The No.1 Website for Pro Audio
Quality Is Back - Keep Your High End
Old 23rd December 2019
  #91
Lives for gear
 

Quote:
Originally Posted by mattiasnyc View Post
The sampling circuit itself doesn't operate at a rate of 44.1kHz though, it's usually in the MHz range. It then gets converted down to whatever sample rate you choose.

I suppose we could hypothesize a system where that wasn't the case and instead it did operate at 44.1kHz, in which case the question would be if you - along with everything mechanical that's involved (keyboard etc) - is capable of executing something at that timing resolution in the first place. I'm 100% you're not. Think about it;

- Suppose we set up a computer to play back audio at a 44.1kHz sample rate, and then we've recorded a click track at say 60bpm or so, and the converter operates at 44.1kHz in this hypothetical example, but we're not recording music or sound as the input, we're recording key presses on a keyboard. Do you really think you would be hitting the key with sample accuracy? If you did you'd be an absolutely astoundingly amazing musician. I bet you that's not what would happen.

And so based on that thought experiment I really think we can conclude that the timing resolution as far as triggering record is concerned is sufficient because you can't execute with any greater resolution anyway.
Got it. So it definitely doesn't have "gaps" in a sense like I am thinking (IE a "resolution grid" in a sense with the ability to inadvertently land a delivery in-between the recordable grid lines). Thanks for clearing that up ! Cool to know. So it basically does record like tape 100% linear not dependent on a *positioning* resolution, but rather another type of resolution ? Thats what I gather from you now.

The hypothesis is irrelevant I guess, but its fun so.. ok - What I was trying to emulate in the explanation was not like your trying to amazingly land your presses on a keyboard in-between the resolution grids. But more that over time it is inevitable that you would sometimes, just with one press even land there without trying. Maybe next 50 times not, maybe 50 times in a row yes. It would be inadvertent random of your timing. See what I mean. So regardless of how small of a space that is, knowing that there is xxxxx amount of those spaces per second means laws of averages will cause you to press the key during a gap just as much as you did within the recordable spaces.

But thanks for clearing this up anyway. Its impossible because the resolution speaking of is I guess really infinite (therefore creating true analog like linearity) but then it samples it down to your chosen sample rate. So = no resolution dependent non-recordable gaps like I imagined.

That said - still, my point does stand as 96k being superior for timing as far as nudge editing within your session after recording. I guess more is indeed more in this case. BUT maybe this will be wrong one day too if a DAW finds a way to allow you to nudge audio at a finer resolution than the sessions sample rate. So a 96k session means at most I got 96,000 different positions per second that I can place/end/nudge a audio clip, which is a finer movement per second than a 44.1k session which would top out at a max of 44,100 places to start/end/nudge your audio clips.

Curious if this restriction will last forever ? Or maybe I do not know and some DAW has already broken this rule/law ? lol Waiting to be wrong on this one too ! haha But at least I know I used to be right !
Old 23rd December 2019
  #92
The snappy answer, is, of course, that data in the space between the samples would have been describing very high frequencies above the decided upon frequency band limit of the sample rate format. (A sample rate which we might hope would have been chosen to adequately accommodate the roughly 20 kHz bandwidth of human hearing.)

It can be pretty hard to wrap one's head around the way the digital audio math works out... but this reasonably brief (for how much it explains and demonstrates) video goes a LONG way to helping most folks see the practical reality [demonstrated primarily with 'real' analog test equipment] -- if not the mathematical twiddly bits. It has opened a LOT of eyes...

The video's home page: https://xiph.org/video/vid2.shtml (includes multiple download/stream formats)

The vid on YT:




PS... The science in this vid is basically solid; the goofiest thing in it is the host's old time sailor beard. And that's not even so odd these days, I guess.
Old 23rd December 2019
  #93
Lives for gear
 

Quote:
Originally Posted by theblue1 View Post
The snappy answer, is, of course, that data in the space between the samples would have been describing very high frequencies above the decided upon frequency band limit of the sample rate format. (A sample rate which we might hope would have been chosen to adequately accommodate the roughly 20 kHz bandwidth of human hearing.)

It can be pretty hard to wrap one's head around the way the digital audio math works out... but this reasonably brief (for how much it explains and demonstrates) video goes a LONG way to helping most folks see the practical reality [demonstrated primarily with 'real' analog test equipment] -- if not the mathematical twiddly bits. It has opened a LOT of eyes...

PS... The science in this vid is basically solid; the goofiest thing in it is the host's old time sailor beard. And that's not even so odd these days, I guess.
Thanks for the link blue. Watched it. The part specifically pertaining to this discussion is towards the end. I do like the science through the rest of the video, I get most of it.

