Go Back   Cockos Incorporated Forums > REAPER Forums > Recording Technologies and Techniques

Reply
 
Thread Tools Display Modes
Old 10-18-2015, 03:49 AM   #1
Mr. PC
Human being with feelings
 
Mr. PC's Avatar
 
Join Date: Apr 2010
Location: Cloud 37
Posts: 1,071
Default Oversampling vs. Increased Sample Rate

So, I know this is a hot topic, and has been debated quite a bit already, but after a few hours googling, I'm not seeing any good answer.

Oversampling - Increases the sample rate of a given plugin, so that Aliasing is reduced. The downside is that the sample-rate conversion introduces its own distortions (also aliasing?).

Running the entire project at a higher sample rate - Some say it will achieve the same effect as oversampling. Others disagree. I don't see why not. (Note, I am not talking about *exporting* at a high sample rate, but mixing everything; in other words, oversampling the entire mix to reduce aliasing, without having to constantly upsample and down-sample (imagine you run 10 oversampled plugins).

You can see the results of different resampling softwares. Reapers does better than most but is beat my a few. The top seems to be the new Ableton Live, and R8tbrain.

http://src.infinitewave.ca/

Based on everything I'm reading, it seems that the best quality can be achieved

1 - *Mixing* (Not recording, or publishing, but mixing) in a higher sample rate. For example 192k.

2 - Not using any oversampling.

3 - Rendering from Reaper at said sample rate.

4 - Using R8tbrain to downsample back to 48 (or 44.1 is that's your thing).
__________________
AlbertMcKay.com
SoundCloud BandCamp
ReaNote Hotkeys to make Reaper notation easy/fast
Mr. PC is offline   Reply With Quote
Old 10-18-2015, 04:03 AM   #2
EvilDragon
Human being with feelings
 
EvilDragon's Avatar
 
Join Date: Jun 2009
Location: Croatia
Posts: 24,790
Default

4. Using SoX to downsample.


That's actually what Live 9 is using.


I would argue that for home recording, anything above 48k is a waste of hard drive space and CPU, though. Unless you have some really high-end gear, as well as a well-treated mixing room in your house...
EvilDragon is offline   Reply With Quote
Old 10-18-2015, 04:06 AM   #3
xpander
Human being with feelings
 
xpander's Avatar
 
Join Date: Jun 2007
Location: Terra incognita
Posts: 7,670
Default

Quote:
Originally Posted by Mr. PC View Post
You can see the results of different resampling softwares. Reapers does better than most but is beat my a few. The top seems to be the new Ableton Live, and R8tbrain.
http://src.infinitewave.ca/
Reaper resample modes were updated for v5. There's a new Extreme HQ mode for highest possible quality and other mode names were changed accordingly. User tests suggest really clean sweep, apparently Infinite Wave just haven't tested/updated v5 into their site yet.
xpander is offline   Reply With Quote
Old 10-18-2015, 04:08 AM   #4
Xenakios
Human being with feelings
 
Xenakios's Avatar
 
Join Date: Feb 2007
Location: Oulu, Finland
Posts: 8,062
Default

Effective oversampling may require oversampling factors beyond practically usable sampling rates. For example oversampling at a 8x factor isn't reaching too far, especially distortion type of audio processing can produce additional harmonics quite easily that would in theory require oversampling that much and even more. 8 x 44.1khz is 352.8khz, does it seem practical to run at a samplerate like that? Or why not play it safer and consider 64x oversampling is closer to "good enough". That would require running at 2822.4khz.

IMHO it would be best if people just listened to the results they are getting and not get too caught up with the numbers and visual analysis results.

Do you actually have an audible problem when producing audio that clearly is caused by aliasing, resampling defects or things not being oversampled enough? Audio examples would be nice...
__________________
I am no longer part of the REAPER community. Please don't contact me with any REAPER-related issues.
Xenakios is offline   Reply With Quote
Old 10-18-2015, 05:38 AM   #5
LightOfDay
Banned
 
Join Date: Jun 2015
Location: Lower Rhine Area, DE
Posts: 964
Default

Quote:
Originally Posted by Mr. PC View Post
... Some say it will achieve the same effect as oversampling. Others disagree. ...
yeah, some say the earth is flat. others disagree.

that is all not a thing of opinion but of facts.

1. fact is, that higher samplerates (in a project) than 44.1 khz doesnt gain you anything.

2. let the plugin(developers) decide, if a plugin needs so or so much oversampling.

problem solved.

like Xenakios said, you cant have such a high samplerate that every plugin is satisfied. (if there is at all the from some stated difference in plugins sounding better at higher rates.)

an example for why not gaining anything at higher samplerates is ReaEQ. ReaEQ doesnt oversample, so the hi-shelve filter isnt really a hi-shelve because when its boosted its curve does go down steep and hits 0 at 20.000hz or 24.000hz. so its a band and not a shelf. if oversampled via a wrapper its a shelf. if higher samplerate is put in its again a bell curve and not a shelf.

see, higher samplerate doesnt do anything here. and that is with all plugins. the people talking about higher samplerates in a project do make sound plugins better are talking bs.
LightOfDay is offline   Reply With Quote
Old 10-18-2015, 07:11 AM   #6
EvilDragon
Human being with feelings
 
EvilDragon's Avatar
 
Join Date: Jun 2009
Location: Croatia
Posts: 24,790
Default

Quote:
Originally Posted by LightOfDay View Post
an example for why not gaining anything at higher samplerates is ReaEQ. ReaEQ doesnt oversample, so the hi-shelve filter isnt really a hi-shelve because when its boosted its curve does go down steep and hits 0 at 20.000hz or 24.000hz. so its a band and not a shelf. if oversampled via a wrapper its a shelf. if higher samplerate is put in its again a bell curve and not a shelf.
It's a shelf, but it's band-limited. There's a bit of a difference.
EvilDragon is offline   Reply With Quote
Old 10-18-2015, 08:32 AM   #7
Joaquins Void
Human being with feelings
 
Join Date: Dec 2014
Location: Stockholm
Posts: 206
Default

I export at 96KHz. I don't see any need to record and mix at anything higher than 48Khz. Just the actual export is done at a higher rate.

