The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
Have the higher sampling rate and bit depths cured the digital monster? Virtual Instrument Plugins
Old 12th August 2014
  #181
Lives for gear
 

Quote:
Originally Posted by bogosort View Post
I completely disagree. One cannot digitize an actual audio signal with math, no matter how old the math is. We need physical circuits to sample physical signals, but how do we build such things? We require guiding principles that enable us to design our circuits such that they meet performance specifications. These principles come from information theory. The math is just a tool.
of course it's all math, that's my point.......... information theory is just the governing rules so to speak. The optimal implementation using the math that's been around for centuries.


Quote:
Originally Posted by bogosort View Post
Why are you bringing up implementation details?
just to give examples....you stated originally information theory relating to digtial audio is all maths, that is not accurate. The math to store continuous functions using binary has existed long way before information theory and even computers. It's not a big deal I was just pointing out the inconsistency of your statement.


Quote:
Originally Posted by bogosort View Post
In any case, arrays, vectors, matrices -- these are conceptual containers, a convenient way for humans to think about data. At the hardware level, the computer doesn't have the foggiest notion of what an array is. As for Fourier Transforms, the (mathematical) FFT algorithms that I've seen don't use (mathematical) matrices. Of course any software implementation will use arrays.
I realize CPUs don't use arrays at the hardware level but we were discussing the mathematics that digital audio requires as a function of storage and retrieval. Certainly you don't need arrays or matrices but it makes things a lot easier. Why do you think modern plugins are starting to utilize GPUs? A GPU does one thing... linear algebra native to its hardware level.
Old 12th August 2014
  #182
Gear Guru
 

Quote:
Originally Posted by PorchBass View Post
- DSD seems a very 'pure' concept with no imaginary 'staircase' waveforms.
here in a nutshell is the entire problem with the level of discussion here.

Shallow understanding of the concepts behind it, and lame oversimplified visualizations of "how it works" influence already placebo-ridden listeners to imagine they are - as in the notorious example - hearing the "stairsteps".

So, even if you admit the staircase is "imaginary", you still think something that conceptually doesn't have stairsteps will "sound better".

As for "purity", not everything that "seems pure", IS pure. Not everything that IS pure, sounds better.
Old 12th August 2014
  #183
Quote:
Originally Posted by Lance Lawson View Post
The Bumble Bee is a way of saying that the math of something does not always work out to reality. Bumble Bees fly so the math against them flying is wrong.
No, it just means there's a set of circumstances where further explanation is needed.

Quote:
Originally Posted by Lance Lawson View Post
44/16 is mathematically more than what is needed to "fool" the ear into thinking that the reassembled wave is the same as a genuine wave that's never been reduced to numbers and then returned to a reconstituted waveform. If it was the debates would have never started. The math is wrong. I know the math is wrong when I experience the sound of higher sampling/bit rates.
but you didn't experience the maths. You experienced the physical manifestation of it. There's nothing wrong with the maths - if it were, your converter wouldn't work at all. It's just an imperfect world which creates imperfect replicas. No-one is claiming conversion is perfect; certainly no-one is claiming early CD transfers were perfect!


Quote:
Originally Posted by Lance Lawson View Post
Yes in the early days there were lots of missteps with mastering and recording so when you buy a CD it's a crap shoot in some ways. But it isn't a crap shoot when you can in the comfort of your own studio make a recording at 44/16 then make a recording at 96/24. It is then possible to A/B without the crapshoot. This is what has sold me on 96/24. I proved it to myself without the question marks of who how and where it was mastered. It doesn't get any more direct than than that.
How much clearer do I have to make what I'm trying to say - you're still confusing production with delivery format.

Try this. Make your recording at 24/96. Mix your recording at 24/96. When you have a finished product, maybe with a touch of limiting (or not - but definitely at full scale, so normalise that 24/96 master) - THEN do the properly dithered conversion to 16/44.1 and ABX those two.