But either he didn't take enough time explaining the "falling between the samples" thing or he just doesn't know how to describe it to a musician (lol), because I am left with still not really understanding why thats not possible.

So he scrolled a waveform around, lol means nothing to me. I have no clue at what resolution that scrolling is to prove anything, nor again, does it mean anything to me personally maybe cause I don't understand the connection between what he showed and that its negating the theory/assumption that falling between the samples is possible.

I'd like to see or hear another test/example on the subject, where the tech dude talks about only that for a while as to really help me grasp it/believe it. To me, it just still seems possible. And either way, less gaps per second... why not ? lol

I actually like your quick explanation better.. Telling me the gaps are somehow specifically designed to only be frequency related and that frequency is undetectable. Now THAT is understandable. But still remains, I don't see how they make it only frequency dependent and not timing dependent too,, cause sound is also time, or uses time. If that time is broken up by a limited amount of pictures of the sound per second, HOW is there not timing gaps ? And more importantly how then is more pictures not smoother for timing ?

I understand he tried to explain it, but it was too geeky I think (tech talkish). And IMO didn't really prove this to be wrong. You proved it better, lol Is there any other videos on this specifically ?

Either way, DUDES !!! This is awesome talk. Love it, love you all ! Hope holidays are going great and music stays dope ! Grooves and writing trumps recording quality anyway, so all this talk is just fun debates to me. I still want to understand though and want solid undeniable unarguable proof.
Old 23rd December 2019
  #94
Lives for gear
 
monomer's Avatar
 

Quote:
Originally Posted by Infa View Post
lol, you know your probably right here. After typing what I did, I realized exactly what you just said - for some of these guys, enough is never enough.

I do agree, its not all of them, but *some*, not most, I'd say some of them audiophile nutcases are more about snobbery than the actual sound. Plus bragging rights to their friends about who got what in their set up, etc..
Yeah, but those nutcases are already on these specs. It's now time to see when the 'typical' audiophiles take the bite.
Old 23rd December 2019
  #95
Lives for gear
 
monomer's Avatar
 

Quote:
Originally Posted by Infa View Post
A lot of it is always to do with that I feel too.
Felling is good. But please please please don't fabricate pseudo-technical explanations about what you're experiencing.
Old 23rd December 2019
  #96
Lives for gear
 
monomer's Avatar
 

Quote:
Originally Posted by Infa View Post
That said - still, my point does stand as 96k being superior for timing as far as nudge editing within your session after recording. I guess more is indeed more in this case.
If you're talking about 'nudging' by single samples then you're right.

But we can do better. We can actually compute a 'nudge' that is halfway between two samples and if done correctly it will be pretty much perfect. Tho your DAW won't do that in real time.

With sampling (the theory you were scoffing at) you can basically represent any phase of a waveform, as long as its frequency is lower than half the samplerate.
Well, technically the phase you can represent is limited by the bit depth (!!) but in any case it is much more precise than the samplerate itself.
I think that for 16/44.1 it's in the nanosecond region.
So with some math you can 'nudge' a PCM signal in nanoseconds, which is thousands of times smaller than the time between the actual samples.

Maybe this article will help you understand some of it.

Last edited by monomer; 23rd December 2019 at 03:49 PM..
Old 23rd December 2019
  #97
Lives for gear
 
monomer's Avatar
 

Quote:
Originally Posted by Infa View Post
Curious if this restriction will last forever ? Or maybe I do not know and some DAW has already broken this rule/law ? lol Waiting to be wrong on this one too ! haha But at least I know I used to be right !
DAW's don't do it because it takes some computation.
If you require a sub-sample nudge there are plugins for that.
Old 23rd December 2019
  #98
Gear Guru
 

Quote:
Originally Posted by Infa View Post
But still remains, I don't see how they make it only frequency dependent and not timing dependent too,, cause sound is also time, or uses time. If that time is broken up by a limited amount of pictures of the sound per second, HOW is there not timing gaps ? And more importantly how then is more pictures not smoother for timing ?

I understand he tried to explain it, but it was too geeky I think (tech talkish). And IMO didn't really prove this to be wrong. You proved it better, lol Is there any other videos on this specifically ?
I think at some point you really have to accept that sometimes when you don't understand what someone is explaining to you perhaps they're right and you're wrong, and that any proof that would satisfy you might require you to increase your understanding of "geek"....

