[mike1127]

Audio is an interesting little corner of knowledge about the world: it's an intersection of the objective and subjective.

Something like a musical performance can only be evaluated subjectively. No one would presume to use measurements to tell Yo-Yo Ma he needs to change something about his performance.

Something like rocket science has only objective criteria. Whether a rocket gets to its destination is not something we have "feelings" about. It can be measured and determined objectively.

Audio reproduction is interesting because it combines elements of each. An audio system exists in the real, objective world, and performs in ways that can be measured. We even have a pretty good idea what makes a good system, objectively: flat frequency response and low amounts of known distortion mechanisms.

And yet the purpose of audio is not to reproduce sound, but to reproduce a musical performance. Therefore, the ultimate test of whether a system has created an accurate (or other people would emphasize enjoyable) is subjective.

Systems that measure well can sound like crap. That's because the universe is bigger than our models of it. My view, almost a philosophical stance, is that in areas of aesthetics our models of the universe are poor and incomplete.

In rocket science, our models are very good and comprehensive.

In reproduction of art and music, our models are poor.

I know many disagree with me.

 

[Quinto]

I think that's because our comprehension of aesthetics in relation to our mind is more poor and incomplete than that of the universe..

 

[royalcrown]

I think you make a good point, but the focus of analysis seems to rest on a sort of Kantian assumption that we have no access to the outer world except through our "faculties", to use Kant's terminology (that is, only through processing the information can we make sense of the world). I'm not going to argue against that kind of a statement as it's way out of the realm of audio, however I will say this: technology has gotten to the point where we can analyze, at least to some extent, audio information before it even reaches the ear.

For example, let's assume that I grant the argument that "sound as sound" is different and distinguishable from "sound as music". I'm not really sure where I stand on that issue just yet, but let's assume it's true. Even with that, given two equal waveforms that are exactly identical, regardless how we process this data, we won't be able to extrapolate any differences between the two waveforms, given that the two waveforms are in fact identical in every respect.

In this sense, we have some notion of analysis of data outside of our cognitive faculties, and while I agree that nobody can say that Yo-Yo-Ma needs to change his act, I think that only extends as far as aesthetic appreciation goes, and no further. I don't think the issue is that people have differing aesthetic tastes, as that's trivially true. Rather, I think that the issue is whether or not an actual difference exists between two given things, regardless of which one is better, worse, etc.

In this sense, while I won't question someone who prefers another artwork to the Mona Lisa, given two perfectly identical prints of the Mona Lisa, if someone says that he finds one more beautiful than the other, I have a hard time believing that, let alone making that an issue of aesthetic taste.

I think, if anything, this applies more to tube amplification than it does to differences between cables (yes, I know you never mentioned cables, but let's face it: we're all here to see if x or y makes an audible difference, whether that be cables, 24/96, jitter, etc etc) - however, with tube amplification, we can readily see the audible differences in the measurements. So in this sense, I agree that measurements cannot tell you what something aesthetically sounds like, however it does serve an immensely useful purpose: it delineates whether or not a difference exists in the first place. Without an "actually existing" difference, any differing in aesthetic appreciation is due solely to some psychological phenomenon, and at that point I think some philosophers of art (I'm thinking Kantian scholars in the sense of disinterestedness, but others apply as well) would argue that those psychological phenomena have nothing to do with aesthetic appreciation and that there is no aesthetic dimension at play.

Sorry for the rambling post and inane references, I've studied aesthetics/philosophy of art way too in-depth.

 

[mike1127]

Quote:

Originally Posted by royalcrown /img/forum/go_quote.gif I think you make a good point, but the focus of analysis seems to rest on a sort of Kantian assumption that we have no access to the outer world except through our "faculties", to use Kant's terminology (that is, only through processing the information can we make sense of the world). I'm not going to argue against that kind of a statement as it's way out of the realm of audio, however I will say this: technology has gotten to the point where we can analyze, at least to some extent, audio information before it even reaches the ear.

Thanks for the post---this is actually the kind of discussion I wanted. The intersection of philosophy and science.


