There is system intrinsic noise and there are non-linear distortions....both interact and conspire to raise the dynamic noise floor causing uncertainty and masking, which translates to loss/obscuring of low level detail information, change in positional accuracy/focus and change in 'timbre' and 'stability'.
This system intrinsic noise has individual 'character' according to the components and materials used and interacts with signal and non-linearities to produce individual system 'signature'.
Listening 'deep into the mix' for these artifacts requires training and practice....in the oft quoted 'mystical' ABX testing, how skilled in this is the average subject ?
By application of particular filters I am able to alter both level and nature of system intrinsic noise on the fly, and from hearing and discriminating these fine differences comes skill in quickly/immediately assessing systems both on the technical level and the enjoyment factor level.
Regarding ABX testing, what is the actual point ?.....to determine preference, or actual SQ or what ?
Dan.
Distortion is anything in (or not in) the output that it is in the input. Current measurement tech does not come close to covering all the possibilities.
You use the term distortion as meaning anything we can measure, but that is a small (and somewhat contrived althou useful) subset of all possible distortion.
dave
I was taking graduate level Fourier analysis when CDs were 1st announced. There are many gotchas with it, and the discrete form (FFT) has even more.If using FFT you need to know where, how, and why.
That does not sy that it isn’t extremely powerful & useful tech… you just need to know when you are stepping over the edge.
dave
So far as I can tell, no scientific proof exists that there are any remaining audible signals that are not measurable. The idea seems to be a belief held by some people who don't understand why looking at a page of measurements such as might be found in a Stereophile review does not seem to clearly describe, explain, or predict differences heard when listening to devices that have been measured by such means.
Actually, it could be that what a piece of equipment sounds like perceptually is somehow encoded in a page of graphs and tables, but if so, it is not clear anyone really knows how to decode the measurement information into predicted perceptual experiences. Thus, an assumption is made, a belief is formed, that a page of measurement information does not or cannot somehow encode all that can be heard.
Actually, it could be that what a piece of equipment sounds like perceptually is somehow encoded in a page of graphs and tables, but if so, it is not clear anyone really knows how to decode the measurement information into predicted perceptual experiences. Thus, an assumption is made, a belief is formed, that a page of measurement information does not or cannot somehow encode all that can be heard.
I just want to clarify my position which I've stated before - just so I'm not mis-represented - I never said anything in the sound was immeasurable (can NEVER be measured - is some mystical entity) - of course an. The examplything in the soundfield that i perceived should be measurable.
The point I am making (& I believe Planet10 is also making) is that we aren't using measurements that might reveal what is actually being perceived.
If you know anything about using measurements you would know that they are like a microscope - they are looking at a specific feature of the tissue/specimen. In a lot of cases the specimen/tissue has to be stained to reveal the features - there are different stains for different aspects being examined but only one stain, one aspect can be examined at one time.
Measurements are the same - the test signal is the equivalent of the stain - a jitter test signal is used to show jitter, a sine wave sweep used to highlight frequency response, two tone signal used to show third order IMD, multitone test signal to highlight a set of non-linear distortions.
These discussions are reminiscent of jitter before it was widely accepted after some people were motivated to devise a test signal & do the measurements which showed varying levels of jitter. Jitter wasn't discovered from the stock measurements that were being used at the time - it required motivation of Julian Dunn, in the mid 90s to research & device the J-test test signal which then made the concept acceptable
Unfortunately, we are at the same pivot point here but I can't see any such pivot happening as there is no motivation to investigate this & it is a stickier problem than jitter.
I mentioned multitone test signals before because they are much more like a music signal & they have been shown to better correlate with subjective assessments of good sound. Coming back to what I said at the start, it may be that if multitone tests were done for every device in the replay chain we might see a pattern emerging? Who knows? All I know is that we are not seeing the correlation between measurements & subjective sound quality.
I hope this explains my viewpoint? Just to repeat, I'm not saying anything is unmeasurable, we just haven't found the correct measurement or set of measurements or we are interpreting existing measurements incorrectly (for instance the masking example)
The point I am making (& I believe Planet10 is also making) is that we aren't using measurements that might reveal what is actually being perceived.
If you know anything about using measurements you would know that they are like a microscope - they are looking at a specific feature of the tissue/specimen. In a lot of cases the specimen/tissue has to be stained to reveal the features - there are different stains for different aspects being examined but only one stain, one aspect can be examined at one time.
