Head-Fi.org › Forums › Equipment Forums › Sound Science › Double blind test 128Kbps vs lossless? I'll be amazed if you can tell much difference
New Posts  All Forums:Forum Nav:

Double blind test 128Kbps vs lossless? I'll be amazed if you can tell much difference - Page 6

post #76 of 257
Quote:
Originally Posted by Currawong View Post

First, I have to say that the encoder for the 128k file is surprisingly good. Things have certainly improved very considerably since the warbly sound we're used to with files compressed that much.  

 

I loaded both tracks into iTunes + Amarra and played them through my Stax rig.  The first file that I heard (I wont say which, as I'll PM the OP instead Edit: See edit below as well) was distinctly compressed, with the instruments lacking the complex harmonics that I usually expect with violins, sounding instead flat and artificial. There are a great many classical recordings that don't capture this well, and this is one of them, as when I heard the other track, I could pick these things, but only because I was paying attention.

 

To be sure, I duplicated the files in the playlist a few times and hit the random button without looking.  I could consistently determine which I was listening to every time, soon within the first few seconds of the track starting.  

 

That all being said, I have been a critical music listener for a couple of decades; I have a high-end rig and I'm familiar with musical instruments, so the combination of all those things made it easier.  Again though, it's remarkable just how good the encoding was if that's what a 128k file sounds like now and I highly doubt I would have picked it as easily out of my iPhone or iPod, for example.

 

Edit: I just ran both files through an audio program to confirm that I was indeed correct, the file I picked as being compressed having a hard cut-off at 16 kHz.

Thank you for putting so eloquently into words what I felt! This for me is what I was on about with the 128k file lacking emotion and feeling.

 

I would love for you to listen to the tracks on the ipod. This is how I did it through my various set ups including my portable Stax rig and I did indeed hear the difference.

 

I have been listening again and again since writing last. I said previously that it did not lack the detail, well, especially listening with the stax that was an erroneous statement on my part and how you describe it is perfect. Thank you for having the technical vocabulary I lack!
 

post #77 of 257

yes, I need to do the whole shuffle thing and test myself, I've been looking at what I've been playing and I think that flaws it.

post #78 of 257
Quote:
Originally Posted by stang View Post

Huge difference. You really gotta pick it up right away, or else you will listen to carefully (if that makes sense) and you will trick yourself into hearing the better one. Number 2 definitely has more body to it. Setup: ZeroDAC>MisterX M3>Denon D5000

 

If I am incorrect, I will be shocked. 


Same results. Beyerdynamic DT220 directly into my laptop headphone out. The violins just sounded off in the first one. I am privileged to have a rather talented violinist that lets me hear her practice and point out flaws as one of my best friends though, so that may play a part in it. And we played this song a few too many times in band.

 

Yes, I did it blind. Did both on shuffle 4 times each and guessed correctly each time.


Edited by Ishcabible - 7/8/10 at 9:35am
post #79 of 257

The people claiming huge differences should try something like the Foobar ABX plugin. It makes it easy to compare and totally blind, and you have easily pasted results to show. I don't doubt that there are audible differences (it's a low bitrate after all), but if with the ABX plugin you get more than a few wrong in a 20 trial test, you really have to question if it'll be noticeable in normal listening sessions, or worth the difference in space on portables.

post #80 of 257

It would be quite a revelation if we all eventually learn from the original poster that the two tracks are in fact clones.  I can only seem to perceive differences when I play short chunks of 10-15 seconds back to back.  Listening on AKG 702 straight out of my Gigabyte onboard realtek hd I think I find the second track has more bloom.  You can better appreciate the strings playing behind the first desk violins.  1st track gives Faulty(?) impression of more detail perhaps.

post #81 of 257

So, can we get the results?

 

I would consider using MP3 as an advanced EQ if it turns out that sample 2 is the MP3

post #82 of 257

I thought "movement 2" is the lossless one but probably I'm wrong.

post #83 of 257
Quote:
Originally Posted by sigtransduction View Post

It would be quite a revelation if we all eventually learn from the original poster that the two tracks are in fact clones.  I can only seem to perceive differences when I play short chunks of 10-15 seconds back to back.  Listening on AKG 702 straight out of my Gigabyte onboard realtek hd I think I find the second track has more bloom.  You can better appreciate the strings playing behind the first desk violins.  1st track gives Faulty(?) impression of more detail perhaps.


