Jump to content
IGNORED

24/192 Downloads ... and why they make no sense?


Recommended Posts

"Carpe Diem, until a jerk get ruined it". Mostly from those with AED**, and please don't be confused since it is not ED***

 

In Chicago 1981 CES I was present on the Sony/Philips CD announcement. I was buying on the returning flight to Miami, FL, a nice LP player: Goldmund from France (the very first one). Then I didn't take too much attention to the new kid on the block: The CD.

 

On the following month I have an appointment with my MD (for health insurance matters), since I was very young by that time.

 

The good doctor showed me his new acquisition, some kind o portable CD players, plus some CDs. He was fascinated, and put the beast to play, it was (at that time) horrible. He was, even if not so old, terrible deaf, so that I have to scream in order he listen what I was telling him, the whole people in the waiting room knows everything about me...

 

But that time CD length was 650MB (74 Min.), more resolution impossible, there were no room available. Then they went to 700MB (80 Min.), but not more resolution, but to fit more music inside.

 

Somebody said but that time, that there were more room, but they don't want to use it, for commercial reasons. This, of course is a lie, because CD mechanism suffer the same malady as LPs, they are very hard to read at the inner tracks (close to the center), then the music is stored at the outside border.

 

So, the CD at 16/44 succeed at the industry standard. Some CD players manufacturers made miracles to get a descent sound. New media came latter with better resolution: SACD, DVD-Audio, etc. But the damage, for a lot of years, was done.

 

Carpe Musica,

 

Roch

 

** Acute Encapsulated Digititis

 

*** Er*ctile Dysfunction

 

Link to comment

During recording I am not sure how they would be integrated, other than some filter to account for the small but real integration error that happens because of the time it takes to collect a sample.

 

It is result of the analog antialias filter combined with ADC settling time. So practically each acquired sample is result of continuous sliding integration since last acquired sample.

 

 

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment

I am quite aware of how to align two waveforms to a resolution finer than their sampling interval. It isn't even that complex using DFT representations. And yes, it is an iterative process. That isn't an increase in the temporal resolution, and isn't going to precisely reconstruct data that was not captured.

 

OK, are we now talking about timing accuracy or transient accuracy?

 

Timing (phase) accuracy in general and between frequencies with linear phase filter and RB resolution is in sub-nanosecond range. Except if modified for example by analog reconstruction filter. My own DAC has max 15 degree linear phase shift at 20 kHz, many commercial ones have more.

 

Transient accuracy, not in terms of timing it, but in terms of attack rise/shape is commonly accurate only to ~2.5 ms on RB.

 

 

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment

"OK, are we now talking about timing accuracy or transient accuracy?"

 

Timing accuracy. That has been the issue from the start. For some reason Paul and "One and a half" introduced the transient issue that had nothing to do with the original research paper.

 

"Timing (phase) accuracy in general and between frequencies with linear phase filter and RB resolution is in sub-nanosecond range."

 

Exactly. Total agreement here.

 

 

Link to comment

After all the "new" record stores closed down including Tower Records, we have only "used" record stores in Reno, new music either has to be purchased at department stores such as WalMart or Target. So it is the internet either for physical discs or downloads, that is just the way it is today.

 

I'm no stranger to mail order as back in the 1970's I collected prerecorded Reel to Reel tapes which were impossible to find at record stores. And Muntz Stereo Tapes only had a small selection, the majority of their stock was 8-track and 4-track tapes. Cassettes were still a dictation medium back then. So I had to mail order all of my prerecorded Reel to Reel tapes, now I use the internet to purchase SACDs and downloads. If one doesn't want to use the internet, many venders such as Acoustic Sounds and Elusive Discs still take checks and money orders.

 

Great well recorded high fidelity music has never been easy to acquire but it is worth the effort IMHO.

 

I have dementia. I save all my posts in a text file I call Forums.  I do a search in that file to find out what I said or did in the past.

 

I still love music.

