Why upload bootlegs in high-resolution?

[vince666]

another short "pill" answer (i am on smartphone, will write something more detailed later, when i will be at my PC)...

comparing 16/44 and 24/96 might be pointless... because your digital hardware might be not good enough then the 24/96 playback is just flawed and with no meaningful sound quality differences compared to 16/44.


in fact, if you wish to make a proper comparison (like I did myself) is to individually compare both a 16/44 and a 24/96 capture directly with the analog source they were captured from.
it will be easier to hear which capture will sound closer to the analog source listened without any digital conversion involved.

Also, not any analog gear is good... there are so many cassette decks (but also so many amplifiers) where the signal passes through integrated circuits (the chips) which ruin the sound in a similar way as bad digital usually sounds... such analog audio gear will never do justice to a good analog source but it would still be good enough to listen to CDs.
Why I know it, about IC chips and such?
Well, it's many years I also work on audio gear... do consider that many good/upgraded transfers shared during these latest few years were transferred on some cassette decks I did restore and modify with my own hands to improve their sound quality (i.e. both my own and neonknight's RS-B565-MH).

[YASHA]

I found this topic in the archive, but given the situation, I suggest that everyone find it themselves :rolleyes:
It is not difficult.

[YASHA]

This topic from the archive is exactly 11 years old... 

[vince666]

finally at my PC keyboard.

First off...
Unfortunately, numbers and measurements cannot answer all (or most) question related to sound quality.
There are so many things which "make" the good sound that are not measurable in any ways, then we must use the only "measurement gear" which is (or might be) able to evaluate just every aspect related to sound quality: our ears!
Anyways, there are several things which ruin the sound which are measurable... but the absence of them doesn't automatically means that the sound is great.
With all of this in mind, it would be easier to understand why I've stated that talking only about "numbers" (frequency response, ect) is pointless...
That's only some of the theory... but, while listening to reproduced music, we are dealing with the real world things, not with theory.

A question...
Is digital sound and the gear to record/play it perfect? (as they always advertized it since the first day they've put it on the market a few decades ago)
If your answer is "yes" then you can directly avoid reading the rest of this post and be happy...

The "software" side of digital audio, meaning with it the audio data...
PCM sample is, of course, a sample of the analog signal you are going to capture.
Being a sample, it's subjected to the numbers related to the sample... this also means some kind of approximation is involved and approximation is distortion (here, with distortion i mean the strict meaning of the word, i.e. a distorted sample of a waveform is a sample where the waveform isn't exactly the same as the analog source).
Samples have to deal both in amplitude domain and in time domain and, just from this point alone, it's obvious that the more bits you have the more possible discrete amplitude values you can sample then the least approximation you get, compared to the analog source signal. (hint/question: did you wonder why we need to use dither?)
In the same way, the higher the sample rate, the more times in a second you take a sample of the analog source signal, the greater accuracy... of course, here, Mr Nyquist would not agree... in fact, it's true that, on digital, there is interpolation involved, to the point that only 2 samples (meaning samplerate here) are enough to describe a sound frequency, this leads to the point that the audio frequency limit of a certain samplerate is just half of the samplerate.
In theory, it's all good... but then we have to deal with what happens in practice when we have to translate the theory through A/D and D/A converters... we just cannot directly listen to audio data or to digital audio: our ears are analog!

