without going into quantum, the real world "patterns" such as the surface of clay is the exact. i am talking about a digital representation, that won't be exact. how can you say the original isn't the exact, without talking quantum mechanics?

hellokeith, that sounds great! would we all need our own hrtfs to be calculated in order to use it though? i don't think the generic ones like dolby headphone over the accuracy or perfection people here are after.

This is false. All analog signals that can physically exist have finite resolution. See my previous post where I explained this.



Quantum effects are relevent. They're actually a limiting factor in many 24 bit audio systems. Certainly you would agree that 24 is less then infinity right? So why limit yourself to lower resolution systems where things are purely classical?



Without addressing your silly QM qualification on that statement, I'll just point out that most techno music is originates in digital form. Therefore it is "exact", regardless of QM (and the analog form is never exact, again regardless of QM).

Just answer this, a block of clay, with a contoured top, in order to graph that from a profile for example, for something to follow that contour would take an infinite amount of resolution... or at least down to the atom. and that is much smaller than anything you can store with bits...

theres a reason if you ran a sound simulation through a non atomic modeled environment you get different results with everything else being equal if you somehow look at more macroscopic behaviour. I know that well

Really? Could you please post some evidence in the kind of TOS 8?

First, please attempt to define the terms you're using.

Then show some evidence.

Finally, to use your example of a lump of clay, in fact there is uncertainty in the location of every atom in it. Sorry to burst your bubble.

This is of course nonsense, you can extend bitdepth to whatever is necessary. I think Woodinville and Mike Giacomelli already explained to you why your assertions are wrong, but all you say is "No. They are not!", and then continue with giving flawed examples.

Infinite resolution simply does not exist, because matter and energy have finite resolutions. There is a boundary where you cannot devide either one anymore.

It is very often very convenient to treat the universe as analog, but this does not make it right in itself. But then neither was Newton, and I'm sure most people were taught classical mechanics in high-school (it just works for a lot of things).

obviously it is not infinite but it is extremely large to anything we can hope to store at the moment in digital.

We are taught classical mechanics until only the last years of college, they still govern the macroscopic behaviours well.

Of course the universe is comprised of things that, upon deep inspection, require infinite (or as nearly infinite as quanta dictate) amounts of data to describe. This is not news. If one wishes to get philosophical, you could estimate that the dataset that describes the universe is itself as large and complex as the universe, yada yada yada.

We don't spend our time worrying about such matters because, in a nutshell, it would be bloody expensive to describe objects that way and would yield no practical value.

If we are back to talking about signals, can we say that digital representations are "less complete" than analog? No way. The discrete nature of a digital description limits the data granularity, but so do the distortions and noise issues of analog. You simply trade the shortcomings of one representation for another. Provided that both representations are designed to achieve the same level of fidelity, then this tradeoff is all that is occuring. Note that this argument is separate from perception, for I can define "fidelity" in terms of reference to some original source signal.

Once you add human perception, the problem degrades further and exact representations of signals become less and less important. At that point, worrying about data granularity is like worrying about 1 bottle of beer being spilled into the ocean - no fish will get drunk, I assure you!

Really?

We're talking about sound, yes?

We can (I have, personally) measured the noise level of the atmosphere with a digital instrument. It is very near the threshold of hearing. It is not "extremely large" or "extremely small" compared to what we conventionally store in digital audio systems, it is right there at the bottom end of a standard recording.

Please provide your counterevidence.

P.S. I think you also greatly mistake what can be done with your lump of clay, but since we're talking about sound, let's stick to the subject at hand. Please show your counterevidence, along with ways to replicate and reproduce the evidence. Tests done in a helium atmosphere, or in outer space, need not apply.

Hello all,

I would like to go back and try to address the original poster's question. The OP posed 2 examples that illustrate his question (the AM/FM/CD example, and the 3 pictures example) so I was thinking I would address each of them in turn. But first, let me just throw out some of my thoughts on the matter.

While thinking about the two examples and about the nature of the question being posed, I was trying to find some unifying principle that connected them, and so far I am thinking that it all boils down to the concept of entropy. It seems to me that when we compare things and try to make a judgement on which is "better" than the other, we may be in fact making some unconcious comparison about the relative amounts of entropy of the two things. I also have come to the conclusion that we may have some built in preference for lower entropy, which is where our gut feeling of "this one is better than that one" comes from. Lets look at the two posed example in more detail.

