• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

ABX complex non-linear distortion test, about 5% on a music sample

My result 9/10


foo_abx 2.0.6d report
foobar2000 v1.6.2
2020-12-14 08:07:41

File A: Quintet1.flac
SHA1: 78941da5875a6dc51169e29bbbb909399b75c377
File B: Quintet2.flac
SHA1: 2cea48c9ec2b493807e65e1d2b49deb19b543939

Output:
ASIO : MOTU Pro Audio
Crossfading: NO

08:07:41 : Test started.
08:09:10 : 01/01
08:09:25 : 02/02
08:09:40 : 03/03
08:10:03 : 04/04
08:10:42 : 05/05
08:10:58 : 06/06
08:11:34 : 06/07
08:12:26 : 07/08
08:14:47 : 08/09
08:15:05 : 09/10
08:15:05 : Test finished.

----------
Total: 9/10
p-value: 0.0107 (1.07%)

-- signature --
7fec2ca562c35ed808875f085546c109154b2e65
 
  • Like
Reactions: pma
  • Like
Reactions: pma
First attempt below.
Thank you for setting up this test!

foo_abx 2.0.6d report
foobar2000 v1.4.8
2020-12-14 18:36:03

File A: Quintet1.flac
SHA1: 78941da5875a6dc51169e29bbbb909399b75c377
File B: Quintet2.flac
SHA1: 2cea48c9ec2b493807e65e1d2b49deb19b543939

Output:
DS : Primärer Soundtreiber
Crossfading: NO

18:36:03 : Test started.
18:36:57 : 01/01
18:37:24 : 02/02
18:37:47 : 03/03
18:38:10 : 04/04
18:38:37 : 05/05
18:38:54 : 06/06
18:39:16 : 07/07
18:39:35 : 08/08
18:40:31 : 09/09
18:40:56 : 10/10
18:41:15 : 11/11
18:41:39 : 12/12
18:43:17 : 13/13
18:43:35 : 14/14
18:44:16 : 15/15
18:44:42 : 16/16
18:44:42 : Test finished.

----------
Total: 16/16
p-value: 0 (0%)

-- signature --
a34ad58cb8b4f55f1bc673b9a44e6bdb544782da
 
I couldn't tell the two FLAC's apart, let alone make an AB selection, so you can count me in as a "fail."
I also don't think THD is as audible as some people think it is.
 
Just realized it's only doing 2 trials. Editing a new test that will have more,updated link

I've tried a couple of times. Headphone directly into my Mac but are unable to find any differences. Anyone succeeded with your link yet? :)
 
I've tried a couple of times. Headphone directly into my Mac but are unable to find any differences. Anyone succeeded with your link yet? :)
Screen Shot 2020-12-14 at 3.52.17 PM.png
 
I am 20 years old and cannot hear any difference between the 2 files through my JDS Element, except the difference file which sounds noticeably worse. Which one is the original?
 
I am 20 years old and cannot hear any difference
I guess we young people are not suited for critical listening. Lmao
I'll try a lowpass filter at 15 kHz (6dB/oct? more?) to simulate my ears in 20-30 years. Lmao²
 
Well, that was very hard. Half of the time I felt like I was taking the best guess. Had some background noise so had to turn up the volume on my open headphones. Didn't hear any artefact stand out or tell-tale distortion, it was more of a 'feeling' of fullness or pressure being there compared to the other file that did not have this filler.

foo_abx 2.0.6d report
foobar2000 v1.6.4
2021-02-10 17:00:24

File A: Quintet1.flac
SHA1: 78941da5875a6dc51169e29bbbb909399b75c377
File B: Quintet2.flac
SHA1: 2cea48c9ec2b493807e65e1d2b49deb19b543939

Output:
ASIO : USB DAC ASIO
Crossfading: NO

17:00:24 : Test started.
17:03:49 : 00/01
17:04:22 : 01/02
17:05:49 : 02/03
17:06:21 : 03/04
17:08:25 : 04/05
17:09:20 : 05/06
17:10:31 : 05/07
17:11:00 : 06/08
17:11:32 : 07/09
17:12:25 : 08/10
17:12:25 : Test finished.