For some processes that produce aliasing there is no clean way of getting rid of it, and even if the plug is oversampled you can still improve it by oversampling it more. And, an offline down sampling filter can do a much better job of it. (Many built in down samplers will be trivial by comparison.)

A Biquad EQ is a naturally bandlimited process so there is no need to oversample it, where as for example clipping produces an infinite overtone series which will fold back into the audible spectrum when the bandwidth runs out.

To completely remove an infinite overtone series you'd need infinite oversampling, though we'll usually settle for it being inaudible. But that may depend on what came into the process to begin with, so it's hard to predict. I have not met an FM-synth yet that I could not coax into aliasing.

So theoretically, if one is using an assortment of effects and such it makes sense to export at as high a rate as one can and then down sample offline. I've never managed to hear an improvement beyond 96KHz though so I decided to stop there.

But as with everything, whether you'll see an improvement depends on the specifics of what you are doing. You pretty much just have to try and see.
Joaquins Void is offline   Reply With Quote
Old 10-18-2015, 10:08 AM   #8
timlloyd
Human being with feelings
 
Join Date: Mar 2010
Posts: 4,713
Default

Quote:
Originally Posted by Xenakios View Post
IMHO it would be best if people just listened to the results they are getting and not get too caught up with the numbers and visual analysis results.
I would bet money that these days almost nobody has an actual problem with aliasing.

But worrying about stuff like this is 'easier' than worrying about the important reasons our recordings and mixes don't sound as good as we want them to ...
timlloyd is offline   Reply With Quote
Old 10-18-2015, 11:25 AM   #9
Mr. PC
Human being with feelings
 
Mr. PC's Avatar
 
Join Date: Apr 2010
Location: Cloud 37
Posts: 1,071
Default

Quote:
Originally Posted by EvilDragon View Post
4. Using SoX to downsample.

That's actually what Live 9 is using.

I would argue that for home recording, anything above 48k is a waste of hard drive space and CPU, though. Unless you have some really high-end gear, as well as a well-treated mixing room in your house...
I specifically said I am not talking about recording or publishing, so there is no extra hard drive space wasted (maybe temporarily for the partial render, but if Reaper 5's downsampling is as good as r8tbrains, then there's no need for that.)

Quote:
Originally Posted by xpander View Post
Reaper resample modes were updated for v5. There's a new Extreme HQ mode for highest possible quality and other mode names were changed accordingly. User tests suggest really clean sweep, apparently Infinite Wave just haven't tested/updated v5 into their site yet.
Yes, I forgot to mention that in the post.

Quote:
Originally Posted by Xenakios View Post
8 x 44.1khz is 352.8khz, does it seem practical to run at a samplerate like that? Or why not play it safer and consider 64x oversampling is closer to "good enough". That would require running at 2822.4khz.
What if we simply changed the project sample rate to 2822.4 or 3072 (for 48 mixdowns) right before export? No tax on CPU, just a longer export time.


So my new solution.

Mix at 48k, no oversampling, then right before export, switch to 3072k, render with new fancy HQ mode down sampled to 48k. No more aliasing, no re-sampling errors, no extra CPU or hard drive space. (Maybe need to tweak after listening to export, but I do this constantly anyhow).
__________________
AlbertMcKay.com
SoundCloud BandCamp
ReaNote Hotkeys to make Reaper notation easy/fast
Mr. PC is offline   Reply With Quote
Old 10-18-2015, 12:02 PM   #10
The Telenator
Banned
 
Join Date: Dec 2011
Location: Oud West, NL
Posts: 2,335
Default

SoX, as EvilDragon said is best or, I think, tied for best if you really need to alter your files.

It is so, so ... SO tempting to wade (for maybe the hundreth time?) into this twisted fruit issue. So tempting to lay waste to all the insanity, silliness and the things no one really can hear, though they'll swear on their mother's grave they can.

The story so far? Well, I don't give a hoot what anyone thinks they hear, on every (this means on ALL) blindfold tests people can only tell the higher sample rate 50% of the time. You know what that means, don't you? Yeah, it means all testers are WRONG half the time.

Ah, well, this is as much time as I care to waste on this incredible waste of time and talent.

Just please know that while a few punters are spending their brief lives fiddling about with these numbers, speculating, refusing to live in reality, and related masochism and self-torture, the REST OF US are actually getting an awful lot of real work and music done.

I use plugins, many of which handle oversampling within. I only use those VST that are built by truly intelligent devs who know what the hell they are doing in this area, such as Fabien of good ol' Tokyo Dawn. Otherwise, I never give it much of a thought.

I've pretty much been affiliated with and employed in one studio or the other since 1977. Do you know how many times in all those years a real pro musician or real engineer has even requested or mentioned anything about sample rate? ONCE.

Knock yourselves out.
The Telenator is offline   Reply With Quote
Old 10-18-2015, 12:04 PM   #11
Xenakios
Human being with feelings
 
Xenakios's Avatar
 
Join Date: Feb 2007
Location: Oulu, Finland
Posts: 8,062
Default

Quote:
Originally Posted by Mr. PC View Post
What if we simply changed the project sample rate to 2822.4 or 3072 (for 48 mixdowns) right before export? No tax on CPU, just a longer export time.
Complete waste of time and effort. You even neatly avoided the question of "do you actually hear any problem?".

