96Khz vs 192Khz

[Mr. Afternoon]

I cannot, for the sake of my life, figure out what is the need of 192Khz. I did some recording tests. I was not able to get any ultrasonics in that range. Looked at some music. Any content in the 50-96khz response was non-existant. So why, WHY, WHY SHOULD I BE INDULGING IN IT? To me it just seems like its wasting space. Unless people like the sound of ultrasonic noise.
Also a 80 minute 5.1 24/192 file is pretty big.

 

[LuvMyQuad]

I cannot, for the sake of my life, figure out what is the need of 192Khz.

Because they can... The Chicago and Doobies Quadio release were 192kHz I believe. The 4.0 files of the Chicago double albums were outrageously large files.

But isn't there some advantage to using the high sample rates when doing mixing, editing, etc?

 

[IMachine]

For mixing there are differences:
https://www.prosoundnetwork.com/pro...-tech-96-khz-vs-44-1-khz-lets-settle-this?amp

 

[DuncanS]

There are no ultrasonics that we can hear, there is a lot of rubbish written about why higher sample rates are needed.

By increasing the sample rate you spread the quantisation noise (in effect the error from digitising) over the whole bandwidth, so it is reduced in the hearing frequency range, a good thing, but not necessarily audible. By increasing the sample rate you can limit the cut-off slope of the Low Pass Filtering, which does seem to benefit a lot of people. This is what used to be called the Brick Wall effect, i.e. the slope is very steep, some people don't like it because of the phase shifts between frequencies it can introduce. The main advantage of high sample rates is when you decimate (a mathematical operation), so reduce the sample rate and bandwidth, this increases the resolution of the signal.

Resolution, i.e. number of bits, is much more important.

The other thing is that when looking at audio spectrums you need to look at it with the amplitude in dBs, we hear volume logarithmically, so a better indication of what we can hear. BTW I'd be surprised if I can even hear 15kHz.

 

[Mr. Afternoon]

But isn't there some advantage to using the high sample rates when doing mixing, editing, etc?

Not as far as I have noticed. Maybe I'll create a test track in 5.1 24/192 to test it out.

 

[DuncanS]

Not as far as I have noticed. Maybe I'll create a test track in 5.1 24/192 to test it out.

64 bits or Floating Point arithmetic in the mixing would be better

 

[Kal Rubinson]

There are no ultrasonics that we can hear.....

By definition.

 

[Plan9]

I've done tests between both sample rates (actually all the standard sample rates really) and I think the difference mainly lies in how your DAC reacts to a particular rate.
With some older DACs, I felt 192kHz sounded worse -and indeed they introduced noise in the hearing range at that rate-, while others sounded maybe 2%-5% "smoother" and "more analogue" at 192kHz compared to 96kHz.
I also found that the better the DAC, the less difference there were between sample rates.
Personnally I think 88.2kHz/96kHz is enough and hits the sweet spot.
YMMV.

 

[jimfisheye]

I cannot, for the sake of my life, figure out what is the need of 192Khz. I did some recording tests. I was not able to get any ultrasonics in that range. Looked at some music. Any content in the 50-96khz response was non-existant. So why, WHY, WHY SHOULD I BE INDULGING IN IT? To me it just seems like its wasting space. Unless people like the sound of ultrasonic noise.
Also a 80 minute 5.1 24/192 file is pretty big.

I was wondering about that 10 years or so ago. So a test. Let's try put any difference between 96k and 192k out of the range of perception bias and into the obvious. So 100 lossy conversions back and forth between 96k and 192k! If something is going to snowball, this should do it. (Seen those Youtube videos re-uploaded 100 times and it's just noise artifacts at the end? Like that.) Using SOX sample rate conversion at highest settings.

Obviously the first conversion from 192k to 96k will alter the data set and not completely null. We're looking for cumulative damage though. If the premise is that 96k somehow cannot hold a complete audio signal (even if just in the decimal dust), then it should snowball over many iterations. Like wav to mp3 would.

The first 192k -> 96k -> 192k.
A/B those two 192k files and they sound the same. (Subjective and all.)
Null them and the difference is not audible at normal monitor settings. The meter wiggles around -100db.

Cut to the chase: Iteration 100 nulled perfectly with iteration 3. Once that initial ultrasonic margin noise was sliced off, the audio stayed digitally perfectly intact over 100 lossy conversions between 192k and 96k.

My conclusion is that if someone determines there's some restoration processing math that can achieve something with sample rates above 96k, that might be all well and good. Not sure how you might prove that. But at the end of the day a 96k "container" can hold a 100% complete audio signal.

Keeping your audio always 24 bit and using balanced connections for analog signals makes an actual difference. If you have some consumer DAC that sounds more accurate at HD vs SD sample rates, you can always upsample any SD program as a workaround.

PS. No one knows what the sound of ultrasonic noise is because no audio speakers reproduce that high up. Or human hearing. If someone things they're "feeling" something... Well it's not coming out of your speakers. Perceiving transistor aura perhaps?

PSS. In listening tests I've done in the past I could hear a difference in the DAC running at SD vs HD sample rates in my older MOTU interfaces. I can't for the life of me tell the difference in the Apogee units. So exactly what I paid for there! Transparent operation even at SD sample rates. 24 bit at 96k seems like a sweet spot to "set it and forget it" and have perfect audio reproduction no matter what edge case. But any sample rate with 24 bit is a really near perfect container.

