BLIND TEST: 128kbps mp3 vs Lossless
May 21, 2013 at 2:35 PM Post #166 of 180

bonfirehay

100+ Head-Fier
Joined
Jun 21, 2012
Posts
196
Likes
14
Quote:
It's really less of a question how great or expensive your equipment it. Instead, it's ime far more important being trained / know what to listen for.

To quote Sean Olive: "if you do half an hour of training a day, I think in a month or two you probably gonna get up to the minimum level harman requires before listeners are allowed into product tests".
I know, that sounds quite a bit over the top, but take a look at their research results:

 
How to train / get training?
 
EDIT: Never mind, found the software: http://harmanhowtolisten.blogspot.pt/2011/01/welcome-to-how-to-listen.html
 
Jul 6, 2013 at 5:46 PM Post #169 of 180

davidsh

Headphoneus Supremus
Joined
Oct 4, 2012
Posts
5,593
Likes
180
B sounded better on my Stax system, no doubt here. Didn't like the song, though.
 
Sep 7, 2013 at 7:34 PM Post #170 of 180

miceblue

Headphoneus Supremus
Joined
Feb 23, 2011
Posts
16,390
Likes
2,844
Yay, I like 128 kbps vs lossless tests. It's really interesting to see the results of these tests sometimes.
beerchug.gif

 
To me, track B sounded more spacious, and A a bit muffled and artifacty(?) (I don't know if that's the right term; aliasing?). Lowering the bitrate tends to affect the highest frequencies the most from what I've tested, and track B definitely has more detail in the cymbal crashes.
 
 
MacBook -> Audirvana Plus (hog mod, integer mode) -> ODAC -> SRS-2170
 
What song is this if I may ask? It sounds really familiar...
 
I'll probably do a Foobar ABX text later today. I need to finish some stuff in OS X first though.
 
 
Yup, it looks like B is the lossless file after analysing the two tracks in Audacity with the spectrogram viewer
 
Sep 8, 2013 at 8:52 AM Post #171 of 180

hogger129

100+ Head-Fier
Joined
Mar 4, 2013
Posts
309
Likes
32
I did this test awhile back.  I made a playlist with about 10 files as FLAC and 10 files as 128k MP3.  I could tell the difference every time when played back on my home theater system.
 
I also much prefer AAC for lossy encoding over MP3.
 
When I get to 256k AAC, the difference between lossy and lossless is WAAAY harder to tell.  I think for on the go listening, I'd use 192k AAC if you have a limited amount of space to work with.
 
I just make up a playlist of FLAC files in foobar2000 and put it on my SD card in my Rockboxed Sansa.
 
Sep 12, 2013 at 4:07 AM Post #173 of 180

miceblue

Headphoneus Supremus
Joined
Feb 23, 2011
Posts
16,390
Likes
2,844
All righty. I finally booted into Windows and did the [in]famous Foobar2000 ABX test.
foo_abx 1.3.4 report
foobar2000 v1.2.9
2013/09/12 00:46:41
File A: C:\Users\Michael\Desktop\A.wav
File B: C:\Users\Michael\Desktop\B.wav
00:46:41 : Test started.
00:56:57 : 01/01  50.0%
00:57:08 : 02/02  25.0%
00:57:22 : 03/03  12.5%
00:57:28 : 04/04  6.3%
00:57:38 : 05/05  3.1%
00:57:44 : 06/06  1.6%
00:58:18 : 07/07  0.8%
00:58:38 : 08/08  0.4%
00:58:45 : 09/09  0.2%
00:58:55 : 10/10  0.1%
00:59:03 : 11/11  0.0%
00:59:27 : 12/12  0.0%
00:59:57 : 13/13  0.0%
01:00:08 : 14/14  0.0%
01:00:18 : 15/15  0.0%
01:00:33 : 16/16  0.0%
01:00:48 : 16/17  0.0%
01:00:49 : Test finished.
 ----------
Total: 16/17 (0.0%)
 
 
I think the test results are quite indicative that I can hear the difference between the two files.
 
Yup, B is definitely the lossless version. File A sounded really thin in comparison to B right next to each other during the ABX test, and it was pretty obvious to me which was which. File A also sounded pretty grainy in the treble.
 