In any case, many plugins won't even operate correctly at those kinds of sample rates. It's not uncommon for DSP code to be hardcoded to work only at some well accepted sample rates, like 44.1khz,48khz, 88.2khz, 96khz. Some plugins may not care about the sample rates, but some will and then you are in a new mess.
__________________
I am no longer part of the REAPER community. Please don't contact me with any REAPER-related issues.
Xenakios is offline   Reply With Quote
Old 10-18-2015, 12:10 PM   #12
The Telenator
Banned
 
Join Date: Dec 2011
Location: Oud West, NL
Posts: 2,335
Default

Quote:
Originally Posted by Xenakios View Post
... many plugins won't even operate correctly at those kinds of sample rates.
Oh, yeah, did I forget to mention? Good one Xenakios.

Should we also mention that even the best conversion methods have some mathematical probability of introducing ARTIFACTS ...? Oh, no!
The Telenator is offline   Reply With Quote
Old 10-18-2015, 01:34 PM   #13
JamesPeters
Human being with feelings
 
Join Date: Aug 2011
Location: Near a big lake
Posts: 3,943
Default

Consider your hardware and what it's actually capable of, not the quoted specs. See this:

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

In particular, try the test files and see if you can hear the aliasing noise at the higher sample rates. I can, so it says my hardware isn't even adequate to use those higher sample rates. In that case, screw it, I won't. Alternately you could buy something which will use those sample rates properly without aliasing (RME comes to mind).

As for the rest: I agree with EvilDragon and Xenakios.
JamesPeters is offline   Reply With Quote
Old 10-19-2015, 01:00 AM   #14
Joaquins Void
Human being with feelings
 
Join Date: Dec 2014
Location: Stockholm
Posts: 206
Default

https://stash.reaper.fm/v/25489/aliasing_demo.mp3

The first series of tones are rendered directly at 44.1KHz while the second is exported at 96KHz and then converted to 44.1KHz offline.

I made this to illustrate that I'm not talking about something theoretical, but something that everyone can hear.

The synth used is Synapse Audios Dune, but you can make this happen with just about any synth that feature phase modulation or oscillator sync. It's not because the dev is bad or anything. It's just that predicting what comes out of certain processes is borderline impossible. It's also a trade off situation. I.e you could preemptively over sample like mad, but then the synth would be much heavier than it needs to be for most use cases.

But again, it comes down to what you are doing. If for example you only work with recorded material or samples, chances are you'll here no difference when rendering at a higher rate. If you use a lot of synths that generate sounds algorithmically it certainly makes sense to try it out anyway.

I should add that there is a sort law of diminishing returns at work with up-ing the sample rate though. I doubt OP will hear any difference between 3072KHz and 192KHz unless there is something very special going on.

Last edited by Joaquins Void; 10-19-2015 at 01:08 AM.
Joaquins Void is offline   Reply With Quote
Old 10-19-2015, 02:07 AM   #15
Time Waster
Human being with feelings
 
Time Waster's Avatar
 
Join Date: Aug 2013
Location: Bowral, Australia
Posts: 1,638
Default

I think the aliasing adds a bit of character

Seriously though, I presume the problem here is that you don't know what the results will be until you render the file? If that is the case, would having your monitor A/D converters set to 44.1 help? I guessing that if they were, you would hear the aliasing and be able to correct it with LP filtering prior to rendering?
__________________
Mal, aka The Wasters of Time
Mal's JSFX: ReaRack2 Modular Synth
Time Waster is offline   Reply With Quote
Old 10-19-2015, 03:03 AM   #16
Joaquins Void
Human being with feelings
 
Join Date: Dec 2014
Location: Stockholm
Posts: 206
Default

Quote:
Originally Posted by Time Waster View Post
I think the aliasing adds a bit of character

Seriously though, I presume the problem here is that you don't know what the results will be until you render the file? If that is the case, would having your monitor A/D converters set to 44.1 help? I guessing that if they were, you would hear the aliasing and be able to correct it with LP filtering prior to rendering?
The problem is that you can't really cut it out before hand without also cutting into the bits you do want. You'll notice that the aliasing is right down there in the normal highs. But when you up the sample rate, the aliasing moves upwards too and can be cut out by the down sampling filter. This is the point of over sampling.
Joaquins Void is offline   Reply With Quote
Old 10-19-2015, 04:49 AM   #17
Softsynth
Human being with feelings
 
Join Date: Jun 2015
Posts: 8,696
Default

Quote:
Originally Posted by EvilDragon View Post
I would argue that for home recording, anything above 48k is a waste of hard drive space and CPU, though. Unless you have some really high-end gear, as well as a well-treated mixing room in your house...
I would go further and say it is probably a waste of time even if you have the worlds best room acoustics, electronics, loudspeakers and headphones.
That is unless your clients demand it.

Epecially if you are using samples, giant 24 bit files are really getting ridiculous. I'm with Embertone on this:
"We suggest the 16-bit version. Not only will it save you a lot of RAM, but also the difference in quality is negligible AT BEST!
For those who strongly want those 24-bit samples, we've adjusted the price to recoup the extra bandwidth costs. Thank you for understanding"

A lot of us are using SSDs so supporting the 24/96 religion gets expensive on space too.

It would be interesting to see blind tests on software synths too. 48khz vs 96.

Last edited by Softsynth; 10-19-2015 at 04:55 AM.
Softsynth is offline   Reply With Quote
Old 10-19-2015, 05:17 AM   #18
Joaquins Void
Human being with feelings
 
Join Date: Dec 2014
Location: Stockholm
Posts: 206
Default

Quote:
Originally Posted by Mr. PC View Post
(Note, I am not talking about *exporting* at a high sample rate, but mixing everything; in other words, oversampling the entire mix to reduce aliasing, without having to constantly upsample and down-sample (imagine you run 10 oversampled plugins).
Just noticed this in the OPs post. (I think you get it, but to clarify) It is exactly during export that you might want to use a higher rate because that's when it counts. What you listen to during mixing is just a monitoring thing.

