Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: ABXing a lame vbr 0 file to the original FLAC (Read 22514 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

ABXing a lame vbr 0 file to the original FLAC

Someone on Reddit has decided to take me up on a challenge and prove that you can hear a difference between a vbr 0 lame encoded MP3 and the original FLAC it was compressed from.

I've been trying to do this for a long time and have never been able to tell the difference.  This person claims there are certain tracks that he can use to "exploit the weaknesses" of the LAME encoder and hear a difference.

Has anyone heard of anything like this?

ABXing a lame vbr 0 file to the original FLAC

Reply #1 –
I don't know of any specific samples for MP3, but AFAIK there are known problem samples for each encoder. If you can still tell those from the original at V0 is what that someone on Reddit has to prove.

ABXing a lame vbr 0 file to the original FLAC

Reply #2 –
I think the listening tests forum on HA might have some killer samples that the developers have used for tuning LAME in the past, so you might be able to download them to test LAME's ability to achieve transparency on those samples for your ears.

ABXing a lame vbr 0 file to the original FLAC

Reply #6 –
Quote from: lozenge on
Quote from: apastuszak on
Has anyone heard of anything like this?


This post from a couple of years ago perhaps?
"You Can Call Me Al" in V0 from LAME 3.99.

It'd be interesting to see what else they come up with, care to share the link so we can follow along?


http://www.reddit.com/r/headphones/comment...byshw?context=3

ABXing a lame vbr 0 file to the original FLAC

Reply #7 –
Quote from: apastuszak on
Quote from: lozenge on
Quote from: apastuszak on
Has anyone heard of anything like this?


This post from a couple of years ago perhaps?
"You Can Call Me Al" in V0 from LAME 3.99.

It'd be interesting to see what else they come up with, care to share the link so we can follow along?


http://www.reddit.com/r/headphones/comment...byshw?context=3


Hes right.  Look hard enough and you'll find a sample that breaks typical MP3 encoders (and encoders for many other formats too).

ABXing a lame vbr 0 file to the original FLAC

Reply #8 –
Quote from: apastuszak on
Someone on Reddit has decided to take me up on a challenge and prove that you can hear a difference between a vbr 0 lame encoded MP3 and the original FLAC it was compressed from.

I don't think he is able to prove that I can hear a difference, but I wouldn't be surprised that he can prove that he can hear a difference, if only because Mp3 has a problem with temporal smearing resulting from block size limitations.  I don't even think that expert listeners have to look very hard to find samples that aren't transparent.

I think some might find it interesting to see how he does with halb27's lame variant.

ABXing a lame vbr 0 file to the original FLAC

Reply #9 –
Electronic music with sharp transients is where MP3 isn't transparent. Typical sample is EIG
None of MP3 encoders (even the last versions of LAME and all kind of extensions) could not achieve transparency on it. Even at V0 or 320 kbps.  if You throw enough bits to MP3 You can get rid of almost any artifacts like stereo issues, tonal parts etc. But not transients. 

Quality is actually pretty high but pre-echo artifact is still there.

P.S. It's impossible to 'fix' it as it isn't the issue of some particular MP3 encoder (LAME, Helix, etc...) but is a format limitation. MP3 short block isn't short enough.
AAC fixes it by shorter block size (120 or 128. In comparison MP3's short block is 192). Plus AAC has TNS.

AAC, Opus, MPC and Vorbis Ogg have noticeble advantage over MP3 on transients.

ABXing a lame vbr 0 file to the original FLAC

Reply #10 –
Quote from: saratoga on
Quote from: apastuszak on
Quote from: lozenge on
Quote from: apastuszak on
Has anyone heard of anything like this?


This post from a couple of years ago perhaps?
"You Can Call Me Al" in V0 from LAME 3.99.

It'd be interesting to see what else they come up with, care to share the link so we can follow along?


http://www.reddit.com/r/headphones/comment...byshw?context=3


Hes right.  Look hard enough and you'll find a sample that breaks typical MP3 encoders (and encoders for many other formats too).


But do you condemn a lossy encoder and tell everyone to use FLAC for the one or two files in your library that might give you some problems?

ABXing a lame vbr 0 file to the original FLAC

Reply #11 –
Quote from: lozenge on
Quote from: apastuszak on
Has anyone heard of anything like this?


This post from a couple of years ago perhaps?
"You Can Call Me Al" in V0 from LAME 3.99.

It'd be interesting to see what else they come up with, care to share the link so we can follow along?


I just ABXed Call Me Al from the 25ht anniversary edition of Graceland released in 2012.  Made a 16/44.1 FLAC and converted to vbr -v0 MP3 using foobar2000 and LAME 3.99.5 and could not tell the difference.

ABXing a lame vbr 0 file to the original FLAC

Reply #12 –
Quote from: apastuszak on
But do you condemn a lossy encoder and tell everyone to use FLAC for the one or two files in your library that might give you some problems?


Of course not, but that is also not what you said in that reddit thread.

ABXing a lame vbr 0 file to the original FLAC

Reply #13 –
Quote from: saratoga on
Quote from: apastuszak on
But do you condemn a lossy encoder and tell everyone to use FLAC for the one or two files in your library that might give you some problems?


