongue in cheek) you are free to interpret the results as you like.

dextro's picture

I do not want to get too deep into the philosophy of experimentation and the "empiric method", but just would like to add that one reason for experimentation is trying to prove a proposition, and in this case what I say is that your proposition cannot be fully proved with the method used.

On the other hand I definitely do not agree that one is free to interpret results of an experiment. Even if the experiment has no flaws, interpretation of results is limited, otherwise science would be more about opinions and not facts.

Michael Lavorgna's picture

"I don't know much about science, but I know what I like."

I do not see how anyone could draw the conclusion that this was a formal experiment meant to prove a point based on what I wrote. But I would be curious to know how one would apply the empirical method to our personal preferences when listening to music.

Stephen Mejias's picture

In my opinion, too many people confuse "listening to music on the hi-fi" with "science." Why?

Michael listened and he heard differences. That's all I need to know.

Michael Lavorgna's picture

I think it's time for some Holiday beer & cheer and while I cannot prove this, I prefer beer to soda. Even though they both have bubbles. Go figure.

dextro's picture

"In the beginning was the Word, and the Word was with God, and the Word was God"

Guess, this is no discusion forum, or is it?

Michael Lavorgna's picture

And that means you may not agree with every comment.

MVanilli's picture

I consider myself an audiophile, but this is too OCD and bourgeois even for me. I think that this sort of article probably alienates more than it educates. I can see it giving great ammunition to the "audiophool" camp.

If a tweak is so slight that it takes hours of listening to the same song over and over again to discern a difference, then I think that the difference is insignificant, whether it's real or psychological.

In general, though, I like your site and find it great reading! Sorry to be critical for my first post, but anonymous negativity is so hot on the Internet right now. :)

Michael Lavorgna's picture

I'm talking about file formats and whether or not we can hear a difference between them. But OCD and bourgeois? Now that's a stretch unless we want to just admit that hi-fi and a focus on the quality of the experience is inherently OCD and bourgeois.

I wish I'd said this - "On one hand given the ever decreasing cost (and size) of storage combined with increasing bandwidth, the entire compression issue is nearly moot (if that sounds sensible to you, feel free to stop reading)."

Oh wait, I did ;-)

MVanilli's picture

You could also say this:

"Given the ever decreasing cost (and speed) of CPUs and RAM, the entire issue of whether you can hear a difference between FLAC and WAV is nearly moot." As CPU speeds increase, the processing difference becomes negligible and any current difference will disappear.

You did say this:

"[At first] there's no way I could accurately tell the difference between which version of the track was playing unless I got up, walked over and peeked at my MacBook Pro..."

And then you said this:

"I let [FLAC and WAV versions of one song] play over and over all the way through throughout the day and night... After a while, over time I began to hear differences."

Sounds pretty OCD to me!

And yes, I think all hi-fi is a little OCD and bourgeois. Harmlessly so, for the most part, but you can keep diving deeper down the rabbit hole pretty much endlessly. It's healthy to stop at some point. Save yourself a few bucks and hours examining the same song.

Vincent Kars's picture

The most beautiful experiment I have encountered was somebody recording the SPDIF out of a WAV and a FLAC. This is exactly what is send to the DAC.

Of course the result was bit identical.

What is the difference between WAV and FLAC?

FLAC is compressed so you need a bit more CPU to decoded it.

WAV is uncompressed, so you need more CPU to read it from the HD.

Probably the total amount of CPU are undiscernibly in the Taskmanager or so slight making you wonder if this will affect sound quality at all.

Playing digital audio is a clock driven thing. If we assume that system activity affects the clock (yep, jitter again) we cannot rule out that different processes affect our audio in a different way.

If this is true than an audible difference between file formats is not a matter of the file format but a property of the hardware. In fact I do think it is a system error.

What we need is a guy having the gear the measure the jitter on e.g. the SPDIF out playing FLAC and WAV. Then we know for sure.

Another interesting experiment would be  to compare uncompressed FLAC with WAV.

I have heard that DbPoweramp now have an option to generate uncompressed FLAC!

Very interesting would be a comparison between uncompressed FLAC and compression factor 8.