Quote:

For example, let's assume that I grant the argument that "sound as sound" is different and distinguishable from "sound as music". I'm not really sure where I stand on that issue just yet, but let's assume it's true. Even with that, given two equal waveforms that are exactly identical, regardless how we process this data, we won't be able to extrapolate any differences between the two waveforms, given that the two waveforms are in fact identical in every respect.

Okay, I didn't study philosophy in school so I don't know what existing bodies of thought would agree with me, but here are my points:

Let's speak for a moment of objective knowledge: We know the world through models, and reality is always richer than the models. Quote from Einstein: As far as the laws of mathematics refer to reality, they are not certain, as far as they are certain, they do not refer to reality.

Let's say you want to measure the temperature. You must construct a device. The operation of this device rests on several engineering assumptions or approximations, which you control the best you are able. Finally you get a temperature signal but it has some noise and requires some averaging. After applying an averaging algorithm you produce: the temperature. Some people think that measuring something is about, well, just measuring it. You hook up the equipment, measure it, and you're done with it. This is not the case. We can only derive measurement by modeling the world, and our models are not certain.

Some scientists will say, "Sure, our models aren't certain. We used Newtonian physics for a long time before relativity came along, and with Newtonian physics the prediction of planetary orbits was off a bit. Just a bit." In other words, they would say that there is no big deal.

Other people believe that in the cracks between the existing models lie all sorts of unexplained phenomena. In the field of audio, poised in the intersection of the objective and subjective, there are abundant possibilities not well explained by existing models.

First let me talk about measuring a device. Some people believe that you run a device through a few categories of measurements and you've fully characterized it. I disagree, in part because all measurements are models, and your models can never describe the whole system.

To give an example: if you measure the frequency response, you have to assume the system is linear. Frequency response doesn't mean anything for a non-linear system. The system is approximately linear, but not completely. Therefore there is something essential missing from the FR graph.

Okay, so you decide to measure harmonic distortion to get a handle on non-linearity. Harmonic distortion is just one model.

Think of like this. A device is not really a thing, it's a behavior. It is something which exhibits certain behaviors under certain input conditions. Measurements pick a very small set of input conditions and attempt to characterize the behavior of the device. Their results are noisy and inherently dependent on the models assumed by the test equipment designers.

No two things ever measure "identically" for at least the reason that there is noise in the measurement. But beyond that, you've only attempted to study the behavior under a limited set of input conditions.

My opinion here is, therefore, a philosophical stance in which measurements and models are fairly poor in the area of audio.

Everyone will agree there are "cracks" between the measurements---unmodeled effects, noise, etc. But some people feel the cracks are insignificant. They believe you can truly characterize a system by a representative set of measurements. They believe all cracks represent very small deviations from known behavior, deviations way too small to matter to the human ear.

On the other hand, I find that the cracks are incredibly significant. They open up the possibility of cables mattering, CDPs sounding different, etc.

Quote:

Sorry for the rambling post and inane references, I've studied aesthetics/philosophy of art way too in-depth.

Cool!

 

[royalcrown]

Quote:

Originally Posted by mike1127 /img/forum/go_quote.gif Thanks for the post---this is actually the kind of discussion I wanted. The intersection of philosophy and science. Okay, I didn't study philosophy in school so I don't know what existing bodies of thought would agree with me, but here are my points: Let's speak for a moment of objective knowledge: We know the world through models, and reality is always richer than the models. Quote from Einstein: As far as the laws of mathematics refer to reality, they are not certain, as far as they are certain, they do not refer to reality. Let's say you want to measure the temperature. You must construct a device. The operation of this device rests on several engineering assumptions or approximations, which you control the best you are able. Finally you get a temperature signal but it has some noise and requires some averaging. After applying an averaging algorithm you produce: the temperature. Some people think that measuring something is about, well, just measuring it. You hook up the equipment, measure it, and you're done with it. This is not the case. We can only derive measurement by modeling the world, and our models are not certain.

You're right in that modeling is bound to be incomplete to some degree, but we do have statistical analysis available to us that lets us estimate, within a confidence interval, what the actual result is. For example, we may not be able to actually know the exact temperature, but we can estimate the temperature to be between, say, 70 and 71 degrees F with 95% confidence. While this doesn't give us an ultimate answer, it does allow for a certain amount of reasonableness.