 

Teresa

Link to comment

Yes. I live near Southampton, UK. It is a town of about 400,000 people. Probably wetter than Reno. There is only one 'record shop' left, compared to twenty or more in the 1980s and earlier. It is not the Internet I don't like (I was using similar services at work long before the Internet started), but 'browsing the shop' is what you don't get.

 

The two large shops, run by HMV and Virgin, were close to each other in the centre of the town. Two or three large floors, and tens of thousands of CDs, plus vinyl and DVDs. One would go in with nothing particular in mind, probably because you were in the town centre for other reasons, and just browse. My interest is mainly classical, but I looked at pop music too. There were not all these silly 'genres' then, it was (and actually still is) just 'pop music'. But I bought some, and still do.

 

You would come out with maybe none at all, maybe half a dozen or so, plus a couple of DVDs of films you always wanted to see but never did.

 

You can't do that now. It is not being a 'wet blanket' as some critic foolishly said, it is just the truth. The shops are gone. Only one is left, and that is now mainly "shoot 'em up" games for the teenage brain dead with surgically attached iPod.

 

Downloads are hopefully the answer, but not yet. The UK record companies (or more accurately the UK subsidiaries of US, Japanese, and European record companies) said in 2011 that they were going to release their catalogues as downloads starting at the beginning of 2012. It is now March and nothing has happened. Buying from HD Tracks is now difficult (I can't be bothered with all this 'get a US based VPN' tedium) and the likes of Naim and Linn have very limited catalogues of extremely well recorded and produced music from mainly second rate performers you have never heard of. Antonio Forcione and the like? They don't exactly grab your attention.

 

Link to comment

To many of us here, but to Julf and Paul in particular.

 

It's interesting that many of us here claim to hear the finest nuances when it comes to audio, but can't seem to "hear" or even notice the broadest strokes when it comes to each others words. It makes me wonder.

 

-Chris

 

 

Link to comment

"It's interesting that many of us here claim to hear the finest nuances when it comes to audio, but can't seem to "hear" or even notice the broadest strokes when it comes to each others words."

 

A good observation! I can only answer for myself, in that I tend to be an objectivist - I believe what counts is what is said, not how it is said, by whom it is said, or how famous the person saying it is. I know it only works in things like science and engineering, and even there it is a fairly naive and simplistic approach, but we can all have ideals.

 

A couple of relevant links:

 

Merriam-Webster: Integrity

 

As an aside - some people seem to have an issue with Wikipedia. I find that it is usually the people who care more about "authority" of a source than the content and accuracy of the information who tend to try to put down use of Wikipedia.

 

Link to comment

"It's interesting that many of us here claim to hear the finest nuances when it comes to audio, but can't seem to "hear" or even notice the broadest strokes when it comes to each others words. It makes me wonder."

 

I do of course also acknowledge that the behavior you describe is typical of certain somewhat obsessive personality types especially common among the computer and audio communities - so probably doubly so in the computer audiophile community :)

 

Link to comment

The Wikipedia is really pretty cool when you think of it, particularly so (I think) if one's contributed to it (I have).

 

How so? Well because it really brings home the fact that it's just a bunch of bozos adding info to the world--so to speak, no stinking OED suits grinding terms here.

 

What's also interesting, just because one contributes doesn't mean that what one's added remains or isn't altered, even if the subject hasn't changed. I mention this because my main contribution happened several years ago and I went to the Wiki to glean some info from that same subject the other day. When I read the article I couldn't tell if what was written (I'd done about 1/3 to 1/2 of the article) was mine. Weird and cool (for lack of good descriptors).

 

-Chris

 

 

Link to comment

I am trying to follow what is going on with all your ear smear, and this post starts out with a few quotes which are totally irrelevant; I needed them to get where I ended ...

 

J: What I am saying is that that is incorrect. Even if the material has been sampled at 44.1 kHz, a suitable upsampling DAC with a filter placed high enough will make the time smear inaudible - or at least move the time smear to your tweeter instead. So we don't need hi-res sample rates in order to avoid time smear.

