Current main stream isn't worth listening in 16/44 either.
I would be a critical listener with high expectations. Not trained professionally. I don't hear a difference going from 16/44 to 24/96. But I hear a tremendous difference listening to a good sound engineer and producer and a well performing artist. That's where my money goes.
As an end product, I would have doubts that anyone could hear a difference past 24/48. Honestly 16/44.1 is just fine for most people. Especially when you run it through a lossy container.
That said. A lot of people in the audio production space see stuff like this, and apply it to recording, where the loss of definition compounds. Also, aliasing is a major thing.
So just because you don't need to be downloading 24/192 audio from spotify, don't say audio interfaces supporting it is pointless
> As an end product, I would have doubts that anyone could hear a difference past 24/48. Honestly 16/44.1 is just fine for most people.
On both accounts I agree. But since many can hear the difference between 16/44 and 24/48, and all agree 24/48 is aurally superior, it would be nice to have it available when desired. I somewhat disagree on several points with the author of the article, and I think many of his declarative statements are purely opinion the fact of which he neatly obfuscates with some legitimate scientific facts.
Distilling this down to a technical arguement, and going back to basics, this level of resolution is pretty not relevant to 99.5+% of music listeners.
If listening in the car, over Bluetooth, or on headphones, you are going to need tens of thousands of dollars to preserve that signal and not destroy it. RFI, EMI, and current leakage on the power line is going to render anything above Redbook (standard cd quality) useless.
Speakers have a bad frequency response, but can be very detailed, so they aren't going to be an obstacle to hearing high-res audio.
But the room is. The room is the undisputed kicked to this arguement.
The room acts as a "wire" from the speakers to your ears, and there are fundamental problems with overcoming this "wire".
Most people don't even try, and just ignore it altogether.
For example, there is the ASRD, attack sustain release decay. Without sucking up the bass energy which is omnidirectional, you will only ever hear the attack well, and even that is sketchy. The sustain, release, decay will be drown out.
Then there is the timing aspect. Any sounds that reach the ear within approx. 20ms will obscure the detail retrieval from the original source. That's a 20 foot bounce. So you need a LARGE room. Most people have small rooms, and ALL their "first reflections" are within that 20ms range, and even lots of "early reflections"
So yeah, room geometry and physics render high Res audio irrelevant for everyone but the highly obsessed audiophile.
The normies that aren't into audio have no clue, none at all, so they argue about high Res music when it's the LAST thing that is going to improve their musical experience.
With audio resolution, you can record frequencies at half of the refresh rate, so 192k audio codecs can record up to 96khz audio frequencies.
With all but the highest quality ultrasonic tweeters, anything above about 30-40khz simply won't be played back, not to mention how anything above 22khz is inaudible to all but the rarest of humans, and anything above 16khz is inaudible to anyone in their mid-30s.
That being said, 48khz audio files are good enough to deliver 24khz of audible frequency spectrum for the same reason, are ~1/5th the size, and humans are physically incapable of telling the difference between them and a 192khz audio file, and the people who say otherwise are either lying to themselves or they are picking up some artifact that is a consequence of playing 92khz audio through a speaker system that is incapable of reproducing it accurately.
Probably correct. 24/88.2 is probably the perfect compromise between too much and too little.
Then there is the issue of the Nyquist-Shannon sampling theorem living up to its FULL POTENTIAL.
"It establishes a sufficient condition for a sample rate that permits a discrete sequence of samples to capture all the information from a continuous-time signal of finite bandwidth. Strictly speaking, the theorem only applies to a class of mathematical functions having a Fourier transform that is zero outside of a finite region of frequencies. Intuitively we expect that when one reduces a continuous function to a discrete sequence and interpolates back to a continuous function, the fidelity of the result depends on the density of the original samples. The sampling theorem introduces the concept of a sample rate that is sufficient for perfect fidelity for the class of functions that are band-limited to a given bandwidth, such that no actual information is lost in the sampling " (1)
Did you get the significance of that? "Such that no actual information is lost in the sampling"
If a proper filter is used, and infinite computing power is used, the analogue waveform can be PERFECTLY recreated.
This simply doesn't happen, and designers don't even try.
Hard to believe, right? The issue is that for full information retrieval, a filter of infinite tap length is needed. Think infinite ripples or ringing in both directions of the sample point. And that requires computing power getting close to infinite.
Which brings me to a crazy dude exploiting this theory to its fullest, Rob Watts, designer of Chord Qutest and M-scaler.
"What’s the significance of the ultra-long tap-length digital filter and of the WTA algorithms? In seminars given around the world, Watts has suggested that few designers recognize the full implications of digital audio sampling theory, which according to Watts points to an astonishing conclusion. Specifically, Watts contends that garden-variety 44.1kHz digital audio files could, if processed through a digital filter of near infinite tap length, yield analog waveforms every bit as accurate and complete as those produced from higher-res audio files, albeit with slightly higher noise floors. Stop a moment and let that claim sink in. Watts is saying, in essence, that a DAC with a properly designed filter system can deliver ultra-high-res sonic results from conventional CD-quality files. " (2)
All design choices are a trade off. Size Vs processing power to "unpack" and recreate data points between samples.