Quote:

Originally Posted by mike1127 /img/forum/go_quote.gif Some scientists will say, "Sure, our models aren't certain. We used Newtonian physics for a long time before relativity came along, and with Newtonian physics the prediction of planetary orbits was off a bit. Just a bit." In other words, they would say that there is no big deal. Other people believe that in the cracks between the existing models lie all sorts of unexplained phenomena. In the field of audio, poised in the intersection of the objective and subjective, there are abundant possibilities not well explained by existing models.

While my understanding of science, especially physics, is limited, I think a distinction should be made between predictions and measurements. When it comes to planetary orbits, we weren't observing something, we were predicting something, which is different from simply observing behavior. AFAIK, our observations weren't falsified that much by relativity, but our theories to predict something else based on said observations were falsified. Of course I'm no expert so if my understanding is different from the truth please let me know.

Quote:

Originally Posted by mike1127 /img/forum/go_quote.gif First let me talk about measuring a device. Some people believe that you run a device through a few categories of measurements and you've fully characterized it. I disagree, in part because all measurements are models, and your models can never describe the whole system. To give an example: if you measure the frequency response, you have to assume the system is linear. Frequency response doesn't mean anything for a non-linear system. The system is approximately linear, but not completely. Therefore there is something essential missing from the FR graph. Okay, so you decide to measure harmonic distortion to get a handle on non-linearity. Harmonic distortion is just one model.

I'm speaking again without being an expert in the subject, but from the literature I've read, the typical measurements were chosen not arbitrarily, but because they are the parameters that affect audibility to the greatest degree. For example, phase distortion is one measurable characteristic that's not really talked about much with headphones or non-transducer performance because with headphones even large amounts of phase distortion doesn't affect the sound that much in blind, large scale testing. I don't have the study on hand, but it was in one of my previous posts. I'll try to find it and get back to you.

As an aside, I think the vantage point of treating measurements as models asks the question: do our ears merely model data in the same way that measurements do? I would say the answer is yes, if our ears are as inaccurate as our eyes are. A good display of this comes (quite surprisingly) from this image:

http://www.thebestpageintheuniverse....lon_fraud3.gif

Our eyes cannot resolve a difference between the two colors, and yet our equipment can indeed resolve a difference. If the equipment is merely modeling the data and is incomplete to some degree, our eyes are also modeling the data and are incomplete to a larger degree.

Quote:

Originally Posted by mike1127 /img/forum/go_quote.gif Think of like this. A device is not really a thing, it's a behavior. It is something which exhibits certain behaviors under certain input conditions. Measurements pick a very small set of input conditions and attempt to characterize the behavior of the device. Their results are noisy and inherently dependent on the models assumed by the test equipment designers. No two things ever measure "identically" for at least the reason that there is noise in the measurement. But beyond that, you've only attempted to study the behavior under a limited set of input conditions.

Perhaps they can't measure perfectly identically, but within some margin they can be de-facto identically. If, say, two pieces of equipment vary by .001% in THD, it's very likely that nobody will be able to tell the difference, and while one individual in 100 billion may be able to distinguish between the two, I think it's more reasonable to assume that nobody can hear the difference than hold out thinking that you might be that individual (or simply assume that you are the individual because you are have "golden ears," a fairly arrogant estimation that plagues many many people on this forum).

Quote:

Originally Posted by mike1127 /img/forum/go_quote.gif My opinion here is, therefore, a philosophical stance in which measurements and models are fairly poor in the area of audio. Everyone will agree there are "cracks" between the measurements---unmodeled effects, noise, etc. But some people feel the cracks are insignificant. They believe you can truly characterize a system by a representative set of measurements. They believe all cracks represent very small deviations from known behavior, deviations way too small to matter to the human ear. On the other hand, I find that the cracks are incredibly significant. They open up the possibility of cables mattering, CDPs sounding different, etc.