Again - no-one is saying record at 16bit - I don't think we've done that since maybe mid 90s in many cases?

If you can still tell this difference of "thin" sound at 16/44.1...well then I guess you've got a point. But saying "I tracked at 16bit and it didn't sound good" is missing the point completely.
Old 12th August 2014
  #184
Quote:
Originally Posted by PorchBass View Post
I would love to see cheap multichannel DSD ADCs come onto market - DSD seems a very 'pure' concept with no imaginary 'staircase' waveforms.
Love the machine!
It's a shame DSD is so impractical then!

"Staircases" when it comes to PCM audio are exactly that - imaginary - of course. Providing S-N is followed, a theoretically perfect converter puts out what it gets in.

The problem comes making that converter, and also the analogue stages of it.
Old 12th August 2014
  #185
Lives for gear
 
bogosort's Avatar
Quote:
Originally Posted by IanBSC View Post
The thing with high sample rates is not that they reproduce higher frequencies (which we can't hear), it is that they accommodate more gradual antialiasing filters that have a noticeable impact on sound quality. If they could build a 44/48k sample rate converter that didn't distort impulse response and create pre and post-ringing it would sound just as good.
Then this issue should be solved for you. Modern converters use local oversampling to widen the transition bands for the anti-alias/image filters. In other words: both 44.1k and 96k target rates will use the exact same analog filters; the decimation filters will be digital and thus completely specifiable, and it is trivial to create digital filters with passband ringing less than 0.1 dB.

Quote:
DSD also has no filters, and thus sounds really good despite being 1 bit.
DSD players include a LPF (sometimes user-selectable) to prevent too much ultrasonic hash from getting into your playback chain, where it could generate IMD products or blow up a tweeter.

It's puzzling to me why an engineer will balk at the idea of carefully crafted filters in a converter when every single device in his chain is acting as a filter, intentionally or not.
Old 12th August 2014
  #186
Lives for gear
 
bogosort's Avatar
Quote:
Originally Posted by chainrule View Post
of course it's all math, that's my point.......... information theory is just the governing rules so to speak. The optimal implementation using the math that's been around for centuries.
Good way of saying it: information theory lays out the governing rules.

Quote:
The math to store continuous functions using binary has existed long way before information theory and even computers. It's not a big deal I was just pointing out the inconsistency of your statement.
As far as I know, the mathematical framework for sampling a continuous time signal was developed in the 20th century. Can you point me to earlier sources?
Old 12th August 2014
  #187
Motown legend
 
Bob Olhsson's Avatar
 

Quote:
Originally Posted by bogosort View Post
Then this issue should be solved for you. Modern converters use local oversampling to widen the transition bands for the anti-alias/image filters...
The real issue is the quality loss below 20 kHz. from multiple down-sample/up-sample processes in succession. Filters vary greatly in quality and none are really perfect. Taking a single conversion stage out of context is meaningless.

My personal experience has been that recording at 96k using digi 192 HD converters, mixing and mastering at 96 followed by an analog or digital conversion down to 44.1 sounds better than recording at 44.1 in the first place. The entire process must be viewed as a system. Considering the growing importance of video as an income source combined with the fact that contemporary video playback systems are likely to be higher quality than their music counterparts, I think it's ridiculous to record anything below 48k and if at all practical 96k makes the most economic sense by far. Every recording project represents a very significant investment in both time and money.
Old 12th August 2014
  #188
Lives for gear
 
IanBSC's Avatar
Quote:
Originally Posted by bogosort View Post
Then this issue should be solved for you. Modern converters use local oversampling to widen the transition bands for the anti-alias/image filters. In other words: both 44.1k and 96k target rates will use the exact same analog filters; the decimation filters will be digital and thus completely specifiable, and it is trivial to create digital filters with passband ringing less than 0.1 dB.
Most high end converter manufacturers that I know of from Mytek, to DAD, to JCF, Ayre and even Dan Lavry would disagree with you that the issue of digital filters has been solved. Why do you think the Bricasti M1 DAC has 6 selectable filter styles? I believe the DCS 904 has like 5 filter types. Each filter type has different tradeoffs and oversampling has not solved them. The only thing that preserves the pulse perfectly, has no phase shift, pre-ringing, or post ringing is DSD.

