Can you hear a difference? 250hz and 1250hz playing at same time, but 1250 Hz is at 60db down. Vs pure 250hz tone.

[Pdxwayne]

I did the ABX on a pair of Marshal Major II EQed to Harman Curve, so i hope no more ridiculous claims of 'cheating' are thrown out at me again

File A: 250hz_h2_to_h4_120_h5_60.wav
SHA1: b604c1acb2c2cdc3fd9bceba6aef1358de8d1a6e
File B: 250hz_tone.wav
SHA1: 9bedd7cc84145a54cc2979ec4bbf86d51f0f082a

Output:
Default : Primary Sound Driver
Crossfading: NO

18:34:37 : Test started.
18:34:45 : 01/01
18:34:56 : 02/02
18:35:07 : 03/03
18:35:17 : 04/04
18:35:38 : 05/05
18:35:47 : 06/06
18:36:04 : 07/07
18:37:11 : 08/08
22:28:37 : 09/09
22:28:49 : 10/10
22:29:03 : 11/11
22:29:13 : 12/12
22:29:25 : 13/13
22:29:36 : 14/14
22:29:46 : 15/15
22:29:55 : 16/16
22:29:55 : Test finished.

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

-- signature --
76c0d5f91c2cbc3e3584fcdd03cd90c42aebdb05

Thank you for checking!

Would you say it is more difficult than your laptop speakers?

: )

 

[abdo123]

Thank you for checking!

Would you say it is more difficult than your laptop speakers?

: )

Yes it was. but not necessarily difficult either. i did it on my first attempt.

 

[solderdude]

Consider that speaker plots are all smoothed (averaged) and that with single tones, that are very narrow in bandwidth, you can have substantially deep dips and even some sharp peaks that won't be seen in those plots but may be substantially audible.
That could be a reason why speakers (in a room) might be easier in certain cases, positions.

As can be seen well over 10dB is quite possible. When one would have done the test with the speaker in the 'blue trace' conditions there might even have been a 20dB 'advantage' at that specific listening position in that particular room.

 

[abdo123]

Consider that speaker plots are all smoothed (averaged) and that with single tones, that are very narrow in bandwidth, you can have substantially deep dips and even some sharp peaks that won't be seen in those plots but may be substantially audible.
That could be a reason why speakers (in a room) might be easier in certain cases, positions.

this is a reach imo. since 250 Hz is in the statistical region for the majority of rooms.

 

[abdo123]

To add some 'loudness' context into my previous blind test. it was conducted at roughly 80 dBSPL. seems to be the sweet spot that we discussed earlier.

 

[Pdxwayne]

Consider that speaker plots are all smoothed (averaged) and that with single tones, that are very narrow in bandwidth, you can have substantially deep dips and even some sharp peaks that won't be seen in those plots but may be substantially audible.
That could be a reason why speakers (in a room) might be easier in certain cases, positions.

As can be seen well over 10dB is quite possible. When one would have done the test with the speaker in the 'blue trace' conditions there might even have been a 20dB 'advantage' at that specific listening position in that particular room.

I agree. I drilled down one REW plot for sweep captured on my left ear (note that my current speakers positions are not the same, but not too far apart from prior positions when I did the measurements) and the overall volume at 250hz is ~10db lower than at 1250hz.

No wonder I could even sense -70 file. In actuality, I was just testing about -60 dB difference when using -70db file.

 

[Pdxwayne]

Yes it was. but not necessarily difficult either. i did it on my first attempt.

Yeah, for both my abx with AK371, they were first attempts too. 8/8 and 8/8. But I wouldn't call it easy....

Thanks!

 

[Blumlein 88]

Consider that speaker plots are all smoothed (averaged) and that with single tones, that are very narrow in bandwidth, you can have substantially deep dips and even some sharp peaks that won't be seen in those plots but may be substantially audible.
That could be a reason why speakers (in a room) might be easier in certain cases, positions.

As can be seen well over 10dB is quite possible. When one would have done the test with the speaker in the 'blue trace' conditions there might even have been a 20dB 'advantage' at that specific listening position in that particular room.

I was thinking along the same lines a few days ago. I therefore altered the speed of the files to move both tones a few hertz or few dozen hertz up or down in frequency thinking that might help. It didn't cause any change, though obviously by your plots it could be the issue in some cases.

 

[GaryH]

Someone detected it on laptop speakers - well f*ck that for a test, lol - peeps need to know what this is about! It's better if this thread doesn't become a pissing contest for people that lack the understanding of the influence of their hardware, albeit I except that not everyone will get the same results even on the same hardware.