Those who ponder to convert to WAV; tagging support in WAV is lacking a standard so portability between media players is very low.

Michael Lavorgna's picture

I'm sure you've seen this post on AudioAsylum from Gordon Rankin (of Wavelength Audio) but it's worth quoting here:

Gang,

We did double blind testing in the following way;

1) 2 computers MacBook Pro and MacBook (MacBook Air at last RMAF) both dual boot Vista Ultimate/OSX with J River and iTunes (we did not use amarra).

2) We would play a WAV and a FLAC of the same song. We would play an ALAC and AIFF. 10 songs total... most of the time like 30-60 seconds these would be randomly mixed (i.e. sometimes FLAT PCM first [AIFF/WAV] sometimes lossless first). Each person was given paper to record the scores.

Conclusion...

a) 93% could tell the difference with the slower MacBook/Air. This is over 50 people now at more than 4 shows.

b) Using a faster machine in the same test brought that number down to 58%.

Gordon goes on to add:

An interesting note... I did a 1KHz sine wave in two formats ALAC/AIFF as I was on a mac. I then ran every freaken hog application (PS CS4... etc) I could find on my machine. The THD rose up for the ALAC at a much higher rate than the AIFF.

Also interesting.

kana813's picture

+1

Gregor Samsa's picture

"given the ever decreasing cost (and size) of storage"

 

This is ostensibly a site devoted to computer audio.  This would suggest that the people writing here know something about computers.  This comment clearly indicates that you don't.  The price of hard drives is double and even triple what it was a few weeks ago.  If you paid any attention to what's happening in the world of computers, you would know that.  I suggest that you bookmark Anandtech and come back and talk to us in a few weeks.

And another thing.  I'm hardly a "flat-earther".  I bought a Cardas USB cable and am satisfied with the purchase.  But rejecting science out of hand like you and Mejias is the province of creationists, Tea-Party members, holocaust deniers, inquisition leaders and similar analphabets.  Without science, you wouldn't be able to drive to work in the morning, and Lavorgna, at your age, would probably not be alive.

And yes DbPowerAmp does convert to uncompressed Flac. It's not just a rumor, and it's pretty nifty.

Michael Lavorgna's picture

Perhaps I should subscribe to a live hard drive price ticker so I can make accurate statements regarding their price minute by minute. Tell me Gregor Samsa, how do prices of a 2TB drive compare to 1 year ago? And 3?

But for those interested in the reason for this recent price increase, here’s a good article from the Financial Times about the tragic flooding in Thailand which has impacted hard drive production. All that said, you can still find a 2TB drive for under $100 and for someone my age, that is a remarkable price.

Nice rant btw. I always enjoy when someone unwinds what could have been a useful post with puffery. Something must really be bugging you. ;-)

In the spirit of the day, have a wonderful Thanksgiving, have some turney and take a nap.

Simon Chick's picture

Passing quickly over the ranters you seem to have attracted, let's get back to the essence of your topic.

Since we all enjoy this hobby, of course we are interested in ways in which we might gain a hairsbreadth of extra quality from our systems. However, since we do want to enjoy the hobby, having reliable and transferable metadata with our music files is very important.

(Nor do we really want to listen to the same song all day in search of that hairsbreadth improvement, although I for one am glad you were willing to do so in the interests of research.)

As you briefly mentioned, wavs and metadata tend not to stay together very well.

Have you heard of a way in which you can have a flac file which is entirely uncompressed? This appeared as an option (introduced with apparently no fanfare at all) in a relatively recent iteration of dbpoweramp, by which you choose a compression setting of 0 when creating a flac.

My understanding is that it gives you a wav in a flac wrapper, so you get the benefits of a flac (especially reliably saved metadata) without having to worry about the deleterous effects (whether real or imagined) of having to decompress the file on the fly.

It amazes me that this option has not received more discussion, and I am interested in hearing of the helpful opinions or experience of others. I'd also very much like to know whether or not my understanding of the 0 compression option is correct.

Michael Lavorgna's picture

I have not tried the new uncompressed FLAC format from dBpoweramp yet but it sounds very promising for exactly the reasons you offer. It is available in release 14.1.