And another point of clarification: No one here that I can see is arguing that higher rates are inherently better or anything of the sort. With a bandlimited signal there is no difference between 96KHz and 44.1KHz. It comes out the same. Mathematically and audibly. The question is solely about avoiding processing artifacts by increasing the spectral headroom.
Joaquins Void is offline   Reply With Quote
Old 10-19-2015, 07:50 AM   #19
The Telenator
Banned
 
Join Date: Dec 2011
Location: Oud West, NL
Posts: 2,335
Default

In an effort to be understanding and reasonable, whether anyone thinks I ever am or not, I won't deny that, for example, Joaquins Void does make some valid points, particularly in his post #14.

But certain things strike me every time when this sample rate discussion rears its ugly head, namely throughout all the talk talk talk rarely, if ever, does anyone offer to help the inquirer by mentioning specific tools that could apply -- TDR's Ultrasonic Filter for instance:

https://vladgsound.wordpress.com/201...alpha-version/

Nor does anyone name names, such as which VST plugins to avoid or suggest using, those which would still work acceptably, those which should be kept well away from all this high sample rate business (I won't drop names here but a lot of 'usual suspects').

But I think the real elephant in the room in these threads is as follows: You give me practically any project, containing, let's say, a few tracks and instruments at least, and I will comb through it and find perhaps 100, perhaps 200 items that could be improved ... easily in most cases. Yet why all this worry or concern about sample rates when you already have scads of things technically wrong with your mixes already?

Why the constant fervour and (really) obsession with the sample rates so often to the exclusion of so many other features? Where is the balance here?

I've heard the untrained listener say, ooh, the song sounds a little too distorted, or too trebly, or too this or this, but I have never ever heard anyone say, Oh, you should have used a higher sample rate ... or, oh, your synth is aliasing!

Where has your perspective gone in regard to your total product?
The Telenator is offline   Reply With Quote
Old 10-19-2015, 08:12 AM   #20
Joaquins Void
Human being with feelings
 
Join Date: Dec 2014
Location: Stockholm
Posts: 206
Default

Quote:
Originally Posted by The Telenator View Post
Where has your perspective gone in regard to your total product?
It doesn't take much time to export at a higher rate. Or to downsample later on. It boils down to changing some settings, so I'd say that it's attention well spent, so to speak.

And well, it happens to be the topic of this thread which is why we are talking about it. Sometimes you just post in a thread and off it goes. That's just how it goes.
Joaquins Void is offline   Reply With Quote
Old 10-19-2015, 08:54 AM   #21
The Telenator
Banned
 
Join Date: Dec 2011
Location: Oud West, NL
Posts: 2,335
Default

Quote:
Originally Posted by Joaquins Void View Post
And well, it happens to be the topic of this thread which is why we are talking about it. Sometimes you just post in a thread and off it goes. That's just how it goes.
Uuunngghhhh!!!....
The Telenator is offline   Reply With Quote
Old 10-19-2015, 08:57 AM   #22
saddle
Human being with feelings
 
saddle's Avatar
 
Join Date: Feb 2009
Posts: 155
Default

One interesting thing I've noticed...I resample from 44.1 to 48 or 96,using sox.I can hear absolutely no change in the audio,as one would want, aned expect. However I hear a huge change in the sound of some plugins, mostly reverbs. Changes the entire mix. I then render to the session rate and downsample/convert to the desired rate.

I like mixing at 48 and 96 for reasons such as this. I don't hear a change in the individual audio tracks.
__________________
74 Takamine F-450s (Pre-litigation) 62 Martin D-28 : 2004 Fender Squier Tele
Rouge Acoustic Electric (work guitar.) Various other guitars and toys.
Reaper x64. Pro Tools 11.3.1
saddle is offline   Reply With Quote
Old 10-19-2015, 09:40 AM   #23
The Telenator
Banned
 
Join Date: Dec 2011
Location: Oud West, NL
Posts: 2,335
Default

Quote:
Originally Posted by saddle View Post
One interesting thing I've noticed...I resample from 44.1 to 48 or 96,using sox.I can hear absolutely no change in the audio,as one would want, aned expect. However I hear a huge change in the sound of some plugins, mostly reverbs. Changes the entire mix. I then render to the session rate and downsample/convert to the desired rate.

I like mixing at 48 and 96 for reasons such as this. I don't hear a change in the individual audio tracks.
You do hear that maybe the most overall about plugins for reverb and certain time-dependent effects. There is a lengthy explanation for that. But keep in mind what Joaquins Void said, essentially, know your plugins, because some just can't handle higher rates. Some are best left to convert within on their own. And every conversion no matter who, how or where adds to the possibility of introducing very small artifacts that can add up.

More than that, I do want to say something about this 48k, what has always been known as the film industry's standard. I'm mildly an even-number proponent, even though it's truly no sweat at all for a PC to do the maths, despite what some people think. The larger issue, though, is that 48k means you have up to 24k of actual frequency available. Even though absolutely none of us except the family dog can hear anything right in there, I like the extra space and some will tell you that what very little audio information that may end up in there can potentially have some sort of extremely subtle (usually positive) effect on the overall. I'll admit the possibility, even though it's beyond my scope -- i.e., haven't looked into this for years.

Last, there is a slow movement or push for us audio-only people to go to using 48k. It will "unify the industries", as they will say. Whatever. In my opinion, it would be nice to have one standard. We have enough to remember and deal with. Someone says they wanna use your thing in film and you have to run back and mess with things again, if you haven't already worked up a film copy. It's getting silly how many versions of finished work some have to have on hand. Film wants it this way, iTunes wants it that way. I'd love ONE WAV, ONE AIFF, ONE MP3 -- and each in good fidelity, as in 320kbps for MP3s and so on. Wouldn't you?
The Telenator is offline   Reply With Quote
Old 10-19-2015, 09:50 AM   #24
JamesPeters
Human being with feelings
 
Join Date: Aug 2011
Location: Near a big lake
Posts: 3,943
Default

Quote:
Originally Posted by Joaquins Void View Post
And well, it happens to be the topic of this thread which is why we are talking about it. Sometimes you just post in a thread and off it goes. That's just how it goes.
+1

Whereas I don't have a reason (currently) to work at a higher sample rate, you demonstrated a valid one "plain as day". There is no argument when results like this are displayed. It's a matter of choice for the person's application.