Both DCS and DAD/Merging believed 352khz gave them the bandwidth to build the perfect Guassian filter. Some listeners still preferred DSD. Others manufacturers have preferred minimal phase moving average filters, which also require large bandwidth because while they are fast, they don't filter so well.

Much literature already exists about this.

From my experience, analog LPFs sound much more benign. As I understand, the JCF Latte uses no digital decimation filters at 192k and does all the filtering with passive analog components.
Old 12th August 2014
  #189
Lives for gear
 
bogosort's Avatar
Quote:
Originally Posted by Bob Olhsson View Post
The real issue is the quality loss below 20 kHz. from multiple down-sample/up-sample processes in succession. Filters vary greatly in quality and none are really perfect. Taking a single conversion stage out of context is meaningless.
What's meaningless is ascribing an unspecified in-band quality loss to some supposed multiple, sequential SRC processing. Where is all this successive SRC happening? The converter inputs are oversampled, not upsampled. Once the signal is digital, it is decimated to the target rate -- and stays at that target rate until the DAC does its thing. If you're talking about a plugin potentially upsampling behind the scenes, any audible deleterious affects will be, well, audible -- so remove the offending plugin.

No filters are perfect, but digital filters can be made arbitrarily close. Of course I am not saying that every digital filter implementation is equally good, but it seems too often we listen with our imagination rather than our ears.

Quote:
My personal experience has been that recording at 96k using digi 192 HD converters, mixing and mastering at 96 followed by an analog or digital conversion down to 44.1 sounds better than recording at 44.1 in the first place. The entire process must be viewed as a system.
It may well be that digi 192 HD converters perform better at 96k (though it's very likely that there is as much SRC at 96k as there is at 44.1k).

Quote:
Considering the growing importance of video as an income source combined with the fact that contemporary video playback systems are likely to be higher quality than their music counterparts, I think it's ridiculous to record anything below 48k and if at all practical 96k makes the most economic sense by far. Every recording project represents a very significant investment in both time and money.
How does 96k make more economical sense than 48k, which is the standard in the TV and film industry? Doubling the bandwidth doubles the resources required, for at best a tiny, subjective improvement. I believe that there is no fidelity to be gained by increasing the sampling bandwidth, but even if you don't, you surely acknowledge that you're chasing diminishing returns, right? An economic argument for higher sampling rates just doesn't make sense.
Old 12th August 2014
  #190
Lives for gear
 
bogosort's Avatar
Quote:
Originally Posted by IanBSC View Post
Most high end converter manufacturers that I know of from Mytek, to DAD, to JCF, Ayre and even Dan Lavry would disagree with you that the issue of digital filters has been solved. Why do you think the Bricasti M1 DAC has 6 selectable filter styles?
Differentiation in the market place? Creeping featurism? This one goes to 11?

Quote:
Both DCS and DAD/Merging believed 352khz gave them the bandwidth to build the perfect Guassian filter. Others have preferred minimal phase moving average filters.
Because both Gaussian and MA filters are terrible in the frequency domain, literally terrible. If you haven't already, I recommend spending a few hours with Matlab or Octave or similar, loading a familiar WAV file, and playing with various filter designs. It's one thing to look at pictures of IRs and infer what you'd hear, and another thing entirely to actually listen.
Old 12th August 2014
  #191
Lives for gear
 
bogosort's Avatar
Quote:
Originally Posted by IanBSC View Post
From my experience, analog LPFs sound much more benign. As I understand, the JCF Latte uses no digital decimation filters at 192k and does all the filtering with passive analog components.
Analog LPFs sound much more benign to your ears or your imagination? I ask earnestly, because it is easy to create a digital LPF with the exact transfer function of an analog LPF. Going the other way around is an entirely different matter.
Old 12th August 2014
  #192
Gear Guru
 