Look at the first image in the article that monomer just linked to.
Old 23rd December 2019
  #99
Quote:
Originally Posted by Infa View Post
Thanks for the link blue. Watched it. The part specifically pertaining to this discussion is towards the end. I do like the science through the rest of the video, I get most of it.

But either he didn't take enough time explaining the "falling between the samples" thing or he just doesn't know how to describe it to a musician (lol), because I am left with still not really understanding why thats not possible.

So he scrolled a waveform around, lol means nothing to me. I have no clue at what resolution that scrolling is to prove anything, nor again, does it mean anything to me personally maybe cause I don't understand the connection between what he showed and that its negating the theory/assumption that falling between the samples is possible.

I'd like to see or hear another test/example on the subject, where the tech dude talks about only that for a while as to really help me grasp it/believe it. To me, it just still seems possible. And either way, less gaps per second... why not ? lol

I actually like your quick explanation better.. Telling me the gaps are somehow specifically designed to only be frequency related and that frequency is undetectable. Now THAT is understandable. But still remains, I don't see how they make it only frequency dependent and not timing dependent too,, cause sound is also time, or uses time. If that time is broken up by a limited amount of pictures of the sound per second, HOW is there not timing gaps ? And more importantly how then is more pictures not smoother for timing ?

I understand he tried to explain it, but it was too geeky I think (tech talkish). And IMO didn't really prove this to be wrong. You proved it better, lol Is there any other videos on this specifically ?

Either way, DUDES !!! This is awesome talk. Love it, love you all ! Hope holidays are going great and music stays dope ! Grooves and writing trumps recording quality anyway, so all this talk is just fun debates to me. I still want to understand though and want solid undeniable unarguable proof.
I think you seem to be right on the cusp of getting a key concept here... Your forthrightness in describing the parts you don't get and your admitted lack of 'geekspertise' should be helpful in coming up with a way of describing that key concept. I hope. Let me see if I can think of a way to say it that might get past some tricky bits you (if you're like most of us) might not have the math expertise to immediately grasp. As the Terminator said, I'll be back.

[Might be a couple days, I'm having a pre-holiday crunch. Hopefully, someone better at explaining such stuff than me will jump in and save you. ]


PS... the Science of Sound article monomer linked looks very helpful in addressing the associated issue of what we might call 'intersample timing' and the phase of waves being transcribed by the sampling process. You might even find that pondering that will lead you to the answers to your other questions -- though a basic understanding of the math of the sampling process is really fundamental to making sense of digital. You don't need to be able to DO the math, necessarily, as long as you can read and follow a step by step description of what the math does. (It's probably a steep read for a non-geek, but Dan Lavry's whitepaper explanation of the sampling process is a pretty good one and does lead the reader through those math processes in a reasonably hands-off kind of way. http://lavryengineering.com/pdfs/lav...ing-theory.pdf)
Old 24th December 2019
  #100
Lives for gear
 

Quote:
Originally Posted by monomer View Post
If you're talking about 'nudging' by single samples then you're right.
Yes, that is exactly what has always been my argue and point about 96k recording over lower ones. Seems like everyone overlooks or doesn't care that for me anyway, its more about the DAW inner nudging resolution/placement than the actual recording of the file.
Quote:
Originally Posted by monomer View Post
But we can do better. We can actually compute a 'nudge' that is halfway between two samples and if done correctly it will be pretty much perfect. Tho your DAW won't do that in real time.

With sampling (the theory you were scoffing at) you can basically represent any phase of a waveform, as long as its frequency is lower than half the samplerate.
Well, technically the phase you can represent is limited by the bit depth (!!) but in any case it is much more precise than the samplerate itself.
I think that for 16/44.1 it's in the nanosecond region.
So with some math you can 'nudge' a PCM signal in nanoseconds, which is thousands of times smaller than the time between the actual samples.

Maybe this article will help you understand some of it.

DAW's don't do it because it takes some computation.
If you require a sub-sample nudge there are plugins for that.
Wow ! this is some excellent crazy news I didn't know. So a plugin can allow you to place a audio file between the sample resolution of a DAW your working in ? I like it !! lol I am surprised the DAW can even allow that to happen though. As I look at a DAWS sample resolution in this case as its "rules" and how do you color outside the lines ? Seems like the DAW has no existence outside its lines to even allow something else to place it there.

And yes sir !! Going to read that article right now. Thanks !

Quote:
Originally Posted by mattiasnyc View Post
I think at some point you really have to accept that sometimes when you don't understand what someone is explaining to you perhaps they're right and you're wrong, and that any proof that would satisfy you might require you to increase your understanding of "geek"....

