[gregorio]

To answer your question, recording digital audio does require some form of quantization.
[1] But being able to record a signal does not imply that every form of measurement is known.
[2] I am saying that the answer you often give, akin to "if we can record it, we can measure everything" is not true.

1. Agreed but I am not implying we can measure and record everything.
2. No, that's not what I'm saying. What I'm saying is the other way around, that we can only record what we can measure, if there is something that we can't measure, then we can't record (or reproduce it). Digital audio data is just a bunch of zeros and ones which represent the amplitude "quantities" of an electrical current at specific points in time. That's it, nothing else. If there is something other than these measured "quantities", then it cannot exist in a digital audio recording.

G

 

[SoundAndMotion]

Hi G,
OK, the other way around:

What I'm saying is the other way around, that we can only record what we can measure, if there is something that we can't measure, then we can't record (or reproduce it).

You cannot measure something if you don't know what it is, and you cannot measure something if you don't know how to measure it. But you can still record it. How does the lack of knowledge prevent recording?
Cheers, SAM

 

[Whazzzup]

Are we not skirting, if a tree falls in the forest conundrum?

 

[gregorio]

You cannot measure something if you don't know what it is, and you cannot measure something if you don't know how to measure it. But you can still record it. How does the lack of knowledge prevent recording?

But we do know what it is, it's an electrical current. That's the only thing we can digitise (in digital audio) whatever else there is, whether we can measure it or not, we cannot record or reproduce.

G

 

[Phronesis]

There are an infinite of ways to measure a physical object, and the measurements we devise are generally based on our models of those objects. Those models can evolve over time, even though the objects may stay the same.

 

[gregorio]

[1] There are an infinite of ways to measure a physical object
[2] Those models can evolve over time, even though the objects may stay the same.

1. That maybe true but the recording of digital audio allows for ONLY ONE measurement, the quantification (measurement) of amplitude over time.
2. Some models can evolve over time but this one (digital audio) has not and can not! Digital audio is defined entirely by that one measurement and has been since the day it was proposed nearly a century ago, the "model" has not changed or evolved in the slightest and if one day it does, then by definition it will be some new technology and no longer be digital audio. The technology to implement the digital audio "model" has certainly evolved but not the model itself. So, like it or not, everything we can record (and therefore you can reproduce) is defined SOLELY and ENTIRELY by that one measurement!

G

 

[Phronesis]

1. That maybe true but the recording of digital audio allows for ONLY ONE measurement, the quantification (measurement) of amplitude over time.
2. Some models can evolve over time but this one (digital audio) has not and can not! Digital audio is defined entirely by that one measurement and has been since the day it was proposed nearly a century ago, the "model" has not changed or evolved in the slightest and if one day it does, then by definition it will be some new technology and no longer be digital audio. The technology to implement the digital audio "model" has certainly evolved but not the model itself. So, like it or not, everything we can record (and therefore you can reproduce) is defined SOLELY and ENTIRELY by that one measurement!

G

I agree that a digital signal can be fully described as the amplitude versus time. However, that's the signal, not it's measurement. There are still an infinite of ways to measure that digital signal, and if there's more than one way to convert that digital signal to analog (apparently there are), then you're back to all of the issues with how to measure, characterize, and compare analog signals.

 

[gregorio]

[1] I agree that a digital signal can be fully described as the amplitude versus time. However, that's the signal, not it's measurement.
[2] There are still an infinite of ways to measure that digital signal, and if there's more than one way to convert that digital signal to analog (apparently there are), then you're back to all of the issues with how to measure, characterize, and compare analog signals.

1. No, digital audio is just a series of zeros and ones, the data in those zeros and ones (bits) contains ONLY the measurement of amplitude, nothing else! If you assert otherwise, please provide some evidence. From that ONE measurement (quantised amplitude data) we can derive all kinds of other information that we can measure in numerous different ways. For example, we can process that ONE MEASUREMENT with a Fourier transform and derive all the frequency information but EVERYTHING we can measure (or not measure) is derived from that one single measurement of amplitude.
2. No, there are NO OTHER WAYS to convert digital audio! If there were, then by definition it wouldn't be digital audio!

G

 

[Phronesis]

1. No, digital audio is just a series of zeros and ones, the data in those zeros and ones (bits) contains ONLY the measurement of amplitude, nothing else! If you assert otherwise, please provide some evidence. From that ONE measurement (quantised amplitude data) we can derive all kinds of other information that we can measure in numerous different ways. For example, we can process that ONE MEASUREMENT with a Fourier transform and derive all the frequency information but EVERYTHING we can measure (or not measure) is derived from that one single measurement of amplitude.
2. No, there are NO OTHER WAYS to convert digital audio! If there were, then by definition it wouldn't be digital audio!

G

I think you're continuing to confuse the signal with its measurement. A typical digital file of music contains a huge amount of data, correct? The series of 1s and 0s in that file isn't a "measurement" similar to a distortion level, S/N ratio, etc.