You're right in that they do open several possibilities. However, I feel that this distinction, while philosophically interesting, is largely academic in nature. The empirical data may be incomplete, but they do seem to work. To use your example, while our predictions on planetary orbits were off, for many applications they did, and still do, work. I think the more practical question is: as far as our application is concerned (that is, within the realm of audio playback for headphones), do our models work? The answer to that question can only be found in empirical data, and in that regard we have blind testing to help us, at which point we're back to square one.

 

[JadeEast]

I think the objective and subjective categories can be a bit of a problem as many people consider them to be incompatible or opposites. The term/category of "first person perspective" and "third person perspective" can be used to allow for an easier integration of the different points of view.

 

[coldkeith]

or the term intersubjective (look into habermas or wilber for a fleshed ut version of what jadeeast is talking about).

 

[cerbie]

Quote:

Systems that measure well can sound like crap. That's because the universe is bigger than our models of it. My view, almost a philosophical stance, is that in areas of aesthetics our models of the universe are poor and incomplete.

I'd say it's because it was measured wrong by whoever claimed that it measured well, and has nothing to do with the [redundant] limitations of models. IE, if many people agree that it's wrong, you have the same as some equipment's measure above a noise floor. It's not a 0.000x% something from a piece of equipment, but it is most certainly a measurement, and we have the maths to decide the quality and magnitude of such measurements.

A useful measurement must measure for a result that you want. If you want to know if an audio playback chain sounds good, you don't measure THD+N to say whether it sounds good or not. Noise and harmonic distortions, as well as phase distortions that don't neatly show up in there, and others, tell you about the precision, and possibly potential limits of resolution, of the system. For whether or not it sounds good, you must use a measurement that compares it to other things that are known to sound good, on criteria based on what sounds good and what does not.

On models: a model can not be a complete. Missing the totality of information of the real thing it represents is what makes it a model.

Beyond that, would you care to define how you are using, "objective," "subjective," and "knowing"? They seem awfully contradictory to me, so I assume you're meaning them differently than I'm reading them.

 

[mike1127]

Quote:

Originally Posted by cerbie /img/forum/go_quote.gif I'd say it's because it was measured wrong by whoever claimed that it measured well, and has nothing to do with the [redundant] limitations of models. IE, if many people agree that it's wrong, you have the same as some equipment's measure above a noise floor. ..... we have the maths to decide the quality and magnitude of such measurements. .... A useful measurement must measure for a result that you want. ..... For whether or not it sounds good, you must use a measurement that compares it to other things that are known to sound good, on criteria based on what sounds good and what does not.

I disagree with your general idea here that it's so easy to get these criteria about what sounds good.

A really simple example in the visual domain is a comparison of the beauty of two women. What measurements tell us that? Although social psychologists have a few general ideas (like symmetry) those are only general and don't really get down to the nitty gritty of direct comparisons.

Or another visual example. suppose we have these two images:

And suppose we ask viewers which image is "more colorful." And suppose we want to make a measurement that captures this impression. There is no lack of information here we can gather. We have the color of each pixel. The problem is what to do with all that information. How do you organize that information so that it reflects something about a human aesthetic impression?

I think that SET amplifiers and vinyl are higher-resolution than other types of ampliers and digital, respectively. I don't think there are any measurements that explain that.

Quote:

On models: a model can not be a complete. Missing the totality of information of the real thing it represents is what makes it a model.

That's true, but people regard this fact with different degrees of significance.
Quote:

Beyond that, would you care to define how you are using, "objective," "subjective," and "knowing"? They seem awfully contradictory to me, so I assume you're meaning them differently than I'm reading them.

Contradictory?

I use objective to mean something existing in the "external world" that presumably can be observed by anyone, and we can agree as to its behavior.

Subject is something we experience. It's "in our heads" so it can't be directly observed.

You will have to give me a better context for "knowing" because it's such a broad topic (epistemology).

 

[mike1127]

Quote:

Originally Posted by royalcrown /img/forum/go_quote.gif As an aside, I think the vantage point of treating measurements as models asks the question: do our ears merely model data in the same way that measurements do? I would say the answer is yes, if our ears are as inaccurate as our eyes are. A good display of this comes (quite surprisingly) from this image: http://www.thebestpageintheuniverse....lon_fraud3.gif Our eyes cannot resolve a difference between the two colors, and yet our equipment can indeed resolve a difference. If the equipment is merely modeling the data and is incomplete to some degree, our eyes are also modeling the data and are incomplete to a larger degree.