Here’s what dBpoweramp has to say about it:

FLAC encoder wording changed, also includes a FLAC Uncompressed encoding option (which stores audio uncompressed, for those who want WAVE PCM but with better ID Tagging).

Gregor Samsa's picture

"you can still find a 2TB drive for under $100"

No you can't.

http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=100007603+60...

Not even close. 

No need for a price ticker.  There's this thing called Google that can help you find out whether or not what you're saying is true.  It's very helpful.  It's also pretty easy to find.  Just Google it.

Michael Lavorgna's picture

Seagate 2TB GoFlex External Hard Drive with GoFlex Desk Desktop Adapter - $99

(these are still available in stores and I know because I just saw them - didn't even need to use Google!)

And here's one from your source NewEgg. And there are others for $109 (which is close) and $119 and you can get free shipping if you order today. And these are all from one source, the one you pointed me to.

But I will say that it is not as easy as it was a few months ago to find a 2TB drive for under $100 and it appears as if production for some hard drive manufacturers as well as hard drive parts manufacturers will not recover to pre-flood levels until late 2012 according to Seagate Chief Executive Officer Stephen J. Luczo. For more, check out this article on Bloomberg.

Michael Lavorgna's picture

The Seagate drive that was $99.99 at Newegg yesterday is $129 today.

But, this Western Digital drive that was $129 yesterday is $99.99 today.

I think I need that price ticker after all.

deckeda's picture

In practical terms, most people's digital collections are still small enough that they don't need to bother with lossless compression.

And if your collection is large enough to be concerned about space, pretend it isn't for the moment. And then do some listening and do your own concluding.

As for me, I get the rather very strong feeling I would have far more in my system to worry about first, such as room acoustics and would better speakers help. Or, I could be downplaying any differences present. By the time I got the big fish there taken care of I'll likely be using a computer that can not merely play the files well, but "really" well. It's a little like buying LPs even though at my present rate of system improvement I'll die before ever hearing them "really" well with a fantastic turntable setup.

Said another way, I'd wager buying a new, dedicated computer properly setup, one that "unquestionably" has no performance issues, is cheaper in the long run than buying enough storage to stay ahead of uncompressed acquisitions. And kids, don't forget you need to double your disk space if you want just one backup. And do it at least thrice for any measure of redundancy.

We know Rankin is a believer in computer playback. I wonder if in his experiments he's identified which Mac (in his experience, since he uses those) you'd need in order to play lossless compression without perceptible flaw. Would a MacPro with multiple cores and 16GB of RAM, filled with SSDs be enough? Taking these experiments "all the way" so to speak, and not messing around with laptops and consumer desktops, seems to be the implied tool necessary.

Michael Lavorgna's picture

I'd wager buying a new, dedicated computer properly setup, one that "unquestionably" has no performance issues, is cheaper in the long run than buying enough storage to stay ahead of uncompressed acquisitions.

I’ll take that bet! The total cost for an 8-Core MacPro with 512GB SSD and 16GB RAM = $5,499 from Apple (you can get the upgrades for less from OWC for a total of around $4,800).

While memory and processing power is important, a PC or Mac Mini can serve as the foundation for an excellent music server. Or you can buy a purpose-built music server for under $2,000. That leaves well over $2,000 for storage, which can buy you roughly 26TBs worth. And that should keep you well ahead of your uncompressed music acquisitions and if it doesn’t you don’t need to worry about how much money you're spending on storage.

deckeda's picture

What's been proposed, if not established is that lossless doesn't sound as good on lesser hardware, which includes good hardware also being asked to do too much.

I also get the impression that because cheaper computers should be good enough, that they are and so the solution lies elsewhere. But what if you could spend more on the CPU and still get there, and in exchange have fewer storage hassles?

Put another way, if Rankin discovered lossless suffers when a computer struggles, the takeaway is less about lossless's shortcomings than that of the computer's.

There are some very real concerns with dealing with larger storage amounts that ironically are best addressed with a beefier computer for file transfers.