The software side of digital audio, meaning audio editing and processing software...
Is all of this as perfect as many would think about? Not, at all!
The rule of thumb is that, the more you process a sample the more it degrades from its own "raw capture" shape.
Quite any kind of processing has some drawbacks involved, included a simple gain change (not to mention resamplings, aka speed corrections, equalizing/filtering or, even worse, noise reduction aka dehissing).
Also the simple vinyl clicks removal isn't that safe... if you apply it automatically on the whole audio file, there will be so many dynamic transients which will be considered vinyl clicks and will get reduced/smoothened or removed by the click removal (i.e. hand claps close to the taper are a good example and the same goes for any other kind of percussive-alive sounds with a sharp transient)... a good example is to compare my "rev.A" of the vinyl rip of "The best of tour '72" (Rainbow Theatre feb 20th, 1972 gig) with the version previously shared by Neonknight... both version come from the exact same raw transfer but, unfortunately, NK applied automatic declicker and decrackler on the whole thing and this affected so many things on that so good audience recording. On my "rev.A" version, I was careful in cleaning out the noises only when they actually were, by a very long and careful manual cleaning work, to just leave the useful sounds intact... of course, the point that the capture was made in Hi-Res helped a lot in getting a great result. (if you go and read my comments/description on that torrent, you'll find quite some explainations about such matters).
So, digital processing isn't too safe... but the same processors will work a lot better at Hi-Res (meaning also higher samplerate) than at, i.e., CD or DAT quality.
Getting more informations just in the first moment (i.e. higher samplerate and the highest bit resolution available) will make the digital processing doing its work with more precision.
About the bits and why, for example, when I work on a 16/48 DAT source then I make the processed version as 24/48... well, if you go and apply ANY kind of processing on any fixed point resolutions (being them 16bit or 24bit) then their wordlenght will automatically expand to floating point resolution (and if this doesn't happen then it means we are using a software which is a toy!)... so, even when I start from 16bit, if I go and apply any kind of processing (included a simply gain change) then the result would be at, i.e., 32bit float... and, at this point, it's better to make a finished version which from 32bit float goes to 24bit and not back again to 16bit... why throwing away 8bits of data from any single sample of the processed audio?
Unfortunately, no A/D or D/A converters can work at floating point resolution (and, also, FLAC cannot encode floating point) then, after processing, we are forced to go back to a fixed point resolution (and apply dither)... then, making the finished product at 24bit is a way to keep the result as close as possible to the actual processed sound.

