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

It takes a sophisticated design to eliminate pre-ringing...more than just filters.

Also, brickwall filters are not perfect at 96khz, it's just that distortions have been minimized. There is still room for improvement which is why 176 and 192 sound better.

Perhaps part of that is that modern high end gear is so resolving that even lower noise floors add to the experience.

 

How far below the ambient noise level do you think a noise floor needs to be? A super quiet room is something like 35 dba.

 

That number really means nothing, because it depends on the frequency of the noise, and how much that masks signal at other frequencies. Anybody have any data on that? I suspect it is quite a complex relationship, and of course varies a great deal from location to location. The point being... that one can hear below the ambient noise level. For example, right now I am hearing a muffled noise from the refrigerator upstairs, and the dehumidifier downstairs. I doubt this would impede my hearing cymbals or hi hat, even though the level might be quite low. And besides, the ear/brain is designed to distinguish between sound signatures. One type of "ambient noise" may tend to mask another similar sound, but will not mask as much a dissimilar sound.

So, what is the real masking that's going on? It's hard to put a number on that. Another reason why using ears is essential. However, if one were to do any objective testing, to have real world relevance, it should be carried out under conditions with normal ambient noise levels.

 

We're talking about sampling rate, which, as you know, is completely unrelated to noise floor.

I was unaware that filter design was inadequate even at 96 kHz. You're usually pointing to Stuart's paper … what do you think about his assertion that we don't need sampling rates as high as 176 or 192 kHz? It would seem that you don't agree.

If evidence and testing pointed to 176 and 192 as sounding "better," I wonder why Stuart would recommend against it.

 

A couple of points...first, I was referring to hirez in terms of both sampling rate and 24 bits. Second, Stuart wrote the paper a good while ago around the emerging of DVD-Audio. I'm not sure if he likes the higher sampling rates or not now as more science has evolved.

I'm also not sure higher sampling rate does not have an impact on the noise floor. I think that was covered in one of the papers I posted.

 

It doesn't.

In the world of hi-end audio though, it seems anything can have an impact on anything, even if it's scientifically impossible.

 

I think you're confusing 'impossible' with 'no correlation yet detected and accepted by the masses'. I know you don't like to hear this, but the ears hear it all even though we just don't have enough amassed knowledge to definitively understand why some tests or graphical representations do not relate to what we hear.

This is nothing new. Back in the 70 amplifier designers went to great pains to lower harmonic distortion to ridiculously low levels thinking that would make a definitive difference. It didn't and in many ways things got worse due to the side effects of that effort such as TID of SID. The same thing happened with amplifiers with lots of power. Remember that Phase Linear claimed that without 700 watts the dynamics of music would not be properly represented and that explained why sound reproduction did not relate to natural sound. This went on and on yet we still aren't in consensus.

I should also add that some amps restrict bandwidth above 20k and some have bandwidths into the megahertz region. One of my pieces once claimed frequency response up to Ch 6. In many cases these amplifiers with bandwidth far beyond the range of human hearing sounded better than the amps with reduced bandwidth. How would this make sense if nothing above 20k matters?

Of course there could be other reasons that these high bandwidth amps often sounded so good. I've also heard the opinion that the slope of the upper, out of band, rolloff had an effect on what we hear. Just as it looked like we were getting somewhere, along comes a bandwidth restricted amp that sounds great. The bottom line is nobody knows exactly what makes one device sound better than another, but well trained ears are usually able to make that distinction.

I hope this isn't considered a thread crap, but I felt it needed to be said in this thread.

 

I didn't say anything was impossible. It's just not how it works, and we've been using this technology for *quite* some time now.

I'm all for admitting something is understood incorrectly, or trying to figure out the science behind new phenomena. At best it's inefficient to throw our hands up and conclude science just hasn't explained what we're hearing yet every time the ears hear something. We need to verify the new phenomenon first, *then* try to explain it using science. Based on the papers I've read on the subject, several audiophile phenomena have yet to pass through the verification stage (and it's not for a lack of trying). I honestly wonder why that is, since verification is done using those same ears.

I'm not presumptuous enough to disregard the wealth of research and testing that has been done for the better part of a century on digital sampling -- just to say that it can't explain what my golden ears are able to discern and that I must know better. That said, I have never seen anyone claim that raising the sampling rate lowers the noise floor. Perhaps Lee is on to something, and we should throw out our collective knowledge and start working to figure out how this is happening.

As I've said before, I can move my head while in my listening spot and effect a very large change in the sound reaching my ears (due to comb filtering). This change is much, much greater than I have ever heard from different digital formats, etc. This convinces me of the need of verification of phenomena, since there are often so many factors involved. I think it's a good idea to set up the testing in a way where you are ONLY testing the intended thing, such as sample rate. Converters play a big part in this I believe, so all those variables have to be isolated as well. Some could perform better at one rate than another, which would explain why 192 kHz has actually sounded *worse* to some. More testing for human perception of very high frequencies must also be done, since other tests have been inconclusive as far as I know. It's a very complicated topic, and I'd love to see more scientific evaluation of it rather than a dismissal out of hand.