Look at the first image in the article that monomer just linked to.
Oh of course Matt. I do that regularly, and have done that with this. But I like to understand it at the end of the day (at least in my lingo, as I do not really need to understand the thing in overly ridiculous tech talk) that's all I'm saying. And even though I trust and believe, my point was that guy in the first video didn't really prove this subject to my mind. I didn't get the "ahhh haaa" moment. I like it when I get the ahh haa moments, cause that means it clicked and will stay.

For the there subjects, that guys video was awesome and godly. Got it immediately and didn't question because it was explained in length and detail.

Side note - , don't shoot me here, but I also didn't know if per chance it was only this guy trying to prove that and therefore his opinion and cause he is really smart he can get everyone else to believe it, so I just would like the ahhh haaa moment as it stamps it in concrete. (for me). Remember, sometimes even extremely smart people are wrong cause what looks good on paper sometimes doesn't equate to what one would expect from that paper.

The only issue I think was that guy's video just didn't stay on that exact subject long enough for him to properly show/prove his point (IMO), that's all I was saying. And I bet its all I would need.

Have some faith in me brother, I'm not dumb. I understand space to the extreme and the universe from quantum physics, quantum mechanics and quantum entanglement and planck lengths to Quasar Stars, Neutron Stars, Black holes, Pulsars (my favorite), our existence, how we sit here safely on tis planet etc.. (I could go on and on) so I am sure once someone gives me the "god particle" on this one, I'll grasp and get it no prob.

Quote:
Originally Posted by theblue1 View Post
I think you seem to be right on the cusp of getting a key concept here... Your forthrightness in describing the parts you don't get and your admitted lack of 'geekspertise' should be helpful in coming up with a way of describing that key concept. I hope. Let me see if I can think of a way to say it that might get past some tricky bits you (if you're like most of us) might not have the math expertise to immediately grasp. As the Terminator said, I'll be back.

[Might be a couple days, I'm having a pre-holiday crunch. Hopefully, someone better at explaining such stuff than me will jump in and save you. ]


PS... the Science of Sound article monomer linked looks very helpful in addressing the associated issue of what we might call 'intersample timing' and the phase of waves being transcribed by the sampling process. You might even find that pondering that will lead you to the answers to your other questions -- though a basic understanding of the math of the sampling process is really fundamental to making sense of digital. You don't need to be able to DO the math, necessarily, as long as you can read and follow a step by step description of what the math does. (It's probably a steep read for a non-geek, but Dan Lavry's whitepaper explanation of the sampling process is a pretty good one and does lead the reader through those math processes in a reasonably hands-off kind of way. http://lavryengineering.com/pdfs/lav...ing-theory.pdf)
Totally understand, take your time, and meanwhile I'll be catching up on these links you all kindly posted. The Lavry stuff I have read years ago, and Nyquist too of course. But I didn't take anything home with them. Long story.

Anyway - thanks man and thanks to you ALL !!! Merry Christmas to you all, of course this stuff comes second. I already appreciate it all so far.
Old 25th December 2019
  #101
Infa, the answers to most all your questions would be answered if you were to be able to wrap your head around the basic mathematical processes behind the Nyquist-Shannon Sampling Theorem. Again, not talking about actually being able to do the math but at least to follow the processes through sidebar descriptions and understand the basic way they work. For instance, once you understand the basics of PCM sampling that underlie most of our commercial audio production and distribution, you'll understand why even a very high frequency (but in bandlimit) wave's phase can be accurately captured down to MICROseconds (not those big, fat milliseconds were used to dealing with at the 'macro' level). But -- and I've been through this -- as long as you're trying to use metaphoric concepts (for instance, bitmap graphics, often alluded to by those looking for a conceptual 'handle' on digital audio processing), you're going to be swimming uphill. As they say. (But I'm still going to chew more on this and see if I can't figure a way to more readily communicate what I'm getting at.)

Have a great holiday!
Old 26th December 2019
  #102
Lives for gear
 
IanBSC's Avatar
Quote:
Originally Posted by Infa View Post
IMO, for now, until most artists get their head out their a$$es using "HD" (the 16 bit 44.1 level) is fine because that is the most common end result most labels/artists give the digital distribution places. When in "Ultra HD" mode (24 bit / 96-192k), I guess they are upsampling the original content for the artists that are behind the times ? IDK really. Just guessing.

