For those who can hear a difference between Wav/AIFF/FLAC

[Yomanze]

Three pages later and none have at least tried to test it and kill the argument stone dead. I spent a good amount of time a year or so ago using Foobar2000's ABX plugin, using FLAC, AIFF, LAME-encoded 320kbps etc. I couldn't reliably tell the difference between FLAC and WAV/AIFF blind, but I did between lossless and 320kbps, admittedly I was using STAX electrostatic 'phones, and at the time not convinced I could have noticed through the speakers.

[Yomanze]

As it stands I sold my V-Link 192 and stuck with CD playback, if someone could lend me a USB to SPDIF converter then would happily repeat the test.

[Maximum]

I converted a file using lame to 320 mp3. Conducted 35 tests in total, some short, some switching a lot. At the start it was difficult, I had to try and locate a reference point where one sounded slightly rougher than the other and switch between them. All in all it was a pretty tedious session, and tiring on the ears listening to the same 5 seconds again and again (red lorry yellow lorry comes to mind). It was going well (down to 7.6%) then I lost my mojo. Results are below.

19:02:27 : Test started.
19:07:17 : 01/01 50.0%
19:07:43 : 01/02 75.0%
19:10:44 : 01/03 87.5%
19:15:35 : 01/04 93.8%
19:16:04 : 01/05 96.9%
19:20:01 : 01/06 98.4%
19:23:53 : 02/07 93.8%
19:27:36 : 02/08 96.5%
19:29:43 : 03/09 91.0%
19:30:38 : 04/10 82.8%
19:31:24 : 05/11 72.6%
19:31:41 : 06/12 61.3%
19:33:11 : 06/13 70.9%
19:34:17 : 07/14 60.5%
19:36:01 : 08/15 50.0%
19:38:06 : 08/16 59.8%
19:38:35 : 09/17 50.0%
19:40:01 : 10/18 40.7%
19:42:50 : 11/19 32.4%
19:45:08 : 12/20 25.2%
19:46:12 : 13/21 19.2%
19:46:51 : 14/22 14.3%
19:48:09 : 15/23 10.5%
19:58:28 : 16/24 7.6%
20:00:17 : 16/25 11.5%
20:05:13 : 16/26 16.3%
20:06:54 : 16/27 22.1%
20:09:51 : 17/28 17.2%
20:12:28 : 17/29 22.9%
20:13:54 : 18/30 18.1%
20:14:59 : 19/31 14.1%
20:21:37 : 19/32 18.9%
20:23:37 : 19/33 24.3%
20:25:56 : 19/34 30.4%
20:26:43 : 19/35 36.8%
20:26:46 : Test finished.

----------
Total: 19/35 (36.8%)

[Rothchild]

Wow Rob, that's pretty impressive thanks for sharing.

What was your reference track?

I've got find a decent reference to do this, it's been fun playing with Betty Wright but as that track sounds like it was recorded on (soulful) sandpaper at the best of times I don't think it's going to give me a chance of hearing a real difference!

[Maximum]

The track was Trafik - Indestructible. I'm sure there are more suitable better mastered tracks to test with though.

[realysm42]

Thanks Rob.

Do you think you started to score worse because you got tired/lost interest?

[Maximum]

Thanks Rob.

Do you think you started to score worse because you got tired/lost interest?

Possibly yes. At times I even thought A matched up with B and couldn't shake that out of my head which needed more time to jar that loose. It's certainly no way to listen to music.

[Rothchild]

There's an interesting cycle in the numbers though isn't there? It gets significantly worse before it gets better (and then gets worse again). Half way through the test it's at 50/50 (having dragged itself back up from nearly total guess work (at test 8). Indeed it's interesting as much for its symmetry as anything: Test 8 96% chance of guessing, Test 15 50/50, Test 24 7% chance of guessing, Test 32 it's cycling back up towards evens.

If anything it looks more like practice paid off rather than boredom set in, as the run from test 24 to 32 is where you manage to hold the most consistent score.

I'm not a statistician but I am aware that this sort of attempt to interpret is dangerous so I will restrain myself until I've done some more research about how we're supposed to interpret these figures, but I think the point I'm trying to make is that whilst the figure did go down at one point to 7% chance of guess work it did also go as high as 96% chance of guesswork. Nevertheless on the assumption that the test was played fairly (insofar as 35 was decided at the outset and the ongoing results were hidden) I think 36% chance of guesswork represents a significant one in the eye for the doubters.

[AlfaGTV]

This is really interesting and did give it a shot on my Mac... Anyhoots, i compared an AIFF RedBook file to the very same file compressed to 320kbps AAC (.m4a)
And, to my surprise, i didnt score enough correct decisions for it to be a relevant result! I am, however, not good at comparisons om the fly.. I know, after a few days if i LIKE what i hear.
I will give it another shot though, upgraded to Mavericks before and that crippled my M2Tech USB2SPDIF adapter so this was done using only the ABX tester, via optical output to my DAC.
(In the long run i prefer Amarra via the M2Tech adapter, AND uncompressed files. Or to be absolutely correct i prefer uncompressed files via UPnP, and no fecking computer at all)

Regars Mike

[Maximum]

Nevertheless on the assumption that the test was played fairly (insofar as 35 was decided at the outset and the ongoing results were hidden) I think 36% chance of guesswork represents a significant one in the eye for the doubters.

To be honest I never intended the test to go that far. I didn't hide the results but I'm not convinced this should make a difference, either the difference is audible and can be judged, or is not and is ruled by random chance alone, and there is no way you can train for a null difference. Whether it would have got worse I'm not sure, I ended it when I got fed up.

The percentages seem a little misleading anyway, 50% should be the statistically perfect representation of random guessing surely. It makes sense in the context it is used I suppose, but may make a lot of people think 50% is significant.