Head-Fi.org › Forums › Equipment Forums › Sound Science › FLAC vs. 320 Mp3
New Posts  All Forums:Forum Nav:

FLAC vs. 320 Mp3 - Page 31  

post #451 of 504
These last few posts are really spot on. The only reason I even posted in this thread was because I kept reading a few posts from people who were so adamant on their lossy collection sounding the exact same through and through to that guys lossless. There isn't a right or wrong decision on what each person decides, but statements like that just doesn't sit right with me. Nobody's opinion is fact.

People just have to make an informed decision based on the facts, and apply that to their situation. What they come up with probably is better than what someone else thinks they should do. Especially if they've done their research and care about what they are doing.
Edited by Rem1x - 5/2/13 at 4:10pm
post #452 of 504
Quote:
Originally Posted by Rem1x View Post

.... better than what someone else thinks they should do.

I'm trying to imagine this thread without the posts containing "You should do X..." or "You have to do Y..."

It's much shorter.

One hypothesis is that extended exposure to lossy compression turns some people awfully bossy. A supposed remedy is to warm up that tube amp, get the turntable motor idling and ready, put the magic pebbles in the magic places, lay that heavyweight audiophile vinyl on the platter and finally drop the tonearm and listen to the sweet mellow sounds of those sine wave test tones, obscure vocalists and never-heard-of-again jazz combos. And some crackles.

edit:typo
Edited by julian67 - 5/2/13 at 8:45pm
post #453 of 504
Quote:
Originally Posted by julian67 View Post

I'm trying to imagine this thread without the posts containing "You should do X..." or "You have to do Y..."

It's much shorter.

One hypothesis is that extended exposure to lossy compression turns some people awfully bossy. A supposed remedy is to warm up that tube amp, get the turntable motor idling and ready, put the magic pebbles in the magic places, lay that heavyweight audiophile vinyl on the platter and finally drop the tonearm and listen to the sweet mellow sounds of those sine wave test tones, obscure vocalists and never-heard-of-again jazz combos. And some crackles.

edit:typo

I don't know if they are bossy or just really defensive. Some people get upset when they come to realize that their opinions aren't facts. Especially when they have a collection of audio so large, they realize changing it is almost impossible. And other people are understanding and just create reasonable examples, situations and ideas. I don't mean to sound rude or harsh. Just think that is why discussions like this can get tense. I for one have learned a lot in this thread, and have really enjoyed talking to you guys.
post #454 of 504

Lossy compression gets a lot of flak on head-fi and it's mostly completely unwarranted and without giving it a proper chance. Too many times have I heard of people on this forum make jokes about how it's stupid to use MP3's with high end equipment when it's really not true at all. High-end equipment doesn't need lossless files to shine, that's just ridiculous.

 

So there's a need for the advocating of ABX testing and the likes. There's no need to consider it a personal attack. The reason that so many people suggest it is because most are surprised by it the first time they do it. Not to mention if somebody is trying to prove that they can hear differences, it is a way to prove it. And the burden of proof is on those who can hear the difference.


I've heard people say many times that they can easily distinguish between mp3(of high bitrate) and lossless - NONE have provided evidence that they can do so. If it were easy to do so, providing evidence would also be easy. Shouldn't take more than a couple minutes. Does this lack of evidence mean that there isn't an audible difference? No, but it certainly points in that direction. I'm cool with the whole peace of mind thing, as stated before I rip lossless for exactly that reason, but until some people start to provide evidence on audible differences these high bitrate lossy files shouldn't be looked down upon as having inferior audio quality.

 

Another funny thing - there have been multiple blind tests where many users said they PREFER a 128kbps MP3 over the lossless alternative. I think this may be due to the fact that at this level some high frequency grain can be removed making the rest sound more clear.  


Edited by chewy4 - 5/3/13 at 11:52am
post #455 of 504
Quote:
Originally Posted by chewy4 View Post

I've heard people say many times that they can easily distinguish between mp3(of any bitrate) and lossless - NONE have provided evidence that they can do so. 