Some artists did release 24/96k which in that case you will notice a good difference and benefit being in Ultra HD mode. And soon, that's the way it will be everywhere. At least the *option* to. For now just research which songs had a real 24/96k release and make your choice from that.
There are actually a lot of "Ultra HD" releases out there if you add up HDTracks, Qobuz, Tidal MQA, and the old DVD-Audio and Blu Ray formats. Some of those sites, like HD Tracks, have had a lot of pressure to not sell upsampled material, so I'd assume most of the HD stuff out there is pretty legit. After all, the only thing you need is for the ME not to convert to 16 bit and you have "Ultra HD"

I've always thought it is best to get full resolution masters of your projects in case someone wants them down the line. As someone who does listen to 24/96 etc at home, I'm often disappointed at great 90s albums that were recorded on 2", but mixed to DAT and thus sound pretty limited compared to fully analog recordings and modern 24 bit.

One argument in favor of the higher bit depths and sample rates is that it gives your converters more to work with. For example, for casual listening I prefer to convert everything to 11.2mhz DSD because I like the way it sounds. If Monty Montgomery was correct, the 16/44.1 and 24/96 files would sound exactly the same because they are being outputted in the same format with only slight tweaks to the upsampling math. And yet 24/96 and 192 tend to produce better results converted to DSD, as they also do in PCM DACs. (though 24/44 and 48 are pretty impressive as well).
Old 26th December 2019
  #103
Lives for gear
 
monomer's Avatar
 

Quote:
Originally Posted by IanBSC View Post
Some of those sites, like HD Tracks, have had a lot of pressure to not sell upsampled material, so I'd assume most of the HD stuff out there is pretty legit.
Well then i'm sorry to disappoint you.

I downloaded their sample tracks and found out that none of them had a noise floor below 16 bits.
The only thing that seemed to be actually at 24 bits were the fades at the start and end of tracks.
Some tracks had a high acoustical background noise, some had noise from equipment (including power line hum) and some had a 16 bit dither.
I even went so far as removing rumble frequencies (and indeed, some of the tracks sounded like they had vinyl rumble) and still the noise was never better than CD.

Even worse, a lot of the tracks (which were all 96kHz) had pretty prominent high frequency samples in the upper part of the spectrum (30~40kHz) that apparently were not filtered out properly (or left deliberately for whatever reason). This suggests to me that these rely on low pass filtering further down the line, maybe amps, maybe speakers, maybe ears. In any case not a cleanly recovered signal or just junk that no one was aware of because it's above hearing range.

In any case, none of these seemed to have the fidelity needed to justify 24 bits.

I tested by looking up the quietest parts of the track. Usually just after the fade-in or before the fade-out. I checked whether there was still musical content there and used only parts where that was not the case.
If the track had a mismatch between the left and right (happened a few times with apparently analog noise floor) i used the side lower in level.
The numbers i found are more or less ballpark but i think they are within +-6dB.

Here is a list i recovered from their tracks:

Hung up on my baby : -60dB
Moonjogger : -96dB
Good morning heartache : -85dB
Valse-Caprice in A-flat Major : -75dB
Down by the riverside : -73dB
Wayfaring stranger : -60dB
Sona's song : -50dB
Jahrzeit : -66dB
John Barleycorn : -74dB
Rain on me : -90dB
Toybox : -88dB
One for Tampa Red : -55dB

BTW. I'm not saying these are bad and some of them sounded genuinely like nice recordings.
But the whole HD market seems like mostly a make-money-fast-on-gullible-audiophiles scheme. They just don't deliver on the format.

Ooh, and of course seasonal greetings to everyone!
Old 26th December 2019
  #104
Lives for gear
 
monomer's Avatar
 

Quote:
Originally Posted by IanBSC View Post
so I'd assume most of the HD stuff out there is pretty legit. After all, the only thing you need is for the ME not to convert to 16 bit and you have "Ultra HD"
Another thing that bothers me about HDtracks is their 'quality assurance' page.

They claim they check for 24 'active bits' with a bit measuring tool.
But this tool is easily defeated with a small (imperceptible) level change in the 24 bit domain.
So i could basically take, say, an 8 bit recording, upscale it to 24 bits and change the level a tiny amount and it will look on that meter as if all 24 bits are 'active'. It's not a good tool for checking the actual bit depth of the production and it for sure can't guarantee the recording was really 24 bits.

Then they say they also check the spectrum and make sure there are frequencies present above hearing range (), even if it means you get severe frequency spikes that can generate IM distortion (see picture on their site). They don't seem to care about what is actually present in those frequencies or that it may lead to fidelity problems. They just check if it's there and then they call it fine.
Old 26th December 2019
  #105
Lives for gear
 
IanBSC's Avatar
Quote:
Originally Posted by monomer View Post
Another thing that bothers me about HDtracks is their 'quality assurance' page.

