Scott, I am not saying any of that.
I am saying that the ABX plugin is not fit for purpose.
Dan.
Waly you're getting at least at the appropriate experimental design question. I'm not sure that schema is the best, but rather running multiple, orthogonal tests (i.e. perhaps running your true ABX but also control tests where you have a decent idea what the result should be, e.g. running the same speaker at ever so slightly lower and higher volumes and definitely a negative test where both arms are identical).
Then you should be able to get both an individuals' result to a battery of tests against his/her peers and the ensemble result to give you a much clearer idea of both the "true test" (i.e. the ABX in question) and the performance of your listeners on that day, under those test conditions.
I don't think it's practical with audio testing to get a big enough N to know if some skew in a gaussian curve is indicative.
I, SY, Jan and other have used it. "The difference between when a track is played in the player or the ABX utility is astounding.", we don't agree. I mentioned it before making all these differences "astounding" and nothing subtle does not enhance the chance that anyone will listen.
It's simply a playback chopper/randomizer.
Max--simple test: put 4 pieces in the Foobar comparison. Make sure it's two copies of the same thing (i.e. AABB -> ABCD). Can you separate A from B and C from D (even though the fundamental files are identical)?
What you wrote needed correction/clarification which was provided.
Free yourself from stereotypes like "GEB" or something similar. As SY correctly pointed out, at the beginning of any experiment a clear stated hypothesis is mandatory. This hypothesis is the basis for any operationalization and includes the answer to the question if there is at least one human that could hear a difference or if the average Joe will be able to differentiate between the DUTs .
If you want to select a highly qualified listening panel then you have several options for doing so. Surprisingly it can make a difference if your are looking for people who are good in detecting "sameness" or if your are selecting humans who are good in the detection of differences. The signal detection theory with its experiments is a good source of information around those questions.
Please remember that every careful experimenter should be interested in getting correct results but not only those that suits his/her own beliefs.
The authors of the often mentioned ITU-R BS.1116-3 wrote for good reasons:
That's a great piece. I like the material about Bob Dylan, whom I know a little bit.Well, all right.
How you value this compared with an interview found in a HF magazine?
Sam Merril: Mason Hoffenberg Gets in a Few Licks
George
BTW I have no idea why I posted that bit about Candy. It seemed to be connected with something relevant at the time. I hope it had some entertainment value.
Something like that should be easily verified using any of a number of standard tests. Are you able to do the necessary electronic tests? Its difficult to to a loopback test on a PC of separate software on the PC so at least another PC or external instrumentation is needed. From your description it should not need exotic stuff to find.
Its possible the ABX plug-in on your system is corrupt for some reason. However testing to confirm first would be really interesting.
If time permits I'll clutter a system with Foobar and the plug-in in a few days.
I am astounded that there should be any such a difference.
To make things clearer, I am playing back some very old Tango music...scratchy sounding violin etc.... a good test.
I found this by accident...hey I expected the playbacks to be identical...in this case, I find that the ABX plugin does change the sound compared to the original.
Dan.
No bait.
The question revolves not around speakers, but the electronic components. So, perhaps reconsider your comments in that light?
Allegedly, and from what I glean, ANY "blameless" (again to borrow the phrase for simplicity) amplifier ought to sound the same as any other "blameless" amplifier. This is apparently the position that has been taken by many participants here.
If it is not, then say what the "position" actually is.
The speaker is a constant.
Presumably the multiple blameless amps will all sound indistinguishable for any given speaker, or any range of speakers chosen to listen to with the various blameless amps.
Or, does this not happen??
So, either they all sound the same, or they do not.
Which ones sound the same and why or what are these specs, or the limits of the specs.
AND
Which ones will not sound the same and why, and what are the differences in the specs?
(the same idea applies to DACs... they all sound the same, or they do not - given the idea that everything else in the system is held constant, and levels are precisely matched, etc...)
Pretty darn basic.
---------------
So, TomCHR's high feedback vanishing distortion amp OUGHT to sound the SAME to our readers and participants in a system where they are level matched and swapped out with a (properly built) Doug Self Blameless amp??
RIGHT???
This is a test track from SONY's first demo disk introducing the perfect sound forever. Notice the DNL at the MSB (0) in the right channel.
That reminds me of the fairly-early big top-of-the-line CD player of theirs which had manual high-order bit trims. One sample of it was highly prized by a departed speaker designer at Harman, and when measured turned out to have misadjusted bits and severely nonmonotonic behavior.
I bought one from the resident local audiophile and it sounded o.k. until it stopped working. It's in storage now. He had been using it as a transport and outfitted it with a glass fiber interface which fed his Wadia converter---the one with the spline fitting technique for allegedly better transient response.
No, I'm not set up to do so currently.
I expect the software plugin is fine.
The sound difference I hear is reduction in extension of both lows and highs, with a bit of dirt in the mids...symptoms of noise.
You will have to listen closely for the changes, amongst which is a change for the worse in the flow and 'vibe' and clarity of the music.