Quote:
Originally Posted by psycho_monkey View Post

Try this. Make your recording at 24/96. Mix your recording at 24/96. When you have a finished product, maybe with a touch of limiting (or not - but definitely at full scale, so normalise that 24/96 master) - THEN do the properly dithered conversion to 16/44.1 and ABX those two.
But isn't that EXACTLY what Lance is saying ??

Quote:
Originally Posted by Lance Lawson View Post
I record in high rate and I know once it's reduced to CD it is a shadow of the high res master.
I mean, except for the ABX part

People can quibble about the bona fides of this or that study, but in all the studies, anyone who has tried it blindfolded, including those who claim success, admit it was "difficult" and that they were "unsure" of their choices.

This flies in the face of claims that "reduction" to CD of a good sounding master results it being a "shadow" of its former self. That implies a drastic reduction in quality. Not something were people have to crank the volume and concentrate on the reverb tails to 'tell'.

It has become impossible to discuss this stuff when the hyperbole surrounding these distinctions has gotten so far out of hand. When people denigrate blind ABX comparisons, they put the entire thing entirely into an 'unarguable' subjective area where they can pretty much say anything they want, and be forever smug in their perceptions.
Old 12th August 2014
  #193
Lives for gear
 
IanBSC's Avatar
Quote:
Originally Posted by bogosort View Post
Analog LPFs sound much more benign to your ears or your imagination? I ask earnestly, because it is easy to create a digital LPF with the exact transfer function of an analog LPF. Going the other way around is an entirely different matter.
What I can say is that having heard DACs that operate with both analog and digital filters, and those that only use analog filters, NOS PCM and DSD, the ones without the digital filter were my preference.
Old 12th August 2014
  #194
Lives for gear
 
IanBSC's Avatar
Quote:
Originally Posted by joeq View Post
But isn't that EXACTLY what Lance is saying ??



I mean, except for the ABX part

People can quibble about the bona fides of this or that study, but in all the studies, anyone who has tried it blindfolded, including those who claim success, admit it was "difficult" and that they were "unsure" of their choices.

This flies in the face of claims that "reduction" to CD of a good sounding master results it being a "shadow" of its former self. That implies a drastic reduction in quality. Not something were people have to crank the volume and concentrate on the reverb tails to 'tell'.

It has become impossible to discuss this stuff when the hyperbole surrounding these distinctions has gotten so far out of hand. When people denigrate blind ABX comparisons, they put the entire thing entirely into an 'unarguable' subjective area where they can pretty much say anything they want, and be forever smug in their perceptions.
And what would you say if he actually carved out a bunch of his time to set up and conduct an ABX test, and still came out saying the the down converted version was a shadow of its former self?
Old 12th August 2014
  #195
Motown legend
 
Bob Olhsson's Avatar
 

Believe me I would absolutely love to hear no difference between 96k and 44.1production from a financial standpoint.

I and many others do hear a difference and there are a number of good technical reasons why one might hear a difference when you look at recording production as a system using real world converters and software rather than simply a single conversion process.

Theoretically it should be possible to use lower sample rates but that doesn't mean it's always practical or that those of us who do hear a difference are delusional. Audio quality has become especially important because of the amount of lossy encoding that has become standard in today's world. It's easy to demonstrate that the better quality you start with, the better the final product will sound.
Old 12th August 2014
  #196
Gear Guru
 

Quote:
Originally Posted by IanBSC View Post
And what would you say if he actually carved out a bunch of his time to set up and conduct an ABX test, and still came out saying the the down converted version was a shadow of its former self?
the point is he has not

and like you, he never will because that might force him into the uncomfortable position of confronting his placebos. Instead, he will say things like: 'the difference is so obvious, I don't NEED to test it". Sound familiar?

what I find most telling is that the sighted listening types will not conduct such a blind ABX even secretly in the privacy of their own home. They wouldn't even have to tell anybody they did it. Or admit their results if they did poorly. It's not the public reputation they are afraid of, it is the internal cognitive dissonance.

