Composing in 96k, Summing at 192k, Bouncing at 48k?

[Prockamanisc]

Their output is 48k though. They record at 96k (just like us) but deliver in 48k.

Yeah, that's fine. I deliver in 48k. And that video, which was recorded in 96k, sounds obstensibly better than the one in 48k, even though it's delivered at 48k. Even downsampling will keep some of the extra that's in there. But what's in there that we're hearing?

 

[yhomas]

I...just don't know where this thread went. This is the argument that most people seem to be making in this thread:

Premise 1) The human hearing range stops at ~20kHz.
Premise 2) Sampling above 44.1kHz is pointless because we can't hear anything beyond that.
Premise 3) The only two premises that matter are premises 1 and 2.
Premise 4) Anything that would show me otherwise is invalid because premise 3 says that it doesn't.
Premise 5) This video doesn't prove anything, even if it does (it does). ().

The question is: what do we do with this information? What's happening within our audible range that's somehow affected by recording in 96kHz and above?

Math errors are the most reasonable explanation. People tend to assume that everything is “perfect”, but it is quite easy to degrade audio with improper processing—but this was much more the case in the days of 32-bit processing.

 

[Prockamanisc]

That video doesn't sound random, though (as I assume the math errors would be). It sounds much richer and fatter, the way I know real synths to be. To me, that video is showing that recording at a higher sample rate gets the sound closer to reality.

 

[jcrosby]

The second half of that video is so much richer than the first. Do you hear it? I wouldn't even call it subjective. Objectivity is whether it's there or not. Subjectivity is whether we like it or not.

Objectively, it's there. Subjectively, I love it.

You realize you're listening to a downsampled version that has also been converted to a lossy format no?
This reminds of optical illusions that trick your brain into thinking one object is larger than the other when in fact they are identical in size...

It also brings to mind the McGurk effect which illustrates that our auditory perception can be completely altered by what we see. Regardless of whether you "think" it is or isn't true, visual bias is evolutionarily hardwired into us... Vision pretty much always wins in fMRI research...




Downsampling also creates artifacts in the audible range, and not all daws are equal. Ironically Ableton Live, (which I have a love-hate relationship with), has SRC on par with Izotope's SRC, whereas Sadie 6, (a mastering DAW with a hefty price tag shows severely audible garbage when downsampling. Other daws, daws that people swear up and down as having a supremely better sound engine produce visible and/or audible artifacts downsampling 96k to 44.1.

Link below, see how your DAW stacks up in terms of downsampling: https://src.infinitewave.ca/

 

[Alex Fraser]

It also brings to mind the McGurk effect which illustrates that our auditory perception can be completely altered by what we see. Regardless of whether you "think" it is or isn't true, visual bias is evolutionarily hardwired into us... Vision pretty much always wins in fMRI research...

Really interesting. More than once I've started working with a plugin which is placed on a different track to the one I believe it to be on. And it takes me about 30 secs to realise the mistake, as at the time, I'm "hearing" the effects.
Glad I'm not going mad anyway..

 

[jcrosby]

Really interesting. More than once I've started working with a plugin which is placed on a different track to the one I believe it to be on. And it takes me about 30 secs to realise the mistake, as at the time, I'm "hearing" the effects.
Glad I'm not going mad anyway..

We've all done this

 

[Erick - BVA]

We've all done this

And I thought I was the only one!

 

[Prockamanisc]

Really interesting. More than once I've started working with a plugin which is placed on a different track to the one I believe it to be on. And it takes me about 30 secs to realise the mistake, as at the time, I'm "hearing" the effects.
Glad I'm not going mad anyway..

That used to happen to me, until I realized that was the thing, so now I always let me ears be the guide. So now the reverse happens: I'll put a plugin on the wrong track and tweak it. When I don't hear anything changing, I'll start to say to myself "I spent all that money on this plugin and it doesn't do anything?". Then I realize my mistake and adjust.

For instance, when I was just watching that McGurk video, I kept saying to myself "that's weird, that doesn't really sound like an F. It just sounds like what I was just hearing...I wonder if they used a bad mic that didn't pick up the air that escapes between the lip and the teeth just before the plosive." And then they explained it and then it made sense what they were doing.

 

[robgb]

It also brings to mind the McGurk effect which illustrates that our auditory perception can be completely altered by what we see

I wonder if this comes into play when we're watching sample library demos. What if I were to do a video showing a Spitfire instrument in action in Kontakt, but replace the sound with what many consider an inferior product (pick your poison). Would people be influenced by that Spitfire name and think it's the greatest thing since sliced bread? Or would they be able to recognize the sound they had had called "synthetic" the month before when it had a different developer's label on it?

 

[yhomas]

That video doesn't sound random, though (as I assume the math errors would be). It sounds much richer and fatter, the way I know real synths to be. To me, that video is showing that recording at a higher sample rate gets the sound closer to reality.

