Anyone actually done abx testing with hi res music?

[mtrycrafts]

While the ability to store ultrasonics is something a higher sampling rate can do, ..

Are you obsessed with ultrasonics? What is the demonstrable benefit for this?

 

[3rdeye]

Irvrobinson that was a really informative post. I have a question for you, but anyone can answer.

I've read that the higher the sampling rate, the greater the audio quality, as this ensures greater precision in your high notes and low notes. I've also read that the higher the bit rate (also called bit depth), the better the audio quality as well.

I don't understand the relationship between sampling rate and bit rate, but if both of these statements are correct, then both sampling rate and bit rate impact audio quality.

I'm trying to determine which is more germane to the discussion of our capacity to hear music given the established listening ranges of 20 Hz to 20 kHz.

That would make it much clearer what format (I'm assuming an uncompressed one, but maybe not) would be sufficient for the most optimal and scientifically possible listening experience. Because, I'm hoping no one is suggesting that there is a format that allows us to hear outside of the 20 Hz to 20 kHz range.

Sent from my ONEPLUS A3000 using Tapatalk

 

[<eargiant]

Benchmark Media has a lot of good information with technical analysis to support it:

https://benchmarkmedia.com/blogs/application_notes/intersample-overs-in-cd-recordings

https://benchmarkmedia.com/blogs/application_notes/14949325-high-resolution-audio-sample-rate

https://benchmarkmedia.com/blogs/application_notes/14949345-high-resolution-audio-bit-depth

https://benchmarkmedia.com/blogs/application_notes

 

[PENG]

Benchmark Media has a lot of good information with technical analysis to support it:

https://benchmarkmedia.com/blogs/application_notes/intersample-overs-in-cd-recordings

https://benchmarkmedia.com/blogs/application_notes/14949325-high-resolution-audio-sample-rate

https://benchmarkmedia.com/blogs/application_notes/14949345-high-resolution-audio-bit-depth

https://benchmarkmedia.com/blogs/application_notes

Just a couple points:

1. The author is not in an unbiased position.
2. The article does not appear to be a technical paper.

Even with the above in mind, the author said:

"When all of these newer technologies are properly applied, the CD can deliver stunning results that closely rival that of the high-resolution formats."

That seems to be in agreement with what many others are saying that, the format and resolution beyond 44.1kHz/16-bit itself is not the deciding factor, the recording/mastering and other factors could be more important.

 

[Irvrobinson]

Irvrobinson that was a really informative post. I have a question for you, but anyone can answer.

I've read that the higher the sampling rate, the greater the audio quality, as this ensures greater precision in your high notes and low notes. I've also read that the higher the bit rate (also called bit depth), the better the audio quality as well.

I don't understand the relationship between sampling rate and bit rate, but if both of these statements are correct, then both sampling rate and bit rate impact audio quality.

I'm trying to determine which is more germane to the discussion of our capacity to hear music given the established listening ranges of 20 Hz to 20 kHz.

That would make it much clearer what format (I'm assuming an uncompressed one, but maybe not) would be sufficient for the most optimal and scientifically possible listening experience. Because, I'm hoping no one is suggesting that there is a format that allows us to hear outside of the 20 Hz to 20 kHz range.

The higher the sampling rate, the higher the audio frequencies that can be recorded and reproduced. For example, a 192KHz sampling rate can record frequencies up to 192KHz/2 = 96KHz. This is because digital sampling is based on the Nyquist-Shannon theorem, which says that the number of samples required to represent a continuous function at a given frequency (like a sound wave) is twice the frequency, which is why a CD's 44.1KHz sampling rate can accurately reproduce audio frequencies up to 22.05KHz. The upper-most frequency that can be reproduced is called the Nyquist Frequency. But here's the thing, and it is a very important concept, higher sampling rates, like 192KHz, do not increase the accuracy of the detection of audio frequencies that can be accurately detected by lower sampling rates. In other words, Bluray audio which has a 192KHz sampling rate is no more accurate at reproducing a 20KHz sound than a 44.1KHz CD is.

Higher sampling rates do push the supposedly detrimental effects of the digital filters to suppress aliasing about two octaves beyond the audible range, but in reality higher sampling rates do not have audible advantages because of the refinement of the digital filters and a technique called oversampling that eliminates the detrimental effects of excluding aliasing.