First, the example of AM vs FM vs CD...I could be wrong, but I am under the impression that when we analyze the signals from the different sources we will find that at least one difference will be in the amount of noise in the signal, with AM>FM>CD. Another difference may be in the range of frequencies transmitted in the different formats. Noise is a signal with very high entropy, because of its random nature (random frequencies at random amplitudes); if I listen to a 1 second clip of noise, and compare it to the next 1 second of noise, they will sound "the same" to me, even though the waveform of the signal could be different (different frequencies at different times, but still random). This is opposed to the actual musical signal, where there is much lower entropy (certain frequencies are played at a higher amplitude much more frequently than others, and have certain patterns in time). So, I believe that even if you were deaf from birth, and then some miracle happend that allowd you to hear for the first time, and the first things you heard was a song played on FM and the same song played on cd, your brain would detect the lower entropy of the cd recording and conclude that it prefers it over the FM signal.

Now for the three pictures example...Same idea. If you look at the 16 color picture, you will see relatively large areas of the picture that is all one color. If we now compare that same area of the photo to the 32 bit photo, we will see that instead of it all being a single color, it is now, say, 10 different colors. The 32 bit photo has less entropy than the 16 color photo. One way entropy can be defined is the number of ways of arranging a bunch of things and leave the collection of things unchanged. For instance, in the 16 color photo we could take any of the ink particles in the area that is a single color, and transpose it with any other ink particle in the same area, and leave the photo as a whole unchanged. This is higher entropy than the 32 bit image because in the 32 bit image there is much less area to rearrange an ink particle of a given color and leave the photo unchanged. Hence, the 32 bit image is much lower entropy than the 16 color photo. I believe that if you were blind from birth and were then miraculously able to see only these 3 pictures, you would still pick out the 32 bit picture as "better", even though you would have no idea what you were actually looking at, because you have never seen a picture of the earth or the space station, or anything else before.

I think that the determination of what is better than the other also has to do with a topic related to entropy...the amount of "information density" presented to our brain. In AM/CD example, the higher the signal to noise ratio, the more "information density" the signal contains. All we can say about noise is that we are hearing random frequencies at random amplitudes (with in the range that the noise exists). However, the music signal contains much more information (patterns in frequency and amplitude that our brain processes). I believe this is simply another way of talking about entropy though...

So, I believe the answer to the question "What is the science behind what we perceive as better or best?" would be to use the scientific principle of entropy and quantify the relative amounts for the samples we are comparing. The lower one wins .

Of course I could be way off, and I'm sure that there are many counter examples that I have not considered. Anyway, thats my 2 cents (ok, maybe 5 cents )

From my observations high frequency components of music usually are far less likely to follow a simple pattern than lower frequency components. If that means more entropy, why not apply a low pass to reduce the entropy? Would one still prefer this less entropic but also hollow signal?

A high complexity is not the same as high entropy. Simply put entropy is just a measure of disorder, so one should ask if removing an ordered (although complex) pattern will really reduce the entropy of a signal. But for that, one would need a means to measure the entropy of the given signal.

I think the idea is nice enough to ponder about.

EDIT: Perhaps a low pass should be seen as decreasing the resolution of the audio and therefor as an increment of entropy, similarly as to why a 4 bit image has higher entropy than a 32 bit image.

That's why i wrote "If that means more entropy...". I should have added "...for you." Possibly i misunderstood yodalige. If so, sorry, my english is quite limited.

Might I suggest that entropy is not a very good measure of what is perceptually pleasing?

A ride cymbal is pleasing to most people, it has high entropy. (Spectral Flatness Measure of 20dB max)

A flute is pleasing to most people, it has remarkably low entropy. (Spectral Flatness Measure of 70dB)

Well, the elephant's been known about for decades -- at least as far back as the early days of 'talking pictures', when 3-channel playback systems were determined to be better than 2-channel. Two channel was adopted for home use only because it was thought that people would not tolerate three loudspeakers in their living rooms. (Someone should do a study of Spousal Acceptance Factor re: home audio over the years...)

Back as far as Fletcher and Snow, I dare say.

Sorry to resurrect a very old topic, but I started it, so..

