96k is better than 44.1k sample rate. why?

  • Thread starter Thread starter altiris
  • Start date Start date
COOLCAT said:
the question why? why go 192Khz? can we hear the improvement?
after reading this very interesting thread, it comes down to using the best technology available, whether we humans can hear it or not. imo.

from a manufacturing viewpoint, eventually the manufacturers make the decison for us by shutting down the old line, for example the 8bit dacs become almost extinct like the dinosaur as the shelf fills with 24/192's.

Need to be careful about saying that 192 is better than 96. As I think I showed above, 96 is really as good as it needs to get. Since the actual sample rate (above 2mHz) is unchanged whether you set your converter to 96 or 192, the data rate of 192 is based on fewer samples for each data point, making it less precise, not more. Combine that with the penalty in processing power, and it becomes difficult to understand the purpose for 192.
 
COOLCAT said:
it comes down to using the best technology available, whether we humans can hear it or not.
Define "best".

Is simply more always better? Or its it better to pay attention to things like return on inventment and the law of diminishing returns?

To use your HDTV analogy, is it "better" to purchase one of those new 103" wide screen plasma HDTVs if your content is going to be "South Park"? Or is it just wasteful and pointless? Does being a billionaire who can afford it make it any less wasteful and pointless?

Sometimes "better" is really pointless. And if it's pointless, it isn't better.

The concept of "quality" includes the characteristic of "efficiency" and excludes the characteristic of "waste".

G.
 
NL5 said:
Am I reading the graphs correctly? It looks like trhe difference doesn't even start until 19705hz - I can't hardly hear anything at that frequency anyway.

There must be more to it than that???

i think that the resolution is a bit of an issue here.

just think about it, if you're taking a sample at 44.1k of a of a 17K sine wave, there's going to be quite a bit of, all be it very minor, distortion.
 
giraffe said:
i think that the resolution is a bit of an issue here.

just think about it, if you're taking a sample at 44.1k of a of a 17K sine wave, there's going to be quite a bit of, all be it very minor, distortion.

But all the distortion will be at frequencies above 22kHz. I can run a test to prove that. I can also crank the resolution of those graphs up as much as you like :)
 
SouthSIDE Glen said:
Define "best".

Is simply more always better? Or its it better to pay attention to things like return on inventment and the law of diminishing returns?

To use your HDTV analogy, is it "better" to purchase one of those new 103" wide screen plasma HDTVs if your content is going to be "South Park"? Or is it just wasteful and pointless? Does being a billionaire who can afford it make it any less wasteful and pointless?

Sometimes "better" is really pointless. And if it's pointless, it isn't better.

The concept of "quality" includes the characteristic of "efficiency" and excludes the characteristic of "waste".

G.




my vote for post-of-the-month.
 
giraffe said:
just think about it, if you're taking a sample at 44.1k of a of a 17K sine wave, there's going to be quite a bit of, all be it very minor, distortion.

Why is it that you think this?
 
mshilarious said:
Need to be careful about saying that 192 is better than 96. As I think I showed above, 96 is really as good as it needs to get. Since the actual sample rate (above 2mHz) is unchanged whether you set your converter to 96 or 192, the data rate of 192 is based on fewer samples for each data point, making it less precise, not more. Combine that with the penalty in processing power, and it becomes difficult to understand the purpose for 192.

to clarify, I was just using 192 as a marketing type term. Maybe 96K is as good as it gets, technically. thats some great data you posted.
which is better 44.1 or 96...or could we just say "which is better 44.1 or "more"?

which is better 44.1 or 96?
Being a current shopper I was interested in this question.
but there's two opposite approachs to determine "better", imo.
One is I can use my ears and a blindfold test to determine whats "better", Two is using measurement tools and graphs and make the decision off data, not my ears, if 44.1 is better than 96.

So its "better" per context the words used in.
per your data the 96 is "better" than the 44.1, because we can SEE it.
per the ears comments its not any "better" as most can't HEAR it.

I refer to the term "better" in the Manufacturing context.
When electronic chips are compared on the test benches, the data approach is used because its repeatable, consistent, quantifiable per the measurement tools. (and we can't hear voltages and clock speeds..er..I can't anyway).