Our ears and brains do not model data at all the way measurements do. Ears and brains model things like: "what instruments are present, what is the tone color of each, what key is this in, etc.". Typically audio measurements are something else entirely.

Regarding that revlon color example, at least on my monitor I think I can see a difference. This is interesting because it disappears when I try to do "quick-switch."

If I rest my attention in the right color patch for a while, until I'm "used to the feel of it," then switch to the left color patch, the left one is distinctly lighter. If I rest there, and switch back, the right one "feels" darker.

If, on the other hand, I switch back and forth rapidly, all differences become uncertain and largely vanish.

That doesn't disprove their hypothesis about Revlon lipstick because the color values in the images are approximations.
Quote:

You're right in that they do open several possibilities. However, I feel that this distinction, while philosophically interesting, is largely academic in nature. The empirical data may be incomplete, but they do seem to work. To use your example, while our predictions on planetary orbits were off, for many applications they did, and still do, work. I think the more practical question is: as far as our application is concerned (that is, within the realm of audio playback for headphones), do our models work? The answer to that question can only be found in empirical data, and in that regard we have blind testing to help us, at which point we're back to square one.

I'm not an audio engineer, but of those I know, many feel our models of audio do not "work" at all, if by "work" you mean predict what will sound high-resolution, accurate to the source, or aesthetically pleasing.

 

[cerbie]

Quote:

I use objective to mean something existing in the "external world" that presumably can be observed by anyone, and we can agree as to its behavior. Subject is something we experience. It's "in our heads" so it can't be directly observed.

Direct observation would be where my difficulty in figuring it out would be.

Objective (what behavior of the universe we can all agree on with great confidence) does not have a great dividing wall from the subjective (what remains purely subjective as a set of our experiences gets smaller all the time). When it comes to measurement, we came to be able to reach these objective conclusions through consistent application of our subjective faculties. We are each every bit a set of instruments for measurement. As it concerns our perceptions, especially aesthetics, we are the most useful tools for measurement at this time. In as much, many aspects of own workings are as much objective as subjective, and the subjective can be used to reach sufficient confidence as to be used in objective contexts.

Quote:

Originally Posted by mike1127 /img/forum/go_quote.gif I disagree with your general idea here that it's so easy to get these criteria about what sounds good.

If you think X sounds good, and Y sounds bad, and 15 out of 20 other people agree with you, that's a pretty good start. Finding specific dimensions that we can agree on, without the use of controlled survey is difficult. But, while using a non-human tool may be best, to remove bias, it can also remove any usefulness of the measurement. The measurement criteria must be based on what is being measured for.

Saying something will measure bad if it sounds bad, and finding out the kind of objective measurement that will show that, are not the same thing.

Quote:

A really simple example in the visual domain is a comparison of the beauty of two women. What measurements tell us that? Although social psychologists have a few general ideas (like symmetry) those are only general and don't really get down to the nitty gritty of direct comparisons.

No, but is that a matter that it can not be done, or simply not having enough of an understanding our own operation and development?

Quote:

Or another visual example. suppose we have these two images: And suppose we ask viewers which image is "more colorful." And suppose we want to make a measurement that captures this impression. There is no lack of information here we can gather. We have the color of each pixel. The problem is what to do with all that information. How do you organize that information so that it reflects something about a human aesthetic impression?

It is not too difficult to figure out if one digital color image is more colorful than the next, especially if they are using similar hues. With those kinds of images, you could take a difference from grayscale, add up what's left, and compare. With different colors, you'd need to convert to a system like HSV, weigh general directions (in S) of H (much less easy, but doable). Then, compare with impressions.

But, there is reason for this: digital imagery has been been tweaked, from early on, to work on our known perceptions and preferences of color. The data making up the image is not devoid of our understanding of what we see and how we react to it. It would be a good bit different if we relied on recording light waves themselves to handle digital imagery (which is basically how we do sound--we take a great deal of what is there, how it is there, not just what we perceive being there).