Of course not, but that is also not what you said in that reddit thread.


Well, what I said is true.  No one has ever shown me any ABX results that shows they can discern a vbr 0 MP3 from a FLAC.  Just a "I can definitely tell the difference."  And when I press them, they proceed to tell me how they need to be at home in from of their tube amp and can't just abx them on a PC.

ABXing a lame vbr 0 file to the original FLAC

Reply #14 –
Quote from: IgorC on
Electronic music with sharp transients is where MP3 isn't transparent. Typical sample is EIG
None of MP3 encoders (even the last versions of LAME and all kind of extensions) could not achieve transparency on it. Even at V0 or 320 kbps.  if You throw enough bits to MP3 You can get rid of almost any artifacts like stereo issues, tonal parts etc. But not transients. 

Quality is actually pretty high but pre-echo artifact is still there.

P.S. It's impossible to 'fix' it as it isn't the issue of some particular MP3 encoder (LAME, Helix, etc...) but is a format limitation. MP3 short block isn't short enough.
AAC fixes it by shorter block size (120 or 128. In comparison MP3's short block is 192). Plus AAC has TNS.

AAC, Opus, MPC and Vorbis Ogg have noticeble advantage over MP3 on transients.


I don't usually listen to electronic music, so I would not have picked up on this.

Right now, I'm streaming from my house using Ampache, which is converting FLACs on the fly to vbr 0 MP3s.  My biggest concern really is which lossy codec will convert the fastest and use the least amount of CPU.

ABXing a lame vbr 0 file to the original FLAC

Reply #16 –
Also headphones are more artifacts-revealing than speakers. So a quite good pair of headphones is needed to hear distortion.

ABXing a lame vbr 0 file to the original FLAC

Reply #17 –
Quote from: IgorC on
Also headphones are more artifacts-revealing than speakers. So a quite good pair of headphones is needed to hear distortion.

Wouldn't the wacky frequency response in some not-quite-good headphones mask some artifacts and exaggerate others? Psychoacoustic models assume flat response from the drivers, right?

ABXing a lame vbr 0 file to the original FLAC

Reply #18 –
Yes, You're right.

ABXing a lame vbr 0 file to the original FLAC

Reply #19 –
Several years back someone made mention of someone being able to hear lossy artifacts more easily because he had fairly substantial high frequency loss.

While this community is generally suspicious of claims as they relate to sound quality, not all of them are manifestations of the imagination. I certainly wouldn't make bets against people being able to distinguish mp3 from the original at any quality level with any musical passages.  I wouldn't do this with DACs or amps either after reading about designs that are often referred to as being pathological.

ABXing a lame vbr 0 file to the original FLAC

Reply #20 –
Quote from: greynol on
Several years back someone made mention of someone being able to hear lossy artifacts more easily because he had fairly substantial high frequency loss.

While this community is generally suspicious of claims as they relate to sound quality, not all of them are manifestations of the imagination. I certainly wouldn't make bets against people being able to distinguish mp3 from the original at any quality level with any musical passages.  I wouldn't do this with DACs or amps either after reading about designs that are often referred to as being pathological.



Well:

no one say there is any lossy codec setting that is always transparent, under every condition

no one says DACs can't ever sound different

no one says amps can't ever sound different


The Amirs of the world use those straw men to ridicule 'objectivists'.

ABXing a lame vbr 0 file to the original FLAC

Reply #22 –
Quote from: apastuszak on
... I don't usually listen to electronic music, so I would not have picked up on this. ...


I don't either. That's why I don't care any more about such stuff being not transparent. I still expect a very acceptable quality even for such music.

Most of the mp3 killer samples with respect to transients belong to this class.

I expect transparency for music originating from natural instruments (including human voice).
The situation is close to perfect here with mp3 at least when using high bitrate and a good encoder.

lame3995o -Q1.7 --lowpass 17

ABXing a lame vbr 0 file to the original FLAC

Reply #23 –
Quote from: apastuszak on
This person claims there are certain tracks that he can use to "exploit the weaknesses" of the LAME encoder and hear a difference.
Has anyone heard of anything like this?

Some people reported succesfully ABX-ing Kraftwerk tracks in Lame MP3 at 320 kbps.


Quote from: greynol on
Several years back someone made mention of someone being able to hear lossy artifacts more easily because he had fairly substantial high frequency loss.

I wonder if this is the reason why people who have been performing musicians for decades, such as Neil Young and Steven Wilson (maybe Flea and Slash too) do actually hear something different when they claim they can easily tell an MP3.
(Talking about both age-induced and exposure-induced hearing loss).

ABXing a lame vbr 0 file to the original FLAC

Reply #24 –
Quote from: Hex144 on
I wonder if this is the reason why people who have been performing musicians for decades, such as Neil Young and Steven Wilson (maybe Flea and Slash too) do actually hear something different when they claim they can easily tell an MP3.

My guess is that they were exposed to mp3 with bit rates that are considered low (128 kbs or lower). Also they never mention proper level matched ABX tests or what encoder was used (iTunes?). I could imagine that comparing a low bit rate mp3 to a studio master might be audible.
In theory, there is no difference between theory and practice. In practice there is.