 

[gregorio]

A typical digital file of music contains a huge amount of data, correct?

No, it contains a relatively small amount of data! Just 16 or 24bits representing ONLY the measurement of amplitude, taken typically 44,100 times a second. That's it, NOTHING ELSE!

G

 

[Phronesis]

No, it contains a relatively small amount of data! Just 16 or 24bits representing ONLY the measurement of amplitude, taken typically 44,100 times a second. That's it, NOTHING ELSE!

G

Um, if I do a little math, that's a huge amount of data in a matter of seconds. Again, doesn't make sense to describe a digital file as a measurement. Measurements characterize the content of that file in various ways for various purposes.

 

[SoundAndMotion]

Okay, I don't want to go back and forth, so I want to summarize my points and move on.

But I can't help picking a couple nits:

But we do know what it is, it's an electrical current. That's the only thing we can digitise (in digital audio) whatever else there is, whether we can measure it or not, we cannot record or reproduce.

G

Typically ADCs measure voltage, not current. Yes, I know Ohm's law and how voltage and current relate. If I find 2 ADCs with the same design goals and basically the same specs, but the input impedance of ADC1 is double the input impedance of ADC2. Under the same test conditions, 1V input to each will give the roughly the same digital output value, but 1mA will give roughly twice the output value from ADC1 as from ADC2. Yes, so what, unimportant in most cases, but...

On to the main point: there is a routine that occurs here very often. I'm sure you know what I mean. Person 1 says "but those units measure the same or the difference is inaudible" to which person 2 answers "but we don't know how to measure everything yet" to which you @gregorio say "Nonsense!. If we don't know how to measure it, we can't record it and therefore can't play it back. So it is irrelevant to this discussion" or something equivalent.

It happens so often, I can find as many examples as you need (older or more recent?). Let's see, here's another one from Sept. 2017:

Again, simple logic! If there were something present that we are unaware of, then we can't measure it and therefore can't record it, digitize it, undigitize it or reproduce it!

G

I think we need an example. Take Transient Intermodulation Distortion(TIM). It was discovered in the early 60's by Matti Otala (completely unknown before that). I think he first described it in his 1970 IEEE paper, but maybe not how to measure it. Several papers by Jung et al., Curl et al. and Otala himself describe how to measure it in the 70's. It then appears in the spec sheets of many/most amplifiers. Are you saying that before the 70's it could not have been recorded? Why not? Hook up a recorder with adequate specs to the amp and hit the big red button. I hope you see that it's ridiculous to deny one could record a signal with some TIM on it before 1970, before it was known.

 

[sonitus mirus]

The process involved with recording and playing digital audio is not relevant to how the audio is being created. TIM always existed and what could be recorded digitally has not changed because of the discovery of the mechanics responsible for this noise.

 

[castleofargh]

Hi G,
OK, the other way around:

You cannot measure something if you don't know what it is, and you cannot measure something if you don't know how to measure it. But you can still record it. How does the lack of knowledge prevent recording?
Cheers, SAM

I don't think so. the very concept of recording involves picking up a variable and quantifying it to store the data. analog or digital, it's the same thing at this point. what a mic is picking is air pressure variation over time at a given position, it converts that into electrical signal with a changing amplitude over time. it's a unique and very clear variable and that's what we'll work with. what caused the sound and how, is entirely irrelevant as that's not what is being recorded.
if the recording process or playback introduces distortions, well then we'll have them in the signal and the fidelity goes down. but it doesn't take knowing what form distortion or how it happens to find out that the signal isn't identical to the original. just checking amplitudes over time(what we record) will show such variations being introduced. they have to, otherwise it would mean that they aren't transmitted.
knowing how to identify distortions and define them specifically, that will hopefully help us make gears with fewer disto. but it's not like people will be oblivious of a change in the signal just because it hasn't been properly defined. looking at the raw samples will always show a change in the signal and that change will always accurately characterize how the signal was altered. because those samples are about the one variable we record. amplitude over time.

I personally find very interesting all the work done trying to characterize nonlinear distortions. the math is beyond me and that pisses me off, but I have high expectations that it will all end up in some DSP making everything better, most interesting is going to be for transducers(at least I hope it will end up there soon). getting a more accurate recording by being able to define and compensate at least some of the distortions of a mic, or a headphone \o/ go math people, go!!!!

 

[bigshot]

You certainly are an audio politician. You basically admitted it to @Phronesis a little while back. You often don't respond directly to questions or comments, rather you use them as a springboard to get on your soapbox and provide your talking points.

I'm afraid you misunderstood my comment. I said that I don't hit REPLY with people who have proven themselves to me to be either disingenuous or who refuse to argue logically and on point. When I hit reply in this board, it sends people a notification by email to come back and reply to my reply. If the person I'm speaking with is incapable of fair discourse the conversation just goes in circles. When I see a clear pattern of ad hominem attacks or argumentativeness or use of logical fallacies or complete refusal to support their points, I address the group with my replies, not the individual. It saves everyone's time and reduces clutter in the threads.