 

[jimfisheye]

64 bits or Floating Point arithmetic in the mixing would be better

Yeah, that lets you preserve the full 24 bit resolution of your source audio with the fader pulled way down. In fixed point, the bit resolution decreases with volume level. Record a signal at -48db into 16 bit fixed. Result is an 8 bit recording. Record the same into 24 bits. Result is a 16 bit recording. 24 bit is pretty bulletproof! 64 bit floating point preserves all the decimal dust no matter what you do. Most modern DAW apps (eg Reaper) have 64 bit floating point mix engines.

 

[Marplot]

I always enjoy these technical conversations and appreciate the input and lack of rancor.

I like 96 source only because what I own so far seems not to be mastered as hot as most CDs are today.

I do not suffer from the belief that I can hear outside the normal human range; actually as an 11C and later on a mortar plt ldr I can pretty much guarantee I do not have full band hearing.

 

[jimfisheye]

It's pretty easy to gaslight yourself with some of this!
You can buy 100 CDs in a row and end up with literally every one being volume war mastered. You might conclude 16 bit at 44.1k sample rate sounds shrill and lo-fi. How could the actual industry with studios and recording engineers turn out consistent product that is so damaged it sounds like some amateur facsimile operation? Yet, here we are.

We have perfect recording containers now. We also have a bunch of nearly perfect formats. Almost all of these can take lossy conversions back and forth and still not do obvious damage. The garbage in, garbage out rule still accounts for the garbage you might hear.

I highly recommend taking inventory on multiple released versions on your favorite albums. Revealing some of the more novelty releases like this is an eye opener! (Ear opener?)

I see people make comments that they'd never listen to an mp3 but then give a shrill loud volume war master a 9 or 10 in a poll because it was in a lossless format. Sure, that destroyed master was lossless! Woohoo!

Digressing a bit. This is a golden age of audio with these modern digital formats. Trust your ears.

 

[atrocity]

With some older DACs, I felt 192kHz sounded worse -and indeed they introduced noise in the hearing range at that rate

Funny that you'd mention this...I just had to prepare Chicago's "25 or 6 to 4" for radio use and started with the 192k version. When I opened it in Audition there was a really annoying whistle/whine/squeal that made me think I had a hardware issue. But that noise was completely gone after SoX downconversion.

 

[dadregga]

It's useful for mixing, but beyond like (maybe, this is still debatable) 48khz really all you're doing with PCM is sampling and storing ultrasonics you can't hear and that your equipment simply won't reproduce.

It's possible some older DACs, due to internal sampling logic and bad implementations, will respond differently based on the input sample rate - but that's a DAC problem, vanishingly rare with modern DAC chips and well-designed implementations, and not a problem with the sampling rate of the source data anyway.


If you can audibly tell the difference between 48khz and 192khz when you're using the same mastering, with the same sample path, with volume-matched output, then either your DAC has a problematic implementation, or you're fooling yourself and an ABX test would demonstrate as much.

We know the range of human hearing, and human ears don't register ultrasonics. Anyone arguing otherwise is at variance with basic physics, and bears a large burden of proof.

Mastering and mixing matters. Final samplerate simply doesn't, past a certain point.

 

[dadregga]

Will say one thing though -

Labels that sell the original files that were the master output (e.g. DXD) or close to it does provide me with some degree of assurance that nobody has monkeyed with the levels or mastering or done any half-assed downsampling of the files in between me and the engineer who created them - so given the choice, I'd rather have those. Not because I can hear a difference, but simply because I want the highest quality digital copy I can get with the least generational loss, all other things being equal.

And that is a kind of value.

 

[DuncanS]

I've done tests between both sample rates (actually all the standard sample rates really) and I think the difference mainly lies in how your DAC reacts to a particular rate.
With some older DACs, I felt 192kHz sounded worse -and indeed they introduced noise in the hearing range at that rate-, while others sounded maybe 2%-5% "smoother" and "more analogue" at 192kHz compared to 96kHz.
I also found that the better the DAC, the less difference there were between sample rates.
Personnally I think 88.2kHz/96kHz is enough and hits the sweet spot.
YMMV.

Some of the earlier ADCs & DACs had degraded linearity (so added distortion) as you pushed the sample rate up, in fact a lot had 16-bits of resolution but may be only 12-14bits of linearity at the optimum rate, and this degraded with increased sample speeds. Luckily things have improved

 

[salsdali]

Because they can... The Chicago and Doobies Quadio release were 192kHz I believe. The 4.0 files of the Chicago double albums were outrageously large files.

the Neil Young Archives: Volume 1 BD box set is 192kHz also

 

[jimfisheye]

Any 192k release I rip to my library I convert to 96k with SOX. My SOP without even listening to the program at 192k. Assuming that some higher end system in some VERY well treated room might reveal something to me in the future (which I believe I proved false)... I just don't care. And I collect for sound quality!

It's funny too. You'd think collecting for best quality sounding release would have been a thing of the analog past. Now there'd be one digital master to rule them all. Nope! More novelty releases with damaging mastering work than ever before to wade through! Some albums never even see the unmolested master released. But we have perfect delivery formats now for anyone who actually wants to release something in full quality.

 

[Mr. Afternoon]

Okay, I did manage to achieve some sort of 192Khz response from my mic...when I distorted the output with my singing!
But yeah, everything seems to be capping out at 96Khz at my end (with the exception of the triangle, but I don't use that in my music regularly) so I'll stick with 24/96.

 

[kfbkfb]

https://tams.informatik.uni-hamburg...beitung/high-quality-audio-coding.pdf#page=19^^^
... a channel that attains audible transparency...
58kHz sampling rate
20-bit representation in a flat noise floor


This research dates from the mid-1990s.


Kirk Bayne