(Aliasing is the creation of digital samples for sounds that have a higher frequency than the Nyquist Frequency. These samples do not accurately represent input, so they are garbage samples, and you have to filter them out of the sample stream. This is what the digital filters discussed earlier do. Oversampling is a technique that duplicates the number of samples per frequency, in most modern converters by eight times or more, which allows the more shallow aliasing digital filters enabled by higher sampling rates without having to increase the digital file sizes like high sampling rates do.)

So, to sum it all up, sampling rates higher than a CD's 44.1KHz really have no technical or audible advantages over modern DACs with 8x oversampling or greater.

 

[Irvrobinson]

Benchmark Media has a lot of good information with technical analysis to support it:

https://benchmarkmedia.com/blogs/application_notes/intersample-overs-in-cd-recordings

https://benchmarkmedia.com/blogs/application_notes/14949325-high-resolution-audio-sample-rate

https://benchmarkmedia.com/blogs/application_notes/14949345-high-resolution-audio-bit-depth

https://benchmarkmedia.com/blogs/application_notes

I'm really surprised at Benchmark Media for this sort of marketecture double-speak, especially that inter-sample overs discussion. To my knowledge, there's no such thing in digital audio as an inter-sample over. In fact, if there was such a thing, I'm wondering how PCM digital audio would work at all. That improperly-designed DAC circuits can clip, and digital clipping is catastrophic, is true, and that recording and mastering can induce digital clipping is true, but to extend the notion to some discussion that Nyquist-Shannon theory is wrong about two samples being insufficient to plot a waveform - which is what the blog post implies - seems really silly. Unless I'm misinterpreting the blog post.

 

[Bucknekked]

I'm really surprised at Benchmark Media for this sort of marketecture double-speak, especially that inter-sample overs discussion. To my knowledge, there's no such thing in digital audio as an inter-sample over. In fact, if there was such a thing, I'm wondering how PCM digital audio would work at all. That improperly-designed DAC circuits can clip, and digital clipping is catastrophic, is true, and that recording and mastering can induce digital clipping is true, but to extend the notion to some discussion that Nyquist-Shannon theory is wrong about two samples being insufficient to plot a waveform - which is what the blog post implies - seems really silly. Unless I'm misinterpreting the blog post.

you have been giving really detailed answers on this subject. I'm enjoying them.
Its one of those situations like reading an article on mathematics or physics.
I can follow along and get the concepts as you explain them. I would not be able to find a flaw or see alternate views. I feel like you're on the job with the answers so I feel like I'm getting some good scoop about my hobby.

 

[PENG]

you have been giving really detailed answers on this subject. I'm enjoying them.
Its one of those situations like reading an article on mathematics or physics.
I can follow along and get the concepts as you explain them. I would not be able to find a flaw or see alternate views. I feel like you're on the job with the answers so I feel like I'm getting some good scoop about my hobby.

I have been a fan of Nyquist and Fourier since university days, so naturally I agree with Irv.

 

[Verdinut]

Here a link to a detailed article entitled "24/192 Music Downloads and Why They Make No Sense". It is well documented:
https://people.xiph.org/~xiphmont/demo/neil-young.html

Any comments on this?

 

[Johnny2Bad]

Getting the brickwall filter well outside the audible range is desirable, as most filters exhibit pre-ringing which obscures detail.

Using a bitrate higher than 16 is desirable because with 16/44.1 data, no 16-bit DAC can resolve all 16 bits (13 is considered excellent). A 24 bit DAC can resolve up to 19 bits (I know of no 24-bit DAC that can do 20 or better) which means you get all the resolution a 16 bit file can offer.

I find true CD-qualtity files to sound fine with upsampling to 96 and a 24-bit DAC. Hi-Rez files that are natively higher resolution than 16/44.1 might offer less chance of oversample errors, a lower noise floor (which won't make the file quieter, but might offer more detail) and less chance for data error in general, but I find it's very file-specific, and quite subtle at that. In many cases it makes no discernible difference provided you are upsampling the CD version to the same resolution.

On the other hand I don't see HiRez downloads as a waste of money. Digital files should be stored at the highest resolution possible. It's only a confluence of circumstances that makes this viable today; storage capacity is not a limit, the ADCs and DACs are common and inexpensive, and after 35 years of promises, the 16/44.1 file format is finally starting to sound organic instead of steely, and DACs need not cost in the four figures to perform at a high level.

Five years ago I would not have said that. I much prefer that people can download music in an uncompressed format without breaking the law versus only having lossy compressed files available online. The Labels have the usual motivation, of course ... to re-sell you your music library. So consider the entire exercise with eyes wide open.