The hardware side of digital audio, meaning the A/D and D/A converters, the soundcards of any kinds and any possible (cheap or expensive) players/devices...
This is the most overlooked part of the whole story...
I am quite sure that most people think they can buy some (not that expensive) consumer product and they will think it's going to sound GOOD, because digital is perfect and such...
Unfortunately, that's totally false.
If you only go and test your own digital device, while stressing it by investigating what happens close to the actual limits (in both frequency and level) , you will soon see weird things happening on just so many devices.
Most converters do struggle to handle amplitude level which are close to their own "0dB FS" amplitude limit... then, i.e., capturing or also simply playing a sound which hit peaks close to the 0dB is going to make them mis-behave just considerably.
A good rule of thumb in the studios is to capture/record while staying at least 10dB (if not even more) under the 0dB and, hey, there in the studios they are using TOP QUALITY converters!
(but it's true that they also need to keep some healthy headroom because they need to process the audio a lot, during their work).
So, for the common people at home with consumer or semi-professional gear, it's wise to capture analog sources with the loudest peaks which hit -3dB FS to -6dB FS at max, to just stay away from the trouble zone.
Do consider that, since we have the samples (meaning the dots along the time domain of the samplerate) , two consecutive samples at 0dB are just meaning that, most likely, in the middle of them the waveform was going over the 0dB... and this happens both in input (because the analog source had its own amplitude value also in between those two samples) and in output (because the digital interpolation will automatically create an overload in between those two samples).
These points alone are enough to let you understand why all those "loudness war" (official) recordings are sounding like crap, with a too hard sound due to (not detected by the PPM meters) clipping distortion and also because most of the time (but also all of the time) the poor D/A converters are working in their own danger zone (the loudest couple of bits).
And I didn't even take into consideration the reduced dynamic range of such loudly compressed recordings, made through those devilish processor known as brickwall limiters or maximizers! (maximizers of crappy clipping distortion, actually).
The sad thing is that, since digital audio has extremely low noisefloor, there would be no true need to push the recording so hot.
Now, let's talk about frequencies...
The most you go and force the converters to work close to their high frequency limit (sorry, Mr Nyquist, your theory is good but it's simply theory because the men weren't still able to make converters which are good enought to make your brilliant theory apply "as is" in the real world) the most weird things happen...
The most common issues are getting distorted waveforms (compared to their source/analog counterparts) and a lot of aliasing and intermodulation distortion.
So, using higher sample rates (meaning much higher than the actual audio bandwidth we wish to listen to) is a way to keep such problems away from the useful/hearable sound.
Aliasing is a terrible kind of digital artifact which works like a sort of bouncing mirror... when you try to capture a sound which exceeds the Nyquist limit of your system (and you should also keep all those nice harmonics/overtones which are a critically important part of any sound) , the part of the sound which is over the Nyquist limit will simply bounce back in frequency...
And this happens not only while capturing but also while processing or by simply listening... and also if they say they added anti-aliasing filtering.
An extreme example of aliasing, to make you understand how this kind of artifact does really sound, is when you listen to a low bitrate mp3 file with a cut at, say, 12Khz to 14Khz (i.e. a 128kbps or less file)... can you hear that modulating crap at treble, with drums cymbals which sound like a confused/modulated metallic shhhhh(it) sound? That's a good example of aliasing!
So, using higher samplerates (meaning the ones which can handle frequencies which are well over the 20Khz) is a good way to avoid aliasing happening and getting into your precious audio.
And I end this long post while telling you in better details my experience with the analog oscilloscope, which I just mentioned above...
A while ago, I was here below at my older brother's laboratory (he is a professional technician who repairs smartphones, etc) to help him trying and testing a few measuring devices he had bought.
In particular, he's got a brand new digital oscilloscope (with inbuilt tone generator) and also a good oldschool used analog oscilloscope (a Tektronix 2235, if I remember the model number correctly, which can work up to 100Mhz)... I was there mostly to test and have a bit of fun with the old analog one, to see if it was still working as intended (and it was!).
To test the analog oscilloscope, we attached it to the sound output of a notebook PC (then it was one of those typical/widespread intergrated soundcards) while running a test tone generator software... the notebook audio system was working at the classic 44.1Khz samplerate, btw...
So, I played a 400Hz sinewave test tone on the computer and the oscilloscope showed a nice sinewave on the screen...
You have to consider that sinewaves are a good choice to test things because the sinewave is, by definition, a waveform which does contain only the first harmonic (the base tone/note) while other waveform shapes do contain also harmonics.
Then, I was gradually raising the frequency of such sinewave tone in the computer while watching the sine waveform at the oscilloscope screen...
1Khz, still a beautiful sinewave... 2Khz still OK...
At some point, between around 3Khz or 4Khz (don't remember exacly, but it was still a tone into the mid-hi frequency sounds area) the oscilloscope started to show waveform which weren't a beautiful sinewave anymore.... in other words, it was detecting harmonics (read: distortion!)... then I kept raising to like 10Khz and the oscilloscope showed a ugly waveform with sort of phased copies of it superimposed on the main waveform (something I would expect when aliasing chimes in)... but CRAP! I though the oscilloscope was defective...
So, we simply attached it to the tone generator of the other (brand new) digital oscilloscope for comparison...
Well, the old analog oscilloscope was perfectly good, in fact, while injecting into it (from the good signal generator of the other oscilloscope) a sinewave of 1Mhz, it did still show a beautiful sinewave...
This means that, of course, the crappy distorted wave, mixed with aliasing, was coming from the notebook soundcard just from 3Khz to 4Khz with a pure sinewave tone...
Not to mention what would happen with a complex waveform, full of overtones/harmonics, like music actually is and which also reaches a lot higher frequencies than 3 or 4Khz.
This experiment easily explained to me why listening to music with cheap/consumer devices does produce a crappy/harsh/flawed/fatiguing sound... it's full of distortion and aliasing!

Then, trying to stick to HiRes digital and some kind of decent playback device is a way to get closer to what you hear directly from a good analog source.