Woodinville,

Do you agree that, all other things being equal, people will unanimously choose/prefer the CD playback over AM radio playback of the same music passage?

No, I won't. It depends on where you're listening.

In a home setting, yeah, most likely. But what's your point? Entropy isn't what you're measuring here.

If you say most likely, then what is your scientific basis? I agree with you, but I want to know what is the metric that makes it so.

The reactions of actual people. Now, it would appear that you are wanting to be combative or argumentitive, or something, and all that, so ciao, dude.

The answer to "better" is a subjective test. There is no "best". Bye.

So we have observable data. Do we have a theory as to why a majority or unanimously people choose the CD over the AM radio playback?


Not at all. I am simply on a quest for knowledge, and I believe you to be one of the most experienced and knowledgeable people in the audio field, hence I am asking you the questions.

Yeah, that didn't seem argumentative to me.

We cannot establish better/best, because the definition of "better" and "best" is constantly changing.

For example, what if there was a comprehensive study 5 years ago consisting of DBX testing that confirmed that audio data stored in a lossless format sounded no better than the same data stored in MP3 format (with encoding option set X) when played back on a home computer? One could conclude from such a study that it makes sense to use MP3 files forever, and the basic facts of the study (lossless is not "better" than MP3) will never change.

But, in reality, the study doesn't tell us much about how these things compare in 2008. Five years ago most people were using soundcards that did a terrible job of resampling 44K data to 48K (actually, a lot of people are still using sound cards that do this) and tiny computer speakers...so, now, very few people would care what people thought about which encoding format was "better" with that kind of setup. What would any set of generic results mean to someone who has spent a lot of time creating a setup with "professional" quality audio components and software?

Generalizations about people and their abilities (which are often the basis of statements containing the words "better" and "best") are only meaningful for a generalized set of people at a specific time.

My apologies, then.

The problem is that "best" is personal preference. Everyone has a few. Everyone's is theirs, they don't have to justify it, and they can't generalize it.

Now "transparent" or "accurate" can be dealt with via things like ABC/hr tests. You can find out how different something is. But we assume that "accurate" only has one perceptual dimension when we do that, and this is not really how the world or the human works. Still, it's pretty repeatable.

For "preference" you CAN do something like a Mean Opinion Score test, but especially with preference, one finds out that the "scale" is a bunch of scales, and everyone has different likes and dislikes, and you wind up wth a great mess of data, signifying not much at all.

Best is personal, per-individual, or even more diverse than that.

For instance, I can listen to "Sympathy for the Devil" in an old Peavy guitar amp, and it sounds "right".

But when I want to hear Murray playing the organ at St. John the Divine, sorry, my C-5's are hardly good enough.

It all depends.

Oh, and playing right now: Sweet Avalon, played by Chet Atkins, Les Paul, and Grandpa Jones. (yes, really)

It's a 1950's recording. The word "bandwidth" hadn't been quite invented yet ( ) but it sounds GREAT. For what it is.




#include "blankety_blank_resamplers_rant.h"

So at most we can say is that "a typical majority of listeners will prefer CD over AM radio, given the same source and playback environment", or is that even going to far?

Is there any research you can point me to that would help to explain that typical majority, if it exists?

What do you mean "typical majority"? IT's not even that simple.

Is my whole premise wrong? I guess I am trying to discover if "better/best" ever hinges solely or even partially on engineering specifications.

Take for example the room correction available in Vista. Is the goal of the room correction to create stable imaging of the soundfield, or is the goal to make the audio presentation sound better? If the room correction does end up making things sound better, surely that was intended? Otherwise, how would a scientist/engineer sell the idea to the bean counters in the first place?

(I hope it doesn't sound like I'm trying to split hairs here.)

For engineering purposes and delivering what was intended, engineering specs are entirely relevant.

For what a given individual would LIKE (as oppose to regard as accurate) in a given environment, sometimes yes, sometimes no.

The room correction offers you the option of interchannel matching (not good microphone) or match-to-flat (good microphone).

Match to flat is NOT the best interchannel match. The two problems are quite different, and the same solution does not fit both.

There are other issues in room correction, if you'll see my ppt decks over at www.aes.org/sections/pnw/ppt.htm you'll see some of the other considerations.

After all, lots of people like the distortions in LP's. They get to prefer that. It's all preference at some level.