POLL: Can you actually hear the difference between high bit-rate mp3 and lossless?

[krmathis]

Quote:

Originally Posted by evilking /img/forum/go_quote.gif "the guys" at HA didn't really care about my findings I posted early last year, even with sample tracks and ABX results. The thread died with a few responses of "well, I don't hear it", "256kbps is mostly transparent", "MP3 will always fail on certain samples" and some other nonsense.

Perhaps because they don't trust your result? They sure have a reason to be critical...

I found a way to reliably CHEAT in foobar2000's ABX utility! posted by the user 'evilking'

 

[tds101]

Quote:

Originally Posted by krmathis /img/forum/go_quote.gif Perhaps because they don't trust your result? They sure have a reason to be critical... I found a way to reliably CHEAT in foobar2000's ABX utility! posted by the user 'evilking'

WOW!!! Total cred bust!!! FAIL!!!

 

[knights]

i cant barely hear dif between 192kbps and lossless, but on some complicated tracks i can...

 

[progo]

I can hear the difference between 320 mp3 and the original, if the song is very familiar. There are few such songs. In most cases mp3 does a good job.

 

[moriez]

Nope. Well, cant hear the difference between 128, 192 or 320. Never tried lossless really.

 

[evilking]

Quote:

Originally Posted by ILikeMusic /img/forum/go_quote.gif I'd like to see how they dismissed your findings but I can find only one thread you participated in and it wasn't about the above topic, although maybe my search technique is faulty. Can you help guide us to the Hydrogenaudio thread you are referring to?

It was under a different username that I can't quite remember (I tried evil_king, evil-king, kingofevil, evildude, etc. I'm sure it was evil something or something evil

). I had to re-register to post the ABX cheat...


Quote:

Originally Posted by ILikeMusic /img/forum/go_quote.gif And considering that the one thread I did find was concerning how easy it is to cheat an ABX test one can see see why they might be skeptical... (just kidding, no inferences there...)

Quote:

Originally Posted by krmathis /img/forum/go_quote.gif Perhaps because they don't trust your result? They sure have a reason to be critical... I found a way to reliably CHEAT in foobar2000's ABX utility! posted by the user 'evilking'

Quote:

Originally Posted by nywytboy68 /img/forum/go_quote.gif WOW!!! Total cred bust!!! FAIL!!!

Yeah, well done guys, but I just found that out a couple weeks ago...

foo_abx 1.3.1 report
foobar2000 v0.9.5
2008/07/08 08:40:54

File A: C:\FLAC Music\The Black Mages - The Black Mages\03 - Force Your Way (Final Fantasy VIII).flac
File B: C:\FLAC Music\The Black Mages - The Black Mages\03 - Force Your Way (Final Fantasy VIII).wav

08:40:54 : Test started.
08:41:53 : 01/01 50.0%
08:42:02 : 02/02 25.0%
08:42:12 : 03/03 12.5%
08:42:18 : 04/04 6.3%
08:42:25 : 05/05 3.1%
08:42:31 : 06/06 1.6%
08:42:37 : 07/07 0.8%
08:42:43 : 08/08 0.4%
08:42:47 : 09/09 0.2%
08:42:53 : 10/10 0.1%
08:43:00 : 11/11 0.0%
08:43:05 : 12/12 0.0%
08:43:10 : 13/13 0.0%
08:43:14 : 14/14 0.0%
08:43:24 : 15/15 0.0%
08:43:32 : 16/16 0.0%
08:43:46 : 17/17 0.0%
08:43:54 : 18/18 0.0%
08:44:00 : 19/19 0.0%
08:44:06 : 20/20 0.0%
08:44:15 : 21/21 0.0%
08:44:20 : 22/22 0.0%
08:44:28 : 23/23 0.0%
08:44:35 : 24/24 0.0%
08:44:41 : 25/25 0.0%
08:44:46 : 26/26 0.0%
08:44:52 : 27/27 0.0%
08:44:58 : 28/28 0.0%
08:45:15 : 29/29 0.0%
08:45:24 : 30/30 0.0%
08:46:07 : Test finished.

----------
Total: 30/30 (0.0%)

EK

 

[ILikeMusic]

You can't find your own one-year-old thread on Hydrogenaudio..?

Anyway, no one is saying that there is never ever a single sample that will cause an encoder to fail, just that they are getting to be extremely rare these days and not really a factor. It's not your ability to find one or two examples that is being doubted (assuming that you didn't cheat the results... kidding, just kidding!

), but rather your (and others') claims that you can easily discern a difference. Under the current state of the art well-trained listeners can only rarely uncover audible encoder failures, and even those cases are rapidly diminishing. People seem to love, even need to claim that they can casually hear a difference, maybe because they think it boosts their street cred. In reality it only reduces it.

.

 

[ds_t]

You'd want to be using compressed lossless instead of WAV though, for smaller file size and tagging capabilities.

(Unless you're the type that believes WAV sounds better/different than FLAC/ALAC etc.; that's a whole different can of worms altogether and one that's not really worth discussing seriously.)
QUOTE]

wav flac - no i can't tell the difference!
The hours i spent with mp3 idv1 idv2 tags - it would be a pleasure to have a player that could just read the filename.