The problem I have with the long list of comparisons you say you've performed is the risk of expectation bias or confirmation bias.

I find that bias is much more likely to be a factor in sighted, uncontrolled tests of sounds that are very, very similar if not identical. That's the sort of claim we regularly hear around here. Here is an example of that... "I bought my new DAC and played my favorite album and it was like a veil was lifted. The soundstage was three dimensional and I could feel the energy being conveyed by the musicians stronger. My old DAC can't compare." No level matching, not blind, no direct switching. Just an impression created by listening to music with no controls. Listening like that is fine for listening to music, and casual listening is the ultimate context we're aiming at, but testing like that leaves one wide open for bias and perceptual error.

I realize that my testing standards aren't up to what is expected from the AES, but I'm not looking for the same sorts of results. I know that there are measurable differences between DACs. I have no problem acknowledging that. Under extreme conditions, like using test tones or at very loud volume levels, those differences may even be barely audible. But I'm not looking for barely audible differences. I am looking for a clear difference that would affect my listening when I put on a CD of Beethoven or Prince or Sonny Rollins in my living room and sit down and listen to it for pleasure. I'm looking for a difference that matters.

A clear difference would show up in a test done with my controls. Bias can't make you confuse day and night. It can only influence you to mistake very similar things. If I level match and juggle the inputs and switch back and forth and I can't hear a difference, that is evidence enough for me that if a difference exists, it just doesn't matter in practice. My ears naturally compensate for very tiny differences. If I can't hear the difference in direct A/B switching, I'm never going to notice it when I sit down and listen to Beethoven's 9th from beginning to end. If you tell me that you have performed a level matched, direct A/B switched, blind comparison of two DACs yourself and heard clear differences, I will sit up and listen to you. That is exactly what I'm looking for- a DAC which sounds clearly different under moderate controls. I'd love to be able to verify that.

Performing level matched, direct A/B switched, blind comparison between two samples of the same music is more than enough to determine if a DAC is transparent for the purposes of listening to music casually in my home. I'm specifically looking for differences that are clearly audible under controlled conditions. If I can't detect a difference with controls, I'm never going to detect a difference without controls. My controls are actually a little into the range of overkill, but that is fine "just to be safe". Everyone has a certain amount of OCD. I just don't let it overwhelm my logical decision making process. I keep my eye on the prize... which is the natural sound of a symphony orchestra in my living room.

As I said before, I have a very precise equalization curve on my system. It's fine tuned to a dB or two across the entire spectrum to suit my listening room. If a piece of equipment is not transparent for the purposes of listening to music in the home, I don't want it, because I would have to apply corrections for each individual source. I'm not willing to do that. I need equipment that is audibly transparent for my purposes. Your purposes might be different... you may be testing for a recording studio or a mixing stage. That would require more stringent controls than I use. But my testing procedure is perfectly capable of determining transparency for the purposes of listening in the home.

One other thing I'll mention... When I see someone take my posts and reply to them line by line or phrase by phrase, I just read a sentence or two, then skip on by to the next person's post. It takes me way too long to parse out what they are talking about if they tear the context up into tiny little scraps and argue each little detail. My time is limited and I prefer to read clear paragraphs where someone states their case in the first sentence, supports it with with evidence in the next few sentences, then sums up with a conclusion at the end. Complete thoughts that are well organized get ideas across more effectively than scraps of contextless arguments piled up in a heap. I always try to reply that way myself.

There are other posters who don't mind line by line replies. You can feel free to communicate that way with them, but I don't have the patience for that particular argumentative technique. I apologize for not reading the last section of your post. If you want to reorganize it into a clearer format, I'd be happy to read it.

You often misstate the equivalence of measurement and recording.

I'm trying to think of a reason why it would be important for an audio component to be able to reproduce sound that isn't present in the recording. Maybe oversampling in DACs? But that is more of a processing thing than something that is measured as opposed to recorded. Do you have any examples of that? Perhaps I misunderstand the point you are making.

Personally, I don't see any reason for audio equipment to need to reproduce sound human ears can't hear. Audibly perfect is good enough. I suppose their can be psychological reasons why someone would feel the need to control things they can't physically perceive, but I'm reasonably well balanced and practical. It isn't a problem for me.

I don't see how small differences in frequency response could even potentially account for perception of differences like bigger stage, more precise imaging, better instrument separation, more articulate bass, much more detail, etc. There would either need to be something else besides frequency response

Aha! That is PRECISELY where bias rears its ugly head! People who do controlled tests generally are unable to consistently detect differences like "imaging" and all those other vague things you mention. If you see someone using words like that, you can pretty much be assured that they haven't taken steps to eliminate bias from their comparison. Here in sound science those sorts of things are big red flags that make us ask politely about what sort of controls the person applied to their comparison test.