I have done ABX testing with different file formats. With different components, it's not my idea of a valid test since there are genuine problems with an ABX format component comparison (not the least of which is many are not true ABX, and almost all use reproducing gear that I know isn't as resolving as my own home system, let alone what I can't afford)*. With the built-in speakers of my MacBook Pro I get about 40% correct; with my IEMs it's about 60%, and on my home system it's about 80%; I have achieved 100% occasionally.

I also find with different components (vs different file formats) that if I can spend some time listening in a relaxed state (like anyone would at home) for a while to the candidates a few weeks at least with music I enjoy, I can much more easily identify them in ABX with close to or at 100% accuracy. Which flies in the face of those who dismiss subjective evaluations, so right off the hop the ABX guys are hostile, but that's my experience.

* If you look up the history of audio, the systems of the 1920's were considered "perfect" by those who first heard them, the systems of the 1960's were considered "perfect" by those who first heard them, and there are people today who anoint some systems "perfect". It's a moving target. The one guarantee is that your children will be listening to higher quality audio than you are.

 

[Irvrobinson]

Getting the brickwall filter well outside the audible range is desirable, as most filters exhibit pre-ringing which obscures detail.

Using a bitrate higher than 16 is desirable because with 16/44.1 data, no 16-bit DAC can resolve all 16 bits (13 is considered excellent). A 24 bit DAC can resolve up to 19 bits (I know of no 24-bit DAC that can do 20 or better) which means you get all the resolution a 16 bit file can offer.

I find true CD-qualtity files to sound fine with upsampling to 96 and a 24-bit DAC. Hi-Rez files that are natively higher resolution than 16/44.1 might offer less chance of oversample errors, a lower noise floor (which won't make the file quieter, but might offer more detail) and less chance for data error in general, but I find it's very file-specific, and quite subtle at that. In many cases it makes no discernible difference provided you are upsampling the CD version to the same resolution.

Alright, I'll bite, what do you think "an oversample error" is?

Are you aware that Texas Instruments rates their 16bit DACs to within 1 LSB? What is your evidence that 13 bit resolution (an error of 3 LSBs) is typical?

You are mistaken that recording strategies of greater than 44.1/16 are required to get excellent performance. Look at the measured performance of the Benchmark DAC3 as measured by Stereophile. There are several characterizations of 44.1/16 performance, and it is flawless. The DAC3 does use a 32bit DAC, but you start out talking about high bitrate files. What are you talking about?

https://www.stereophile.com/content/benchmark-dac3-hgc-da-preamplifier-headphone-amplifier-measurements

And, drum roll, JA calculates the DAC3's 24 bit resolution as 21 bits.

 

[Verdinut]

The quality of recordings today may be better for pop music today than it ever was, but as it concerns classical music and opera, we have a different picture.

Many of today's recordings and many transfers of analog recordings to CD are recorded too loud and digital overload of the signal, as you know, gives edginess to the sound which is a lot more irritating than some of the slight analog distortion of the past.

Some recording engineers do an excellent job, notably most of the ones involved in movie soundtrack and classical concert recordings. But there are some who don't have the experience and don't seem to care about the dynamic range demands in classical music and then, we are subjected to that silly distortion.

I have over 2500 CDs and over 500 DVDs/Blu-rays. I haven't noticed many DVDs or BDs with distortion but many CDs and even some SACDs have that edginess that I could do without. I have good equipment: OPPO BDP-95 and Marantz SR5010 feeding QSC Cinema amps with DIY 3-way floor standing speakers featuring a 15 inch sub in each cabinet for the 3 front channels.

Nowadays, I prefer to buy Blu-rays and Blu-ray Audio discs. IMO, the DTS Master Audio surround format gives me the most satisfaction.

 

[Beave]

Getting the brickwall filter well outside the audible range is desirable, as most filters exhibit pre-ringing which obscures detail.

Pre-ringing looks cool on scope pictures, but humans don't hear scope pictures. We hear sounds. And pre-ringing is one of those audiophile FUD words. Pre-ringing is a time domain issue. The ripple in the passband of an analog filter can be audible if the filter is poorly designed such that the ripple is large enough.

Using a bitrate higher than 16 is desirable because with 16/44.1 data, no 16-bit DAC can resolve all 16 bits (13 is considered excellent). A 24 bit DAC can resolve up to 19 bits (I know of no 24-bit DAC that can do 20 or better) which means you get all the resolution a 16 bit file can offer.

