Jitter vs no-Jitter Part III

[bibo01]

Dear Friends,

After our first test "Jitter vs. no-Jitter" here it is another test, actually two, on our ability to recognize jitter.

Following the advice of the user "mitchco" We have prepared a file containing two songs by Marilyn Mazur & Jan Garbarek from the album "Elixir" - "Bell-Painting" and "Elixir" - for high and low range respectively, with high DR.

At the moment I do not want to give too much technical information, which will be given at the end of the survey. The two tests seem the same, but obviously they are not.
Both songs were recorded with two equal converters, but one was defective and produced much jitter.

In both tests we have to try to recognize the song with LESS jitter.

The poll will last one week only and it is published on three forums: NextHardware, ComputerAudiophile, WhatsBest.

For "Jitter vs. no-jitter Part II" the tracks can be downloaded from:
dropcanvas - instant drag and drop sharing - canvas view
or
https://www.dropbox.com/s/agbq0wizwa301au/Test 1.zip?dl=0

For "Jitter vs. no-jitter Part III" the tracks can be downloaded from:
dropcanvas - instant drag and drop sharing - canvas view
or
https://www.dropbox.com/s/pis3kp818onp168/Test 2.zip?dl=0

Good listening!

 

[bibo01]

Many shy users here

 

[lateboomer]

I have voted.

 

[bibo01]

Thx

 

[bibo01]

Poll result

We can call the double poll - Part II and III - as closed.

I thank all those who have taken part with their listening.
I thank Tom "Gefrusti" for the essential technical contribution.

------------------------

The password to open the file with the answer can be downloaded here: dropcanvas - instant drag and drop sharing - canvas view

The answer is:
Part II
Track A: Jitter OFF
Track B: Jitter ON

Part III
Track C: Jitter ON (like B but renamed)
Track D: Jitter ON + extra jitter.

------------------------

Bearing in mind also four votes that were given to me in private, the cumulative result of votes from three participating forums NextHardware, ComputerAudiophile and WhatsBestForum, is:

Part II
A: 13
B: 10
X[SUB]1[/SUB]: 6

Part III
C: 10
D: 16
X[SUB]2[/SUB]: 2

 

[bibo01]

For the record, files A and B were obtained in the same way of poll Part I, that is with two equal DACs but one with bad clock that produced a lot of jitter.
File C, as mentioned above, is the same as B, just renamed.
File D is like C but with added white noise up to -70dB. So file D is a mix of periodic and stochastic jitter.

 

[lateboomer]

I got part II correct, but part III wrong. Trying to make sense out of it. The tests were hard. I give it to you bibo01, well done.

 

[bibo01]

In the two parts poll the tracks with less jitter were A and C - I try to give my own interpretation.

In Part II there is a slight majority for A, but not decisive; in fact, it tends to disappear when you consider as well that on the other side six votes did not perceive difference. Basically, this test was not easy to "pass", in the sense that the difference between the two files is not high.

In Part III the difficulty to perceive difference is reduced further (only 2 votes for X[SUB]2[/SUB]. However, most of the listeners preferred file D, the one with more jitter and with opposite characteristics to A.

The position of Tom "Gefrusti" - my tech expert - about it is:
"... I think the differences are so nuanced that everyone (regardless of the time) interprets according to the "particular section" that is analyzing in the sound of the songs.
As it was a comparison that "flows" in real time, our perception system makes some frames at different times ... so the micro-variations in timbre//tone become misleading in what it should be a comparison of the same mirror segment of sound that you are listening.
I personally stopped playback in several places and I compared that with a mirror image of the other songs ... I have tried to listen carefully to all the details that make the attack, decay and tonal//timbre yield ... They seemed to me so alike that ... I just could not make repetitive and demonstrable what I was forcibly trying to perceive.
Is preference expressed at random? If the brain stores the spectrum into different time segments ... we will fight with spectrum yield completely at random.
If you notice, in our tests there has never been an overwhelming prevalence of a particular file ... instead it would happen if the differences went up gradually."

My personal theory, however, is that there is a difference in the way in which people express the pleasantness and/or the correctness of a piece of music.
Normally a listener does not listen in ABX. While listening, a comparison is made internally with listener's own idea of ??music that is based on various personal references: experience, aesthetics, knowledge ...
IMHO, some people have a more cerebral approach to decision-making, perhaps combined with a knowledge of music, that establishes the correct sound of a song, and therefore determines the way they establish the perceived pleasure.
Other people, however, have an approach "with the belly" to decision making that is the opposite of the previous one: if the song is perceived as more pleasant, therefore it is the more correct.

I report my experience in regards to the tests which somehow corresponds to the "theory" expressed above.
I began with the assumption from the first poll of a few weeks ago to see if there was direct consequence between the two tests, that is, the "thinner" song is the original with less jitter, whilst the song with more body has more jitter although subjectively "better".
My partner and I listened to the songs separately for 3 times in a row, "in the dark", that is, one was listening (only the first part - plin plin) while the other would change the song. Answering the question "the song with less jitter", we both made the same choices and, apparently, we have even guessed right because we voted A and C. For the record, however, if she had to express a subjective preference, my partner would have chosen B and D.

 

[esldude]

Well what you have in these tests is an alternative choice test. The statistics for that are different than ABX. If you have enough responses (30) or so, 75% correct is the threshold for a 5% chance the result is not chance. That with 2 choices. Some will handle a test like this when you have choice of no difference by dividing the no difference votes equally between the two choices. Doing that your tests don't show the results reaching the 75% correct threshold.

The results would be cleaner if you had only two choices and not a no difference choice.