I looked at the tracks in wave form and they do seem to have differences.  

post #84 of 257

Success!!!  I can DBT both tracks day in day out without failure.  I am also willing to wager $10,000 dollars on anyone willing to take me up on the challenge.

 

I still cannot tell you with 100 percent certainty which is the MP3 - but my challenge is laid down, I can differentiate both tracks blind.

 

I used my D7000's again, this time with my usual EQ of 60hz down 2db and 170hz down 1db.  I didn't use it before because I felt that eq would mask differences - in fact it exposed them.

 

The difference I heard, real and DBT'd, was in the warmth - warm overtones.  Without the EQ - these overtones were always boosted and not differentiatable.  Once the EQ was in place most of the warmth was EQ'd away on track 1.  Track 2 retains a significant proportion of that warmth.

 

Track 1 was leaner

Track 2 retained warmer overtones leading to a fuller character

 

I was looking for differences in high freq energy, once the EQ was in place it was very obvious that the differences - audible to me was in the lower mids.

post #85 of 257


yup. I've got pretty much the same results through ABX-ing it. But the moment I tried to listen a couple of times, I'd screw it all up . I'd say in track 1 the violin appeared thinner though. I'll ABX again later for the heck of it just to confirm though.

Quote:
Originally Posted by SP Wild View Post

Track 1 was leaner

Track 2 retained warmer overtones leading to a fuller character

post #86 of 257

Hello,

ABX 8/8 without replaygain (must not be used for lossy encoding ABX because the encoding mostly deals with masked sounds) on Sennheiser HD600 headphones. The mp3 artifacts are quite audible on the first note at high volume.

post #87 of 257
Quote:
Originally Posted by Pio2001 View Post

Hello,

ABX 8/8 without replaygain (must not be used for lossy encoding ABX because the encoding mostly deals with masked sounds) on Sennheiser HD600 headphones. The mp3 artifacts are quite audible on the first note at high volume.


Could you better explain why Replaygain isn't supposed to be used?

post #88 of 257

Unless you specify the "scale" option in the encoder parameters, or change the volume during the decoding, the volume of an mp3 is exactly the same as the original.

 

Replaygain, however, can show differences. But what are the differences between an mp3 and the original ? The highest frequencies has been removed, and signal has been replaced with noise where and when it is not supposed to be audible.

 

Replaygain might correct the overall loss of volume caused by the removal of high frequencies by boosting what's left. As a result, everything that is left will sound slightly louder than the original.

 

The presence of noise might also alter the replaygain estimation, which is eaxctly the goal. However, since the difference is mostly concentrated where and when it is not audible because of masking effects, replaygain will, in this case, apply an unsuited correction. It does not make the difference between apparent frequencies and masked frequencies.

post #89 of 257

Replaygain only changes the file's overall volume. It's just turning a volume knob. It doesn't increase any one frequency more than another. As long as the amount Replaygain changes the volume by is the same or close, all of the flaws will still be there and just as audible.

 

I tried with one of my files, R.E.M.'s Radio Free Europe. The FLAC is Replaygained to -7.31dB. The V0 MP3 is -7.32. The V9 MP3 (69kbps) is -7.48. I doubt you can hear a volume difference of 0.17dB, but if you could DBT it on the same file then I'd believe it. But I don't think you can convince me that you'd hear a difference of 0.01dB.

 

Besides, wouldn't you have to scan them first for Replaygain to have any effect at all?


Edited by Head Injury - 7/9/10 at 5:08pm
post #90 of 257

I don't think I can ABX that. My best score is ABXing a 0.25 dB volume difference.

 

But why introduce a difference ? When high frequencies are removed, I find it more natural to compare the medium frequencies at the same volume.

 

Without replaygain, you'll have medium-bass at, say -15 dB on both files, and treble at -25 dB on file A, and -infinity dB in file B.

With replaygain, you end up with medium-bass at -15.2 dB in file A, treble at -25.2 dB (whole volume decreased by -0.2 dB), while file B has medium-bass at -15 dB (unchanged).

 

EDIT : yes, files must be scanned and tagged for replaygain to work. Therefore you need to convert them to flac first.


Edited by Pio2001 - 7/9/10 at 5:39pm
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › Double blind test 128Kbps vs lossless? I'll be amazed if you can tell much difference