Audibility of speaker nonlinear distortion - test

Status
Not open for further replies.
Jump to Latest
#21
The hardest to me was the 1khz, and the easiest the 10+12khz.
Code: 
foo_abx 2.0.4 report
foobar2000 v1.3.16
2017-11-13 23:37:45

File A: 1000Hz_gen.wav
SHA1: b303226639a6f0633023c06bccc5dcc3711f2742
File B: 1000Hz_rec.wav
SHA1: 774114f7416ddbdc290faece1c8b02d9ec2f676c

Output:
WASAPI (event) : Headphone (VIA HD Audio), 24-bit
Crossfading: NO

23:37:45 : Test started.
23:38:22 : 01/01
23:38:32 : 02/02
23:38:40 : 03/03
23:38:49 : 04/04
23:39:01 : 05/05
23:39:13 : 06/06
23:39:39 : 07/07
23:39:48 : 08/08
23:39:56 : 09/09
23:40:07 : 10/10
23:40:26 : 11/11
23:40:35 : 12/12
23:40:35 : Test finished.

 ---------- 
Total: 12/12
Probability that you were guessing: 0.0%

 -- signature -- 
59ef2a10457073939030f79ab9000dcdd5395f7a
Code: 
foo_abx 2.0.4 report
foobar2000 v1.3.16
2017-11-13 23:44:47

File A: 2kHz gen.wav
SHA1: 83cdde7dd2d9e166fa7f92a0a8b69526ce83357f
File B: 2kHz rec.wav
SHA1: 63db87c63b3d51d113d6de5279dfd179b1858643

Output:
WASAPI (event) : Headphone (VIA HD Audio), 24-bit
Crossfading: NO

23:44:47 : Test started.
23:45:19 : 01/01
23:45:34 : 02/02
23:45:39 : 03/03
23:45:49 : 04/04
23:45:53 : 05/05
23:45:58 : 06/06
23:46:02 : 07/07
23:46:06 : 08/08
23:46:10 : 09/09
23:46:13 : 10/10
23:46:17 : 11/11
23:46:21 : 12/12
23:46:21 : Test finished.

 ---------- 
Total: 12/12
Probability that you were guessing: 0.0%

 -- signature -- 
782be79c93083a040623da0114db20dffca4d6d8
Code: 
foo_abx 2.0.4 report
foobar2000 v1.3.16
2017-11-13 23:42:19

File A: 10+12khz_gen.wav
SHA1: 821894330fe457aa2c82286c7f323c34d0dab4ac
File B: 10+12khz_rec.wav
SHA1: 271b9efc65be91ca113cfc1248032534c9f88463

Output:
WASAPI (event) : Headphone (VIA HD Audio), 24-bit
Crossfading: NO

23:42:19 : Test started.
23:42:53 : 01/01
23:42:59 : 02/02
23:43:03 : 03/03
23:43:07 : 04/04
23:43:12 : 05/05
23:43:16 : 06/06
23:43:25 : 07/07
23:43:29 : 08/08
23:43:33 : 09/09
23:43:37 : 10/10
23:43:41 : 11/11
23:43:45 : 12/12
23:43:45 : Test finished.

 ---------- 
Total: 12/12
Probability that you were guessing: 0.0%

 -- signature -- 
d0d5ff625726dc6427338c2a5e4fc32c0e676f3e
 
Last edited:
#22
Well, this should settle the matter. Scott and Geddes and any others who dont seem to believe one can hear much of anything in the way of distortion. I would guess they have not spent much time critically listening.

I will not be back to my home music system for another month. Would love to take test now. Wondering also how an excellent headphone use would change any results or make it even more sensitive.


THx -- Richard Marsh
 
Last edited:
#24
I am curious to hear your thoughts on this...
Ive been developing a ribbon for the last 6 months. Out of 3 diaphragm designs 1 has stood out and while everything looks like a go I am torn on its release. The "problem" is making the desision between the "high" distortion version and the "low" distortion version. The reasion there is a dificult decision is two fold. 1- the higher distortion version has a nice 90 db sensativity vs an 86 db sensativity on the lower distortion version. 2- Neather I nor anyone else so far can tell the difference between the two listening to music.
some details...
This ribbon is designd to be used as low as 1khz. In the lower distortion version, 2nd through 5th HD are all down around .3%-.5% @90 db 1 meter. In the higher distortion version 2nd,4th,and 5th are all below .5% but 3rd starts climbing at around 3khz reaching 4% by 1khz.
I was quite suprised when I started listening tests. I thought for sure I would easily hear a difference. I simply do not.
While my "test" is simple and im sure we could, in theory anyway, tear it down from a technical standpoint, it seems an interesting example of two basically identical transducers measuring quite different distortion levels yet sounding identical.
 