Which brings me to my main point, it may be unnecessary to have a sample rate higher than 44100x per second, but it is practical to do so. By doubling the sample rate and storage requirements, it takes far less computational power to play the song as it would to mathematically calculate that extra sample between each sample.
That's interesting and it makes sense. Early and mid-stage audio courses focuses so much on the numerical differences between levels of digital integrity, 44.1 vs 48 vs 88.2 vs 96, etc., and they mention that if you encode an audio sample and compare it with a waveform analyzer that you can see the differences between the audio waveforms.
Which is true, but the differences are incredibly minor and mostly in the inaudible range, and those differences are really nothing once you consider how the audio is going to be altered and mangled to some degree by the DAC, amplification, and speakers they are played back through, the rooms they are played in, and by the listener themselves.
Fidelity is always lost, so the smart thing would be, unless it is business critical with real money on the line, to find a balance. 96k is half the size of 192k but twice DVD audio rate. I would say that's a good place to be unless your paycheck depends on something better, and if they aren't played side-by-side on a tens of thousand dollars audio system within an expertly acoustically treated room you would never be able to detect the difference.
While I can hear the difference between 128 kbit mp3 and flac, I sometimes can't between 320kbit mp3 and flac. But that being said, on principle, I want as much information there as possible, no not as much as "makes sense", because common sense is debatable. Since sound is analogue, there's a new data-point every Planck time, at a resolution of how many Planck lengths I don't know.. Obviously, that's idiocracy.. But I'll take "this is provable many times more precise than you're physiologically able to hear" any day.
I agree that it's certainly overkill for listening sake, and that decent mp3/aac/opus encodes are
generally indistinguishable in quality to a proper "lossless" format.
That said, I still really do like having high bitdepth/sample rate audio available for download.
Storage space is cheap and as physical CDs are getting phased out I don't want to be in a world where the best you can acquire is the minimal viable quality. If space is at a premium lossless audio with overkill quality can be transcoded to a smaller compressed format at 10s to 100s times realtime. Additionally it means should better compression schemes come out in the future you can reap the benefits of them. I had a lot of old CD rips from the early 2000s in 96kbps WMA and that was regrettable.
As a scifi speculation for fun if in the future there were augmentations that improved human's hearing ability or effective brain links which could transmit audio more effectively than our ever deteriorating ears we'd want the best possible audio signal to feed into that.
All about use cases. In the open back, those are hard to beat. You are obviously a person of good taste as these are generally considered a flat and true headphone. I would however recommend buying used, and that might get you something nicer, like 800's. The other hifi headphone that everyone in the "know" seems to love is the Dan Clark aeon 2, open back. $900 new, $600ish used.
The other thing to consider is amplification, but at the $500 price point you can probably get away with feeding it from a phone or entry level DAP.
Once you get above this price point, there is a balancing act of signal and amplification Vs headphones, and you are into the thousands of dollars range.
I would recommend an older used astell and Kern DAP to pair with a headphone like the HD650. Being free of the noise on your home electrical circuit and using battery will get you very close to ideal for very little money.
15 comments
[ 2.9 ms ] story [ 44.9 ms ] threadI would be a critical listener with high expectations. Not trained professionally. I don't hear a difference going from 16/44 to 24/96. But I hear a tremendous difference listening to a good sound engineer and producer and a well performing artist. That's where my money goes.
24/192 Music Downloads and why they make no sense (2012) - https://news.ycombinator.com/item?id=19318898 - March 2019 (314 comments)
That said. A lot of people in the audio production space see stuff like this, and apply it to recording, where the loss of definition compounds. Also, aliasing is a major thing.
So just because you don't need to be downloading 24/192 audio from spotify, don't say audio interfaces supporting it is pointless
On both accounts I agree. But since many can hear the difference between 16/44 and 24/48, and all agree 24/48 is aurally superior, it would be nice to have it available when desired. I somewhat disagree on several points with the author of the article, and I think many of his declarative statements are purely opinion the fact of which he neatly obfuscates with some legitimate scientific facts.
If listening in the car, over Bluetooth, or on headphones, you are going to need tens of thousands of dollars to preserve that signal and not destroy it. RFI, EMI, and current leakage on the power line is going to render anything above Redbook (standard cd quality) useless.
Speakers have a bad frequency response, but can be very detailed, so they aren't going to be an obstacle to hearing high-res audio.
But the room is. The room is the undisputed kicked to this arguement.
The room acts as a "wire" from the speakers to your ears, and there are fundamental problems with overcoming this "wire".
Most people don't even try, and just ignore it altogether.
For example, there is the ASRD, attack sustain release decay. Without sucking up the bass energy which is omnidirectional, you will only ever hear the attack well, and even that is sketchy. The sustain, release, decay will be drown out.
