Page 1 of 3

24/192 Music Downloads ...and why they make no sense

Posted: Wed Aug 30, 2017 2:44 pm
by Michael Willis
I just read this article about sample rates and bit depths; after having spend some time fretting about sample rate, and then deciding to do everything in 48000, I'm glad to read something that seems rooted in reality that corroborates my choice.

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

EDIT: It also makes me consider setting JACK's bit depth to 16 instead of 32; I'm curious what opinions I can get from other people on that.

Re: 24/192 Music Downloads ...and why they make no sense

Posted: Wed Aug 30, 2017 4:08 pm
by ssj71
g
Michael Willis wrote:EDIT: It also makes me consider setting JACK's bit depth to 16 instead of 32; I'm curious what opinions I can get from other people on that.
Well, requirements when processing audio are different than just playing it back. The lower noisefloor when manipulating audio is helpful to prevent artifacts. For samplerates, really it makes most sense for plugins/processes that need it implement proper oversampling rather than running the whole system at a high rate.

Re: 24/192 Music Downloads ...and why they make no sense

Posted: Thu Aug 31, 2017 2:58 pm
by folderol
The recommendations I see from various people on respected sites like Sound On Sound is to work at least 96k 24bit for all processing and only convert down as the very last step and at the same time applying dither.
They also advise retaining the original in case of later remixes.

Edit. Brain was in neutral 96 not 48 :(

Re: 24/192 Music Downloads ...and why they make no sense

Posted: Fri Sep 01, 2017 3:36 pm
by ssj71
42low wrote: If you have 16/44 you never ever can deliver 24/96. Once lower you can never go back up.
well, you can resample and add empty bits, but you are right that you can't regenerate information that is lost.
42low wrote: How to deliver to those with 'superhearing' who demand 24/96 flac?
EDIT: I just realized you are probably not agreeing they have superhearing only trying to deliver what they want. Still I think they should be educated about why you don't provide useless bits.
Previous response: Now hold on there a second. Did you read the article? It pretty clearly states that there is no such thing as superhearing. Those who demand such listening material are flat out wrong and NO appropriately conducted study has ever been able to prove otherwise.
42low wrote: My personal opinion.
24/96 recordings and projects have more information inside than 16/44.
Effects done over those recordings will also be done over 'more information', so IMO will be deeper in sound. Clearer.
And my gear can handle it, so why should i do less?
They do have more information but that extra information is useless for human ears. Its useful for keeping fidelity through various processes like summing/mixing signals where noise adds together and can reduce SNR or waveshaping which can cause aliasing. Aliasing can be prevented by upsampling before the waveshaping or other non-linear process and downsampling immediately after. This upsample/downsample will typically be less overhead than doing all processing at a higher samplerate. Also any signal can still alias, and if you have information above human hearing, and do a naive waveshaping process on it (and many plugins don't do much testing at 96kz) then you could have worse aliasing. To do something like waveshaping, you should filter out any high frequency data that would alias before the process anyway. So if you are removing all that ultra-sonic frequency information anyway, why carry it around?
42low wrote: But i know. I'm on another recording forum were they worship and advice 16/44 too and there they are totally closed for other options and will never give in (as encountered). So i know many use those settings. This discussion turns up regularly.
Man i've laught about those discussions so many times. :mrgreen: *rofl* They buy tons off gear for tons of money. All must be 'the best'. Must have 'the best' and most expencive condensor mics on the market. All for 'the deep and wide sound'. And they they are going to produce in low quality compressing that quality. How stupid. :mrgreen:
And when i attended them on the contradiction of their vision they don't like it (and me :wink: ) at all. Strange, isn't it? :roll: :mrgreen:

All can do what they want, but i record in 32f/96. Release each song in different qualities and file types to fulfil everyone's wishes. And i save and store my projects in 32/96 without compression or dithering for later use.
Thats cool, do what you like, and you are of course welcome here, but there IS science to this, and I'm going to promote the scientific facts. :)
Actually its not even science, its just math.