You have to ask yourself: "What is the relevance of this test?". Is it to be able to determine the difference at any cost? In other words do you do the test on some silly hardware that is completely not "flat" and therefore sounds like balls in everyday life when you listen to music on it or do you do the test on something that sounds good or that is approaching some kind of standard. The goal is not to run the test on "shit enough" hardware with crazy enough frequency responses in order to hear a difference, but I assume the purpose of the test is to see if psychoacoustically people can recognise the true difference between two tones - and that can only be done accurately if the hardware is producing those two tones at correspondingly accurate SPL levels (ie the frequency response graph). Essentially, it's invalid to test it on hardware that is not flat (or it's taken into consideration based on the known frequency response). I suppose if being lenient you could say that it could be interesting to see if people are more likely to notice the difference on certain types of hardware that fall away from the accepted target standard, but it's very loose & really I don't what conclusions you could draw. As I see it this is supposed to be a test of a person's psychoacoustic ability to seperate two tones or recognise their combination, and if the hardware is not controlled (ie not ideal "flat") then it ceases to be a test of the person but is instead influenced by the hardware. One thing is for sure, the hardware definitely influences the result if it's not controlled - which is the point you seemingly didn't understand.

Well said. Sometimes the lack of basic scientific literacy astounds me. Control for potential confounding variables. Always. Otherwise you might end up convincing yourself (despite not even able to even hear up to 12 kHz well) that you can hear the difference between hi-res and 16/44.1, when what you're most likely hearing is downward IMD from the ultrasonics into the audible band due to doing your listening test using a laptop (DirectSound and not WASAPI/ASIO playback at that) with ultrasonic junk and relatively high IMD with IEMs such as these with themselves relatively high nonlinear distortion (approaching 1% 3rd order THD at 94 dB SPL, so even higher for louder peaks). Be very wary when anyone says something like, 'but I passed the ABX with just a laptop and earphones, so it must be easy, I could pass it with anything!' It doesn't work like that. You need a known, confirmed clean playback chain if you want to test your ability to distinguish the actual difference under test in the files and not their interaction with your hardware.

 

[Maki]

10/10 ABX(Y?), easy. After about the 5th trial I stopped listening to both X and Y samples and was able to hear the answer based on X alone. Test run on STAX 009 with EQ to neutral. This log is the second test because I forgot to save the text of the first, which is why the times are faster. 1250hz is cut by roughly 3dB to bring it in line, I also have a mild warmth/bass boost which results in +2dB at 250hz over the stock FR which should in theory make it a bit harder.

foo_abx 2.0.6d report
foobar2000 v1.6.7
2021-10-26 21:18:56

File A: 250hz_h2_to_h4_120_h5_60.wav
SHA1: b604c1acb2c2cdc3fd9bceba6aef1358de8d1a6e
Gain adjustment: -17.16 dB
File B: 250hz_tone.wav
SHA1: 9bedd7cc84145a54cc2979ec4bbf86d51f0f082a
Gain adjustment: -17.16 dB

Output:
Default : Primary Sound Driver
Crossfading: NO

21:18:56 : Test started.
21:19:01 : 01/01
21:19:04 : 02/02
21:19:07 : 03/03
21:19:09 : 04/04
21:19:12 : 05/05
21:19:14 : 06/06
21:19:17 : 07/07
21:19:23 : 08/08
21:19:26 : 09/09
21:19:32 : 10/10
21:19:32 : Test finished.

----------
Total: 10/10
p-value: 0.001 (0.1%)

-- signature --
c56a2f479f8e5ee46fccc57e15930cea5291f5fe

 

[xaviescacs]

After about the 5th trial I stopped listening to both X and Y samples and was able to hear the answer based on X alon

I have been doing all tests on ASR only listening to X each time after some training before the test.... I must say I found easier this way, listening to A and B each time confuses me in the end. Besides, I feel like listening to A and B each time makes it 10 tests combined, instead of one test with 10 tries, but it's more a feeling than a probability statement.

 

[Maki]

I have been doing all tests on ASR only listening to X each time after some training before the test.... I must say I found easier this way, listening to A and B each time confuses me in the end. Besides, I feel like listening to A and B each time makes it 10 tests combined, instead of one test with 10 tries, but it's more a feeling than a probability statement.

I listen to A and B at the beginning to make sure which is which. I'm not actually sure if foobar randomizes those though. Also foobar glitches out every once in a while and adds really nasty static bursts when switching sometimes. That was unpleasant.

 

[xaviescacs]

I listen to A and B at the beginning to make sure which is which. I'm not actually sure if foobar randomizes those though. Also foobar glitches out every once in a while and adds really nasty static bursts when switching sometimes. That was unpleasant.

ABX in ubuntu does not randomize A and B (I think ) but it's kind of a toy program.

 

[Jimbob54]

Well said. Sometimes the lack of basic scientific literacy astounds me. Control for potential confounding variables. Always. Otherwise you might end up convincing yourself (despite not even able to even hear up to 12 kHz well) that you can hear the difference between hi-res and 16/44.1, when what you're most likely hearing is downward IMD from the ultrasonics into the audible band due to doing your listening test using a laptop (DirectSound and not WASAPI/ASIO playback at that) with ultrasonic junk and relatively high IMD with IEMs such as these with themselves relatively high nonlinear distortion (approaching 1% 3rd order THD at 94 dB SPL, so even higher for louder peaks). Be very wary when anyone says something like, 'but I passed the ABX with just a laptop and earphones, so it must be easy, I could pass it with anything!' It doesn't work like that. You need a known, confirmed clean playback chain if you want to test your ability to distinguish the actual difference under test in the files and not their interaction with your hardware.