In short, its not the chips fault we can't HEAR the improvements due to other limitiations such as crappy playback deck or earwax-I-buds.

maybe 96 is where the buck stops as the Ears gang doesn't buy it, and the Data gang doesn't see any more improvements at 192 or whatever's next?

If this is true, we can all start upgrading all of our playback systems to 24/96...then maybe we'll hear it? :confused: :p
 
SouthSIDE Glen said:
Define "best".

Is simply more always better? Or its it better to pay attention to things like return on inventment and the law of diminishing returns?

To use your HDTV analogy, is it "better" to purchase one of those new 103" wide screen plasma HDTVs if your content is going to be "South Park"? Or is it just wasteful and pointless? Does being a billionaire who can afford it make it any less wasteful and pointless?

Sometimes "better" is really pointless. And if it's pointless, it isn't better.

The concept of "quality" includes the characteristic of "efficiency" and excludes the characteristic of "waste".

G.

"best" is based off our own senses/perceptions OR, using test measurement tools...hell, maybe both. I have a split personality so I use both when I can. Ears and Data tools. :p

I don't know on the ROI question... 44.1 or 96, what will be the expectations to upgrade? is this the return, that it meets your expectations?
So you drop $300 on a new 96 and you do a recording and you can't hear a difference...expectations determine success.

the pointless term, is more for shopping and purchasing and applications I'd guess your refering to.
from a engineering Dilbert data approach...studying data and crap is what many do. If not for that its pointless to get out of bed. :p

on HDTV, its visibly better to most people. What consumers want is a bang for the buck. In the beginning HDTV sets just wasn't enough "bang". At $50,000 a hdtv only the Speilberg and Lucas crowd owned them. Now their $800-1000 and there's more and more HDTV material and its selling like crazy. Its a pretty good bang at $1000.
Data shows DLP HDTV chip announced 1.4 % of the market share = a billion dollars a year business...at 1.4%! simple math approx 1%=1 billion$$$, 100%=100billion dollar market. Its growing apparently alot of people are giving it a thumbs up.

will people give a thumbs up on 96Khz...will they plkug it in and be blown away by the pure awesomeness of improvement over 44.1?????

I want to know this before I buy .......so I don't waste money. :D
 
COOLCAT said:
will people give a thumbs up on 96Khz...will they plkug it in and be blown away by the pure awesomeness of improvement over 44.1?????

I want to know this before I buy .......so I don't waste money. :D
Short answer: NO. :)

Longer answer: If you want to look at bang for the buck: the answer to that should be pretty self evident. Every thread and every discussion and every post on this subject tells us one thing: any advantage to 96k is debatable and marginal at best. In other words, 96k yields, at worst, questionable, and at best, limited "bang". How much buck is questionable or limited bang worth? Wouldn't you have to say that extra buck for it is therefore also limited or questionably worth it?

There's another more practical angle to look at: what do people listen to? At this point it's either CDs, MP3s or streamed audio, with DVD-A and SACD so far looking like fairly major flops as far as market penetration and the cares and needs of the buying public. This means a vast majority of distribution format that's at 44.1K or worse. That's the target audience and the target medium. Does one need 96k to deliver top-notch content to these media? Nope.

Let's take a stand on the other side of the coin, let's look at best case. Lets say that through some kind of market manipulations that several things happen more or less at the same time: that DVD-A or SACD replaces CD as the new disc standard, that MP3 and standard streaming are replaced by a losless compression format that improves the reproduction quality greatly without much if any bandwidth cost, and that you are producing Telarc-worthy classical or acoustic productions that are both the domain of the audiophile end user with the golden ear and are also the type of content where smaple rate makes the most difference. In that case you want to go to 96k, right?

Well, sure...with a big "BUT". That "BUT" is that you had *first* better upgrade everything that falls in your signal chain before the converter, and that you had better not only get a 96k converter, but you'd better get an A-list one, otherwise just upgrading to 96k is going to be a huge waste of time, money and bandwidth.

It's been said here before many times, but it's worth repeating now, I think: 96k does not make content sound better, it only - at best - reproduces it more accurately. Again, look at the wide screen DLP HDTV. If you are watching an analog broadcast of the original "The Honeymooners" from a station 30 miles away with a thunderstorm in the area, your HDTV isn't going to make the signal look any better than it does on your 12" portable b/w TV set, except maybe the tape smear and broadcast static will be in higher resolution. That is exactly analogous to recording a heavy metal garage band recorded by someone with limited recording experience through SM57s and MXL990s plugged into Eurorack preamps. Whatever converter you use, setting it to 96k isn't in and of itself going to make the recording sound any better, except that the shortcomings in the recording will be more apparent because they are more accurately reproduced.