Quote:

I think that SET amplifiers and vinyl are higher-resolution than other types of ampliers and digital, respectively. I don't think there are any measurements that explain that.

In technical terms, there is evidence that you are wrong. That's where making useful measurements comes in. Regardless of technical merit of technology, you are perceiving more musical something coming from that kind of setup. Measuring that some specific waveform (that is not a reproduction of your perception of the music, which would be the ideal baseline) is more accurately reproduced on system X than system Y does not a measurement of whether you will hear more from the playback of some musical recoding or not. It may measure that is greater potential to, but about as far as that can go, without having measurements based on what you actually want to know.

Quote:

You will have to give me a better context for "knowing" because it's such a broad topic (epistemology).

OK: Quote:

We know the world through models, and reality is always richer than the models.

How do you know that reality is richer than the models, if we know reality through the models. If we know it through the models, then reality is no richer than the model, as the model is how we see reality. In that case, how can you even be sure that reality is richer than the model? Or, is there something else about, "know."

 

[mike1127]

Hi Cerbie,

Okay, first of all I think we disagree in our philosophical ground. I think that people make certain deep choices about how the world works... choices that come before we can gather evidence. You and I just disagree. That's fine. I will just simply articulate my disagreement but recognize you could be right.

Quote:

Originally Posted by cerbie /img/forum/go_quote.gif Direct observation would be where my difficulty in figuring it out would be. Objective (what behavior of the universe we can all agree on with great confidence) does not have a great dividing wall from the subjective (what remains purely subjective as a set of our experiences gets smaller all the time). When it comes to measurement, we came to be able to reach these objective conclusions through consistent application of our subjective faculties. We are each every bit a set of instruments for measurement. As it concerns our perceptions, especially aesthetics, we are the most useful tools for measurement at this time. In as much, many aspects of own workings are as much objective as subjective, and the subjective can be used to reach sufficient confidence as to be used in objective contexts.

I agree with you that we should not pre-impose limits on our knowledge. Perhaps, someday we will gather direct evidence about what's going on "in our heads." We will understand consciousness better---there's plenty of research on that.

Yet some philosophers speak of "qualia"---the quality of an experience in the first person. I think it's important to acknowledge how far removed from objective evidence this is. Currently, a brain FMRI cannot explain how I experience beauty, or whether two people experience beauty in the same way.

And yet some people act as if we are fairly close, or as if that brain FMRI is very useful. I will not try not to rigidly hold to the view that its NOT useful, but I certainly don't think it's very useful.

I practice mindfulness meditation and have some very good teachers. You may be aware there have been studies on the brains of meditators. From my view, these studies are primitive and filled with misconceptions about meditation. And yet some researchers have declared that certain meditative experiences correlate with certain brain states. This is premature in my view.


Quote:

If you think X sounds good, and Y sounds bad, and 15 out of 20 other people agree with you, that's a pretty good start. Finding specific dimensions that we can agree on, without the use of controlled survey is difficult. But, while using a non-human tool may be best, to remove bias, it can also remove any usefulness of the measurement. The measurement criteria must be based on what is being measured for. Saying something will measure bad if it sounds bad, and finding out the kind of objective measurement that will show that, are not the same thing.

I don't quite follow you, but refer again to my color example. The problem is finding a way of organizing or processing the available information to produce numbers that correlate with subjective experience. There's no lack of information.

As I said, I'm not a recording engineer. I've communicated with a few. Some believe that measurements are good predictors of the accuracy of recorders, amplifiers, speakers, microphones, etc. Others believe we have no measurements that usefully correspond with subjective perception of accuracy. The latter engineers produce the most accurate recordings, from my view. So I side with them.

Quote:

It is not too difficult to figure out if one digital color image is more colorful than the next, especially if they are using similar hues. With those kinds of images, you could take a difference from grayscale, add up what's left, and compare. With different colors, you'd need to convert to a system like HSV, weigh general directions (in S) of H (much less easy, but doable). Then, compare with impressions.