Let's not pretend this was a scientific test at all. Zero controls in anything and zero stipulations by the OP. Dont expect the subjects to put the work in. Hence he cant draw the conclusions from the test that he'd like to.

 

[Robbo99999]

Let's not pretend this was a scientific test at all. Zero controls in anything and zero stipulations by the OP. Dont expect the subjects to put the work in. Hence he cant draw the conclusions from the test that he'd like to.

Yeah, but on a point of view of knowledge/rigour it's important to understand the potential impact of the hardware, but most of us are probably aware of that now after the discussions in this thread already. If you're aware of the variables, you can decide to control them as best you can.....and that would go for any tests you do whether in audio or anywhere else.

 

[Pdxwayne]

My understanding of Solderdude"s statements are
*My test is not an example of masking
*It is very easy
*Anyone can hear the difference

How true are those statements, now that we have a variety of results?

 

[Jimbob54]

Yeah, but on a point of view of knowledge/rigour it's important to understand the potential impact of the hardware, but most of us are probably aware of that now after the discussions in this thread already. If you're aware of the variables, you can decide to control them as best you can.....and that would go for any tests you do whether in audio or anywhere else.

Agreed. And the person doing the experiment can't control the hardware in a test like these. You and I are the subjects, tasked only to do an ABX test on those two files. Which we did and reported our results. All Wayne can conclude is some people can't hear it in some circumstances.

All I can conclude from my small run of tests is I can hear it, easier in some circumstances than others.

 

[xaviescacs]

Agreed. And the person doing the experiment can't control the hardware in a test like these. You and I are the subjects, tasked only to do an ABX test on those two files. Which we did and reported our results. All Wayne can conclude is some people can't hear it in some circumstances.

All I can conclude from my small run of tests is I can hear it, easier in some circumstances than others.

Don't you think that there is not a way to formulate properly the null hypothesis of this experiment if we expect to validate that everybody can hear the difference? I mean, I believe it's impossible to reject the null that there is no subject that can't hear it. What can be tested, and has been proved, is that a human being can hear it, and therefore we can reject the null that it's not audible.

 

[solderdude]

My understanding of Solderdude"s statements are
*My test is not an example of masking

Clearly you understood wrong.
This came from another thread which was about masking of harmonics in music and your thesis that low bass harmonics most likely elevated the treble by 2dB.
Your 'test' clearly is (by accident ?) right at the border of audible masking and had nothing to do with the disagreement about distortion, masking etc.

*It is very easy
*Anyone can hear the difference

How true are those statements, now that we have a variety of results?

I meant if you can hear it others can too.
Theory predicted that you can detect -60dB 5th harm. with tones (again not with music).
You are no special snowflake (nor am I) despite your thesis that you could be 1 in 100000 that possibly could do this and later found out that your speakers gave you an unfair advantage.
The fact that your wife and kids could not hear is might be lack of interest, training or being fed up with tests they do not care about and are not obvious.

Yes, the 'anyone can hear it' statement was a bit bold but not as bold as the claims you constantly make.
Clearly, people that don't care, circumstances that are not correct, people without training won't pass your test.

The fact that I and some others can AB this with headphones with clear results and you could not so in your case at the SPL levels you tried masking occurred.

 

[dominikz]

For me it took a bit of time in the first trial to key-in on the distortion, but after that it was reasonably easy to pick out quickly. I used Sennheiser x Drop HD58X Jubilee with RME Babyface (gen 1) soundcard and Head'n'Hifi Objective2 DIY kit headphone amp.

foo_abx 2.0.6d report
foobar2000 v1.6.7
2021-10-27 13:17:27

File A: 250hz_h2_to_h4_120_h5_60.wav
SHA1: b604c1acb2c2cdc3fd9bceba6aef1358de8d1a6e
File B: 250hz_tone.wav
SHA1: 9bedd7cc84145a54cc2979ec4bbf86d51f0f082a

Output:
Default : Primary Sound Driver
Crossfading: NO

13:17:27 : Test started.
13:18:11 : 01/01
13:18:41 : 02/02
13:18:53 : 03/03
13:19:04 : 04/04
13:19:15 : 05/05
13:19:28 : 06/06
13:19:33 : 07/07
13:19:39 : 08/08
13:19:43 : 09/09
13:19:51 : 10/10
13:19:56 : 11/11
13:20:02 : 12/12
13:20:11 : 13/13
13:20:17 : 14/14
13:20:22 : 15/15
13:20:30 : 16/16
13:20:30 : Test finished.

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

 -- signature --
9d46d12b1a8dfeb2f75e83d5824cef7c4e6c1174

I imagine most would be able to hear this, but perhaps not many would have the patience to initially listen for and make out the distortion component. It is a pretty faint distortion, IMHO.