Comparing more than 2 file is a well known way to reduce listener sensitivity to small differences that are audible.
Rule of thumb: Keep it as simple as possible.
Mooly has already explained a couple of times why the master file is being held back until the results are in.
I agree with his reasoning as I'm fed up of having results spoiled by people using analysis programs on the files whilst listening and voting is still open.
The alleged explanation has a few issues of its own, but I guess I'll have to let this one play itself out. Or not.
I'm not trying to be credible, I'm just trying to help others master logic 101.
Just because there is an explanation doesn't mean that it makes sense if you understand little side issues like logic or reason.
If the UUTs are good enough, using analysis programs is futile.
These appear to be very good UUTs. ;-)
Doing preference tests on files that are not first known to sound different is illogical.
Doing listening tests without a fixed reference is illogical.
Why spoil what could be a good test with preconditions based on bad logic and irrelevant facts?
The answer to that question should be obvious. ;-)
BTW the UUTs are fine, but the time synching of the files is far from it. This is readily apparent from the files provided, so much for the false secrecy.
Last edited:
Attachments
If a reference file of CD player to AD were provided, it would instantly be identifiable by those with the knowledge to find minute differences in file stuctures.
We've been in this situation before unfortunately, where many hours work was ruined by someone posting a complete and correct key to twelve totally random files, not by listening but by applying their knowledge as an IT expert.
So that is the reason why. If you were sat down listening to the files in a controlled environment then I wouldn't hesitate to include the original.
This is still easy to cheat, it just takes a different method. That's the problem with "tests" like this when you allow comments and you have known cheaters as participants.
I like "A" the most. "C" (or "D") the least. C/B/D are "high end" but from all 6 I experienced fatigue only with C and D. C had an issue with fatigue PLUS it doesn't sound right musically, so "C" is the least prefered.
All true unfortunately.
Doing a closed test is certainly an option although I personally think that it would probably stop many from joining in. Everyone enjoys all the comments and discussion at the end of the day.
Really?
If so, it is trivial to create a file that contains the same binary data as the reference file but also has the same data structure as the rest of the test files.
That was not done by the means you allege to be applicable here because the correct key applied to all of the test files. All that little trick does is enable picking out the reference file from the rest.
The usual means for identifying test files without listening involves FFT analysis and knowledge of the watermarks (whether intentional or not) applied by the means used to create them.
I've done little exercises like that, and the means I used don't work here.
But, I was able to easily exploit the lack of care that was used to create them, no reference required.
Hopefully, all of these files were created by the same program, have the same format, and therefore can only be distinguished by their content. The same can be done with any other wav including the reference file by putting it through the same file save process.
These files are badly enough edited that at least one of them has a distinctive pattern, anyway. I can identify it in an ABX test in about one heartbeat. This is strange because in general, the editing is very good.
However that does not address the basic problem that preferences are meaningless until the ability to hear differences is established.
These files appear to be similar enough that any serious attempt to identify most of them should result in random guessing, which has already been reported.
This is true as the test has been structured. By jumping the gun to preference testing which is just a public opinoin survey, social means can be used to skew the results.
Were it structured as I would suggest, not so much.
May be there are many motives for joining a listening test, or for creating a listening test. Of those many motives, I can see 2 major ones:
1) Having fun, it's just a game. It doesn't matter what the result are as long as it is fun then that's what matters.
2) Learning about audio. Pure interest in studying the subjective/objective aspects of audio. XRK has conduct one listening test that is considered useful for others who wants to know the quality of drivers they don't have access to listen or try out. We know how good 10F, TC9 and TG9 are from that listening test. Very useful!
I myself have purely the second motive. I have learnt a lot from these tests conducted by Mooly, PMA and XRK. I may be the oned who benefit the most, because I can hear differences, and by being able to hear that I can relate with the technical aspect. For example, I know for sure how high slew rate opamp is critical for producing complex music. I know for sure how 2nd order harmonics sounds and affects violin and vocals. I wouldn't know that if not from the test provided by Mooly and PMA.
I have to say thanks because things like that that I'm trying to learn from test like this. It is so important for me that I don't have the slightest interest to cheat or to analyze files. Tho I'm not a cheater, doing file analysis could be fun if I have no better objective
I think you forget one motive for cheating: it is FUN! (as I have mentioned in my previous post).
I have never cheated since in University and in my professional works. The level of morale and dignity is probably low around here, so believe me it is very frustrating to keep maintaining your morale standard much higher than the environment. There are a LOT of opportunities that you will miss, and even more than just that.
But in Junior High School, I have cheated during quizes just for fun, proofing to myself that I know a lot of ways to cheat (Just like I loved to put my name to appear at the top of every game machine in town). I didn't do that because I was afraid I would fail if I didn't (I was too smart for my friends for a fair competition anyway
).So believe me, I'm not surprised (not I even look down on them) or even care if many here cheated. It should annoy me if it will skew the result and ruin my conclusion. But I don't depend on that to draw a conclusion and learn something useful.
arny... I just don't know on that. Its not my field of expertise, I wouldn't have a clue where to begin on that tbh.
All I can say about these (and all the files in all the tests I've run) is that the files are made consecutively one after the other and with only the device under test being changed.
I listen to all tracks and I must say this is very hard for me.
I will try more latter in the evening.
Listen on semi old ears - Sen HD580 - O2 like amp - 9023 DAC - PC
PS
Both Foobar ABX & Lacinato ABX supports multiple files in ABX training mode
My recommendation: try Lacinato - more features than foobar
offset & gain configurable per file
I will try more latter in the evening.
Listen on semi old ears - Sen HD580 - O2 like amp - 9023 DAC - PC
PS
Both Foobar ABX & Lacinato ABX supports multiple files in ABX training mode
My recommendation: try Lacinato - more features than foobar
offset & gain configurable per file
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.