Would you mind trying the same test again though, with that Ultrasonic Filter plugin?

https://vladgsound.wordpress.com/201...alpha-version/

For those of us wondering if this will solve that sort of potential problem you demonstrated, the test could be valuable.

Thanks in advance!
JamesPeters is offline   Reply With Quote
Old 10-19-2015, 09:52 AM   #25
James HE
Human being with feelings
 
James HE's Avatar
 
Join Date: Mar 2007
Location: I'm in a barn
Posts: 4,467
Default

never ever render at differ rate than you are mixing in REAPER.


All the pitchshifting / timestretching just gets FUBAR.
You'll get away with it sometimes, when you aren't doing any processing like that, and you don't have any effects that care, like most reverbs...

But still, just don't do it.
James HE is offline   Reply With Quote
Old 10-19-2015, 10:13 AM   #26
zyisrad
Human being with feelings
 
Join Date: Sep 2015
Posts: 280
Default

Just like everyone pines for that analog sound while using ITB plugs now, in the future people will be nostalgic for that 44.1khz sound.
zyisrad is offline   Reply With Quote
Old 10-19-2015, 10:18 AM   #27
kenz
Human being with feelings
 
Join Date: Aug 2013
Posts: 339
Default

Quote:
Originally Posted by The Telenator View Post
You do hear that maybe the most overall about plugins for reverb and certain time-dependent effects. There is a lengthy explanation for that. But keep in mind what Joaquins Void said, essentially, know your plugins, because some just can't handle higher rates. Some are best left to convert within on their own. And every conversion no matter who, how or where adds to the possibility of introducing very small artifacts that can add up.
Indeed, some plugins even have bugs for higher sample rates or simply sound different due to different algorithms (mostly unintentional, i.e buggy) at higher samplerates. It has nothing to do with the samplerate itself.

To be fair, aliasing is indeed the samplerate's fault and not the plugin. Unless you think lack of oversampling is the plugin's fault. But I am talking about the proper implementation here of the math/code, not extra features for more precision. (if we consider aliasing as discrete lack of time precision, so to speak)

BTW, changing samplerate just for rendering won't work with some plugins. For example, many EQs store a specific % of the entire band range for their parameters. If you double the samplerate, the EQ will be completely different, because now 10% is in a completely different spot than the previous 10% of a lower band range.
kenz is offline   Reply With Quote
Old 10-19-2015, 10:23 AM   #28
Joaquins Void
Human being with feelings
 
Join Date: Dec 2014
Location: Stockholm
Posts: 206
Default

Quote:
Originally Posted by JamesPeters View Post
...
Would you mind trying the same test again though, with that Ultrasonic Filter plugin?
...
For those of us wondering if this will solve that sort of potential problem you demonstrated, the test could be valuable.

Thanks in advance!
It wont. When you render to 44.1 the aliasing is baked into the signal from the VSTi. No getting it out.

When you render to 96 you get a lot less of it baked in because most of it can be accommodated with in that bandwidth as ultra sonic content.
And the down sampling filter is steep enough to cut everything ultra sonic out, when you go to a lower rate. (I guess the super sonic filter might help a little, but either way it wont fix the problem with rendering to 44.1.)

The Ultra sonic filter is for using between plugs when running at higher rates, to sort of protect them from having to deal with frequencies they were not designed for.

Quote:
Originally Posted by James HE
never ever render at differ rate than you are mixing in REAPER.


All the pitchshifting / timestretching just gets FUBAR.
You'll get away with it sometimes, when you aren't doing any processing like that, and you don't have any effects that care, like most reverbs...

But still, just don't do it.
You can alway try. It's not like you break anything by doing a render, is it?
Joaquins Void is offline   Reply With Quote
Old 10-19-2015, 10:23 AM   #29
Judders
Human being with feelings
 
Join Date: Aug 2014
Posts: 11,044
Default

Quote:
Originally Posted by zyisrad View Post
Just like everyone pines for that analog sound while using ITB plugs now, in the future people will be nostalgic for that 44.1khz sound.
Well you already get chip tune music that is nostalgic for 8bit music chips.
Judders is offline   Reply With Quote
Old 10-19-2015, 10:48 AM   #30
serr
Human being with feelings
 
Join Date: Sep 2010
Posts: 12,562
Default

There's a really simple way to look at this without losing your mind.

The precision we can record at with 24 bit 96k format is nearly perfect.

To the point that you can even play a little fast and loose and convert formats downstream during production and have a number of other things go wrong and the audio STILL comes out with almost no perceivable damage even if you're trying really hard to hear it.

Some of you might think this is weird but to me that's a good thing!

It should be pretty obvious that a good mix on the crudest midrange audio band only portable POS device is still more enjoyable than a bad mix on I don't care what system.

There's no argument that the quality of mix work trumps full fidelity.
But that's certainly not an argument to dumb down the recording format! WTF?!


As for format conversions like sample rate conversion - only do this when necessary. Don't throw away fidelity (even the smallest amount) for no reason. That's lazy at best.

Absolutely NEVER upsample a final mix or master!!!
No one wants a lower sample rate master masquerading as HD and using twice the hard drive space to boot! (Probably get the most agreement here even among the 'sour grapes' crowd regarding HD audio.)

There ARE plenty of legit reasons to decide to work with lower sample rates.

Working with someone overseas with a poor internet connection?
48k or 44.1k is just fine. Get back to work!

Lightweight Windows computer that can't handle HD audio?
48k or 44.1k is just fine. Get back to work!