The analog itself has its own problems, like noisefloor and W&F... but, for sure, it doesn't have any of the aforementioned issues, which are typical to digital.

I hope I answered to some doubts about the Hi-Res matter...

Cheers,

Vince.

[Son of Nothing]

I think vince666 has given a comprehensive answer, and I won't compete with or deny that. I will just add my own perspective (which is based on opinion and not science, let's make this clear from the outset).

My perspective is that I used to believe the same as the original poster here, and specifically I used to believe in Nyquist. And while the Nyquist theorem is absolutely correct from the digital signal processing perspective, my opinion has been adjusted on another forum where an expert I have trust in pointed out that while the theory is fine, in the real world digital to analog conversion in particular is far from perfect. And part of dealing with those imperfections is the need to filter the converted signal. And these filters are also imperfect, especially those with a steep cutoff. And the closer you get to the Nyquist frequency the steeper the filter cutoff you need. So if you put your Nyquist frequency close to the limits of human perceptibility, then you need a filter with a steep cutoff, and that introduces its own distortion. If you choose a Nyquist frequency well above the limits of human perception (which is where the 96k sample rate comes in), your analogue stage can use much smoother cutoff curves in their filters and the end result supposedly is a more accurate analogue reconstruction of the signal. This is sounds plausible to me. That, plus Vince's discussions about the potential losses resulting in multiple rounds of digital processing in software convinces me that it's reasonable to at least process recordings at high sample rates and bit depths.

But...

... for, me personally, I'm going to horrify Vince and others by saying sometimes I still do downloaded 16/44 or 16/48 versions in preference to 24/96. And that's because my aged ears are far worse than even a crappy computer audio interface. So unless it's a recording of special interest to me that I'm likely to listen to more than once, sometimes I will save the space and download bandwidth (sorry, even in 2023, not everyone has North American or European or East Asian bandwidth, especially in rural areas which I where I now spend much of my time) and go for the smaller file set. But that is now a consciousness compromise I make, not a default position.

[vince666]

Son of Nothing : first off, no worries... I am not going to be horrified if you happened to prefer downloading 16/44 or 16/48 in preference of 24/96.
On average/consumer listening gear and amplification/speakers setup, the hearable difference between CD/DAT quality and HiRes is slim or non existant, since the whole system isn't even able to reproduce the CD/DAT quality as it really deserves.
Of course, for archiving purposes (or to listen to it on some high quality gear and HiFi system), HiRes is the way to go... and, even better than that, it's listening directly to the analog source with some truly good reproduction gear... but, unfortunately, only those who own the analog sources can do that.
As for myself, for most of my own music listening (excluding most of PF live recordings which I only have as digital captures) my own main way to enjoy music is to listen to vinyl and cassettes.

About your other point... yes, Mr Nyquist was totally right with his own theory... in fact, the problems arise when people think they can apply his theory to the kind of technology we have until nowadays.
Moreover, if I am correct, Nyquist made his own theory (or started making it) like during the 1920's... and, back then, he was a true visionary indeed! But, of course, especially back in 1920's when also analog recording wasn't still "advanced", thinking about sampling theories was like pure science fiction.
I've found this document by Mr Nyquist himself which is from 1928 and which it sorts of being related to his (now well famous) sampling theory:
https://web.archive.org/web/20130926...gs/nyquist.pdf

So, that theory is GOOD... just, the technology is still not good enough to apply it to the real world just "as is".... so, the best way to overcome some of the most annoying sound artifacts related to the real world digital sound and gear is to simply stick to Hi-Res.

And, yes, your point about the steep filtering, rather than the A/D and D/A stages themselves, is good.

Cheers,

Vince.

[Son of Nothing]

I forgot to make that point, but yes I would totally agree that for archiving you want to preserve that in the highest sample rate and bit depth as is feasible.

[mark2114]

A 16 bit CD/DAT no need for HRes, an analog master or 24 bit master HIRes yes please.