Re: 24/192 Music Downloads ...and why they make no sense

Posted: Sat Sep 02, 2017 9:12 am
by barbouze
There is a serie of articles recently released on http://www.earlevel.com/main/ about sampling theory that goes great along this discussion. It has given me a better view on what are actually those samples. :D

Re: 24/192 Music Downloads ...and why they make no sense

Posted: Sat Sep 02, 2017 9:52 am
by sysrqer
42low wrote: So i don't feel the need to discuss at all. I explain what a real and involved professional explained about it, and i have no problem at all with those who tend to believe theoretic philosophers.
It's hardly philosophy. Everyone now has access to all the information your expert had, and possibly more these days considering how much research has been done since, so discrediting everyone else and being a self-professed expert is a bit silly.

Re: 24/192 Music Downloads ...and why they make no sense

Posted: Sat Sep 02, 2017 10:43 am
by Lyberta
Michael Willis wrote:EDIT: It also makes me consider setting JACK's bit depth to 16 instead of 32
As a developer, I don't think is possible. JACK is pretty much hardcoded to 32 bit float and all plugins expect that.

I think you only need high sample rate if your effects produce very high harmonics (> 20 kHz) that you then transpose into the hearing range. Otherwise, high rate is useless.

Bit depth, on the other hand, is very important but everyone uses 32 bit float for effects so you don't really have a choice here. As a developer, it would be idealistic to use 64 bit floats inside the plugin but you would still need to convert them to 32 bit to pass to other effects.

Re: 24/192 Music Downloads ...and why they make no sense

Posted: Sat Sep 02, 2017 11:44 am
by barbouze
42low wrote:
barbouze wrote:There is a serie of articles recently released on http://www.earlevel.com/main/ about sampling theory that goes great along this discussion.
For all information. I'm NOT talking theory.

[...]

So i don't feel the need to discuss at all. I know how it works, and what too do and choose for good/best results.
I only explain at those who are interested what a real and involved professional explained about it, and i have no problem at all with those who tend to believe theoretic philosophers.
My advice will stay as in my first reaction, produce at high ranges. To play it then it can always be lower. If one don't do it like that? Then simply don't.
Theory was meant to be understood as sampling theorem i.e. everything related to ADC<->DAC. I thought it would be of some interest for all readers of this topic and was not meant to disprove your explanations. Sorry if you took it the wrong way :?

Re: 24/192 Music Downloads ...and why they make no sense

Posted: Tue Sep 05, 2017 3:38 pm
by ssj71
I haven't read every word here, but those images only depict part of what sampling theory tells us. They're just examples of aliasing. They do not really represent the significant part of the discussion of sample-rate selection (what information is useful) and don't have any detail about bit depth (which is an orthogonal issue). I think neither one of us wants to argue, so I won't worry about it here, but if anyone has additional questions or wants more explanation on why I disagree with many of these points I'll provide it.

Re: 24/192 Music Downloads ...and why they make no sense

Posted: Tue Sep 05, 2017 5:37 pm
by ufug
It's a challenge for those of us who are not engineers to not react to audio esoterica with a raised eyebrow.

This is a great example of why. People who seem like experts do not reach consensus, and yet both sides approach the arguments with authority. What are the rest of us to think if we can't actually hear the difference?

To raise the eyebrow even higher, when Steve Jobs advocates something I immediately suspect it is some kind of hype (in this case: a new file format, new hardware players, new prices), which makes me a little wary that we can effectively hear the difference in real world environments.

It's all very interesting though, both the article and the discussion. For now, I have so many areas to work on with recording/mixing that sample rates can safely move waaay to the back of the line. :)

Re: 24/192 Music Downloads ...and why they make no sense

Posted: Tue Sep 05, 2017 7:37 pm
by ssj71
ufug wrote:It's a challenge for those of us who are not engineers to not react to audio esoterica with a raised eyebrow.