Last edited:
#26
lowmass said:
The reason there is a difficult decision is two fold. 1- the higher distortion version has a nice 90 db sensitivity vs an 86 db sensitivity on the lower distortion version. 2- Neither I nor anyone else so far can tell the difference between the two listening to music.
Click to expand...

Maybe ask yourself who, or what market, you are planning to sell them to? Will it be people who care about specifications? Will it be mastering engineers with very keen ears? (if so, you might want to see if some of the better ones can hear any difference.) Or, will you be selling it to a market that cares more about sensitivity and maybe maximum achievable SPL? If you are sure it won't make any difference no matter who the target market consists of, then why is the decision so hard?
 
#28
Mooly said:
I found both the 1 and 2 kHz files are 100% distinguishable using just a volume level of '2'
Click to expand...

Thanks Karl, I tried the 2kHz test again with DacMagic Plus and even lower volume level than before and got 10/12 and 11/12.

Code: 
foo_abx 2.0.2 report
foobar2000 v1.3.7
2017-11-14 11:32:48

File A: 2kHz gen.wav
SHA1: 83cdde7dd2d9e166fa7f92a0a8b69526ce83357f
File B: 2kHz rec.wav
SHA1: 63db87c63b3d51d113d6de5279dfd179b1858643

Output:
WASAPI (event) : Speakers (Cambridge Audio USB Audio 1.0), 24-bit
Crossfading: NO

11:32:48 : Test started.
11:33:52 : 01/01
11:34:15 : 02/02
11:34:35 : 03/03
11:34:50 : 04/04
11:35:08 : 05/05
11:35:15 : 06/06
11:35:39 : 06/07
11:35:47 : 07/08
11:35:54 : 08/09
11:36:07 : 09/10
11:36:23 : 10/11
11:36:32 : 11/12
11:36:32 : Test finished.

 ---------- 
Total: 11/12
Probability that you were guessing: 0.3%

 -- signature -- 
78d0069405a3c06480f9f9295b14c19c351713ed

So not perfect yet, but much better than before in this test 🙂.
 
#29
mimi123 said:
When you know the distortion profile for some speaker, have you thought about creating digitally the test tones instead of recording and processing them?
Or do you think we lost something this manner?
Click to expand...

Yes. In fact in the past years I have posted several test threads with mathematically added distortion here at diyaudio, but the files had some mistakes that were detectable by abx switching and those mistakes were not originated in distortion. I have learned quite a lot from these mistakes and maybe I might be able to prepare a test with mathematically added distortion, similar as the one measured, into a music sample.
 
#32
Markw4 said:
Maybe ask yourself who, or what market, you are planning to sell them to? Will it be people who care about specifications? Will it be mastering engineers with very keen ears? (if so, you might want to see if some of the better ones can hear any difference.) Or, will you be selling it to a market that cares more about sensitivity and maybe maximum achievable SPL? If you are sure it won't make any difference no matter who the target market consists of, then why is the decision so hard?
Click to expand...

Well if it were cheap to make a ribbon that outperforms the present offerings in some meaningful ways then it would be a much easyer decision. The higher sensativity version would goto the wider market. However due to cost the units will cater to a crowed where many are convinced that such distortion numbers indicate poor performance. This is enough to make many question it.
We have test half to death with a number of people and are confident that while there maybe someone on the planet that can hear a difference we have reason to be confident that there seems to be more to the distortion argument than we know.