Bitrate higher than 16??? Don't you mean bit depth? As for 13 bit resolution being excellent, even the cheapest DACs have been able to do that and better for 20+ years now. Even so, the difference between around 14-15 bit equivalent and 19 bit equivalent is not a real-world issue, ie, it's not going to be audible. You're talking going from a dynamic range well beyond any real-world recordings to a dynamic range even farther beyond any real world recordings. Not to mention that your listening environment isn't even to being capable of reproducing such a large dynamic range.

I find true CD-qualtity files to sound fine with upsampling to 96 and a 24-bit DAC. Hi-Rez files that are natively higher resolution than 16/44.1 might offer less chance of oversample errors, a lower noise floor (which won't make the file quieter, but might offer more detail) and less chance for data error in general, but I find it's very file-specific, and quite subtle at that. In many cases it makes no discernible difference provided you are upsampling the CD version to the same resolution.

What are 'oversample errors'?

 

[Irvrobinson]

Many of today's recordings and many transfers of analog recordings to CD are recorded too loud and digital overload of the signal, as you know, gives edginess to the sound which is a lot more irritating than some of the slight analog distortion of the past.

Digital overload, meaning that the amplitude is too high for the word depth to represent, does not result in edginess, it results in gross distortion. Out of the thousands of recordings I have there are only two examples of overload, and both are live performance recordings of classical solo piano. The "edginess" you think you're hearing could have a lot of causes, from simply lousy recordings, to equalization, to distortion in the analog domain (which is also unlikely in professional recordings), to frequency response problems in your own speakers, or even your room, but it isn't due to digital overload.

 

[Bucknekked]

Here a link to a detailed article entitled "24/192 Music Downloads and Why They Make No Sense". It is well documented:
https://people.xiph.org/~xiphmont/demo/neil-young.html

Any comments on this?

Verdinut
I appreciate the article you posted. Now there's a fellow who can represent a point of view and be both thorough and entertaining. I copied and saved that entire article to my local machine so I can always reach out and read it. Sometimes URL's disappear.

For all those in the hobby who would continue to support and toss in the pseudo-scientific, jargon laced , chicken little ideas on why the RBCD Standard is no good and why only higher bit rates and higher sampling rates will fill the bill, I think this is yet another fact based explanation of why we already have it pretty good.

He makes the point that the preponderance of industry authors supports the 44.1khz/16 bit CD standard and its the fringe that continues to rattle and shake over "something better". In every endeavor, there's always a fringe element. Sometimes its the fringe that produces the radical change in a pursuit. Most of the time however, its just noise.

Case in point, we read some articles over the long weekend about a guy who has talked a group of folks in to funding his building of a new rocket which he will launch himself in. His goal? To prove the earth is flat. He ran out of money getting rockets to launch on his own nickel. He found the "flat earthers" at a convention and they offered (or he convinced them) to fund his next rocket launch to prove the earth is flat. He is happy to take their money because he's gets a free rocket ride out of the deal. Good science? Or snake oil at its best.

Some folks think the issue of "flat earthers" was settled a long time ago. Nope. There are those still rattling for it. After all, according to the rocket man in this article, Buzz Aldrin and Neil Armstrong are just actors. And according to this guy, the only thing you really need to know about them is that they are "freemasons". Enough said after knowing that.

Such is pseudo-science. One claim after another. And a new sucker born every minute.

Thanks for the article. I enjoyed it and got some good data points from it

 

[Irvrobinson]

I've been doing some reading on this "inter-sample overs" concept, and I've got to say that it's obscure and difficult to follow. As near as I can tell, the concept involves some sort of overload in the D-A conversion, and sometimes the discussion seems to be about momentary word-depth overflow. Then the discussion always reverts to some graphic showing two samples for a sine wave that are either side of a peak, and that the peak exceeds the amplitude limit (which will vary by word depth, obviously), and that the peak, the "inter-sample over", is somehow causing distortion. Like in this article:

https://www.head-fi.org/threads/benchmark-talked-about-headroom-for-intersample-peaks-in-dac-does-it-really-matter.834975/

Since the peak from the graph will be in the analog domain, it either implies an absolute voltage limit in the DAC or complete bullshit. Benchmark Media seems to say (it is not entirely clear to me) that they solve by problem of these inter-sample overs by building more headroom into the D-A process, which makes it smell like bullshit.

It would be nice if John Siau, whose team at Benchmark apparently designed one hell of a DAC, explained exactly, in engineering terms, what inter-sample overs are. I'll be the first one to volunteer to eat crow here and admit stupidity if it really is a D-A problem inherent in PCM audio, and not an implementation detail in DACs.