 

Define "time smear" here. To me, at this stage of the thread (http://www.computeraudiophile.com/content/24192-Downloads-and-why-they-make-no-sense#comment-134105), it looks like balony.

 

P: Time smearing in the filters is one reason

 

Here Time Smear fits perfectly.

 

J: Do you agree with his statement about 44.1 kHz systems being incapable of sub-10 microsecond timing precision?

 

This too looks correct to me. But timing precision is different from time smear.

 

P: 44,100 samples per second = 1 sample every 22.67+ microseconds.

 

P: No, I do not think this is capable of accurately recording / replaying a sub 10us transient.

 

Keep in mind what a transient is; it is a transition from one (level) to the other. It is NOT a frequency.

A transient has no length in the time domain, or at least for this discussion we better think it hasn't. BUT :

When in real life the sampler samples 44100 times per second, while the transient will happen at any random time, it will be missed for its exact point in time of the happening.

So what ? big deal. It will be "seen" when the next first sample (at sampling the signal) comes about.

Wat is time accurate ? NO.

Will a 192KHz sampler have caught it time accurately ?

NO. But better.

 

When normal filtering is applied, the transient will be literaly smeared away from its original position. It will be flattened also.

 

When the signal is upsampled with a means of real interpolation, this will happen :

The peak of the transient is connected to its base (I'm looking backward in the time domain), and it will go via a slope the amount of injected (new) samples imply. The transient will be as steep per time unit (read my before post about this !), but, its starting point is put backwards in time. The peak (top) of it remains as inaccuracte as it was in the time domain, but its envelope starts earlier than it did.

Same will happen with any decay when there, but this is not part of the story (or otherwise our transient would start to be a frequency indeed).

 

With described means, the transient will be averaged along the time axis for 1. its dynamics and 2. the point in time it really happened and when looking at the average amplitude. This latter is a moot thing, because we will be perceiving the peak only, BUT it will have become less dynamical. So, point in time is as wrong, but the emphasis was taken away.

Remember, this assumes analogue can follow, because when not all is moot, and any discussion is useless.

 

Thus, when analogue can follow, the 192 sampler, being live taken or upsampled, will smoothen the transient for its steepness.

 

True ?

No. The 192 live sampler will have an as steep transient as the 44.1 live (or decimated) sampler, because we assumed it infinitely steep (again, read my earlier post about this).

The 192 transient now has no chance anymore in averaging out its actually still wrong point in time. Unless it is upsampled again of course.

 

It is here where 44.1 material, properly upsampled may start to sound better than 192 etc.;

Precicely denoted points in time WHICH ARE OFF are spread towards their more original position AND towards the more wrong position just the same, but since the steepness is less it is more comfortable for many things. Amongst which :

... when analogue CAN NOT FOLLOW.

 

... And not any analogue can follow infinitely steep steps ...

 

Hey, and I said my NOS1 can easily follow that (earlier post) ?

 

Yes. But what I did not tell -but which is obvious for those who get the grasp of it in the first place- is that everything, including the transient(s) we talk about here, is upsampled 16x. So, the transient now is not happening by means of one sample hence one step - no, it happens in 16 steps. And now it forms a nice slope to everything which can't deal with inifinitely steep slopes (which is all analogue !). Peak remains at the exact point in time (as off as it was), but the "harshness" of it being off is smeared over 16 samples now.

 

Being where I am, now compare with the super-smearing common filtering. You wouldn't see any transient anywhere. Its energy now is smeared (oh yes) over numerous samples (think 128 for your comfort, or 1024 for more), BUT, is included in all the adjacent samples. No peak will stick out anywhere, but you should be able to see the general wave going up slightly.

This is perfect for analogue.

It is also perfect for super vagueness or the opposite of freshness. Additionally that little lump with still its peak somewhere, may happen milliseconds later in time now.

 

 

Before we rattle about how 44.1 can be time-accurate into the microsecond level, we better KNOW that nothing of the kind will even happen with that commonly used filtering.

