"Things that cannot be measured"

[BluesDaddy]

I must admit "accurately reproducing the input signal" would seem to be the only valid definition of high fidelity.

Certainly the only truly relevant one to the home consumer. Until Mr. audio2desgin brings to market his holographic recording system that perfectly recreates the actual music production we'll have to settle for it.

 

[BluesDaddy]

Maybe you're right, but ask yourself if you are not sometimes commenting something that wasn't explicitly written ?
Enough futile arguing.

Or perhaps you miss the implications of the things you assert.

 

[audio2design]

There are fundamental differences between images and audio. For images, the "equivalent" to time resolution of audio is spatial resolution. Spatial resolution always matters for images because you can always zoom in (or magnify) to get the benefits of higher resolution images (or notice the insufficient resolution). You can't do that with audio. It doesn't make sense in music preproduction to "zoom-in", i.e. by time dilation. There is no benefit for resolution higher than what we have in audio recordings made (properly) in the last 40+ years. There is no missing details that we can perceive or benefit from.

You are taking a too simplistic view. It is too early to know where this can go. For example, what could be done w.r.t. extracting timing information in a highly intelligent fashion and then directing the sound to more than two speakers so that the full breadth of timing information embedded in a recording could be appreciated?

 

[StevenEleven]

People probably said that about what is now regularly done with AI and image processing.

You are probably applying a rather simple view of what information is in a recording. If I can identify the instruments used in the recording, I can create more robust models for what is likely noise or not, and then replace it with an intelligent extrapolation based on the instruments playing. It is way too early to know where this is going to go.

In my very unimportant opinion you raise a very interesting discussion and illustrate how this place may become a bit too much of a simple-minded shark tank that even eats its own at tImes.

I guess my next thought would be can you measure the fidelity that is added (which I do agree is a valid concept that is going to become reality, I will go to the trenches with you on that one, as an amateur I have used both pro audio and pro photography software to improve what I believe to be the fidelity of signals)? My guess is that you will answer yes, this added fidelity could easily be measured.

 

[NTK]

You are taking a too simplistic view. It is too early to know where this can go. For example, what could be done w.r.t. extracting timing information in a highly intelligent fashion and then directing the sound to more than two speakers so that the full breadth of timing information embedded in a recording could be appreciated?

You've got anything to backup your theories?

 

[audio2design]

Because my friend you are wrong, as our members have tried to tell you over and over.

No, just some members who I consider to have simplistic views, and not very forward looking views, based on what they wrote. That could be the result of purely looking at signals and not actually spending a significant amount of time doing actual research in psychoacoustic fields.

Recently some of those similar members lost their minds when I said all cables are inherently directional due to mfg flaws, whether this was audible or not, and it was due to the signals being AC. (I never claimed, and even distinctly said it would not be audible) I was insulted, made fun of, and had a lot of really bad use of engineering thrown at me including misuse of superposition and misuse of Lorentz Reciprocity, even by senior members. My statement was absolutely correct, but technical people can be as dogmatic if not more than not technical people.

While prices for similar quality are coming down, audio overall is pretty stagnant. View such as this are not going to change that.

 

[StevenEleven]

There are fundamental differences between images and audio. For images, the "equivalent" to time resolution of audio is spatial resolution. Spatial resolution always matters for images because you can always zoom in (or magnify) to get the benefits of higher resolution images (or notice the insufficient resolution). You can't do that with audio. It doesn't make sense in music preproduction to "zoom-in", i.e. by time dilation. There is no benefit for resolution higher than what we have in audio recordings made (properly) in the last 40+ years. There is no missing details that we can perceive or benefit from.

But you can take poorly recorded instruments and improve the fidelity of those instruments in the recording, and there is no doubt in my mind that AI will play a very important role in just that. You are free to disagree of course, I get what you are saying. Same with excess noise, or pops and cracks, etc., in a recording. Those can be viewed as deviations from fidelity, in my view, and will in the long run be addressed by more and more powerful AI. I get what you are saying and I think in the end audio will take a broader perspective.

 

[BluesDaddy]

You are taking a too simplistic view. It is too early to know where this can go. For example, what could be done w.r.t. extracting timing information in a highly intelligent fashion and then directing the sound to more than two speakers so that the full breadth of timing information embedded in a recording could be appreciated?

I agree this is an interesting concept, but your model would also have to include the characteristics of the particular instrument used, not simply in generic, but the exact instrument used in the recording. You'd also have to have all the characteristic practices of the recording and mastering engineers to know what they tend to inject into the recording. This seems to be way beyond the amount of information even in a blu-ray today and could not be merely hardware centric, it would be more like Atmos (or some similar multi-channel, objective oriented sound system) for music only more so. Of course, high fidelity would then be reproducing all that information as transparently as possible, would it not?

 

[NTK]

But you can take poorly recorded instruments and improve the fidelity of those instruments in the recording, and there is no doubt in my mind that AI will play a very important role in just that. You are free to disagree of course, I get what you are saying. Same with excess noise, or pops and cracks, etc., in a recording. Those can be viewed as deviations from fidelity, in my view, and will in the long run be addressed by more and more powerful AI. I get what you are saying and I think in the end audio will take a broader perspective.