One of the most interesting things that has developed through all this is how easily we were able to hear the differences between diaphragm designs. There originally were 5 different designs. pure foil types, sandwich constructions, laminated foil/film types and some unique constructions I will have to stay quiet on. All were constucted in a variety of masses, with a variety of magnet configuations and face plate geometrys as well as all sorts of transformer wind arraingments and core types. While the distortion profiles of many were lower than the version that is winning this war, their sound quality was simply not as convincing or often had some specific coloration that in extended listening became abvious.
 
Last edited:
#34
RNMarsh said:
Well, this should settle the matter. Scott and Geddes and any others who dont seem to believe one can hear much of anything in the way of distortion. I would guess they have not spent much time critically listening.


THx -- Richard Marsh
Click to expand...
i guess you've not encountered the "no usable metric" mantra.
it is a troubling problem from a measurement/quantitative perspective, but to simply say it's irrelevant and a non issue does not make the phenomenon/problem go away.
this is a debate that has been going on for a long time and what dismays me the most is when polarization occurs rather then engage in open discussion in order to work towards enlightenment.
 
#35
RNMarsh said:
Well, this should settle the matter. Scott and Geddes and any others who dont seem to believe one can hear much of anything in the way of distortion.
Click to expand...
To be fair, I think that Earl is on recent record here saying that speaker low level distortion is audible on pure tones, but not in music playback.

That would be the next step in the test, music playback of the same device where the sine distortion is easily heard.

Remember some years ago when Art did a huge series of horn driver tests?
 
#37
Pano said:
To be fair, I think that Earl is on recent record here saying that speaker low level distortion is audible on pure tones, but not in music playback.

That would be the next step in the test, music playback of the same device where the sine distortion is easily heard.

Remember some years ago when Art did a huge series of horn driver tests?
Click to expand...
High Frequency Compression Driver Evaluation

Lots of good info in there.
 
#38
1K test

First try and I found it easy. Dell laptop soundcard, 6V6 SEP, Vistaton fullrange drivers.
Code: 
foo_abx 2.0.4 report
foobar2000 v1.3.12
2017-11-14 10:32:00

File A: 1000Hz_gen.wav
SHA1: b303226639a6f0633023c06bccc5dcc3711f2742
File B: 1000Hz_rec.wav
SHA1: 774114f7416ddbdc290faece1c8b02d9ec2f676c

Output:
DS : Speakers / Headphones (Realtek High Definition Audio)
Crossfading: NO

10:32:00 : Test started.
10:32:14 : 01/01
10:32:32 : 02/02
10:32:39 : 03/03
10:32:50 : 04/04
10:32:59 : 05/05
10:33:11 : 06/06
10:33:23 : 07/07
10:33:34 : 08/08
10:33:44 : 09/09
10:33:55 : 10/10
10:34:06 : 11/11
10:34:13 : 12/12
10:34:21 : 13/13
10:34:32 : 14/14
10:34:32 : Test finished.

 ---------- 
Total: 14/14
Probability that you were guessing: 0.0%

 -- signature -- 
a892693f55c57bbbc78156de6cb572f9eb0085e8
But with headphones and my M-Audio soundcard as headphone amp, much harder, not sure I could pass. Strange.
 
#39
10K + 12K

Again, easy. The recorded file had an extra tone.
Code: 
foo_abx 2.0.4 report
foobar2000 v1.3.12
2017-11-14 10:42:37

File A: 10+12khz_gen.wav
SHA1: 821894330fe457aa2c82286c7f323c34d0dab4ac
File B: 10+12khz_rec.wav
SHA1: 271b9efc65be91ca113cfc1248032534c9f88463

Output:
DS : Speakers / Headphones (Realtek High Definition Audio)
Crossfading: NO

10:42:37 : Test started.
10:42:44 : 01/01
10:42:53 : 01/02
10:43:03 : 02/03
10:43:12 : 03/04
10:43:24 : 04/05
10:43:33 : 05/06
10:43:40 : 06/07
10:43:51 : 07/08
10:44:02 : 08/09
10:44:12 : 09/10
10:44:19 : 10/11
10:44:29 : 11/12
10:44:39 : 12/13
10:44:49 : 13/14
10:44:49 : Test finished.

 ---------- 
Total: 13/14
Probability that you were guessing: 0.1%

 -- signature -- 
1fd217ec079cac40bc5494cdb00f9614754ade5d
 
Status
Not open for further replies.
Top Bottom