That is simply untrue.

How is it possible to discuss the subject if people fail to even acknowledge the self evident existence of anything that they don't approve of?

It's impossible to "distinguish between mp3(of any bitrate) and lossless"???? Of any bitrate? This is such nonsense that it surely arises from dogma or belief because it has no connection with reality, experience or reason. I don't think there is even one developer or knowledgable advocate of lossy encoders who would claim that "mp3(of any bitrate) and lossless" cannot be distinguished.

There are numerous published and freely accessible listening tests which demonstrate very definitely that lossy audio can be distinguished in abx tests from lossless easily at low bitrates, less easily at medium bitrates and only with great difficulty or not at all at high bitrates.

Anyone can make a set of samples at different bitrates and abx them against lossless and find exactly that. It's hardly a secret or a controversy.

I can only speak for myself but I haven't made any claim that I can always or easily distinguish high bitrate lossy from lossless in an abx.

What I have said:

Some samples/passages still defeat lossy encoders. These aren't common but do exist and any large collection is likely to contain some. Some "killer" samples are well known and freely available (several used to be distributed as part of a training pack for abx testing). If you listen to your music long enough you might even notice you have a few of your own.

abx testing is really hard. Again I'm speaking for myself; other people may find it a breeze. I find it really hard to produce a result in an abx test except on samples which are tonally simple and ideally also of short duration. As I described in a recent post I easily identified an unsatisfactory encode during normal listening and then found it extremely difficult but not impossible to abx it. Note the contrast: easily vs extremely difficult.

Null results are null.

Subjective and relative tests are subjective and relative.

A void does not constitute evidence, data or proof.

High quality lossy encoding might often be as good as the lossless source in listening. Of course it might. But it can never be better! And as source for making files with different encoders it cannot be as good.

The only advantages of lossy encoding are:
reduction in file size or reduction in bandwidth required.
some personal devices only support lossy format X or lossy format Y.

If you can afford 3 pennies per GB for disk storage and you're not constrained by the limitations of a personal player then those advantages evaporate.

There is nothing you can do with a lossy file that you can't do with lossless.

There are things you can do with lossless that you cannot do as well with lossy i.e. proper back ups, highest quality transcoding.

Playing lossless files means always hearing the music at best possible quality on your playback equipment. Guaranteed. Playing lossy files means it might do. No guarantee.
Edited by julian67 - 5/3/13 at 11:49am
post #456 of 504
Quote:
Originally Posted by julian67 View Post


That is simply untrue.

How is it possible to discuss the subject if people fail to even acknowledge the self evident existence of anything that they don't approve of?

It's impossible to "distinguish between mp3(of any bitrate) and lossless"???? Of any bitrate? This is such nonsense that it surely arises from dogma or belief because it has no connection with reality, experience or reason. I don't think there is even one developer or knowledgable advocate of lossy encoders who would claim that "mp3(of any bitrate) and lossless" cannot be distinguished.

There are numerous published and freely accessible listening tests which demonstrate very definitely that lossy audio can be distinguished in abx tests from lossless easily at low bitrates, less easily at medium bitrates and only with great difficulty or not at all at high bitrates.

Anyone can make a set of samples at different bitrates and abx them against lossless and find exactly that. It's hardly a secret or a controversy.

I can only speak for myself but I haven't made any claim that I can always or easily distinguish high bitrate lossy from lossless in an abx.

What I have said:

Some samples/passages still defeat lossy encoders. These aren't common but do exist and any large collection is likely to contain some. Some "killer" samples are well known and freely available (several used to be distributed as part of a training pack for abx testing). If you listen to your music long enough you might even notice you have a few of your own.

abx testing is really hard. Again I'm speaking for myself; other people may find it a breeze. I find it really hard to produce a result in an abx test except on samples which are tonally simple and ideally also of short duration. As I described in a recent post I easily identified an unsatisfactory encode during normal listening and then found it extremely difficult but not impossible to abx it. Note the contrast: easily vs extremely difficult.