Want to hear what kind of loss you're dealing with?

A/B/X tests with such a small difference in fidelity that it's within variations in perception aren't going to do anything for you.

We already know that 16 bit 44.1k (CD) makes a damn fine recording.

But is it perfect?

Do 100 generations between 44.1k & 96k:

Nope not perfect. Damn good. Have to get obvious like this test to hear loss but there it is.

100 generations of 48k - 44.1k:

There's the ringer! (ie worst case)

100 generations between 96k & 192k:

Well look, the two files still null perfectly!

So... 96k can contain audio pretty much perfectly. Even over 100 generations.

(Digital sample rate conversion generations that is in the above examples.)


We hit the line in the sand of a really perfect digital container for audio about 15 years ago.

None of this will help you make a good mix. And that certainly SHOULD be pointed out! If you have a problem with a mix - it has zero to do with the recording format (unless we're talking about mp3 destruction).

But the thing that drives engineers crazy - making a copy of something where no matter what, it sounds exactly exactly the same in every detail - is possible now. Someone doesn't want this? This task comes up a lot...


I mean, use an old 8-track tape deck or a wire recorder if it inspires you artistically. All opinions on what drives artistry are fair game. But the sour grapes arguments against HD audio are just silly.


You folks doing the A/B/X tests...

You converting all the test tracks back to 96k (or 192k) before the listening test so you can use the same DA converters and not introduce other variables?

Or are you switching between programs at different sample rates and actually critiquing the sound of either Reaper's or Core Audio's on the fly sample rate conversion?

Or have you actually set up multiple DA converter units, have their outputs calibrated within .1db of each other, and have an analog switching system set up to compare?

Oh, and are the test tracks actual dynamic full fidelity recordings or squished pop music?


Oh yeah, and we can deliver surround sound mixes in perfect lossless condition to the consumer now because of these formats. How about less fussing over portable quality audio containers and more 5.1 mixes!


I haven't extensively tested some of the different heavy duty plugins at 96k vs. 192k. I've worked on one 192k analog restoration project (audio transferred from mag film stock) and it seemed like the noise reduction plugin (iZotope RX) and the multiband compressor (UA Precision Multiband) liked 192k. (Just like the owners manuals said they would.)

I didn't have the opportunity to compare to a 96k transfer of the analog program though. I can only speculate that the fact that said plugins 'seemed' to be more flexible than I remember them being at 96k was real and not imagined.

But the above 100 generation test is the reason I keep audio at 96k most of the time and sample rate convert every single bluray music disc I purchase from 192k down to 96k for my library. Anything I produce and release will be 96k or 88.2k at most.

16 bit is the thing to avoid at any sample rate. Just provide that format for the CD version but keep everything 24 bit even if you 'only' use 44.1k.


The only thing sillier to me than all that is the DSD crowd.
Right, so 24/96 or 24/192 PCM audio isn't good enough...
We all need to re-purchase DSD format converters now...
One of the most expensive pieces of gear in our entire systems...
Because HD sample rate PCM just doesn't cut it...
And then the all the consumers will happily go and buy SACD players or standalone DSD converters...

Don't know what kind of rooms and speaker systems these guys have...
Must be something pretty special!

Last edited by serr; 10-19-2015 at 11:47 AM.
serr is offline   Reply With Quote
Old 10-19-2015, 12:06 PM   #31
Mr. PC
Human being with feelings
 
Mr. PC's Avatar
 
Join Date: Apr 2010
Location: Cloud 37
Posts: 1,071
Default

Quote:
Originally Posted by The Telenator View Post
I use plugins, many of which handle oversampling within. I only use those VST that are built by truly intelligent devs who know what the hell they are doing in this area, such as Fabien of good ol' Tokyo Dawn. Otherwise, I never give it much of a thought.
So you believe oversampling makes a difference; how is that different from using a higher sample rate?

Quote:
Originally Posted by Xenakios View Post
Complete waste of time and effort. You even neatly avoided the question of "do you actually hear any problem?".

In any case, many plugins won't even operate correctly at those kinds of sample rates. It's not uncommon for DSP code to be hardcoded to work only at some well accepted sample rates, like 44.1khz,48khz, 88.2khz, 96khz. Some plugins may not care about the sample rates, but some will and then you are in a new mess.
I hear a big difference using oversampling; that's what's triggered my whole quest for best samplerate/oversampling option. I haven't yet tested my latest theory. I'll do a mixdown comparing

1 - No oversampling, Mixed at 48k
2 - Oversampling, Mixed at 48k
3 - No oversampling, Mixed at 3072k
4 - Oversampling, Mixed at 3072k

Quote:
Originally Posted by The Telenator View Post
Oh, yeah, did I forget to mention? Good one Xenakios.

Should we also mention that even the best conversion methods have some mathematical probability of introducing ARTIFACTS ...? Oh, no!
That's why, rather than allowing multiple plugins to do sample rate conversion every time we want to compress etc. it makes sense to upsample once, and downsample once using a high quality resampler.

Quote:
Originally Posted by JamesPeters View Post
Consider your hardware and what it's actually capable of, not the quoted specs. See this:

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

In particular, try the test files and see if you can hear the aliasing noise at the higher sample rates. I can, so it says my hardware isn't even adequate to use those higher sample rates. In that case, screw it, I won't. Alternately you could buy something which will use those sample rates properly without aliasing (RME comes to mind).

As for the rest: I agree with EvilDragon and Xenakios.
I'm not talking about making a final render at a nigher sample rate. I specified this in my original post, twice. Higher exported sample rates actually case *more* noise. There's no controversy that above 48k has no real benefit for final export. I'm talking about *mixing* in at a higher rate. Just like we mix at 32bit and export at 16. 24/16 is fine IMO for final export. Again, this is only about using a higher sample rate in lieu of oversampling.