They claim they check for 24 'active bits' with a bit measuring tool.
But this tool is easily defeated with a small (imperceptible) level change in the 24 bit domain.
So i could basically take, say, an 8 bit recording, upscale it to 24 bits and change the level a tiny amount and it will look on that meter as if all 24 bits are 'active'. It's not a good tool for checking the actual bit depth of the production and it for sure can't guarantee the recording was really 24 bits.

Then they say they also check the spectrum and make sure there are frequencies present above hearing range (), even if it means you get severe frequency spikes that can generate IM distortion (see picture on their site). They don't seem to care about what is actually present in those frequencies or that it may lead to fidelity problems. They just check if it's there and then they call it fine.
The earliest gripes from customers were that some 96khz recordings had sharp drop-offs around 20khz which indicated 44.1khz recordings that were upsampled. Whatever it means for IMD (I personally think it's pretty irrelevant, or at least I haven't been able to hear it in any capacity) consumers interpret ultrasonic filtering as being the product of a bogus file.

My assumption is that the -60db noise floors comes from mostly tape sources, as even 16 bit will have a lower noise floor. That's fine with me, I think 24 bit reproduces tape sources better. Most of my analog equipment ranges from -75 to -100, so I'd think a 24 bit album on a Neve board etc, will not have the full 20-21 bit range. I'd definitely have no idea why a pure digital recording would have such a high noise floor, (esp if they were recorded with a minimal signal path) but perhaps that is what you get when 192khz or 352khz aren't A-weighted? Personally, I would give acoustic background noise a pass as that is part of the recording.

Since nobody is actually listening at 120db, this is probably ok. Obviously, I can't demonstrate the reasons scientifically, but I've found that 24 bit recordings have a little better quality even listening at 70db, and even with recordings that are not particularly dynamic. In theory you would have to listen at over 100db to tell the difference between 16 and 24 bit, but that has not been my experience. Whether end consumers understand this, if they like the way the files sound then they pay for them. I think all that really matters is that the audio was captured at 24 bit and left that way, even if the recording has a high noise floor.

The work flow behind what you are describing seems a little odd. My assumption would be that the mastering engineer captures at 24/96 etc, and that is the end of it. That you would get something like 24/96 converted to 16/96 and then back up to 24/96 for fades, or brick walled at 20khz all seems very strange. Not saying it doesn't happen ever, but I have no clue why it would.

IMO the biggest problem with these recordings is that a lot of (remasters) are heavily compressed and hot, largely defeating the purpose of higher audio quality.
Old 27th December 2019
  #106
Lives for gear
 

Quote:
Originally Posted by theblue1 View Post
But -- and I've been through this -- as long as you're trying to use metaphoric concepts (for instance, bitmap graphics, often alluded to by those looking for a conceptual 'handle' on digital audio processing), you're going to be swimming uphill.
Man, this here might be the best thing ever explained ! lol Funny and simple as it seems, I get this. My issue on this subject is exactly this. Even smart people (actually especially smart people) have this issue. The world is blanketed in likeliness. So many things are relative. So its nature to relate things to other things for intelligent people. I'm not ashamed. But I have been mistaken.

Here is what I was doing - Not allowing the word "resolution" to have different bases in my mind. I hear resolution related to sample rate and couple it with resolution for how my DAW resolution (grid/sample grid, etc..) works cause it is piggybacked to the resolution that sample rate uses. In a sense the sequencer and DAW resolutions are similar to video resolution/pixels, etc.. as well, so this mistake is easily made.

BUT from what I am gathering, this particular time, when dealing with Sample Rate resolution (44.1/96k, etc..), its not the same thing conception wise as what I was relating it too of what I *did* understand. THIS is what made me think its only logical that a lower sample rate resolution would cause for bigger gaps and broader "auto quantizing" (in a sense for lack of better words), because it does this on all sequencers/DAWs, etc.. So its not a dumb thought. Its just doing as you said, mistakenly relating two things that even though share a same name as resolution, shouldn't be compared.

I will now first start off by removing any preconceived thoughts on trying to relative the concept to analogies or metaphoric concepts of things I do know, and instead learn what it does from scratch. As a new concept, related to nothing.

Anyway, we got another Christmas down brohams, now for New Years ! Yay, taxes !! haha
Old 27th December 2019
  #107
Lives for gear
 
monomer's Avatar
 

Quote:
Originally Posted by IanBSC View Post
The earliest gripes from customers were that some 96khz recordings had sharp drop-offs around 20khz which indicated 44.1khz recordings that were upsampled. Whatever it means for IMD (I personally think it's pretty irrelevant, or at least I haven't been able to hear it in any capacity) consumers interpret ultrasonic filtering as being the product of a bogus file.
Yes, i can see why people think that.
But it it easily defeated by even just adding noise in the ultrasonic region.