Null results are null.

Subjective and relative tests are subjective and relative.

A void does not constitute evidence, data or proof.

High quality lossy encoding might often be as good as the lossless source in listening. Of course it might. But it can never be better! And as source for making files with different encoders it cannot be as good.

The only advantages of lossy encoding are:
reduction in file size or reduction in bandwidth required.
some personal devices only support lossy format X or lossy format Y.

If you can afford 3 pennies per GB for disk storage and you're not constrained by the limitations of a personal player then those advantages evaporate.

There is nothing you can do with a lossy file that you can't do with lossless.

There are things you can do with lossless that you cannot do as well with lossy i.e. proper back ups, highest quality transcoding.

Playing lossless files means always hearing the music at best possible quality on your playback equipment. Guaranteed. Playing lossy files means it might do.

Not sure why I put of any bitrate there, but I meant high bitrate. That's what nobody has provided evidence of there being a noticeable difference.

 

I'll edit that now.

 

post #457 of 504

I'm pretty sure I recall seeing (multiple) people post 10/10 ABX logs (or better) for 320 kbps mp3 vs. lossless, with a modern encoder, say LAME 3.98 or whatever.  You know, there are ways to cheat if you want, but I didn't really see a reason not to take it at face value.

 

Anyway, people are free to use whatever they want.

post #458 of 504
Quote:
Originally Posted by mikeaj View Post

I'm pretty sure I recall seeing (multiple) people post 10/10 ABX logs (or better) for 320 kbps mp3 vs. lossless, with a modern encoder, say LAME 3.98 or whatever.  You know, there are ways to cheat if you want, but I didn't really see a reason not to take it at face value.

 

Anyway, people are free to use whatever they want.

Where at?

 

I believe I've only seen one but the user admitted they weren't listening under normal conditions - cheating by messing with an equalizer to bring up the artifacting frequencies and whatnot. But if there are some people who did it legit I stand corrected, although I'd like to see them first before passing any judgment. At any rate I mostly just take issue with those who say that the difference is easy to detect. People who say that are either unaware of how hard it actually is, or have a very bizarre definition of easy.

post #459 of 504
Quote:
Originally Posted by chewy4 View Post

Where at?

 

I believe I've only seen one but the user admitted they weren't listening under normal conditions - cheating by messing with an equalizer to bring up the artifacting frequencies and whatnot. But if there are some people who did it legit I stand corrected, although I'd like to see them first before passing any judgment. At any rate I mostly just take issue with those who say that the difference is easy to detect. People who say that are either unaware of how hard it actually is, or have a very bizarre definition of easy.

Like this? (posted my result at page6 btw)

 

So far I have 2 success case in abx-ing 320mp3 vs lossless, had to tried really hard to detect the differences though.

post #460 of 504
Quote:
Originally Posted by chewy4 View Post

Where at?

 

I believe I've only seen one but the user admitted they weren't listening under normal conditions - cheating by messing with an equalizer to bring up the artifacting frequencies and whatnot. But if there are some people who did it legit I stand corrected, although I'd like to see them first before passing any judgment. At any rate I mostly just take issue with those who say that the difference is easy to detect. People who say that are either unaware of how hard it actually is, or have a very bizarre definition of easy.

 

You need to find the right track, right part, maybe listener.  Most probably still not obvious.

 

Note that for people posting ABX logs and quoting probabilities, if they are viewing the results while testing, and/or restarting and/or trying multiple times, that's a more "complicated" test statistically, certainly a way to increase false positives in general unless you are accounting for those things.  The probability doesn't really apply to somebody going through until they hit a streak of X/Y that crosses "p < 0.05" mark.**  But still, anyway, you can interpret the data.

