-30dB IS A HUUGE Difference.....
I can describe the differences when asked.
Over the next few days I will try the recordings on my other systems, and my Stax electrostatics.
Dan.
Me too, on a netbook with 1/2" speakers, and within the first bar....didn't bother to listen to the rest of the recording yet....not really my kind of music.
I can describe the differences when asked.
Over the next few days I will try the recordings on my other systems, and my Stax electrostatics.
Dan.
Last edited:
Didn't you get the memo?
Stop talking about as you are skewing the results for everyone else!
Duh!
Thank you.
All, please do not post results of your file analysis, like -30dB etc. It also confuses the issue, in addition the file analysis may not be straightforward and may be erroneous.
You could make this test only in case that the files were sampled simultaneously at the same times!
Now it confuses the readers, it is impossible to compare naked samples that are not acquired at the same time.
It wasn´t intended to be. 😉
It was just a clarification wrt to A/B tests.
A/B is fine, but one might encounter a problem if the preference in the population is segmented.
The afore mentioned negative control could help even in that situation.
From my personal experience i recall that listeners do have more problems getting used to the ABX-protocol compared with an A/B.
So, under that assumption you´ll need much more trials overall to keep the risk of a beta error low.
I am nevertheless quite confident that one can get very impressive results with any protocol after spending some time with training.
There is simply no free lunch; each testing strategy has its own strengths and weak spots.
Last edited:
For those who have foobar installed, you may try ABX Comparator component
foobar2000: Components Repository - ABX Comparator
and then to be non-biased at all.
foobar2000: Components Repository - ABX Comparator
and then to be non-biased at all.
THAT's why the 3rd presentation that MIGHT be a repeat of one of the others is so important ... and you have to juggle the presentation numbers so the 1st 2 presentations might be 'exactly' the same. with subtle tricks & shifts that should not affect sound but should fool the cheaters
SIGH ...
Last edited:
"I can't hear any difference" is the most important result of all.
This is what is used to weed out the deaf Golden Pinnae in an ABC test that MAY include repeated presentations.
For a paired comparison like this, especially when no ABX protocols are used, you need to repeat the test (at least once) for those who express a preference. This should give you about the same statistical significance as a single ABC test.
The 'third' repeated presentation could be simply recording the same OPA again. If done in 44.1 or 48 kHz, this should be sufficiently different to fool those who cheat by looking at the samples.
Last edited:
Richard, I am trying to avoid possible digital mess that my system is able to produce. That's why I have chosen 48/24, my system works nice at 96/24 and 48/24. 44.1/16 creates some digital artifacts that I dislike.
After closing the poll, I will post all possible measurements of "blue" and "green" chains, regarding frequency responses, gain, FFT and complete Rightmark tests.
After closing the poll, I will post all possible measurements of "blue" and "green" chains, regarding frequency responses, gain, FFT and complete Rightmark tests.
Your present test is 'paired'; ie it has 2 'presentations', blue & green.
I'm suggesting you add a 3rd file, eg red, which might be one of the OPAs in blue or green recorded again. Then juggle the names again. Then the test is to 'rank' or give opinions on 3 'presentations'.
IF 2 of the the presentations are of the same OPA (without telling the listeners), a true golden pinnae will give 'similar' rankings and comments to these.
A deaf Golden Pinnae might ascribe Chalk & Cheese attributes to the 2 recordings of the same OPA. You ignore his comments even if he sometimes prefers 5534/4562 to 741.
If you do the test (with different name sequences) 3x with a true golden pinnae and he ascribes similar rankings/comments to the 2 recordings of the same OPA and also the odd man out, then this guy is almost certainly a TRUE golden pinnae and I would listen very seriously to his comments ... even if he prefers 741s to 4562 or OP627
I've just described exactly the procedure I've used to determine stuff which only the very best ears can pick up on music.
The proportion of 'fails' ie deaf Golden Pinnae is also important as it tells you how many people won't be able to tell the difference.
___________________
ABX is a 'paired' test. The listener has a 50% chance of getting the 'right' answer if guessing
The above test I call ABC, and the listener has 33% chance of getting the 'right' answer if guessing.
You need to do many more ABX tests to get equivalent statistical validity compared to properly conducted ABC tests.
______________
There's loadsa other important stuff on DBLT I could go on about but I'm trying to concentrate on the simplest that will result in good validity for your poll.
I apologise to da statisticians among yus for my gross simplifications but this is all about getting useful (for designing better sounding stuff) info from very limited trials.
Last edited:
Got it, thanks for the explanation. We can do it as a next test, according to your description. Current poll has been already running and collecting votes.
- Status
- Not open for further replies.