Calling something a "shadow" is strong words. I would expect 10 out of 10 score in multiple trials. Even if someone scores perfectly, there is also the matter of whether he had to 'sweat' to do so.


Quote:
Originally Posted by Bob Olhsson View Post
Believe me I would absolutely love to hear no difference between 96k and 44.1 production .
I am not talking about producing your entire project at 44.1. I am citing Lance's claim about producing your project at high-res, mixing at high res, and hearing a drastic loss in quality when the final product is put on CD.
Old 12th August 2014
  #197
Lives for gear
 
IanBSC's Avatar
Quote:
Originally Posted by joeq View Post
the point is he has not

and like you, he never will because that might force him into the uncomfortable position of confronting his placebos. Instead, he will say things like: 'the difference is so obvious, I don't NEED to test it". Sound familiar?
How do you determine what doesn't need blind testing? What, if anything, is safely not a placebo?

I agree with what he "would" say. I don't need a blind test to hear the difference, and so taking up my time to set up a proper blinded record and playback system would be a waste. It would be interesting, but not worth my time.
Old 12th August 2014
  #198
Motown legend
 
Bob Olhsson's Avatar
 

For "the public at large" the standard is actually 48x24 or lossy encoded 48x24.
Old 12th August 2014
  #199
Lives for gear
 
bogosort's Avatar
Quote:
Originally Posted by Lance Lawson View Post
24 bit audio in general is a huge improvement across the board but I'm not sure most consumers even know to seek 24 bit out.
There's no good reason for music listeners to seek out 24-bit recordings.

Quote:
In any event this is what I'm talking about. These samples are full wave one is 96/24 the other is 44/16. No alterations were made with regards to eq compression etc.
How are you listening to these, at 44.1 or 96? I tried listening at both rates and couldn't hear a difference between the files, though both sound really harsh so I had to stop listening.

Interestingly though, the 96k file has zero energy above 24 kHz. (The CD version tops out at the expected 22.05 kHz.) Was it tracked at 48 kHz?
Old 12th August 2014
  #200
Lives for gear
 
bogosort's Avatar
Quote:
Originally Posted by IanBSC View Post
How do you determine what doesn't need blind testing? What, if anything, is safely not a placebo?
When there's no controversy or discrepancy, there's no need for ABX. But when someone says "night and day difference" and someone else says "I don't hear it", ABX is appropriate.

Obviously everyone is free to do and think what they like, but in a public discussion about sampling rates, I consider it a matter of intellectual and professional integrity that its participants at least try a private, personal ABX.
Old 12th August 2014
  #201
Gear Guru
 

Quote:
Originally Posted by bogosort View Post
When there's no controversy or discrepancy, there's no need for ABX. But when someone says "night and day difference" and someone else says "I don't hear it", ABX is appropriate.

Obviously everyone is free to do and think what they like, but in a public discussion about sampling rates, I consider it a matter of intellectual and professional integrity that its participants at least try a private, personal ABX.
thank you!
Old 12th August 2014
  #202
Lives for gear
 
IanBSC's Avatar
Quote:
Originally Posted by bogosort View Post
There's no good reason for music listeners to seek out 24-bit recordings.
As a consumer as well as engineer I have to emphatically disagree.

Not all music or listening environments are appropriate 24 bit, but on well recorded music in quiet listening focused environment, and hopefully a nice system, it absolutely is an improvement in terms of microdynamics, seperation/masking, smoothness and dynamic range. Not that PCM sounds ideal to me.
Old 12th August 2014
  #203
Lives for gear
 

Quote:
Originally Posted by IanBSC View Post
No ... what he means is that there's no good reason for music listeners to seek out 24 bit releases (this is a true thing ...). Better for you?