I certainly think we could try to do it. But the situation is analogous to this. Suppose we come up with a bunch of measurements to gauge the accuracy of prints (ink) reproductions of paintings. And suppose that I view some of these reproductions which are supposed to be accurate, and they are not. They simply are off the mark, even though the measurements say they should be.

That's the situation in audio, to me.

This may seem surprising, but don't forget measurements are like viewing the world through a peephole. An amplifier is not an object, it's a behavior of great complexity. Measurements take a little peek at that behavior in a few limited circumstances.


Quote:

[regarding SET amps and vinyl] In technical terms, there is evidence that you are wrong. That's where making useful measurements comes in. Regardless of technical merit of technology, you are perceiving more musical something coming from that kind of setup.

In my view, the measurements that predict more accuracy from transistor amps and digital recorders are just wrong. They sound distorted to me, therefore they are distorted. We are measuring the wrong thing.


Quote:

OK:How do you know that reality is richer than the models, if we know reality through the models. If we know it through the models, then reality is no richer than the model, as the model is how we see reality. In that case, how can you even be sure that reality is richer than the model? Or, is there something else about, "know."

This is very interesting. The main evidence is that our models keep evolving and changing. A friend of mine wrote this:

Quote:

Originally Posted by friend Another of my favorite examples comes from Thomas Kuhn, who points out that many 18th century physicists were against the study of (static) electricity because everyone "knew" that objects could interact only by being in physical contact so the idea of attractive and repellant forces acting at a distance was obviously superstitious nonsense.

And my view is that when scientists create the models, they are aware they are making assumptions.

Furthermore, the view that reality is richer than our models is a kind of philosophical ground. It's a choice that is made before gathering evidence and influences how one gathers and interprets evidence. I can't really justify it. I believe it's a choice we all make and there's no way to say one choice is better than the other, except through intuition and experience.

For example, if your personal experience is that transistor amps and digital recorders are very accurate, then for you the measurements are fine.

 

[mike1127]

Quote:

Originally Posted by royalcrown /img/forum/go_quote.gif You're right in that they do open several possibilities. However, I feel that this distinction, while philosophically interesting, is largely academic in nature. The empirical data may be incomplete, but they do seem to work. To use your example, while our predictions on planetary orbits were off, for many applications they did, and still do, work. I think the more practical question is: as far as our application is concerned (that is, within the realm of audio playback for headphones), do our models work? The answer to that question can only be found in empirical data, and in that regard we have blind testing to help us, at which point we're back to square one.

Whether the models work depends on the field of inquiry. For planetary orbits, they work well enough to get spacecraft to the right place on Mars, etc. My view of aesthetics, to the contrary, is that our measurements are poor or totally wrong predictors of accuracy. Remember that measurements are just a tiny peek at the behavior of a system. Only in listening to that system do we perceive the gestalt.

Now I recognize that other people, with other ways of perceiving musical information, find the measurements work for them.

An interesting question is: do I like SET amps and vinyl because of the distortion, rather than because of a technical accuracy, and if so, should that be called euphonic distortion?

I have an idea, which is actually purely theoretical at this time, but interesting to me (at least):

http://www.head-fi.org/forums/f133/u...ortion-430573/

 

[coldkeith]

As an aside, I think the vantage point of treating measurements as models asks the question: do our ears merely model data in the same way that measurements do? I would say the answer is yes, if our ears are as inaccurate as our eyes are. A good display of this comes (quite surprisingly) from this image:

http://www.thebestpageintheuniverse....lon_fraud3.gif

Our eyes cannot resolve a difference between the two colors, and yet our equipment can indeed resolve a difference. If the equipment is merely modeling the data and is incomplete to some degree, our eyes are also modeling the data and are incomplete to a larger degree.



I am a painter and can easily tell the difference between these two colors.

 

[royalcrown]

Quote:

Originally Posted by mike1127 /img/forum/go_quote.gif Whether the models work depends on the field of inquiry. For planetary orbits, they work well enough to get spacecraft to the right place on Mars, etc. My view of aesthetics, to the contrary, is that our measurements are poor or totally wrong predictors of accuracy. Remember that measurements are just a tiny peek at the behavior of a system. Only in listening to that system do we perceive the gestalt.