Then there is the timing aspect. Any sounds that reach the ear within approx. 20ms will obscure the detail retrieval from the original source. That's a 20 foot bounce. So you need a LARGE room. Most people have small rooms, and ALL their "first reflections" are within that 20ms range, and even lots of "early reflections"
So yeah, room geometry and physics render high Res audio irrelevant for everyone but the highly obsessed audiophile.
The normies that aren't into audio have no clue, none at all, so they argue about high Res music when it's the LAST thing that is going to improve their musical experience.
With all but the highest quality ultrasonic tweeters, anything above about 30-40khz simply won't be played back, not to mention how anything above 22khz is inaudible to all but the rarest of humans, and anything above 16khz is inaudible to anyone in their mid-30s.
That being said, 48khz audio files are good enough to deliver 24khz of audible frequency spectrum for the same reason, are ~1/5th the size, and humans are physically incapable of telling the difference between them and a 192khz audio file, and the people who say otherwise are either lying to themselves or they are picking up some artifact that is a consequence of playing 92khz audio through a speaker system that is incapable of reproducing it accurately.
Then there is the issue of the Nyquist-Shannon sampling theorem living up to its FULL POTENTIAL.
"It establishes a sufficient condition for a sample rate that permits a discrete sequence of samples to capture all the information from a continuous-time signal of finite bandwidth. Strictly speaking, the theorem only applies to a class of mathematical functions having a Fourier transform that is zero outside of a finite region of frequencies. Intuitively we expect that when one reduces a continuous function to a discrete sequence and interpolates back to a continuous function, the fidelity of the result depends on the density of the original samples. The sampling theorem introduces the concept of a sample rate that is sufficient for perfect fidelity for the class of functions that are band-limited to a given bandwidth, such that no actual information is lost in the sampling " (1)
Did you get the significance of that? "Such that no actual information is lost in the sampling"
If a proper filter is used, and infinite computing power is used, the analogue waveform can be PERFECTLY recreated.
This simply doesn't happen, and designers don't even try.
Hard to believe, right? The issue is that for full information retrieval, a filter of infinite tap length is needed. Think infinite ripples or ringing in both directions of the sample point. And that requires computing power getting close to infinite.
Which brings me to a crazy dude exploiting this theory to its fullest, Rob Watts, designer of Chord Qutest and M-scaler.
"What’s the significance of the ultra-long tap-length digital filter and of the WTA algorithms? In seminars given around the world, Watts has suggested that few designers recognize the full implications of digital audio sampling theory, which according to Watts points to an astonishing conclusion. Specifically, Watts contends that garden-variety 44.1kHz digital audio files could, if processed through a digital filter of near infinite tap length, yield analog waveforms every bit as accurate and complete as those produced from higher-res audio files, albeit with slightly higher noise floors. Stop a moment and let that claim sink in. Watts is saying, in essence, that a DAC with a properly designed filter system can deliver ultra-high-res sonic results from conventional CD-quality files. " (2)
All design choices are a trade off. Size Vs processing power to "unpack" and recreate data points between samples.
Which brings me to my main point, it may be unnecessary to have a sample rate higher than 44100x per second, but it is practical to do so. By doubling the sample rate and storage requirements, it takes far less computational power to play the song as it would to mathematically calculate that extra sample between each sample.
(1) https://en.m.wikipedia.org/wiki/Nyquist-Shannon_sampling_the...
(2) https://www.theabsolutesound.com/articles/chord-electronics-...
Which is true, but the differences are incredibly minor and mostly in the inaudible range, and those differences are really nothing once you consider how the audio is going to be altered and mangled to some degree by the DAC, amplification, and speakers they are played back through, the rooms they are played in, and by the listener themselves.
Fidelity is always lost, so the smart thing would be, unless it is business critical with real money on the line, to find a balance. 96k is half the size of 192k but twice DVD audio rate. I would say that's a good place to be unless your paycheck depends on something better, and if they aren't played side-by-side on a tens of thousand dollars audio system within an expertly acoustically treated room you would never be able to detect the difference.
That said, I still really do like having high bitdepth/sample rate audio available for download.
Storage space is cheap and as physical CDs are getting phased out I don't want to be in a world where the best you can acquire is the minimal viable quality. If space is at a premium lossless audio with overkill quality can be transcoded to a smaller compressed format at 10s to 100s times realtime. Additionally it means should better compression schemes come out in the future you can reap the benefits of them. I had a lot of old CD rips from the early 2000s in 96kbps WMA and that was regrettable.
As a scifi speculation for fun if in the future there were augmentations that improved human's hearing ability or effective brain links which could transmit audio more effectively than our ever deteriorating ears we'd want the best possible audio signal to feed into that.
https://danclarkaudio.com/aeon-2.html
The other thing to consider is amplification, but at the $500 price point you can probably get away with feeding it from a phone or entry level DAP.
Once you get above this price point, there is a balancing act of signal and amplification Vs headphones, and you are into the thousands of dollars range.
I would recommend an older used astell and Kern DAP to pair with a headphone like the HD650. Being free of the noise on your home electrical circuit and using battery will get you very close to ideal for very little money.
Something like this: https://www.usaudiomart.com/details/649760711-astellampkern-...