 

[evilking]

Quote:

Originally Posted by ILikeMusic /img/forum/go_quote.gif You can't find your own one-year-old thread on Hydrogenaudio..? Anyway, no one is saying that there is never ever a single sample that will cause an encoder to fail, just that they are getting to be extremely rare these days and not really a factor. It's not your ability to find one or two examples that is being doubted (assuming that you didn't cheat the results... kidding, just kidding! ), but rather your (and others') claims that you can easily discern a difference.

I guess I shouldn't throw those terms around so "casually" then. When I said "easily discernable", I meant "eventually heard a difference" and I've already specified the type of track (much presence between 15-19khz, unusual dynamics, intentional noise, etc). So, no, I wasn't "casually" ABX'ing high bitrate lossy files...

I've tried searching HA for the thread but there are only so many combinations of "mp3, 320kbps, ABX, high bitrate" that I can try and going through many thousands of posts is not something I want to do.


Look, I've read some of your previous posts and this topic seems to be your favourite for discussion (recently anyway). I'd like to continue this one but something tells me I wouldn't find it very fulfilling.

EK

 

[ILikeMusic]

Well if you can reliably resolve a difference for a significant number of tracks (easily or not) then I still think you might have something to contribute to the development effort. If you can really prove it that is, and I have a feeling that might be why you didn't get a good response on the other forum. Of course we'll never know since the thread seems to be unrecoverable.

And yes, exposing audiophile BS is one of my more favorite topics.

 

[evilking]

Quote:

Originally Posted by ILikeMusic /img/forum/go_quote.gif A strong +1. I don't know if I have 'Golden Ears' but they're pretty good and just for fun I was trying to ABX LAME 3.97 128 kbps from the source the other night and found it brutally difficult, when I even could. I always get a good belly laugh at some of the claims I read here.

See, there are two sides to this problem. You find it hard to believe that 320kbps ABX'ing is regularly possible, I find it hard to believe that there's music almost transparent at 128kbps MP3...


Maybe low frequency single instruments? Solo piano?



EK

 

[FrederikS|TPU]

ILikeMusic what headphones / sources are you using? I have a hard time believing that you cannot hear the difference between 128 kbps and lossless.

Sometimes I have a hard time telling the difference between FLAC and 320 kbps on my portable, but some records really points out compression flaws. I think it has a lot to do with the mastering of the record as well, some beautifully mastered tracks makes it easier to spot.

 

[ILikeMusic]

To be precise I was talking about LAME 3.97b -V5, which creates a VBR file with an average bitrate of around 128, although being a VBR file it can throw a considerable amount of extra bits at difficult samples so it may be audibly superior to 128 CBR. I can't say for sure since I haven't tested with LAME 128 CBR... don't see much point to it, VBR being almost always superior overall with a similar size file.

Don't assume that any file with the evil characteristic '128' must be terrible. If you think that LAME 3.97b -V5 is easily resolvable from lossless (not impossible, but difficult most of the time) then all I can say is try it in a true blind test. Foobar 2000, etc., make it easy. I think you may be surprised.

That's one problem when people talk about '128 kb MP3'... it can mean 128 CBR with an old Xing encoder (yes, terrible) or LAME -V5 (surprisingly good.) MP3 has been around a long time and is more-or-less open so there are many both good and bad encoders in existence... you can't really talk about MP3 quality unless you specify the bitrate and encoder. When I refer to 'high-bitrate' MP3 I am referring to LAME 3.97b or 3.98 at -V2 or above, and I haven't yet met anyone who could reliably ABX this from the original in person.

P.S. I test with Foobar 2000 and a Sen HD590, Ety ER4P, and/or Shure SE530 (if I think I can hear a difference I might try them all.) Source is a PC with one of the higher-end Creative soundcards (not in front of that PC right now and I forget the model.)

.

 

[evilking]

Quote:

Originally Posted by ILikeMusic /img/forum/go_quote.gif To be precise I was talking about LAME 3.97b -V5, which creates a VBR file with an average bitrate of around 128...

Actually -V5 averages around 145-150kbps for most of my music...


I'll bite and ABX some tracks at -V5, maybe upload some samples.


EDIT:
Killswitch Engage - Breath Life, chorus (1:10), there's an odd wavering quality to both cymbals.
Ladytron - Runaway, 34 seconds in, one of the faint synth sounds that pans left and right simply doesn't in the MP3. Another synth sound is slightly distorted.
Ladytron - Deep Blue, 58 seconds in, voice is simply different, also some faint sounds are missing from the MP3 (eg. just after the word blue there is a high pitched click/beep in the right channel in the flac but not in the MP3).
Prodigy - Fuel My Fire, every voice sounds different throughout the song. Really obvious, this one.

Samples on next page.


EK

 

[ILikeMusic]

Quote:

Samples on next page.

I'll be interested to hear them, thanks for the effort. But... we've kind of digressed in that the thread is about high-bitrate MP3 vs. lossless, not -V5. Try the same same samples again at -V0 and we'll be more on target. Should be interesting now that we have a low-bitrate baseline.

.