I am sure one day AI can have Frank Sinatra doing hip-hop. But that's not music reproduction, which is what we generally talk about here.

 

[StevenEleven]

I am sure one day AI can have Frank Sinatra doing hip-hop. But that's not music reproduction, which is what we generally talk about here.

Music reproduction in real time will be addressed by AI to improve fidelity, IMHO. Read this in five years and maybe we’ll have this talk again.

The improvements in chess and photography software in the last two years by way of AI have been staggering. I cannot conceive that there is any roadblock to applying the same power in audio. That’s how it looks to me. Of course I could be wrong, but that’s how I see things based on my experiences with AI, neural networks, etc. The acceleration of the technology is absolutely staggering.

 

[Raindog123]

People probably said that about what is now regularly done with AI and image processing

You cat train modern AI (eg, a neural net) on a bunch of faces, so the AI can compose a realistic-looking face following the trained pattern, or fill in missing pixels... But if I delete my mustache from a photo, it will never guess and reconstruct its shape or color. The same with audio - there can be (are?) some pattern-based [reconstruction] filters, and the AI now can even compose music pieces - again by identifying and following patterns... but it cannot recreate missing uncorrelated, non-redundant sound information.

 

[NTK]

You cat train modern AI (eg, a neural net) on a bunch of faces, so the AI can compose a realistic-looking face following the trained pattern, or fill in missing pixels... But if I delete my mustache from a photo, it will never guess and reconstruct its shape or color. The same with audio - there can be (are?) some pattern-based [reconstruction] filters, and the AI now can even compose music pieces - again by identifying and following patterns... but it cannot recreate missing uncorrelated, non-redundant sound information.

AI can make "educated" guesses.

 

[Raindog123]

Recently some of those similar members lost their minds when I said all cables are inherently directional due to mfg flaws, whether this was audible or not, and it was due to the signals being AC. (I never claimed, and even distinctly said it would not be audible) I was insulted, made fun of, and had a lot of really bad use of engineering thrown at me including misuse of superposition and misuse of Lorentz Reciprocity, even by senior members. My statement was absolutely correct, but technical people can be as dogmatic if not more than not technical people.

No it was not. I was there (remember, the one misusing the Reciprocity principle). It’s a dead horse, but the folks let you off. With a statement equivalent to “a ‘pint of beer’ can't be called so because it’s never exactly a pint”. [Nicer] folks pointed out nicely that your model of a cable with asymmetric distribution of resistance across it is not/might not be the ‘practical, subject-of-the-initial statement“ truth. Just like ”a pint“. I was wrong about ‘generalizing’ the Reciprocity principle, and I openly admitted it, but I did not see anyone ‘losing their minds’. The original opponent of yours was probably pointing out that non-reciprocal effects (like diodes) in a practical symmetric cable that would get affected by the DC current flow, are not an issue for AC. But you chose to escalate (to the ‘smart Alec’ level, remember?) I wonder if there is a ‘double smart Alec‘ out there?

 

[BluesDaddy]

AI can make "educated" guesses.

That's cool, but a "guess", even an "educated one", would not be "high fidelity". It might be extremely engaging, on the whole people might enjoy it far more than faithfulness to the original, but that's sort of like a sex doll isn't it? In any case, it would still not be "high fidelity".

 

[Wes]

Given that I have actually worked in the audio "sciences", who are you to state that I am in the wrong place. AUDIO is not an electrical signal or bit in a data stream. If recreation of an electrical signal, data stream is all this website is about, then it is incorrectly named. That is but one portion of audio.

what sort of work have you done in the audio "sciences"??

I don't understand your defn. of HiFi as something other than accurate reproduction. Granted, you might not want hiFi and prefer euphonic distortions - I like the even order harmonics myself. But those are departures from fidelity.

 

[Raindog123]

AI can make "educated" guesses.

Sure! But this might come handy for my next argument.

”Realism” - yes. “Fidelity to the source” - a different story...

 

[Wes]

Let's hear more about those sex dolls...

 

[Raindog123]

I agree this is an interesting concept, but your model would also have to include the characteristics of the particular instrument used, not simply in generic, but the exact instrument used in the recording. You'd also have to have all the characteristic practices of the recording and mastering engineers to know what they tend to inject into the recording

We can take it - the concept - even further. At some distant future, thanks to progress of the AI, everyone will have a personal copy of “pocket Genesis” (Peter Gabriel‘s Genesis that is), “pocket Pink Floyd” (don’t get me started, whose Pink Floyd), “pocket Beatles” (maybe we can settle on just George)... Upon demand, we would summon those to play a tune-or-two for us. And it will be undeniably, unquestionably “true to the source” high-fidelity experience... Someday.

 

[maverickronin]

Let's hear more about those sex dolls...

So...one time at this insurance office where I worked in IT...

This guy came in for a meeting with his agent, and brought along his RealDoll, fully dressed, in a wheelchair.

 

[David Harper]

maybe instead of reproducing accurately the input signal we're talking about reproducing realistically the sound of the original performance? Would these two be different? I don't know. But it seems to me that a Sony walkman might accurately reproduce the input signal. But none of us would consider that high fidelity.