Quote:
Originally Posted by Joaquins Void View Post
I should add that there is a sort law of diminishing returns at work with up-ing the sample rate though. I doubt OP will hear any difference between 3072KHz and 192KHz unless there is something very special going on.
Yes, I agree with this, and only used 3072KHz in response to the earlier comment about Reaper's 64x oversampling. (So according to this, 64x oversampling is also useless. We could just go for 4x.

Quote:
Originally Posted by Softsynth View Post
Epecially if you are using samples, giant 24 bit files are really getting ridiculous.
Again, I'm feeling that every person in this thread, except Joaquins Void, has misunderstood me. I am *not* talking about using HQ samples.

Quote:
Originally Posted by The Telenator View Post
Nor does anyone name names, such as which VST plugins to avoid or suggest using, those which would still work acceptably, those which should be kept well away from all this high sample rate business (I won't drop names here but a lot of 'usual suspects').
Actually, I'd love to know more about which plugs to avoid, though really, I'm sticking these days to Reaper, Voxengo, and try to be as minimalist as possible. Still, I find I end up using a lot of EQ, and quite a few 'light' instances of compression, distortion, etc.

I would like to know if these plugins specifically are able to run at a higher [project] sample rate.

Quote:
Originally Posted by kenz View Post
BTW, changing samplerate just for rendering won't work with some plugins. For example, many EQs store a specific % of the entire band range for their parameters. If you double the samplerate, the EQ will be completely different, because now 10% is in a completely different spot than the previous 10% of a lower band range.
You mean the EQ frequencies will be altered? But so many people will swap between 44.1 and 48 etc. so that will seriously wreck a lot of mixes. Is there a way to know which EQs do that? (For me it only really matters if Voxengo does that).

Concerning the TDR Ultrasonic. Is it still useful if working at a very high sample rate? And if plugs are doing their own conversions regardless, then the aliasing will be 'baked in' to the plugin's output, so a filter wouldn't do much. Either way I'll try it of course. Could a conventional LPF do the same?
__________________
AlbertMcKay.com
SoundCloud BandCamp
ReaNote Hotkeys to make Reaper notation easy/fast
Mr. PC is offline   Reply With Quote
Old 10-19-2015, 12:41 PM   #32
kenz
Human being with feelings
 
Join Date: Aug 2013
Posts: 339
Default

Quote:
Originally Posted by Mr. PC View Post
You mean the EQ frequencies will be altered? But so many people will swap between 44.1 and 48 etc. so that will seriously wreck a lot of mixes. Is there a way to know which EQs do that? (For me it only really matters if Voxengo does that).
Yes, although the difference between 48k and 44100 isn't that big since most EQs use log-scale, thus a small change in band range won't change the sound much.

I don't know about other Voxengo EQs but I know HarmoniEQ does that (or did? it was long ago on possibly old version when I remember it)
kenz is offline   Reply With Quote
Old 10-19-2015, 01:09 PM   #33
Softsynth
Human being with feelings
 
Join Date: Jun 2015
Posts: 8,696
Default

MR PC,
I did understand your original post so sorry if my response appears to not address what you are saying. The 24 bit sample comment is part and parcel of the debate for a lot of people.
Usually the same people that get stuck on high bitrates are the same people that will obsess about 16 bit versus 24bit recordings/samples too.

If you're going to be freezing tracks with your 96khz rate then your projects will get much bigger than they would be at 48khz.

It's pretty easy to tell if you are listening to a good mix, even with lowly lossy, compressed 128k MP3 on Soundcloud. Even though this is significantly inferior to the potential of CD quality.
Worrying about the minutiae of difference that you might imagine between those sample rates is most likely a waste of time.


Recording at 96khz isn't going to make the difference between a great and a mediocre mix for anyone.

Joaquins Void's exporting projects at 96khz when the audio was recorded at a lower rate is at best a confidence trick, since that can gain you nothing other than additional artefacts that (luckily) are unlikely to be audible.
Since this does not even take advantage of the lower noise floor of high resolution anyone doing this is merely exposing problems above the audible threshold that could at worst cause minor artefacts in the audible range.
Exporting at 96khz hasn't done anything to the aliasing in that scenario!
Softsynth is offline   Reply With Quote
Old 10-19-2015, 01:22 PM   #34
Dstruct
Human being with feelings
 
Join Date: Jul 2006
Posts: 12,480
Default

Quote:
Originally Posted by kenz View Post
I don't know about other Voxengo EQs but I know HarmoniEQ does that (or did? it was long ago on possibly old version when I remember it)
If it does it is a bug.
Dstruct is offline   Reply With Quote
Old 10-19-2015, 01:24 PM   #35
The Telenator
Banned
 
Join Date: Dec 2011
Location: Oud West, NL
Posts: 2,335
Default

Mr.PC: "So you believe oversampling makes a difference; how is that different from using a higher sample rate?"

Well, since you asked, higher sample rates can make a difference, and oversampling can make a difference, of course. But it depends on what you or the plugin designed to make use of them are doing with them. I want to say more but risk confusing some reader again, which would be unfortunate considering how much 'straightening out' just got posted above here while I was out getting a sandwich.

A thread about the entire subject can be helpful, and I'll never state that knowing about this ball of wax is pointless, whether anyone makes much use of it or not. Unfortunately, what happens or did happen in the other 197,000 threads on this subject deteriorated into flame wars and food fights, and many of the sample-rate unschooled came away learning nothing much.

I've already said enough for my part, but let me just repeat something to show the subtlety or even dichotomy in play here: People who think a higher sample rate, such as 96k makes music sound better than 44.1 FAIL to guess correctly which is the higher-rate file when both versions are played back 50% of the time in blindfold tests. This testing has been done over and over and forever with the same results.

