Why 96kHz is better than 192kHz

MarkH · Post by **MarkH** » Fri Dec 24, 2004 7:41 pm

This makes for an interesting article. Thought it would be worth sharing. Enjoy!

http://www.gearslutz.com/board/showthre ... adid=25038

hugolp · Post by **hugolp** » Fri Dec 24, 2004 11:47 pm

That is plain stupid.

It is true that if you have 192Khz converters they will work slightly better at 96Khz. But the difference is small. If they are good converters they will sound good and if they are not so good, they will sound kind of the same at any rate. You get a little less performance from the converters but you get better sound because of the better bandwith. And what happen if you have 96Khz converters? Should you only use 48Khz maximum? So if I want to drive my car at 100mph I should get a car that gets to 200mph, cause of the security improvement I am going to get?!?!??!!?!?

And what if you dont record on to the computer? You will get better results with 192Khz(if your computer can handle it). Less jitter, and more frequency bandwith.

At the end I wont hardly notice the difference from 48Khz and above. I work at 96Khz because its too much for my computer to record at 192Khz, but his last and main reason is bollocks.

Hugo

PS:Sorry for my English, when I get technical its still dificult.

hugolp · Post by **hugolp** » Fri Dec 24, 2004 11:51 pm

And if you keep on reading the thread he gets answered

Hugo

drumroll57 · Post by **drumroll57** » Sat Dec 25, 2004 4:20 am

Here's another point of view.

Sound is very difficult to differentiate depending on the size of the physical location you are listening in.

- A properly encoded VBR MP3 sounds quite convincing at my computer's speakers.

- Nothing wrong to my ears with a 44.1 kHz - 16 bit CD on an average-size studio's control room high-end monitors.

But neither of those sound good at all in a room large enough to accomodate 2,000 listeners. However, an analog tape recording of the CD above, or a vinyl cut of it will sound incredibly better in the same large room. (smoothing out the waveforms and reconstituting the broken stairsteps, especially in the high frequencies)

I have performed a number of A/B tests that indicate that these characteristics only come into play in bigger-sized rooms. There, 24 bit - 96 kHz is making a HUGE and IMMEDIATE difference to 16 bit - 44.1 kHz in the perceived quality of sound. No need to be a 'Golden Ear™'!

I call it The Magnifying Glass Effect.

As an analogy, try printing a .jpeg of a picture you downloaded from the web onto a 2 meter X 3 meter poster. The picture looked fine on your monitor screen, but now you see all those jagged and pixelized digital artifacts. But I bet you that the original 125 Megabyte version of it will look just perfect on that giant poster. It's a matter of scale.

I cannot comment on 192 kHz, as the last time I ran those tests I didn't have such a portable unit available. (...and still don't) However, those are things that you need to test in the proper and appropriate environment. Theoretically, the higher sampling rate should yield smoother highs. That is why DSD (Sony's proprietary Super Audio CD format) is sampling at above 1 GHz!.... But then they compress it a bit, because their perceptual studies have indicated that this higher rate is the threshold of the human ear. The same way you can set find out the actual treshold of resolution between digital imaging and 35 mm film. A megapixel camera picture certainly doesn't cut it when projecting your image on a movie-theater sized screen. Maybe 15 -20 megapixels will do the job.

Who cares what a race car does in your living room? All that matters is how it performs on the race track. (if you are a race car pilot, that is...)

...control-room race car drivers on that other board, sound like to me....yano?

D.

ethios4 · Post by **ethios4** » Sat Dec 25, 2004 8:08 am

I read something a while back about the desirability of the sampling rate used in production being a whole-number multiple of the desired end sample-rate. In other words, if your mixing for CD, its best to use 88.2 rather than 48 or 96, because it's 2x44.1, thereby reducing mathematical inaccuracy.
Any thoughts on this?

special ed · Post by **special ed** » Sat Dec 25, 2004 8:11 am

this might be a stupid question but,

if my recordings are at 44.1 k and 24 bit, could i record those back to or transfer them to a 96k 24 bit rate for better results on a club system?

drumroll57 · Post by **drumroll57** » Sat Dec 25, 2004 9:10 am

special ed wrote:this might be a stupid question but,

if my recordings are at 44.1 k and 24 bit, could i record those back to or transfer them to a 96k 24 bit rate for better results on a club system?

Passing them through a tape machine, or something that will 'recompose' the smoothness of the analog waveform (to get rid of the stairsteps), yes!