 

Peter

 

 

 

 

 

 

 

Lush^3-e      Lush^2      Blaxius^2.5      Ethernet^3     HDMI^2     XLR^2

XXHighEnd (developer)

Phasure NOS1 24/768 Async USB DAC (manufacturer)

Phasure Mach III Audio PC with Linear PSU (manufacturer)

Orelino & Orelo MKII Speakers (designer/supplier)

Link to comment

Referring to his graph :

 

J: So as you can see, it is quite possible to reproduce time differences smaller than one sample.

 

P: Those look time synchronous to me, but with differing amplitudes. Is the scale too small? I would think a 5us delay would show as half a division.

 

Paul, Julf for sure is correct in stating the 5 microsecond thing. But Julf doesn't seem to be able to clearly state what it is about, and you are fairly closed to the gestes this is about. Ok, at least I seem to see them.

 

Although Julf talks about the time domain, literally it isn't. It will be merely about the amplitude domain. However, the time domain implies it. So, because in the time domain 5us is shifted, the sound will be different. All amplitudes have changed. For the 44.1 sample rate, I'd say that 11025 different possibilities exist, all sounding different.

But it can not be explicitly utilized; not with/for normal music.

 

Notice that when a recording would be played back and recaptured from analogue into digital (say at 44.1), the chance that two recordings sound the same will be 1:11025 because of this phenomenon.

 

And FYI : The most similar happens when you start your (ripped) CD one second from the normal start; The filtering will workout differently throughout. Nice eh ? So now you can play all your albums again from off the 1 second mark, and they will all sound different. Well, they should, because the net result now is different.

But this is off topic.

 

 

 

Lush^3-e      Lush^2      Blaxius^2.5      Ethernet^3     HDMI^2     XLR^2

XXHighEnd (developer)

Phasure NOS1 24/768 Async USB DAC (manufacturer)

Phasure Mach III Audio PC with Linear PSU (manufacturer)

Orelino & Orelo MKII Speakers (designer/supplier)

Link to comment

Keep in mind what a transient is; it is a transition from one (level) to the other. It is NOT a frequency.

 

Well, of course it is frequency too. Steepness of the step defines frequency content. You can seamlessly transform between time and frequency domains, back and forth.

 

Frequency content determines the transient. When you AD-convert a transient, the analog antialias-filter defines the rise times which cannot be faster than sample period, otherwise you'd get aliasing again.

 

And now it forms a nice slope to everything which can't deal with inifinitely steep slopes (which is all analogue !).

 

There are no infinitely steep slopes in analog domain. If it's from microphone, it is limited by the mic frequency response and if it's from analogue electronics, it is limited by the rise times (V/µs). And naturally in analogue, rise times can be translated to gain bandwidths too.

 

Here's frequency content of a single step:

step.png

 

 

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment

There are no infinitely steep slopes in analog domain. If it's from microphone, it is limited by the mic frequency response and if it's from analogue electronics, it is limited by the rise times (V/µs).

 

Of course the physical world is "analog" also (at reasonably macro dimensions, anyway), so whatever is making the transient, whether it is the stick and drum with a rimshot, or speakers with a synthesizer attack, won't produce a transient with infinitely steep rise time. Some of them can be quite steep, though! :-) I'm not acquainted with what human audio sensitivity is to such transients versus pure tones at various frequencies - whether it differs at all, and if so, how.

 

One never knows, do one? - Fats Waller

The fairest thing we can experience is the mysterious. It is the fundamental emotion which stands at the cradle of true art and true science. - Einstein

Computer, Audirvana -> optical Ethernet to Fitlet3 -> Fibbr Alpha Optical USB -> iFi NEO iDSD DAC -> Apollon Audio 1ET400A Mini (Purifi based) -> Vandersteen 3A Signature.

Link to comment

But it can not be explicitly utilized; not with/for normal music.