And finally, back to the practical side of things. Let's say that in 5 years, 96k does become a standard. Upgrade *then*, not 5 years ahead of time. Why? because 5 years from now you're not going to be using 90% of the non-vintage gear you buy today, it'll be too obsolete. I say this as I stare at my "state of the art" ADAT-XT and look at the $250 dongle that came with my original Cubase/VST, both of which are just collecting dust now.

One almost never catches the technology rabbit. When they do, the rabbit is near the end of it's life and dies too quickly to make the chase worthwhile.

G.
 
COOLCAT said:
So its "better" per context the words used in.
per your data the 96 is "better" than the 44.1, because we can SEE it.
per the ears comments its not any "better" as most can't HEAR it.

Sorry to be overly technical. Somewhere back in another thread I had done listening and analysis on 48 vs 44.1. As a result, there is no question in my mind that 96 is better than 44.1, so is 48 and 88.1 for that matter, as I can hear the difference, and analysis verifies that there is indeed a difference.

I went on to explore the reason behind the difference. It's pretty simple and it exactly matches the theory: there is a slight attenuation of very high (over 18.5kHz) frequencies when using 44.1 as compared to any higher standard rate.

Is 96 better than 88.1? In my analysis, no. Is it better than 48? Only very slightly; it shows up on the FFT, but my monitors/ears cannot discern.

Is working at 44.1 a handicap? Not really, if one can't make a decent mix at 44.1, 96 will not help. However, if resources are available, I would record and mixdown at a higher rate. Of course, the penalty of 44.1 will be visited on the final mix, but at least the raw data exists at a higher rate for a higher resolution future. There are further argument in favor of higher rates with respect to porcessing, but I do not consider those here.

I would oppose a statement that 44.1 was inferior for any reason other than attenuation. I have found no evidence, nor am I aware of any theory that holds that 44.1 causes harmonic distortion or any other distortion except for high frequency attenuation. That may be a feature of a crap 44.1 converter, but it is not a mere theoretical result of using 44.1.


PS I still watch broadcast TV from 70 miles away!!! 14 ft VHF and 4 ft UHF antennas :cool:
 
SouthSIDE Glen said:
It's been said here before many times, but it's worth repeating now, I think: 96k does not make content sound better, it only - at best - reproduces it more accurately

That's a great point! When I did the 44.1 vs 48 listening test, to me 48 sounded better, with less smeared highs. But another listener preferred 44.1! Maybe it sounded "warm" :D ;)

I think in 2006, if you are shopping for converters, the argument is somewhat moot. Just about every converter on the market can do 96, and if it's good at 44.1, it's probably good at 96, and if it sucks at one it probably sucks at the other.

If you are using older quality gear, I wouldn't worry too much. Work at whatever rate you can manage. Remember most of the benefit comes from the step from 44.1 to 48.
 
Damn G...
A $250 dongle? you had a $250 "state of the art" dongle! :eek: ...hahahahahaa

I think 16/44.1 will be replaced though, if nothing else because the BIG Companys will shuv it up our dongles and thats all that will be on the shelf.

and we'll have to go buy new Car Decks to play the "new" BluRayStreaming F MUSIC SADVD 3's with fiber optical switching heads that float.

Dongles will be optional. :p
 
simple math (17K/44.1K)
there are only (rounded) 2.6 samples per cycle, and straight lines in between them. see any straight lines in a sine wave? it takes about 5 points to plot a cycle with what i would consider to be a high degree of accuracy (see 96K)
minor distortion, but tons of it.

you can take issue whit my definition of "high degree of accuracy" but not the numbers.
 
giraffe said:
simple math (17K/44.1K)
there are only (rounded) 2.6 samples per cycle, and straight lines in between them. see any straight lines in a sine wave? it takes about 5 points to plot a cycle with what i would consider to be a high degree of accuracy (see 96K)
minor distortion, but tons of it.

you can take issue whit my definition of "high degree of accuracy" but not the numbers.