This is a great example of why. People who seem like experts do not reach consensus, and yet both sides approach the arguments with authority. What are the rest of us to think if we can't actually hear the difference?
I understand that and am sad to be contributing to it. I highly recommend you read the originally posted article to better understand the principles under discussion. Its very well written, even if mostly targeting the context of files for playback.

Re: 24/192 Music Downloads ...and why they make no sense

Posted: Tue Sep 05, 2017 9:39 pm
by ufug
ssj71 wrote: I understand that and am sad to be contributing to it. I highly recommend you read the originally posted article to better understand the principles under discussion. Its very well written, even if mostly targeting the context of files for playback.
FWIW, I did read it and everyone's comments here as well. I sincerely admire the reasoning on all sides, and I have a basic (loose?) grasp on the subject. It's a fascinating article and it covers a lot more ground than just sampling rates.

Perhaps my comment was off-topic, which was not my intention. I suppose I was thinking that if there was consensus that it would have some non-theoretical applicability, but it's certainly not a difference I can discern. Please proceed!

Sometimes I miss my Portastudio.

Re: 24/192 Music Downloads ...and why they make no sense

Posted: Wed Sep 06, 2017 2:51 pm
by ssj71
42low wrote:
But i do think most of us will hear it too.

Then i don't point at the difference between an 24/96 export and a 16/44.1 export of one project.

But what i point at is two different total projects in both 24/96 and 16/44.1.
Test it. Try it.

Start a project in 24/96. Record within it. Import (also 24/96) in it.
Edit it serious. Add reverb, add compressor or limiter, EQ. And do some editings on the master.
And export this project to a 24/96 flax.

Do the same with a 16/44.1 project. Record within it. Import (also 16/44.1) in it.
Edit it serious. Add reverb, add compressor or limiter, EQ. And do some editings on the master.
And export this project also to a 24/96 flax (so there can't be the difference).

Then load both in Audacity each in a seperate track, and mute both alternately to try to hear if difference is there.

I'm not discussing. I only want to share my experiences with it.
I did this and heard the difference (and i have no superhearing or whatever). I would advice to try it once. If you then don't hear it ... ok!
This does sound like a good exercise and I'd expect the results to be better with 24/96. But I postulate that you'd have just as good results with 24/44.1 and exporting to 16/44.1.

Re: 24/192 Music Downloads ...and why they make no sense

Posted: Wed Sep 06, 2017 3:07 pm
by ssj71
42low wrote:By accident looking for another one within my bookmarks i found another interesting one about this subject.
http://www.trustmeimascientist.com/2013 ... n-it-isnt/

Didn't read it again at full now. But i looks like i found it interesting enough to bookmark it.
So i'm not going to defend or attack the content of it. :mrgreen: Just for more visions and information. He has some interesting statements too.
I've read that article before and absolutely agree with it. Its another good read.
I think we agree on the end results, maybe just have different interpretations of the explanation.
42low wrote:And he refers tot his intersting pdf too.
http://www.lavryengineering.com/pdfs/la ... _audio.pdf
If i understand this one well the best option would be 96khz, were the bitrate 16 or 24 doesn't matter?
"At 96 KHz sampling rate the theoretical bandwidth is 48 KHz."
I haven't read that one, but from the previous article its mostly a matter of hardware design. Designing hardware to accurately play back 96khz files is easier and therefore cheaper, but you start to get intermodulation distortion at those rates. They state several times numbers around 50-70Khz as the best balance of easier hardware design vs distortions, but there are no standard rates in that band. So you either overshoot and spend a bit more CPU on processing or undershoot and hope you hardware is well designed. IMHO most hardware is designed around CD rates anyway so I just record at that.

But we're totally in agreement about working in 24 bits.

Re: 24/192 Music Downloads ...and why they make no sense

Posted: Wed Sep 06, 2017 5:09 pm
by Jack Winter
I think 24/192 makes perfect sense!

* The people selling high res downloads can sell the music again.
* The audiophiles are happy to buy and listen to it.
* The people that like to argue about it can have a field day, time and time again.

Really, what is not to like about it? 8)