 
Setup:
MacBook Pro -> Windows 8 -> Foobar2000 -> WASAPI (event) -> ODAC (24/96 mode, exclusive mode checked) -> STAX SRS-2170 (SRM-252S amp + SR-207 earspeaker)
 
Sep 12, 2013 at 6:12 AM Post #174 of 180
Joined
Jan 30, 2011
Posts
13,315
Likes
24,307
  All righty. I finally booted into Windows and did the [in]famous Foobar2000 ABX test.
 
I think the test results are quite indicative that I can hear the difference between the two files.
 
Setup:
MacBook Pro -> Windows 8 -> Foobar2000 -> WASAPI (event) -> ODAC (24/96 mode, exclusive mode checked) -> STAX SRS-2170 (SRM-252S amp + SR-207 earspeaker)

 
Did you volume match (apply replay gain) before taking the test?  Most people aren't.  But the two files are different by ~ 0.4 dB.  "B" is the louder one.  Just makes the abx that much easier.
 
Sep 12, 2013 at 11:05 AM Post #175 of 180

UltMusicSnob

100+ Head-Fier
Joined
Aug 14, 2013
Posts
177
Likes
11
The MP3 algorithms are complex, with lots going on (it's astonishing and impressive that MP3 128 can be so accurate with less than 90% of Redbook bit rate).
 
On the basic psychoacoustic theory, the above responses on cymbals and percussive sounds generally, this is a nice demonstration of an inherent weakness of the model. Since psychoacoustic masking depends on identification of particular strong frequencies (in a given frame) and consequent masking of **nearby** frequencies, it makes sense that sound sources which have a larger noise component (as opposed to a pure sine wave, or say a flute or soprano's voice) will translate less well. The pick attack on strummed guitar strings is what I noticed here--less well-defined, blurred into a noise sound instead of a reproduced pick strike on a string.
 
A sound source which produces a relatively uniform spread of frequencies doesn't provide the differentiation that MP3 needs in order to identify the psychoacoustically masked components. I basically never listen to MP3's, period, so I hadn't looked at this before. Seems like it would be a pretty severe limitation, given the importance of the full drum set in popular music.
 
Sep 12, 2013 at 12:31 PM Post #176 of 180

miceblue

Headphoneus Supremus
Joined
Feb 23, 2011
Posts
16,390
Likes
2,844
 
All righty. I finally booted into Windows and did the [in]famous Foobar2000 ABX test.

I think the test results are quite indicative that I can hear the difference between the two files.

Setup:
MacBook Pro -> Windows 8 -> Foobar2000 -> WASAPI (event) -> ODAC (24/96 mode, exclusive mode checked) -> STAX SRS-2170 (SRM-252S amp + SR-207 earspeaker)


Did you volume match (apply replay gain) before taking the test?  Most people aren't.  But the two files are different by ~ 0.4 dB.  "B" is the louder one.  Just makes the abx that much easier.


Hm no I didn't. How does one volume match using Foobar2000?
 
Sep 12, 2013 at 12:43 PM Post #177 of 180

UltMusicSnob

100+ Head-Fier
Joined
Aug 14, 2013
Posts
177
Likes
11
Hm no I didn't. How does one volume match using Foobar2000?

With the equipment I have, I'd probably scan both in SoundForge 10, then process and then Save the file to get as close an RMS match as possible (by reducing one--increasing volume on one is likely to put the file into clipping).
 
Could do that in Audacity (free sw). I think foobar has a function to match, or at least a plugin you can get.
 
Sep 12, 2013 at 1:11 PM Post #178 of 180

stv014

Headphoneus Supremus
Joined
Jul 17, 2011
Posts
3,493
Likes
271
Simple solution with foobar2000:
- select the two WAV files
- right click, then ReplayGain -> Scan per-file track gain
- click "Update File Tags" (this only adds gain information as a tag to the files, but does not actually change the PCM data)
- when starting the ABX plugin, make sure that "Use ReplayGain" is checked
Note that this method will make the sound quieter by about 11 dB (for this particular track), so you need playback hardware that has enough dynamic range to handle that without audible noise.
 
You can remove the ReplayGain tag with "ReplayGain -> Remove ReplayGain information from files".
 

Users who are viewing this thread

Top