 

[Bucknekked]

I've been doing some reading on this "inter-sample overs" concept, and I've got to say that it's obscure and difficult to follow. As near as I can tell, the concept involves some sort of overload in the D-A conversion, and sometimes the discussion seems to be about momentary word-depth overflow. Then the discussion always reverts to some graphic showing two samples for a sine wave that are either side of a peak, and that the peak exceeds the amplitude limit (which will vary by word depth, obviously), and that the peak, the "inter-sample over", is somehow causing distortion. Like in this article:

https://www.head-fi.org/threads/benchmark-talked-about-headroom-for-intersample-peaks-in-dac-does-it-really-matter.834975/

Since the peak from the graph will be in the analog domain, it either implies an absolute voltage limit in the DAC or complete bullshit. Benchmark Media seems to say (it is not entirely clear to me) that they solve by problem of these inter-sample overs by building more headroom into the D-A process, which makes it smell like bullshit.

It would be nice if John Siau, whose team at Benchmark apparently designed one hell of a DAC, explained exactly, in engineering terms, what inter-sample overs are. I'll be the first one to volunteer to eat crow here and admit stupidity if it really is a D-A problem inherent in PCM audio, and not an implementation detail in DACs.

irvrobinson
Although I must admit I didn't understand one freakin' word regarding the "inter-sample overs" problem, I do understand an underlying principal. The underlying principal is that jargon is one of the first refuges of scoundrels. If explaining a problem or issue in plain english gets one shot down in flames, the scoundrel will always dress it up in complex jargon and make it seem like only fools don't understand it.

I think deliberate obfuscation is a staple amongst audio folks when it comes to pet peeves and things like why the standard CD format just isn't good enough.

I think you are right on target with your explanations and questions. I don't think you misunderstand much in this area of audio. If someone can't explain it to you plainly, there is much to be suspicious of.

I explain arcane and complex issues in plain english for a living. Its why I have a living. I look at explanations written in techno goggley gook and I explain them to business executives in plain english. If you can't do that, then often the techno-speak is really obfuscating and not illuminating a problem area.

Keep up the good work, irv.

 

[Johnny2Bad]

Pre-ringing looks cool on scope pictures, but humans don't hear scope pictures. We hear sounds. And pre-ringing is one of those audiophile FUD words. Pre-ringing is a time domain issue. The ripple in the passband of an analog filter can be audible if the filter is poorly designed such that the ripple is large enough.


Bitrate higher than 16??? Don't you mean bit depth? As for 13 bit resolution being excellent, even the cheapest DACs have been able to do that and better for 20+ years now. Even so, the difference between around 14-15 bit equivalent and 19 bit equivalent is not a real-world issue, ie, it's not going to be audible. You're talking going from a dynamic range well beyond any real-world recordings to a dynamic range even farther beyond any real world recordings. Not to mention that your listening environment isn't even to being capable of reproducing such a large dynamic range.



What are 'oversample errors'?

Am I to understand that you hear no difference in digital filters?

Please name a single 16-bit DAC that can resolve 14 bits of data. While you're at it, name a single 24-bit DAC that can resolve 20 bits of data.

And I appreciate the typical response when measurable differences are defined; that "it doesn't matter". Clearly it doesn't matter to YOU.

See:
https://service.tcgroup.tc/media/Level_paper_AES109(1).pdf

 

[Irvrobinson]

Please name a single 16-bit DAC that can resolve 14 bits of data. While you're at it, name a single 24-bit DAC that can resolve 20 bits of data.

You don't read very well. See my earlier post. Every TI 16bit DAC is rated at better than 1 LSB accuracy. Or don't you know what that means? I also posted a link to the Benchmark Media DAC3, which by measurement resolves 21bits out of 24.

BTW, do you know what that paper is really saying, and it is referencing 17 year old equipment?

 

[Beave]

Am I to understand that you hear no difference in digital filters?

Did I say that? I don't think I said that. In certain circumstances, differences may be audible. But just because the impulse response looks different doesn't mean that there is an audible difference.

Please name a single 16-bit DAC that can resolve 14 bits of data. While you're at it, name a single 24-bit DAC that can resolve 20 bits of data.

Perhaps you should define what you mean by 'resolve' first. Because every cheap 16 bit DAC I evaluated from the mid 90s to the early 2010s could meet or exceed the dynamic range and SNR that would be expected from 14-15 bit resolution. And higher bit depth DACs could do even better.

And I appreciate the typical response when measurable differences are defined; that "it doesn't matter". Clearly it doesn't matter to YOU.

So how exactly do you KNOW you can hear differences? What's your test methodology?