192kHz/24bit vs. 96kHz/24bit "debate"- Interesting revelation

If I'm following you, I think I DID make that comparison.

The comparisons I made... and, again, this was to simply test the Meridian's processor vs. my Marantz DV9500 (which I wrongly thought COULD output 192/24 were the following-

Regular Redbook cd ripped into my hard drive and played through Meridianvs same original redbook cd played through my Marantz... Marantz wins.

Regular Redbook cd ripped into my hard drive and played through Meridian vs. same original redbook cd burned to disc and played through Marantz... Marantz win.

"Audiophile" redbook cd (DCC or Audio Fidelity) ripped to my hard drive and played through Meridian vs. same cd played through my marantz. Marantz wins.

"Audiophile" redbook cd (DCC or Audio Fidelity) ripped to my hard drive and played through Meridian vs. same original cd burned to disc and played through Marantz... Marantz wins.

Music downloaded from HDTracks at 192/24 and played through Meridian vs. same album burned to disc as a DVD Audio and played through my Marantz... Meridian wins.

The only two tests I didn't do were:

Music downloaded from HDTracks at both 192/24 and 96/24 and played through Meridian (because remember I wasn't testing that difference)

Music ripped from a DVD A that I own (for example LA Woman or Sinatra at the Sands) and then played through the Meridian vs. original DVD A disc through the Marantz.

I realize that my results don't really prove anything. I just found it interesting and POSSIBLY revealing.

 

You ain't seen nothin' yet. There are some good folks in here already, but wait till Black Elk, Ivan Wemple, Don Hills, and crapfromthepast show up.

 

How much recording is done at 24/192, versus 24/96? The two labels whose hi-res downloads I've bought in the past (Chandos and Hyperion) seem to record at 24/96 only (or, occasionally, 24/88.2).

 

This is a really good article. Anyone can make assertions about how they think things work, but Monty is actually supporting it with scientific information.

I've never thought of 24/192 as being potentially "worse" than 24/96, but it is an interesting premise that has some scientific basis as far as I can tell.

 

At audiophile labels, quite a bit. At mainstream labels, not very common. A lot of Pro Tools users record at 24/44 or 24/48.

Still, if you like the recordings that are available, and there are some blockbusters, then there is definitely improvement to be found.

 

Nick, the dcs paper from Mike Story showed that the energy smear was less at 24/192 which is evidence that there are sonic advantages to 24/192. I guess you rejected that scientific argument since it doesn't support your view.

You accused me of picking and choosing but it seems to me that you have done exactly that.

 

When you say Marantz "wins" or Meridian "wins" what do you mean exactly. Obviously you feel there was a difference, and one was preferable to the other. But I'm curious as to exactly what qualities led you to declare one a winner.

My experience with mods and tweaks is that sometimes they affect one or two particular qualities, but most often there is a kind of across-the-board improvement in virtually everything. This is not a tweak or mod, but I'm just curious what your criteria were.

 

How right you are. Some of these guys have had a hand in creating the technology we're often talking about (SACD, optical disc technology, etc.), and I hope they are still reading the hardware forum. I would love to hear more about the science behind this stuff!

 

In looking at the Xiph Neil Young page, there are lots of problems. Here are some observations:

1. They don't present any actual scientific evidence themselves. There are no controlled listening tests either proving or disproving the benefits of 24/192 hirez.

2. They do present papers via links and their own spin on papers which suggest no discrimination between 16/44 (!) and 24/96.

3. They ignore any links to the Ohashi study which showed listeners discriminating hirez from 16/44.

4. This is the biggest one for me. No rebuttal of Bob Stuart's Coding2 paper or Mike Story's paper or Dan Lavry's paper. Mike Story's work is not even mentioned. These don't support Xiph's view but are written by very well respected digital researchers. If Xiph had credibility, those well-known papers would have been discussed in detail.

5. The entire page is pretty much a litany of digital artifact discussion but not really a structured argument against 24/192.

 

It's too bad we can't "tag" posts, the way you can in Facebook, to summon their spirits.

 

Just using the word "wins" to simplify. These tests were conducted awhile ago so I can't really describe the difference right now. I just remembered, as I say in my initial post, that the only test where the Meridian outperformed (i.e sounded better) than my Marantz was with HDTracks stuff played through Meridian at 192/24 vs. same album burned to disc as a DVD A.

 

And I agree with your comment about mods or tweaks.