 

I think it's interesting that you seem to pick and choose points you like from the "papers" you like to post. Stuart definitely doesn't feel that 44.1 kHz sampling is sufficient, but he does make many well-reasoned points about sampling rates that fly in the face of the audiophile trend of going higher and higher. He also thinks (again, apparently based on actual tests) that audible transparency can be attained with a "14-bit representation with appropriate noise shaping." Nothing in his paper that you keep referencing seems to indicate 24/192 is even useful at all as a delivery format.

I'm not sure which paper you are referencing that says sample rate affects the noise floor, but it is definitely something I've never heard claimed before. I'd like to see the reasoning behind that.

 

I surely haven't dismissed anything. I just think we're not advanced enough to have all the answers.

Take my mom. She's had stomach troubles for the last 4 months and can't eat much. She's had every test the doctors can think of and they've all come up with nothing. Does this mean she's imagining her troubles?

Of course nobody would say that. They just can't seem to come up with a test that can explain how she feels, quite like they haven't yet come up with tests to definitively quantify what we hear.

 

Spek, there's no need for this kind of vitriolic statement. We are trying to have a civil discussion here.

 

I have never claimed here that Stuart is supporting 24/192 as he is not in this particular paper. But given what I know about Meridian I think he has evolved his thinking a bit since then. They include 24/192 playback for instance on some of their products.

As for audiophile trends, he is supporting hirez so he is at a minimum in the hirez sounds better camp. And since he say you need at least 60khz, he really is backing at a minimum the nearest available format at 96khz (88.2 not being as common).

 

P.S. I'm not sure a paper yet exists on the merits of 24/192 over that of 24/96 so again I'm not sure you will get satisfied here.

But I can certainly tell you from my pro recording work that 24/176 and 24/192 sound better than 24/96.

 

If you read the lower half of page 10 of the Coding2.pdf, you come across a discussion of noise floors for adequate transparency where Stuart suggests you need 14 bits at 58khz:

 

Also, if you closely read the Mike Story paper on pre-ringing, you find a very interesting graph titled "Figure 4 - Frequency Responses of Anti-Alias Filtering for Different Sample Rates".

That illustrative graph essentially makes the case for 192 over 96khz from a pre-ringing standpoint.

So I guess that is at least one factor in why 24/192 sounds better.

 

Further backing this up is the chart "Figure 6 - Energy vs Time for various sampling rates"

That chart shows the energy smear associated with different sample rates. Or as Mike states:

 

No, trust me, the same stuff gets trotted out across all price bands!

 

I have an Esoteric sa-50 which has an "apodising filter", sort of, with the FIR and S/DLY mode. What is the real purpose of this pre-ringing filter in the simplest terms of what the different modes do. Thankyou, I know this is slightly off the topic but please explain if possible, John M.

 

It depends on how you make the room quiet, but in general low frequencies are hard to block, and high frequencies easy, but iirc around 35 dba you hear your own heartbeat, blood flow, breathing, movement related noises, and anything not out in the boonies will have subsonic noise leaking in.

Pure tones can be detected about 20 db below the noise floor, other types of sounds need higher levels. Its one of the reasons 60 hz hum is so annoying even at very low levels.

Vinyl records are around 65 db signal to noise, microphones around 110 db, but those figures are from memory.

 

This the first time I've ever felt too dumb to comment on a thread I started. I'm really impressed with the engineering minds so many of you possess. Whichever side of the brain engineering is, I'm the other side.

 

I only know that apodising filters (also in Meridian products and the minimum phase upgrade on Ayre players) effectively eliminate the pre-ringing.

Benchmark does find them to have other issues though: http://www.benchmarkmedia.com/discuss/forum/general-conversation/apodizing-filters

 

Does any of this relate to audibility? Has there been any real testing?

Funny...in testing, we aren't nearly as "golden-eared" as the industry would have us think.

 

Were the other discs all redbook? (You mentioned gold CDs.) It is possible that burning redbook would have no effect, but burning 24/192 woudl have an audible effect. In the second case there is a lot more data to read, so we might expect the benefit of a hard disk over an optical disc to be more obvious.

To eliminate the burn process, you need to have compared hard disk with optical at 24/192 through the Meridian. (I realise you can't do that now.) You can't take a previous comparison (at redbook or 24/96) and transpose the result to 24/192. So... unfortunately, you can't draw any firm conclusions because you changed too many variable at once and you can't disentangle them.

 

http://xiph.org/~xiphmont/demo/neil-young.html

 

http://forums.stevehoffman.tv/threads/24-192-music-downloads-and-why-they-make-no-sense.278967/