 

I've been half-following this via e-mail, so I haven't looked at any of the graphs, and thus I may be misunderstanding everything. But my more or less vague impression is that what Paul was describing might be applied to transients, while what Julf was describing might be applied to differentiate between two audio sources, one slightly time-shifted from the other. In terms of application to "normal" music, I think the fact that music contains transients is pretty apparent. Regarding two audio sources, it seems to me sensitivity to time-shifting is probably part of what helps us to determine, for example, that a vocalist's voice on a recording has been "doubled" (he/she has dubbed over his/her voice to make it sound fuller) - it would be near impossible to have such layered recordings timed so exactly that we would think it was a single vocal performance.

 

One never knows, do one? - Fats Waller

The fairest thing we can experience is the mysterious. It is the fundamental emotion which stands at the cradle of true art and true science. - Einstein

Computer, Audirvana -> optical Ethernet to Fitlet3 -> Fibbr Alpha Optical USB -> iFi NEO iDSD DAC -> Apollon Audio 1ET400A Mini (Purifi based) -> Vandersteen 3A Signature.

Link to comment

Well, of course it is frequency too.

 

Yea, well, I referred to my other post for a reason.

So it (for sure) is, but it is unrelated (up to "it is not") to what we talk about here.

 

Especially when you show a sheer frequency (which you did), it is not.

 

Nothing wrong with the contents of your post, but it confuses.

 

You can seamlessly transform between time and frequency domains, back and forth.

 

I mentioned/implied this many times in that other post.

 

There are no infinitely steep slopes in analog domain. If it's from microphone, it is limited by the mic frequency response and if it's from analogue electronics, it is limited by the rise times (V/µs). And naturally in analogue, rise times can be translated to gain bandwidths too.

 

True. But it is also not the subject. The subject is, per my transformation of the whole subject, that no matter what the transients are, they are there in digital (mind you, "Hires"). And then I don't care whether the quote above is true or not (no matter it is :-);

We have to deal with those transients and ...

Well, I don't what to repeat what has been said already (post from yesterday).

 

That this all urges for the even higher sample rate is ...

logic.

The sheer fun of it could be that I seem to have reasoned out (but merely implied) that

 

a. the sample rate should be high enough so that no other stepping emerges than the smalles step possible (DSD seems okay to me for this)

 

or

 

b. the not at all high sample rate is upsampled such that enough "smear" will happen so the 100% vertical digital transient is turned into acceptable slopes hence in analogue sines can do the job needed.

 

Something like 192 does not fit in either a. or b. although b. can be applied to it again.

Still upsampling from 44.1 will smear the vertical digital transient better than the same one-sample transient in 192.

 

And so I now think I am on to something ...

 

 

Lush^3-e      Lush^2      Blaxius^2.5      Ethernet^3     HDMI^2     XLR^2

XXHighEnd (developer)

Phasure NOS1 24/768 Async USB DAC (manufacturer)

Phasure Mach III Audio PC with Linear PSU (manufacturer)

Orelino & Orelo MKII Speakers (designer/supplier)

Link to comment

Although it has been all in my post from yesterday, it seems good to emphasize a part from it. Look here :

 

 

 

This is part of a 176.4 track as it was recorded from vinyl by a PMII.

Look at the smallest amplitude steps (vertical) versus the largest (or larger) ones;

The largers ones are there because the sample rate (thus 176.4 here) are not sufficient to turn these larger steps into the smallest.

 

Remember, this is a perfectly legal filtered file and nothing needs to be done to it at D/Aing it.

 

It is these unnecessary large steps which I call "transients" here. So, nothing like a rimshot, but micro detail of what happens in there and how it works out. Ehm, squares.

 

These transients are fed to analogue, and since they are completely vertical, some nasty things should happen to the analogue parts.

While this is a most mild example, I know that half of recordings contain transients which cover for more than half of the voltage range. So, that would be something like 1.4Volt and more. This happens (must happen, should happen) in a time frame which is infinitely small. Digital depicts that. And analogue can't do it. It is only the extend of how it can be done which determines how good the result will be (sound wise).