I used to think converting audio to digital was as simple as plotting voltages over time and you get a kind of line graph representation of your audio. But after reading some of Dan Lavry's papers I have found that there is other math involved to render and reconstruct the waveforms. Sinc functions and decimation and all kinds of goodies. Not so simple math at all...
http://www.lavryengineering.com/documents/Sampling_Theory.pdf
 
COOLCAT said:
Damn G...
A $250 dongle? you had a $250 "state of the art" dongle! :eek: ...hahahahahaa

I think 16/44.1 will be replaced though, if nothing else because the BIG Companys will shuv it up our dongles and thats all that will be on the shelf.

and we'll have to go buy new Car Decks to play the "new" BluRayStreaming F MUSIC SADVD 3's with fiber optical switching heads that float.

Dongles will be optional. :p
Yeah, I honestly don't remember for sure what I paid for Cubase/VST, but yeah, I think it was $249. Long story behind why I never upgraded to VST24 and beyond, but it was mostly because I was using other stuff or doing other things by then.

No question the industry will shove something else down our throats sooner than later - and manage to talk us into believeing that its for our own good by bombarding us with advertising while they simply phase out the old stuff.

Unfortunately I got my movie collection up to date just in time to see VHS phased out for DVD. And that was after I used my VHS collection to replace my Betamax collection :( . Needless to say that with only about a half-dozen exceptions - most of which were gifts from other people - I have not bothered to build a DVD collection. Which I'm glad of, what with Blue Ray and HD-DVD starting to duke it out as early as this Christmas.

I don't know whether you're a young-un or an old fart like me, Coolcat, but after a while one realizes that the technology chase gets one nowhere faster than one thinks, and that all one can rely upon when the dust settles is their own skills. Experience never obsoletes; as long as one stays in the game experience only grows bigger and more useful. And gear can never replace skills. Harvey can probably do more better with one old RCA ribbon mic and a surplus Revox tape recorder than most of us can do with a whole garage full of modern digital gear because he's got the experience and the skills.

G.
 
giraffe said:
simple math (17K/44.1K)
there are only (rounded) 2.6 samples per cycle, and straight lines in between them. see any straight lines in a sine wave? it takes about 5 points to plot a cycle with what i would consider to be a high degree of accuracy (see 96K)
minor distortion, but tons of it.

you can take issue whit my definition of "high degree of accuracy" but not the numbers.

Yes, it will look ugly in digital form, but again those distortions are all ultrasonic frequencies removed by the anti-imaging filter. That is the crux of the Nyquist theory.

Here is a plot of a 17kHz sine wave generated at 96kHz (blue), then downsampled to 44.1kHz and upsampled back to 96kHz (yellow). Note the frequency of the resulting distortion, it's 27kHz.
 
I do know one rock solid fact why 96khz is *sometimes* better than 44.1khz.

If you pitch a sound far down, in 44.1, you will have less high end than in 96. The air of the human voice is in the highest audible register. Lets say you are recording a scream, there's gonna be stuff in the 10 to 20k range. Now pitch this a full octave down. The 10 to 20k will become 5 to 10k, and because 44.1 has no information above (give or take a khz) 24khz, you have nothing at all about 12khz or so. On the other hand, if you recorded at 96khz, and pitch that down a full octave, you will still have stuff in the 10 to 20 khz range.
 
mshilarious said:
Here is a plot of a 17kHz sine wave generated at 96kHz (blue), then downsampled to 44.1kHz and upsampled back to 96kHz (yellow). Note the frequency of the resulting distortion, it's 27kHz.

Question- in these cases does the fact that the blue and yellow lines don't overlap anymore after resampling constitute distortion, ie a difference between input and output? Distortion doesn't always mean fuzz, or some spike somewhere, as I understand the definition.

Not trying to muddy the waters, nor am I trying to imply anything about sound, merely a technical question. :)
 
boingoman said:
Question- in these cases does the fact that the blue and yellow lines don't overlap anymore after resampling constitute distortion, ie a difference between input and output? Distortion doesn't always mean fuzz, or some spike somewhere, as I understand the definition.

Yes that is correct. It's the attenuation I've been talking about. But that is a function of the filter. In theory, if it were possible to make a true brick wall filter, you wouldn't see that. That is why 48 is better than 44.1--there aren't many more samples available, but the filter is just a little farther away from 20kHz to make the difference.
 
Back
Top