Come on guys; this sort of test is commonly known as a conservative test.
As it is routinely not shown, that the tests are valid, one could come to the conclusion that this might be intentional.
Wishes
As it is routinely not shown, that the tests are valid, one could come to the conclusion that this might be intentional.
Wishes
Most double blind tests that I've read about, done by AES and such don't work and shouldn't given how we hear. It's not that complicated.
Edit: more specifically they take lots of subjects and do short A-B tests that "Prove" for instance that CD is no different sounding that SACD.
Edit: more specifically they take lots of subjects and do short A-B tests that "Prove" for instance that CD is no different sounding that SACD.
Last edited:
Which particular ones have you read? (not read "about') What leads you to the assertion that they don't work? Analysis? Evidence?
BTW, AES doesn't "do" tests, it publishes journals.
Well SY, you have discredited 3 Phd's (all professors) this morning, already, it seems hopeless to try further.
Well, if you cite their work to support something that they didn't say, that's discreditable. The idea that this reflects on them is bizarre.
edit: I'm trying to figure out who #3 is. You cited Leventhal, who doesn't support your claim, you've cited Rees, who published a purely speculative piece in a hobby magazine, but I can't figure out who the Third Tenor is.
edit: I'm trying to figure out who #3 is. You cited Leventhal, who doesn't support your claim, you've cited Rees, who published a purely speculative piece in a hobby magazine, but I can't figure out who the Third Tenor is.
It is possible to draw the conclusion that Leventhal supports the claim, without having said it explicit.
Leventhal has shown (the somewhat basic mechanism) that if the probaliity of a participant to detect a difference is below 0.7 than the risk for error type 2 will be quite high for the usual 16 trial ABX.
So if experimentators refuse to care for the detection capabilities of their participants, but still use a 16 trial ABX ......
They might have not known before Leventhal brought it to their attention but not afterwards.
Wishes
Hmmm. Not sure I appreciate that you put a word ('vital') in a quote of mine so it looks as if I said it. Please don't.
jd
Saying "type 2 error increases with n" is not even vaguely the same thing as what John asserted.
Typo, you mean "decreases with n" (or did you mean increases with _r_ ? ).
Isn´t that a construction of a test in favour of a negative result (means not rejection of the null hypothesis) ?
Escpecially if one knows about that fact; before Leventhals paper it might have been due to oversimplification of the task, but after his articles it is intentional to use the same routine as before.
Wishes
P.S. AFAIR you stated yourself, that your test protocol wasn´t meant to detect audible differences per se, but more to test, if the differences were as big as stated by listeners before.
That is a different test goal and nobody should be surprised if it gives negative results.
Last edited:
Yeah, the test was published in a AES journal, which I read about in AudioXpress. But tests like these never will work, because we all have different ears, and experience with sounds. You have to learn the difference to identify it, or perform well on a test.
Blind testing takes away a lot of sensory information the we use to process sounds. So, it makes sense that we wouldn't perform as well on blind tests. It's a new experience. And new experiences require some learning or relearning.
The conditions under which we hear a sound do matter. Which objectivists a quick to point out to support their side of the argument.
Last edited:
Can you give me a for-instance with respect to sound? AFAIK, no-one requires the test subjects to wear earplugs.
So I take it that you haven't actually read any of the actual papers?
Yes, thanks.
I'd refer readers of the Leventhal paper to also read the fascinating discussion which followed up in JAES v35 pp567-572. Again, this paper does not at all say what John claims it does.
I spoke to Les himself. I know what he was getting at. ONCE he has to put it in the language approved by Lipshitz, et al, no wonder that one cannot find the forest for the trees.
I won't waste anyones time further here, on this, 3 full professors have been knocked down already, this morning. Leventhal, Rees and Shaffer. (You should read the article titles, closer SY)
I won't waste anyones time further here, on this, 3 full professors have been knocked down already, this morning. Leventhal, Rees and Shaffer. (You should read the article titles, closer SY)
John,
You're getting on a slippery slope here. If you claim that DB tests don't work because they take away the 'other' (presumable non-auditory) sensory inputs, you imply that the reason sighted, non-DB test, DO find a difference is due to non-auditory sensory input, iow, inputs other than the sound. That in turn could imply that there are no sound differences.
jd
No professors were knocked down whatsoever. What was knocked down was your claim that they said something specific. Apparently they didn't. And your claim that he (Les) didn't say it but he really meant to is another non-verifiable claim.
jd
I'm assuming that the series resistor won't cover for switching distortion on a solid state output stage (higher frequency stuff), it'll certainly give a more bloated bass response due to a lower damping factor though.
I can imediately see how amplifier to speaker cabling can make a difference due to the reactive nature of loudspeakers.
Getting back to the possible interconnects has anyone considered the amount of RF that is kicking about these days? Shielding on cables must surely play a part. I don't live that far from a 5KW FM radio transmitter, on top of that there are mobile phones, wifi & broadband via RF, anything that is switched on in the local area causing some kind of RF interference etc.
Perhaps i'm a mile off, but with a few hundred millivolts kicking about driving power amps & speakers i think it might well end up being significant. I guess this is more of a problem with single ended connections as balanced connections should render this a non entity due to cancellation?
This is a problem with differential interconnects also. If you look at the characteristics of the circuitry used to get the difference you will see that the ability to reject common mode signal falls off with frequency. A second limitation is that the circuitry can clip when the common mode voltage goes too high (About 6VRMS.)
A differential circuit is not the same as a balanced circuit. In the most common balanced connections a transformer is used to both produce a balanced signal and one this is isolated from the other circuits, ground, etc. A balanced line should have identical source and load impedance. The interconnect should have good twist to insure uniform behavior to an external field. The transformer acts as a very good EMI (RF) filter in most cases.
Vacuum tubes are usually the least affected by EMI followed by FETs.
In my sensitive circuitry passing cars and especially trucks cause observable interference. The most common source of EMI is your computer. My unmodified Sony CD player sounds different (Less grungy by personal non-documented observation) when the EMI it is experiencing is lowered. Some EMI comes in through the power line.
A properly constructed coaxial cable will minimize the outside influence above a few hundred khz. The equipment should be able to handle the rest by using proper signal levels, low output impedances and power supply filtering in all but the extreme cases. In those case extra shielding may be useful, but then why would you want to expose yourself to those levels?
So the answer is yes cable shielding is very important.
Just saying the Blind testing is a new experience for most people. You're most likely to get a positive result if the person learns the difference first. If you test a person on a system they are unfamiliar exactly what difference should they be listening for.
We learn sounds Jan. A blind test should be done with sounds we are familiar with. That's all I'm saying.
And I think there is absolutely no reason you can't learn a sound by hiding the component brands, and have them labeled A or B or C. And then take the test, to see if you can identify these differences in a random blind test.
But, I don't know how important this is in taking a blind testing, if the person is unbiased.
Jan, if you closed you eyes could you echolocate to navigate? Some blind people do. If not why? You have bad ears? How long do you think it would take to learn these differences?
So how do you think taking a sample of people and running them through blind tests proves that there is no audible difference, when they are unfamiliar with the sounds? What difference should they be listening for?
Last edited:
- Status
- Not open for further replies.