 

I apologize -- I haven't read that paper. I have only referred to the Stuart paper. I did ask you if any listening tests had been done to see if this "energy smear" relates to audibility at all. I'm sure there are theoretical advantages of 192 kHz sampling, but I'm not convinced they would be audible (such as extended frequency response to 96 kHz). I'm not trying to pick and choose anything, seriously. If (properly-conducted) listening tests show the audible advantages, I'd be on board with it.


The xiph.org article has a section about relevant listening tests, but there may be others I'm not aware of:

Please note I am not asserting that 16/44 is the highest quality playback possible...just that listening tests conducted thus far have shown it to be as transparent as "hi-res." That could change in future listening tests, of course, and then we would know differently.

 

Reading further....Xiph agrees with Bob Stuart's article when he says the following:

However, this is in conflict with the statement of belief above which is quoted here:

So which is it?

Is Stuart correct when he proves a scientific basis for hirez improving playback?

Or is 16/44 good enough as there is no empirical evidence (Ohashi anyone?) that more resolution matters?

So Xiph seems to be deeply conflicted on any hirez benefit. Hard to accept Xiph as an authority on 24/192 in that event. Or at least until they get that basic point straightened out.

Also, they don't address the later Lavry paper that dives deeper into the benefits of hirez.

P.S. I won't even get into the deep flaws of the Meyer-Moran test. That's been well covered.

 

But we do have at least two tests that did show audibility: the Pras/Gaustavino study and the Ohashi study. Both of these tests showed discrimination of hirez samples.

 

They link to ABX tests (although some require third-party subscription) and there's plenty more out there. Most of the objection to these ideas appears to start at "la la la la I can't hear all that evidence" - yes you can, if you're actually interested in the truth and seek it.

And?

There are always outliers. If you have a mountain of evidence suggesting A and a small amount of evidence counteracting it, A is still considered true. The fallacy that a single piece of evidence can completely counteract established theory is rather popular amongst the general public but is clearly not sensible. Most, if not all of the studies showing the opposite of established theory are considered flawed in some way (homoeopathy is a great example) - those that aren't tend to slightly (!) alter thinking, not overthrow it.

If more and more ABX tests start to show positive results then things need re-considering. Until then, they do not.

I haven't read all of those papers. From what I remember though, Dan Lavry's report* actually agrees with Xiph that superior sampling rates do not affect fidelity in the audible range. Bob Stuart's report also says that 24-bit/96 kHz audio is pointless, so I really don't know what you're trying to suggest here.

*I hate the term "white paper" because people always shorten it to "paper", which, in this context, should be reserved for those published in scientific journals that are peer reviewed.

I don't agree. Have you not read the page title?

 

My statements stand Dragon. They are clearly cherry picking the studies that support their view that there are no hirez benefits while ignoring good work by actual audio researchers like Pras and Gaustavino (unlike the Boston Audio Society). Then they are later claiming that 24/96 is good enough in part based on Stuart and Lavry.

Xiph or Monty or whoever don't really add anything to the discussion except for a controversial title.

It would have been nice to see Xiph actually do some tests on 24/192 versus 24/96 discrimination. They do work in audio coding it appears. THAT would have been useful to the discussion.

 

Lee, I think you quoted the way Monty feels about Stuart's paper:

I understand what he means. His paper is very reasonable, although obviously Stuart comes to different conclusions about the sample rate needed.

Another interesting point that I think often gets skipped when talking about digital audio is transparency. The goal is to be "transparent," i.e. to faithfully transport the input signal. Once that point is reached, no further improvement is needed (even if we *can* achieve higher specs). I'm not saying for a certainty that 16/44 is transparent, but that's the case Monty is making. I do fully believe in benefits of hi-res for recording and production work, especially 24 bit for its lower noise floor which allows much more headroom.

Can you link to these tests? I can't seem to find the Ohashi study you keep referring to. There are a couple I've heard about but I'm not sure if it's that one or not.

 

I find the Meyer/Moran test quite interesting. I'll borrow this post from SA-CD.net since it sums up my impressions of the Meyer/Moran test:

 

192! I don't encode below 320.

 

Pras is here. Will look for the Ohashi.

http://www.aes.org/e-lib/browse.cfm?elib=15398

 

You are missing some zeros.

 

Thanks. That darned paywall.

 

I think he's mistaking kHz for kbps.

 

I think he's being facetious.