Measurements are the same - the test signal is the equivalent of the stain - a jitter test signal is used to show jitter, a sine wave sweep used to highlight frequency response, two tone signal used to show third order IMD, multitone test signal to highlight a set of non-linear distortions.
These discussions are reminiscent of jitter before it was widely accepted after some people were motivated to devise a test signal & do the measurements which showed varying levels of jitter. Jitter wasn't discovered from the stock measurements that were being used at the time - it required motivation of Julian Dunn, in the mid 90s to research & device the J-test test signal which then made the concept acceptable
Unfortunately, we are at the same pivot point here but I can't see any such pivot happening as there is no motivation to investigate this & it is a stickier problem than jitter.
I mentioned multitone test signals before because they are much more like a music signal & they have been shown to better correlate with subjective assessments of good sound. Coming back to what I said at the start, it may be that if multitone tests were done for every device in the replay chain we might see a pattern emerging? Who knows? All I know is that we are not seeing the correlation between measurements & subjective sound quality.
I hope this explains my viewpoint? Just to repeat, I'm not saying anything is unmeasurable, we just haven't found the correct measurement or set of measurements or we are interpreting existing measurements incorrectly (for instance the masking example)
You need to widen your horizons, lose the blinders.
For instance, despite being widely shown to be a useless predictor of sound quality. THD is still widely used even when more accurate predictors have been devised and proposed, but THD is so easy to perform its momentum keeps it from being tossed.
dave
IOW, you don't have any example to cite. Dodging of pertinent question gives it away.
If we had a laundry list of examples one would hope that we would be developing tests for them.
One example, already mentioned, is DDR. We need a measure of how much information is lost as opposed to typicl distortion measures which attemt to measure what is added.
The same message i gave to BV applies to you.
The choice to say that existing measures tell us how something sounds is a subjective choice to ignore all that shows it is not. ie being an “objectivist” is a subjective POV. So no different than the ardent “subjectivist”. The 2 ends of the spectrum are both in the same leaky boat. Like most things in life, being somewhere in the middle makes the most sense, and that also means keeping an open mind and pushing forward.
dave
One example, already mentioned, is DDR. We need a measure of how much information is lost as opposed to typicl distortion measures which attemt to measure what is added.
The same message i gave to BV applies to you.
The choice to say that existing measures tell us how something sounds is a subjective choice to ignore all that shows it is not. ie being an “objectivist” is a subjective POV. So no different than the ardent “subjectivist”. The 2 ends of the spectrum are both in the same leaky boat. Like most things in life, being somewhere in the middle makes the most sense, and that also means keeping an open mind and pushing forward.
dave
If one were to give two wav files to a measurement guy, would any of them be able to tell with certainty if the files are audibly distinguishable to some human? Or, in what way they might perceptually sound different? Probably not or not in much detail, would be my guess. So far as I am aware, measurements are not currently available to do that, but I could be wrong? I can say this, they contain music and one of them passed through an extra stage of power amplification. No loudspeakers or microphones were used. Just D/A > A/D, or D/A > amp > A/D.
Last edited:
Part of the problem is that you could have two 100% bit identical files, and there are some who would still say they sound different, and therefore discard your measurement (which in this case is a file compare). Right at the end of the audio chain is the mot unreliable, and often inconsistent, part by far!
I will assume for the case given a proctored ABX test would be required for verification.
EDIT: Stipulation of that nature would require more than one set of files, but we could work out practical details. The question remains as to what we can predict about audibility from measurements? Is it more like rough generalizations, or specific details -- what can we do, where is the technology at now for that?
EDIT: Stipulation of that nature would require more than one set of files, but we could work out practical details. The question remains as to what we can predict about audibility from measurements? Is it more like rough generalizations, or specific details -- what can we do, where is the technology at now for that?
Last edited:
It's not worth your time. You'll just get the same old poorly constructed arguments spit back at you.
In order to make any progress at all with such claims, those claiming differences would need to describe them in such a way that other listeners could either confirm or deny they're hearing the same things. But I wouldn't hold my breath for any such progress
On what do you base this comment? It's easy to make these glib toss off comments just like I could say every thing that matters has already been measured. Taking a step back, a stereo signal has two voltages vs time quantized in two dimensions this information content is limited. It has nothing to do with the measurement and all to do with interpreting the information. As I pointed out months ago the THX algorithms on a Kindle give a scary realistic 3D sound field, with what 1/2" full range speakers.
Last edited:
- Status
- Not open for further replies.