The 'dirtier' the source track the better.
Dan.
No, you seen to be equating vanishing THD with "sameness". But they still might in fact sound the same to someone in their favorite system or a dozen different folks might each prefer a different one over the other.
Let us try and with some civility discuss a theoretical test to determine what is the minimum audio bandwidth required to make no perceptible changes in recorded music to be played back on a reproduction system.
Now this is often given as 15 - 15,000 or 20 - 20,000 hertz. Some acknowledge that younger subjects may have better high frequency hearing.
The obvious method is to use a set of headphones and a level controlled oscillator. Some claims have been made that high frequency perception is not accurately measured by this technique. One valid point is that issues such as room reverb and even some noise level may actually improve the perception of audio. Of course there is also atmospheric loss of high frequencies which don't occur with headphones.
So we can first start with the music source. Musical instruments may actually produce energy to several hundred thousand hertz. Our ears have a dynamic range generally acknowledged of around 140 dB.
To be sure we have no bandwidth or dynamic range issues we can aim for a 200,000 hertz bandwidth and 150 dB or 25 bit dynamic range.
The test room should of course have noise levels of -10 dBa or so to not allow any known influence.
The room reverberation is another issue. A dead room may not be the best, so multiple rooms may be needed. Perhaps dead, .5 S, 1 S, 1.5 S, 2 S, 3 S, 5 S & 7 S reverb times should be used to allow for classic reverberation. That of course leads to direct to reflected ratios, and more rooms to examine those contributions. Perhaps we can get away with only 20-30 test rooms. particularly if we allow for artificial noise level adjustments upward.
Then there is the reproduction system. Harmonic distortion probably should be at least 60 dB down for upward masking. Downward masking is not as effective, try 100 dB for those distortions or noise.
Test subjects? Figure a minimum of 10, with 30 or more better.
Makes earphones seem quite attractive for a simplified test.
Even with all this the issue bites us hard that you can't prove a negative.
So demonstrating something exists with headphones is fine. For example I can hear from 15 to 18,000 hertz. But concluding based on headphone measurements that a loudspeaker system only needs to cover 15 to 18,000 hertz probably ignores the kids who can hear to 22,000 hertz and maybe much more!
Now this is often given as 15 - 15,000 or 20 - 20,000 hertz. Some acknowledge that younger subjects may have better high frequency hearing.
The obvious method is to use a set of headphones and a level controlled oscillator. Some claims have been made that high frequency perception is not accurately measured by this technique. One valid point is that issues such as room reverb and even some noise level may actually improve the perception of audio. Of course there is also atmospheric loss of high frequencies which don't occur with headphones.
So we can first start with the music source. Musical instruments may actually produce energy to several hundred thousand hertz. Our ears have a dynamic range generally acknowledged of around 140 dB.
To be sure we have no bandwidth or dynamic range issues we can aim for a 200,000 hertz bandwidth and 150 dB or 25 bit dynamic range.
The test room should of course have noise levels of -10 dBa or so to not allow any known influence.
The room reverberation is another issue. A dead room may not be the best, so multiple rooms may be needed. Perhaps dead, .5 S, 1 S, 1.5 S, 2 S, 3 S, 5 S & 7 S reverb times should be used to allow for classic reverberation. That of course leads to direct to reflected ratios, and more rooms to examine those contributions. Perhaps we can get away with only 20-30 test rooms. particularly if we allow for artificial noise level adjustments upward.
Then there is the reproduction system. Harmonic distortion probably should be at least 60 dB down for upward masking. Downward masking is not as effective, try 100 dB for those distortions or noise.
Test subjects? Figure a minimum of 10, with 30 or more better.
Makes earphones seem quite attractive for a simplified test.
Even with all this the issue bites us hard that you can't prove a negative.
So demonstrating something exists with headphones is fine. For example I can hear from 15 to 18,000 hertz. But concluding based on headphone measurements that a loudspeaker system only needs to cover 15 to 18,000 hertz probably ignores the kids who can hear to 22,000 hertz and maybe much more!
Actually, the issue of ADC and DAC doesn't come into play in Mike's tests. When the listener played vinyl that was genuine vinyl. When the listener thought he played vinyl (actually lowering the cartridge on the record etc) what he heard reproduced was either genuine vinyl or a digital file.
Some records didn't contain analog audio but a sort of time code that cued the digital replay without the test person having any idea what it was what was played. Add a bit of run-in groove vinyl noise to the digital tracks and presto! A DBT vinyl vs. digital test.
PS: On the 'blameless'- DS did NOT mean it was an amp that was perfect, without shortcomings. He used Blameless in the sense that all known distortion mechanisms in that specific topology had been minimized. Interesting, again, how people re-interpret other people's terminology for personal agendas.
Jan
Last edited:
What could possibly lead to a "preference" in a DBT/ABX??
Explain?
Do they NOT sound the same in all cases with "reasonable" loads?
What is the point of departure?
- Status
- Not open for further replies.