I think you're missing the purpose of measurements. The measurements don't exist to show whether or not something sounds good, or is perceived as accurate. Audio measurements do two things that are applicable to us:

1. They tell you how accurate the reproduction of a system is with relation to the original source (CD, etc).
2. They tell you whether or not a difference is appreciable.

For example, the measurements may tell you that two components will sound the same, regardless of how bad or good the two components may be. Think of two amplifiers, that are for the sake of argument exactly equal except for total harmonic distortion. Say component one has 33.0000000000000000000000000000% distortion, and one has 33.0000000000000000000000000003%. Regardless of whether or not the measurements will tell you, subjectively, how "good" or "bad" either one is, the measurements will tell you that the two will be audibly equivalent. Now imagine the same thing, but the difference in THD is between 33% for amp 1 and 3% for amp 2. Again, the measurements won't tell you which one someone will prefer (although in this exaggerated of a situation it's trivial), but they will tell you that the two will be easily distinguishable from each other.

Quote:

Originally Posted by mike1127 /img/forum/go_quote.gif Now I recognize that other people, with other ways of perceiving musical information, find the measurements work for them. An interesting question is: do I like SET amps and vinyl because of the distortion, rather than because of a technical accuracy, and if so, should that be called euphonic distortion? I have an idea, which is actually purely theoretical at this time, but interesting to me (at least): http://www.head-fi.org/forums/f133/u...ortion-430573/

To be perfectly honest, I read your thread and didn't really understand your first abstraction at all. That being said, with regard to the visual analogy, no doubt some algorithms make the text more visually appealing, I fail to see how that changes the situation at all. I don't think anyone will argue that some edge enhancement algorithms are visually pleasing, but that doesn't make them more accurate.

I think a similar analogy would be taking some famous portrait (I usually use the Mona Lisa because it's easily recognizable) and adding some photoshop filters to them. For sure, some filters will ruin the picture, but I think some touch ups, if properly applied, will make the Mona Lisa look better to most, especially if doing so removes any inconsistencies in brush strokes, etc. That being said, I think a lot of art aficionados would prefer the original than the touched up one. Neither are wrong in picking which they prefer, but even if the touch ups made the Mona Lisa look stunningly beautiful, or hell, even reveal some otherwise obscured details in the background that are otherwise unnoticeable, we wouldn't call the touched up Mona Lisa more accurate than the original. In fact, it would seem logically inconsistent, and possibly contradictory, to call anything edited "more accurate" with respect to the original than the original itself.

Quote:

Originally Posted by mike1127 /img/forum/go_quote.gif Our ears and brains do not model data at all the way measurements do. Ears and brains model things like: "what instruments are present, what is the tone color of each, what key is this in, etc.". Typically audio measurements are something else entirely. ... I'm not an audio engineer, but of those I know, many feel our models of audio do not "work" at all, if by "work" you mean predict what will sound high-resolution, accurate to the source, or aesthetically pleasing.

See above. The whole purpose of audio measurements is not to find out what's most aesthetically pleasing, or which is perceived by some to be accurate - it tells you what is objectively more accurate, though not objectively more aesthetically pleasing, because no such thing exists.


Quote:

Originally Posted by mike1127 /img/forum/go_quote.gif Regarding that revlon color example, at least on my monitor I think I can see a difference. This is interesting because it disappears when I try to do "quick-switch." If I rest my attention in the right color patch for a while, until I'm "used to the feel of it," then switch to the left color patch, the left one is distinctly lighter. If I rest there, and switch back, the right one "feels" darker. If, on the other hand, I switch back and forth rapidly, all differences become uncertain and largely vanish. That doesn't disprove their hypothesis about Revlon lipstick because the color values in the images are approximations.

Quote:

Originally Posted by coldkeith /img/forum/go_quote.gif I am a painter and can easily tell the difference between these two colors.

I think you guys are proving maddox's point for him: The human eye cannot distinguish between the two shades of color. This isn't a matter of experience or training, this is a biological limitation of the eye itself. While changes of delta E can be as large as 10 while still being acceptable to most people, it's largely accepted in the literature that differences of less than delta 4 cannot be detected.