 

NOTICE : The example from the picture is NOT what I am talking about, because a frequency is shown. The transients which cover for the 1.4V mentioned, IS what I talk about because it can't be a frequency (unless it would be a high frequency test signal). It even will be so that this can only be about a transient in the lower regions of the spectrum, because it -jumping like that- can only be about the base wave form (the higher frequencies riding on it).

 

When such a transient happens as steep at the higher sample rate, it better had been in the lower sample rate, because it could have been "smeared" better by means of upsampling.

 

Maybe a number for you :

In XXHighEnd there is a self-check for it (me) doing things wrong, and when static would emerge. This checks for these kind of wild transients, and initially it trips when 2/3 of the total voltage range is exceeded. But, still 2000 of those excessions are allowed or otherwise nothing would play.

Going lower than this 2/3 would let trip many albums/tracks.

 

Back at the time I had this checked by a recording engineer, and besides he completely agreed with me ("no, you are not looking at wrong things") he told me that already a woman voice can incur for such transients. Mind you, not as a frequency, but as a first "burst" (and preceeded by a little swing-up because it ever was analogue, and analogue does that / needs that).

 

Lush^3-e      Lush^2      Blaxius^2.5      Ethernet^3     HDMI^2     XLR^2

XXHighEnd (developer)

Phasure NOS1 24/768 Async USB DAC (manufacturer)

Phasure Mach III Audio PC with Linear PSU (manufacturer)

Orelino & Orelo MKII Speakers (designer/supplier)

Link to comment

Linn (UK) speakers, with the exception of the 'Isobarik' which design they purchased from someone else and never fully understood, used to focus on fast transients to the exclusion of almost everything else. Aided and abetted by Naim's (who worked with them closely) attitude at the time.

 

As a result most people found them earsplittingly unlistenable. They certainly grabbed your attention in the showroom, but if you took them home even the cat left after a few minutes.

 

Link to comment

"my more or less vague impression is that what Paul was describing might be applied to transients, while what Julf was describing might be applied to differentiate between two audio sources, one slightly time-shifted from the other."

 

Exactly. Because the Kunchur paper that "one and a half" quoted is talking about the ability of human hearing to pick up time differences. Somehow that was extrapolated to the false claim that a 44.1 kHz sample frequency system wouldn't be able to reproduce those time differences. The transient discussion is another issue altogether...

 

Link to comment

"Julf for sure is correct in stating the 5 microsecond thing. But Julf doesn't seem to be able to clearly state what it is about"

 

Fair enough, we definitely seem to have a communication issue. So let me try to one more time, if you can bear with me.

 

So, let's again look at the picture:

 

 

 

The blue wave is a 3 kHz sine wave. The reddish brown staircase is that wave sampled using a 44.1 kHz sample frequency.

 

The purple wave is also a 3 kHz sine wave, but it starts 5 us later than the first (blue) wave. The green staircase is the sampled version.

 

As we continue to sample both waves every 22.7 us (44.1 kHz), our ADC will see slightly different values because the second wave is "lagging behind" by 5 us - so the samples will be taken at slightly different points along the wave.

 

When we reproduce the sampled signals, again at 44.1 kHz, what will be reconstructed is the two 3 kHz sine waves, with a 5 us time difference between them. Both will be reproduced OK, with the 5 us time difference preserved.

 

I again repeat the suggestion I gave for testing this out in practice: record a signal at 44.1 kHz. Upconvert it to 192 kHz. Shift it by one sample interval (5.2 us). Downsample it back to 44.1 kHz. What will you see? The original signal, but shifted by 5.2 us.

 

I hope this helps clarify my point.

 

Link to comment

"with the exception of the 'Isobarik' which design they purchased from someone else and never fully understood"

 

As a result, every "improvement" that Linn did to the Isobariks during their lifetime was a backwards step, and the successor, the Keltik, was pretty much a failure...

 

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...