Michael Lavorgna's picture

We'd have to define the size of the library in question to determine whether or not "storage hassles" would be an issue.

While this is an interesting topic, if we're talking about storage needs well in excess of a few TBs, then there are additional issues to address. And this is not the place to cover that, imo.

satkinsn's picture

The AudioAsylum note is interesting.

I have a hard time believing that machines as competent as the 'slower' ones used could make a difference, but maybe so.

If that's the case, then there's something about flac playback we (or at least I) don't understand.

I have always assumed I don't like computer playback as much because something else in my chain was bad; a crappy dac, not great headphones, my own emotional issues with computer audio.

Scott A.

Watertown NY

 

Michael Lavorgna's picture

I like that, it sounds like a diagnosis.

I doubt very much that FLAC playback would result in a dislike for computer playback but that's easy and free to put to the test - convert a few tracks to an uncompressed format and see.

slim's picture

 

about this thread, but did not quite realize what it is until now:

We all seem to agree that the musical content (as bytes in the file) is the same between WAV, AIFF, FLAC, ALAC - converting forth and back confirms this - while ID3 tags may / will differ.

Given that, any difference heard will have to be attributed to (a non-mutually exclusive combination of) the following factors:
- biological (e.g. state of mind of the listener)
- physical (e.g. temperature, air pressure conditions in the listening space)
- biophysical (e.g. the electromagnetic field of the playback computer interacting with the listener's neurons)
- technological, i.e. the playback chain.

While a full record of the three former situations is not feasible, the technological one IS - but we readers are left out on that. And that is why I consider the original post fishy: it mentions Decibel as playback program, but nothing thereafter.
- Is the music sent through core audio and transported to an external DAC via SPDIF?
- Or does Decibel talk directly to some other audio device (which would most likely be a USB DAC)? If so, does it use hog mode? etc.

From that information, it would be possible to set up adequate technological tests for differences (along the lines that Vincent Kars mentions, footnote 1).

Personally, I don't have the need for any such test. I am convinced that even a modest computer system like Mac mini coupled to a non-crappy DAC which uses its own timing will result in ZERO difference technology-wise.

 
Then I read the cited Rankin post (which is a lot fishier, still, in terms of describing the testing procedure and result interpretation) and take his conclusion to be "You need a 70,000$ system to hear those differences" - great for me, I am likely never in that situation.

Also, I try to take statements like the ones by Cookie Marenco and Todd Garfinkle about the sound detereorating from conversion to FLAC as a joke (although they are probably meant dead serious): I envision the difference in sound to stem from the bytes screaming out their frustation about having been caged in some FLAC-sorta-jail, ouch!

And finally, even though I quite like Oldhams "the letting go", I am bound to get seriously depressed from listening to that same track over and over and over again ...
 

(1)

- If the SPDIF signal is the same, the difference heard would have to be attributed to a crappy DAC (you're right, Michael, this WOULD be a diagnosis, then).
- If the SPDIF is not the same, it has to be due to the computer and one could try to identify the playback software or the audio hardware as the culprit.
Same testing path feasible with data transport over USB.

Michael Lavorgna's picture

Personally, I don't have the need for any such test. I am convinced that even a modest computer system like Mac mini coupled to a non-crappy DAC which uses its own timing will result in ZERO difference technology-wise.

There's your answer (not mine).

Predetermined conclusions sound fishy to me as does someone who finds so many things "fishy". And I'd be happy to give you the details of the setup but you apparently have no need for it.

slim's picture

in the paragraph which was not worded carefully enough (because rewritten right before sending it off). I mean to say:

I am convinced that even a modest computer system like Mac mini coupled to a non-crappy DAC which uses its own timing SHOULD BE ABLE to result in zero difference technology-wise.

(Add: Of course there are always ways to mess it up, like photoshopping or rendering video parallel to music playback, or using the computer's timing on the DAC's USB section ...)

Therefore, I am of indeed highly interested in your setup.

And I still like the adjective "fishy" in this context: the surface is shiny, but if you try to grab it, it simply keeps slipping through ...

Pages

X
Enter your AudioStream username.
Enter the password that accompanies your username.
Loading