It seems that although oscilloscopes and lab technicians love the straighness of digitally encoded sound, perceptual studies of human hearing indicate that our ear is far more sensitive to those irregularities (stairsteps) than originally thought, and more forgiving when it comes to hiss, signal-to-noise ratio and the likes, as the latter do not mess up the 'continuity' of the flow of information.

D.

MarkH · Post by **MarkH** » Sat Dec 25, 2004 10:14 am

special ed wrote:this might be a stupid question but,

if my recordings are at 44.1 k and 24 bit, could i record those back to or transfer them to a 96k 24 bit rate for better results on a club system?

Apogee

ethios4 · Post by **ethios4** » Sat Dec 25, 2004 6:56 pm

special ed wrote: if my recordings are at 44.1 k and 24 bit, could i record those back to or transfer them to a 96k 24 bit rate for better results on a club system?

Timestretching works better at higher sample-rates, as there is more information to work with.

montrealbreaks · Post by **montrealbreaks** » Sun Dec 26, 2004 4:15 pm

ethios4 wrote:
special ed wrote: if my recordings are at 44.1 k and 24 bit, could i record those back to or transfer them to a 96k 24 bit rate for better results on a club system?
Timestretching works better at higher sample-rates, as there is more information to work with.

Agreed 100% on the Timestretching issue.

If you have 96K samples in Live, that's great - even if you only output 44.1 or 48 KHz. It improves the time-stretching, and the final sample rate conversion happens at Live's outputs.

Plus, even an analog mixer will even out some of those "stairsteps" that you're talking about. Just attenuate it a touch or boost it a touch with the EQ, and bob's your uncle. Remember though, the less analog components in your chain from recording to re-recording will produce a cleaner sound, and only one component is necessary to re-record to get rid of the digital harshness...

special ed · Post by **special ed** » Mon Dec 27, 2004 1:10 am

so if i took a 44.1k 24 bit loop, ran it out into some analog outboard gear that has a vca and ran it back in to record it at 96k, that would reduce some of the stairsteps and enable better time-stretching?

special ed · Post by **special ed** » Mon Dec 27, 2004 2:40 am

well, i just turned a few 44.1k loops into 96k, and i notice a considerable difference (havent even tried time-stretching them yet) the drums sound more punchy yet warm, i guess it all sounds more organic. im stoked. i never recording at 44.1 again. of coarse my old g4 ti book 500 will probably choke on them if i make my live set at 96k. oh well, they sound killer on my desktop. thankx much for the ideas guys.

Moonburnt · Post by **Moonburnt** » Mon Dec 27, 2004 2:45 am

I'm confused about this stairsteps business, because surely the reconstituted analog sound produced by any DAC is going to be more or less splined between samples so there won't be any steps...I thought it would be more of a case of losing all the subtleties between samples due to the fact that the original waveform wasn't perfectly smooth between the sampling points. If that is the case, no amount of re-analogifying or resampling would invent that information again - it might sweeten the sound, but surely it wouldn't make it any more transparent or accurate?

special ed · Post by **special ed** » Mon Dec 27, 2004 3:22 am

while i doubt that the original recording samples were completely smoothed out, they are given more warmth. what i did was i opened up logic, converted everything from 44.1k to 96k, switched logic to record at 96k, ran each loop through an analog vca and back into logic to record it at 96k. i do notice an improvement of sound, to me it does sound more warm and natural.

montrealbreaks · Post by **montrealbreaks** » Mon Dec 27, 2004 4:29 am

Moonburnt wrote:I'm confused about this stairsteps business, because surely the reconstituted analog sound produced by any DAC is going to be more or less splined between samples so there won't be any steps...I thought it would be more of a case of losing all the subtleties between samples due to the fact that the original waveform wasn't perfectly smooth between the sampling points. If that is the case, no amount of re-analogifying or resampling would invent that information again - it might sweeten the sound, but surely it wouldn't make it any more transparent or accurate?

Agreed; It will not recover the original waveforms that were lost due to the digital conversion in the first place but...

resampling through VCA gear has an advantage in that the electronics in analogue gear aren't capable of instantaneous voltage changes - there are SLIGHT delays in those microvoltages that occur between two samples that tend to "smooth" out the steps created by the sample and bit rates.

Think of it like a portamento for voltage, ever so slight, and your analog gear will re-introduce new information, smoothing out the difference between two samples. Naturally, this will not recover lost high end information, BUT it will reduce the square wave super-harmonics that come at the Nyquist frequency +.

At least that's what I THINK. I got an Arts Degree, not a science degree.

Why 96kHz is better than 192kHz

Why 96kHz is better than 192kHz

The larger the space....