There are all kinds of math errors that don’t sound random. But IMO, a simple single pass of good a quality (properly implemented) 48k vs 96k recording (of anything), played back at 48kHz isn’t going to show differences that anyone can hear in blind testing. So any audible differences represent some gross error in the test methodology or in the hardware/software implementation. But IMO, errors are much more common than people think.

 

[Josh Richman]

Just to get some clarity. So are you guys typically working at 44.1k or 48k?

 

[Nick Batzdorf]

Tons of VI's

Whoa whoa whoa - let's stop right there.

Can you hear any difference working at 96K - sacrificing half your computer system's performance - rather than 48 or 44.1? These are VIs, not acoustic recordings.

My very, very strong hunch is no.

And I'm not an audio skeptic. (For example, I've heard the difference with some tweak audio cables, and I do believe they can be a legitimate improvement when you're carrying analog signals.)

 

[Nick Batzdorf]

Not necesarily. Some people set their project rate higher for upsampling plugins (although most processes vulnerable to aliasing should upsample internally anyway). Some also claim that reverbs rendered at higher sample rates sound better. I havent AB that though.
I dont think that high sample rates are necessary. And if you go up dont go down again until the final export.

^ Divico knows wassup. That right there is the answer.

 

[Nick Batzdorf]

*sigh*

This thread is just going to be filled with pedantic people agreeing with each other, but with a tone of disagreement so that they can show how much they gleaned from the one YouTube video they saw on the subject, right? In that case, forget I was here, I'm out.

It's also worth pointing out that Divico is absolutely right!

But look, this is a question for your ears rather than opinions. Prockman, you need to listen yourself.

 

[jcrosby]

I wonder if this comes into play when we're watching sample library demos. What if I were to do a video showing a Spitfire instrument in action in Kontakt, but replace the sound with what many consider an inferior product (pick your poison). Would people be influenced by that Spitfire name and think it's the greatest thing since sliced bread? Or would they be able to recognize the sound they had had called "synthetic" the month before when it had a different developer's label on it?

Although I don't know for sure it certainly wouldn't surprise me if this were the case...

To me the McGurk Effect suggests a few things...
A. Our brains evolved to prioritize visual cues as a means of survival, and probably as an adaptation for social bonding.
B. It also seems to suggest that expectations alone might have the capability to modify our perception of what we hear...
I.E. if something as simple as seeing the lip curled under to form an f hacks your brain into 'hearing' an f, I'd imagine it's just as easy to be led by any number of other factors... The room, and equipment in the room, User Interface etc.
I'd imagine even what someone is saying in a walkthrough can influence your perception of sound.

Speaking of which there's a great EDM mastering series where they basically discuss and demonstrate all of these things... Not only is it a solid, no frills mastering series, it's great fun as well... (And useful across all genres despite being EDM centric.)... The series is actually where I picked up on this video... If anyone's interested post back and I'd be happy to paste a link...

 

[Divico]

And dont forget to do proper blind testing
Had an argue with my dad once that high res mp3s are worse than wavs. Like audibly worse.
Blind tested me. In the end he played me the same file 3 times and #i was like. Yeah this ones definetly the wav and here we have the mp3. Our ears can be fooled so easily.
And yes I can hear the difference on a good system

 

[JohnG]

Just to get some clarity. So are you guys typically working at 44.1k or 48k?

I've met some Real Famous Composers who work at 44.1 for "regular" composing (not live orchestra -- when they are doing stuff based on samples) and upsample at the end to whatever delivery format is needed.

Most guys seem to record live orchestra for films at 96k and work in that until final delivery (if final delivery is 48). I don't know anyone who does full orchestra for TV but I would be surprised if they bother with 96 -- maybe they do so I'm not condemning it.

I work at 48k all the time, ignoring the matter that nearly all my samples were recorded at 44.1. Some people have told me that's crazy but plenty of guys do it.

So how's that for a fuzzy answer?

 

[Divico]

Unless youre not rendering and doing destructive edits there is no wrong. Keep in mind to render at 48 for film and your good to go. I doubt that the difference between 44 and 48 is crucial for composing and probably also for mixing.
Some claim that reverbs sound better in higher sample rate so if you can hear the difference go with a higher one for mixing and rendering and downsample afterwards.

 

[robgb]

I'd imagine even what someone is saying in a walkthrough can influence your perception of sound.

You may be onto something there. I'm reminded of 8Dio and Spitfire walkthroughs where they often talk about the "beauty" of the sound or describe its "silkiness" or whatever. These things may or may not be true, but we tend to want to believe them.

 

[Divico]

That video has its audio encoded with the opus codec, which only encodes frequencies up to 20kHz.

I highly doubt that frequencies above 20k make a difference for us adults. Maybe for young mozarts they do
As I´ve written earlier the point in those high sample rates is not to capture ultrasound but a technical one.
Mainly this has to do with the anti alias filtering