 

Just doing a search:

 

http://www.head-fi.org/t/573974/lossless-vs-mp3-abx-results-among-other-abxs#post_7791681

http://www.head-fi.org/t/431522/abx-test-of-320kbps-vs-flac-results#post_5797757

http://www.head-fi.org/t/431522/abx-test-of-320kbps-vs-flac-results/60#post_6574596

(edit: yeah, I missed searching for ones with actual 10/10, but hopefully you'll forgive me on that.. also some results that include misses are more significant than 10/10 anyway)

 

I remember reading some others too.

 

Seems like encoders have more trouble with tracks going to 0 dBFS, which is kinda expected maybe.

 

 

** I don't read xkcd, but there's a relevant comic here:

http://xkcd.com/882/


Edited by mikeaj - 5/3/13 at 12:32pm
post #461 of 504
I also wondered "where at?" and then used the mighty google and found a few. Some are old enough to now probably be limited to curiosity value i.e. Hydrogen Audio - Finally I can ABX Lame insane or http://www.head-fi.org/t/223612/can-you-hear-the-difference-between-lame-v0-and-lossless/60 but others are relatively recent i.e. http://www.head-fi.org/t/441084/my-lossy-vs-lossless-abx-experiment

Anyway there are various blogs, posts and similar out there which turn up relevant results with a search of "abx lame lossless" "abx lossy lossless" "abx mp3 lossless" and similar. I have only skimmed a couple of these so don't have any opinion on them yet.
post #462 of 504

Thanks for the links guys. I'd say some of those are significant, particularly the deadlylover one.

 

mikeaj you do make a good point about people cherry picking result sets - and unfortunately people do this and don't even see why it's wrong. Anyone can get a 10/10 without even listening given they give it enough tries. I've even trumped someone else's results without listening when they were trying to claim a cherry picked 7/7 result meant that they passed. But the deadlylover one definitely seems honest enough.

post #463 of 504

It would be nice to have a copy of the exact test files used.  I'd like to try the test myself or even look at them with an analyzer to see if there is anything that might show why they might sound different.  While I'm sure my ears and equipment would never pass, it would be interesting to see how others might perform with the same data.  Also, it would be interesting to see if the same song could be encoded by someone else to see if perhaps something was slightly amiss with the initial encoding.  The ABX is a first step, but it needs to be "peer reviewed" in a sense.  

post #464 of 504
Quote:
Originally Posted by chewy4 View Post

some of those are significant, particularly the deadlylover one.

Agreed. That's a very interesting thread and I'm only halfway on page two!

There's someone who successfully abx the difference between high quality modern mp3 and lossless and who says
Quote:
Originally Posted by deadlylover View Post

.....

Yeah I feel like I'm not missing anything at all when using mp3's, I just wanted to see if it was possible to distinguish them. For all intents and purposes, mp3's are more than transparent enough for everyday listening if there aren't any insane compression artifacts.

It's more of a feeling really, rather than hearing. No wonder it's so bloody difficult to ABX them. When I mentioned 'easy' it was in relative terms, I wouldn't want to bet my left nut on being successful every time I try, even on the easiest tracks.

I'm going to take a (hopefully) long break from ABX testing, as it's REALLY draining......

Which I think is really well expressed description of the quality of compressed audio, its experiential difference and similarity to lossless, what listening to music is actually about (feeling, stimulation) and why and how abx testing can be so challenging that ascribing meaning or qualities to null "results" means one has moved from knowing to believing.
post #465 of 504
Quote:
Originally Posted by sonitus mirus View Post

The ABX is a first step, but it needs to be "peer reviewed" in a sense.  

No, it doesn't need to be peer reviewed because it's a purely subjective test.

He used apps that are accepted as a kind of de facto standard because they work as described and he used the version of lame current at the time. Then he published the logs. If someone publishes similar logs from a similar test but the scores show 5/10 or 12/30 would people start to question an obviously knowledagble author's competence to make an mp3 from a flac in foobar or suggest this needed intervention to verify? I doubt it. I think you have to assume good faith unless there is some specific reason not to. The author does nothing to suggest he is anything other than entirely credible.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
This thread is locked  
Head-Fi.org › Forums › Equipment Forums › Sound Science › FLAC vs. 320 Mp3