Re. your edit -- well, OK. He (and I) were generalising, as were you. Now that you've elaborated - very quiet room ... very low noise production, then yes, you might be able to hear 16 bit dither. If it's noise shaped you probably won't even then. So, given that the majority of music isn't dynamic enough to benefit, and the majority of listening environments are too noisy to hear the difference ...

There's no good reason for the vast majority of music listeners to seek out 24 bit releases ... better?
Old 12th August 2014
  #204
Lives for gear
 
IanBSC's Avatar
Quote:
Originally Posted by -tc- View Post
No ... what he means is that there's no good reason for music listeners to seek out 24 bit releases (this is a true thing ...). Better for you?
A true thing?
Old 12th August 2014
  #205
Gear Guru
 
kafka's Avatar
I still think I nailed this thing on page 3. Of the very few people people who think they can perceive the difference between 44.1/16 and 96/24 (let's be fair, and call it 96/20. The other 4 bits are just there for engineering convenience), vastly fewer think the difference is clear, and even fewer care. All of the content that matters is easily contained in 44.1/16. That, to me, is a conclusive argument against any economic reason for a change in technology.

The fact is, there is no consensus for occupying twice the bandwidth, computation, and storage. There's hardly any interest in it at all, even among professionals. There's certainly no economic demand for it. Nobody is witholding purchases because of sound quality. SACD failed because nobody cared. So why bother with it?
Old 12th August 2014
  #206
Lives for gear
 
IanBSC's Avatar
When I do the conversion to 16 bit it ALWAYS sounds different. And a -96db noise floor isn't a big deal for me listening at 75db. The problem isn't the noise.

Sometimes I like 16 bit for a crunchier less dynamic sound, or in the car. Some hip hop and electronic is well suited. But it definitely sounds different.
Old 13th August 2014
  #207
Gear Guru
 
kafka's Avatar
I think there's some improvement in a 24-bit recording also. I think even the numbers bear out that the difference should be audible under careful listening conditions. But if I try to quantify it, it's so small that it hardly matters. And I just think there's no real case for 96k recordings.
Old 13th August 2014
  #208
Lives for gear
 

Quote:
Originally Posted by IanBSC View Post
When I do the conversion to 16 bit it ALWAYS sounds different. And a -96db noise floor isn't a big deal for me listening at 75db. The problem isn't the noise.

Sometimes I like 16 bit for a crunchier less dynamic sound, or in the car. Some hip hop and electronic is well suited. But it definitely sounds different.
There is nothing 'crunchy' about properly prepared 16 bit. Perhaps you are doing something wrong - how do you do it?

It's trivial to demonstrate that the only difference between a 24 bit and properly prepared 16 bit version of a given signal is the dither noise itself.
Old 13th August 2014
  #209
Lives for gear
 
Lance Lawson's Avatar
 

I decided to use a different set of wave files where digital reduction was not in the process. Posted are two wave files taken directly from a mixed analogue studio master. The high res sampling was recorded first and the CD rate recorded second. Output of the tape machine was left the same at 0db. Samples were recorded into SONAR peaking 0db No compression or EQ has been added. I hope these links work.

The signal chain is as follows
The 7.5 ips reel to reel tape dubbed from 15 ips mixed master. Done at studio at the time of original recording sessions. Tape is Ampex 406 baked and treated to allow play.

Teac A-2300SD reel to reel @ 7.5ips into a Digital Audio CDX-01 PCI interface into SONAR.

FileSwap.com : Caroline Solo 9624.wav download free

FileSwap.com : Caroline solo 4416.wav download free
Old 13th August 2014
  #210
Gear Guru
 
Brent Hahn's Avatar
 

The links work fine. But man, that tape is a mess.
Topic:
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