----------
Total: 8/10
p-value: 0.0547 (5.47%)

-- signature --
c1eaeffc3a19c6ead3a5d06ddbd5238ce65e4d05
 
Last edited:
It just dawned on me - what volume are y'all listening at? If the volume is normal, the 5% THD products might actually be below the threshold of audibility at their respective frequencies. In other words, the 5% THD might only be audible when the test track is turned up. That might account for some of the differences in performance with this test.
 
Normally I'm on: 13, to pass this test: 20.
But the samples sounded soft. Possibly due to being classical music and no replaygain. For normal listening I would've upped to 15 or 16.
Not sure this tells you much without reference/scale. Max is 60, but volume control is perhaps not linear.
 
Normally I'm on: 13, to pass this test: 20.
But the samples sounded soft. Possibly due to being classical music and no replaygain. For normal listening I would've upped to 15 or 16.
Not sure this tells you much without reference/scale. Max is 60, but volume control is perhaps not linear.

Audibility would be based on dB at a given freq at your ears. Hard to know what your digital volume settings correspond to.
 
The idea would be to set to a normal or slightly louder than normal level you'd normally listen to this. For most people you likely have a sound level of 70-80 db with some peaks higher.
 
Well, that was very hard. Half of the time I felt like I was taking the best guess. Had some background noise so had to turn up the volume on my open headphones. Didn't hear any artefact stand out or tell-tale distortion, it was more of a 'feeling' of fullness or pressure being there compared to the other file that did not have this filler.

Thanks for participating! Your guess on an original file was correct. I think we may disclose it, regarding the long time this test has been available.

what volume are y'all listening at? I

Not very low and not very high, you need to prevent your ear self-distortion.
 
The idea would be to set to a normal or slightly louder than normal level you'd normally listen to this. For most people you likely have a sound level of 70-80 db with some peaks higher.

Sure and it would be nice to control for that by establishing the listening level as part of the protocol. Otherwise you will get a lot of false null hypothesis (no diff) test results if people listen at a level where humans are known not to be able to hear anything at that frequency, whether it's a distortion product or not.
 
foo_abx 2.0.6d report
foobar2000 v1.6.3
2021-04-02 01:14:13

File A: Quintet1.flac
SHA1: 78941da5875a6dc51169e29bbbb909399b75c377
Gain adjustment: +5.14 dB
File B: Quintet2.flac
SHA1: 2cea48c9ec2b493807e65e1d2b49deb19b543939
Gain adjustment: +4.25 dB

Used DSPs:
Skip Track, Resampler (dBpoweramp/SSRC)

Output:
WASAPI (shared) : Hovedtelefoner (USB-C to 3.5mm Headphone Jack Adapter)
Crossfading: NO

01:14:13 : Test started.
01:16:35 : 01/01
01:18:02 : 01/02
01:18:19 : 01/03
01:20:19 : 01/04
01:20:36 : 02/05
01:21:04 : 02/06
01:21:43 : 02/07
01:22:14 : 03/08
01:22:33 : 04/09
01:22:51 : 05/10
01:23:22 : 05/11
01:23:36 : 05/12
01:23:47 : 06/13
01:24:00 : 07/14
01:24:10 : 07/15
01:24:23 : 08/16
01:24:23 : Test finished.

----------
Total: 8/16
p-value: 0.5982 (59.82%)

-- signature --
9e475d8baa5dee06e8bd4fbaaaa4d4aeccfc71fb
 
  • Like
Reactions: pma
Too lazy to attach the pc to the sound system so I just listened to the files. Thought I could tell them apart, then changed my mind on which was which. Clearly guessing. Should be able to save some money on my next purchase. Thanks for posting the files.
 
So nobody yet, quite as expected. If there is hardly anybody who can hear such high distortion on music, so why there is a challenge of the -120dB THD??

Because a Etymotic engineer on Head fi said the upper limit on when THD was a issue was around 2% with music. So that could be why many still find the ER4XR with 1.65% still clean but might sound wrong on pure sines waves?.
 
Back
Top Bottom