Quote:
My assumption is that the -60db noise floors comes from mostly tape sources, as even 16 bit will have a lower noise floor.
Yes, the results are a mix of tape hiss, acoustic noise and dither. At least, that's what it sounded to me.

Quote:
That's fine with me, I think 24 bit reproduces tape sources better.
Not if the noise floor is at -60dB! You'd be coding the noise with more bits than the actual music!

Quote:
Most of my analog equipment ranges from -75 to -100,
Yeah, same here in my (very) humble setup. Can manage about 80~100dB on the analog side.

Quote:
Personally, I would give acoustic background noise a pass as that is part of the recording.
But why? It is still noise and it will still mask lower level signals. I don't think you'd hear a difference between 16 and 24 bits if the noise floor is at -60dB.

Quote:
Since nobody is actually listening at 120db, this is probably ok. Obviously, I can't demonstrate the reasons scientifically, but I've found that 24 bit recordings have a little better quality even listening at 70db, and even with recordings that are not particularly dynamic. In theory you would have to listen at over 100db to tell the difference between 16 and 24 bit, but that has not been my experience. Whether end consumers understand this, if they like the way the files sound then they pay for them. I think all that really matters is that the audio was captured at 24 bit and left that way, even if the recording has a high noise floor.
The question is whether that is due to the converter itself and not the information contained within the sample stream.

Quote:
The work flow behind what you are describing seems a little odd. My assumption would be that the mastering engineer captures at 24/96 etc, and that is the end of it. That you would get something like 24/96 converted to 16/96 and then back up to 24/96 for fades, or brick walled at 20khz all seems very strange. Not saying it doesn't happen ever, but I have no clue why it would.
Yeah, i dunno how they did those tracks. Basically every one of those demo tracks was different. Some were more open, others more forced and distorted.
It didn't seem mastering overall was the culprit. I'd say that the recordings and mixes were already limited, be it by acoustic, electrical or dither noise.
Well, technically i wouldn't be able to tell if some of the dithering was applied in mastering instead of mixing, but otoh i didn't see many of those.
Most had an acoustical or electrical noise floor so what is IMO mostly at play here is recording technique not matching up with delivery format specs.

Quote:
IMO the biggest problem with these recordings is that a lot of (remasters) are heavily compressed and hot, largely defeating the purpose of higher audio quality.
Yeah, i guess this is something that actually could help quality. The tracks with a more 'airy' dynamic range sounded noticeably better to me. Others sounded 'burned' and had an unpleasant character to them.

Ooh, and did i tell you about the one that faked it?
One of the tracks (don't remember which one) had a false bottom!
So, i examined the fade out of this track and there was a load of harsh noise (sounded like converter or dither and clearly part of the music itself as it faded out with the instruments) that dropped beyond 16 bits and basically left a clean noise-free (well, i didn't check how deep it went) recording of some sort of hall or subway station or something. In any case i heard some acoustical space not related to the recording. So my guess is they put that in there at -90dB just to make bits 17~24 move a little, giving the suggestion the actual music recording was 24 bits.

In any case, upon close examination this HD audio thing looks a huge mess of a marketplace.

Last edited by psycho_monkey; 28th December 2019 at 10:54 AM..
Old 27th December 2019
  #108
Lives for gear
 
monomer's Avatar
 

Quote:
Originally Posted by Infa View Post
I will now first start off by removing any preconceived thoughts on trying to relative the concept to analogies or metaphoric concepts of things I do know, and instead learn what it does from scratch. As a new concept, related to nothing.
Actually, you can still make metaphors using bitmaps and whatnot. The problem is that you actually need to understand sampling theory before you can apply it correctly.

Pictures and sound are just different types of (band limited) signals.
Sampling theorem actually applies to multiple things in movies. It applies to the spatial sampling (the relations between the pixels in one frame) and it applies to temporal sampling (the relation between individual frames).
Old 27th December 2019
  #109
Quote:
Originally Posted by Infa View Post
Man, this here might be the best thing ever explained ! lol Funny and simple as it seems, I get this. My issue on this subject is exactly this. Even smart people (actually especially smart people) have this issue. The world is blanketed in likeliness. So many things are relative. So its nature to relate things to other things for intelligent people. I'm not ashamed. But I have been mistaken.