Meanwhile, as serr pointed out, "The precision we can record at with 24-bit 96kHz format is nearly perfect." Believe it or not, some people think all the world's greatest music, a copy or two of each work, should be rendered into this format and salted away deep in the earth, like the current seed repositories and people who cryogenic-ally freeze themselves or their loved ones for some future time and use. If we destroy ourselves and our planet, aliens finding this ball of dirt millenia from the destruction will have practically perfect renderings of our music to listen to and wonder about us.

Both examples above are true. Why? Connect the dots!
The Telenator is offline   Reply With Quote
Old 10-19-2015, 01:49 PM   #36
JamesPeters
Human being with feelings
 
Join Date: Aug 2011
Location: Near a big lake
Posts: 3,943
Default

Quote:
Originally Posted by Joaquins Void View Post
It wont. When you render to 44.1 the aliasing is baked into the signal from the VSTi. No getting it out.
I guess what you mean by "render" in this case is the output of the plugin (since the project would be operating at a sample rate lower than 96KHz). I'm used to hearing the word "render" in the context of creating a file, so that probably threw me off a bit.

Quote:
Originally Posted by Mr. PC View Post
I'm not talking about making a final render at a nigher sample rate. I specified this in my original post, twice. Higher exported sample rates actually case *more* noise. There's no controversy that above 48k has no real benefit for final export. I'm talking about *mixing* in at a higher rate. Just like we mix at 32bit and export at 16. 24/16 is fine IMO for final export. Again, this is only about using a higher sample rate in lieu of oversampling.
I know you're not talking about that. The article I linked had that in mind. However my point was: if your audio hardware can't even produce 96KHz without aliasing, mixing in 96KHz is going to have the disadvantage of increased aliasing artifacts blended in with the sound you hear. Try listening to those files I mentioned. Most audio devices aren't going to play them without hearing aliasing noise. Granted the files have been normalized to show the effect more clearly, but if your hardware produces these artifacts then you know you have to weigh pros/cons of the aliasing artifacts for the entire mix at 96KHz+ (what you'll hear blended in with the audio due to aliasing, since your converters etc. can't reproduce it properly) vs if you'll hear artifacts from specific plugins which cause worse aliasing when used in projects at lower sample rates (such as that synth example)...or buy hardware which can produce that frequency range without aliasing.

Last edited by JamesPeters; 10-19-2015 at 01:56 PM.
JamesPeters is offline   Reply With Quote
Old 10-19-2015, 04:10 PM   #37
Softsynth
Human being with feelings
 
Join Date: Jun 2015
Posts: 8,696
Default

Going the other way I remember the zero oversampling high end audiophile craze:

http://www.enjoythemusic.com/magazin...mpling_dac.htm
Softsynth is offline   Reply With Quote
Old 10-19-2015, 04:13 PM   #38
karbomusic
Human being with feelings
 
karbomusic's Avatar
 
Join Date: May 2009
Posts: 29,260
Default

Does anyone have any A/B/X examples where nothing has been normalized? I'd like to hear this difference in a real-world situation if someone has a non-enhanced example. The graphs at http://src.infinitewave.ca/ no longer really interest me since they have no ability to distinguish what is audible. I actually quit using it for that very reason.
__________________
Music is what feelings sound like.
karbomusic is offline   Reply With Quote
Old 10-20-2015, 12:02 PM   #39
Mr. PC
Human being with feelings
 
Mr. PC's Avatar
 
Join Date: Apr 2010
Location: Cloud 37
Posts: 1,071
Default

Quote:
Originally Posted by The Telenator View Post
Mr.PC:
I've already said enough for my part, but let me just repeat something to show the subtlety or even dichotomy in play here: People who think a higher sample rate, such as 96k makes music sound better than 44.1 FAIL to guess correctly which is the higher-rate file when both versions are played back 50% of the time in blindfold tests. This testing has been done over and over and forever with the same results.
The test shouldn't be between two sample rates, but two songs of the same sample rate, say 48k, however one has been *MIXED* at a higher sample rate, say 192k. Both mixes should use the same FX, eg. 10 compressors, 10 EQs, a limiter etc. (and these FX should work at the higher sample rate) and not over-sample. I plan to export such a test shortly.


BTW, my 8 year old student saw this thread, and asked me what it was about! It was a long and difficult explanation.

Here's a rough experiment.

1 - http://vocaroo.com/i/s1DyAG0ABYae
2 - http://vocaroo.com/i/s13BEeK3NIp6
3 - http://vocaroo.com/i/s1ycrbdY9ufD
4 - http://vocaroo.com/i/s1aM9FqncDhB

One is mixed at 192k, the other at 48k. Both exported as 16bit 48k flacs without oversampling. Then I have 2 more mixed at both rather *with* oversampling. (Reaper doesn't seem to allow higher than 192k).


I can't trust my own judgement here as I'm biased knowing which is which, but I don't think I'd really pass the double blind test.

However, this version has only a few tracks, and not too many effects. Even the over-sampled versions don't really sound too different (I aim to mix as 'lightly' as possible. I'll try this experiment with something using heavy compression and eq tomorrow (or maybe someone else would like to try).

I'm doing a test in my Daw right now using a sine wave and ReaSynth. The differences are obvious in the Daw, but the Rendered files are exactly the same.
__________________
AlbertMcKay.com
SoundCloud BandCamp
ReaNote Hotkeys to make Reaper notation easy/fast

Last edited by Mr. PC; 10-21-2015 at 05:13 AM.
Mr. PC is offline   Reply With Quote
Old 10-20-2015, 12:25 PM   #40
serr
Human being with feelings
 
Join Date: Sep 2010
Posts: 12,562
Default

Quote:
Originally Posted by Mr. PC View Post
Both exported as 16bit 48k flacs without oversampling.
Why would you add the 16 bit reduction? That's usually far more perceivable than anything to do with sample rate conversions.

I would expect it to skew the results.

Maybe put the 24 bit files out?
serr is offline   Reply With Quote
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT -7. The time now is 01:12 AM.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.