Here is what I was doing - Not allowing the word "resolution" to have different bases in my mind. I hear resolution related to sample rate and couple it with resolution for how my DAW resolution (grid/sample grid, etc..) works cause it is piggybacked to the resolution that sample rate uses. In a sense the sequencer and DAW resolutions are similar to video resolution/pixels, etc.. as well, so this mistake is easily made.

BUT from what I am gathering, this particular time, when dealing with Sample Rate resolution (44.1/96k, etc..), its not the same thing conception wise as what I was relating it too of what I *did* understand. THIS is what made me think its only logical that a lower sample rate resolution would cause for bigger gaps and broader "auto quantizing" (in a sense for lack of better words), because it does this on all sequencers/DAWs, etc.. So its not a dumb thought. Its just doing as you said, mistakenly relating two things that even though share a same name as resolution, shouldn't be compared.

I will now first start off by removing any preconceived thoughts on trying to relative the concept to analogies or metaphoric concepts of things I do know, and instead learn what it does from scratch. As a new concept, related to nothing.

Anyway, we got another Christmas down brohams, now for New Years ! Yay, taxes !! haha
No! Not a dumb thought at all. Trying to use an analogy of bitmap graphics or other such possible parallels is a very natural reach -- a reach for something more readily understandable you could map your developing understanding against. That's an smart thing to do... it's just that sometimes mathematics delivers up some real surprises.

I feel like my brain is still fogged by the holidays but I think I can say the seeds for my expanded understanding of how sampling audio signals works come down to a couple of basic a prioris:

1) the understanding (learned when I was a little kid reading audio books) that all complex audio wave forms generated by sound in air effectively can be broken down into sine (sin/sinus) wave components. (To simplify, this is because of the physical properties of sound waves moving through air.)

2) the mathematical theorem that demonstrates we can accurately reconstruct and represent the shape of a sine as long as we know more than two points on its path. (Sine waves are NOTHING if not predictable. ) From this we derive the understanding that we need to take at least 2.x samples per the shortest period wave we want to capture.

That's nice for sines, I can hear someone thinking, but what about the complex wave forms full of microsquiggles we see in extreme blowups of wave forms? Those microquiggles themselves take the form of sine wave components and so are ALSO determinable by the sampling theorem -- as long as we're sampling fast enough we can determine those 2+ values within a single component...

But what if we can't sample that fast? What's the significance of that? Do we lose that information? Yes, we do -- but the implication is ALSO that such components represent high frequencies that are above the bandlimits we've selected for our sample. When we filter our DAC output to remove above frequency band limit content (to avoid alias error 'misinterpretation' related to that upper frequency bound), we do remove those above band frequency components. (And if we need broader frequency response, we can simply increase the sample rate and raise our anti-alias filter appropriately.)

Last edited by theblue1; 28th December 2019 at 06:36 PM..
Old 27th December 2019
  #110
Lives for gear
 

You guys are awesome ! Thanks.

Anyone remember Jurassic Park, the original one ?

So after reading #2 , your saying that sampling actually "fills in the blanks" successfully with frog DNA and so the body can be remade identical to what it was/suppose to be, even though/if parts are ever missing in places ?

LOL - I know I said I will do no more analogies, but this one was just too funny ! I had to. Plus, in all seriousness what you said there made me think that's kinda what its doing.

?? No ??

Quote:
Originally Posted by monomer View Post
Actually, you can still make metaphors using bitmaps and whatnot. The problem is that you actually need to understand sampling theory before you can apply it correctly.
Totally makes sense. Once one thing is not understood, the ability to use metaphors *correctly* dwindles.

Quote:
Originally Posted by IanBSC View Post
IMO the biggest problem with these recordings is that a lot of (remasters) are heavily compressed and hot, largely defeating the purpose of higher audio quality.
Quote:
Originally Posted by monomer View Post
Yeah, i guess this is something that actually could help quality. The tracks with a more 'airy' dynamic range sounded noticeably better to me. Others sounded 'burned' and had an unpleasant character to them.
Yea guys, this I fully agree with. For years now I already have backed way off the "push" on my masters. Or should I say I tell the mastering engineer to. I like a better balance of loudness vs dynamic range. Since the late 90's to now, music has been horribly too pushed. With about 2007-2015 being the highest.

It has been slowly tapering back down. But still I hear so many people ruining their recordings wanting unnatural volume. I used to fall for it when I was younger, now I don't. Volume is a knob, people can use it, so why do it ? Kids don't even care anymore, they will just use the volume knob and turn it up. So its a moot point to *overly* smash something.

Can't wait for the LUFS standard to be set in stone in our industry. Then this non-sense will be over. Sortof anyway.
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump