Samplerates. Is bigger better?

[joeritson]

I've been recently using 96khz and thinking it's a massive increase in quality from 44.1khz, is this true? I've noticed a small difference in clarity, (could just be placebo) but I just read that Andy uses 24bit/44.1khz. Now, I'm thinkin', if 44.1 is good enough for him, it's good enough for me.
So what's the difference?

 

[Ermz]

No doubt there's a quality increase if you're using a good playback system. Most of the time we're quantizing, editing, resampling, printing, crushing and distorting so much that it makes no difference. So most stick with 44.1kHz. Plus, CDs are still 44 so there isn't much point in tracking and mixing with a samplerate that will take twice the hard disk space.

 

[MetalJonesy]

if i were doing audio at higher sample rates, rather than jumping to 96k (post production) id step up to the 88.2kHz mark. just to do with sampling issues etc from hanging between formats etc

 

[joeritson]

No doubt there's a quality increase if you're using a good playback system. Most of the time we're quantizing, editing, resampling, printing, crushing and distorting so much that it makes no difference. So most stick with 44.1kHz. Plus, CDs are still 44 so there isn't much point in tracking and mixing with a samplerate that will take twice the hard disk space.

I completely understand what you mean here. But if you've got twice the flour, the cake will be thicker, won't it?
What I mean is, you've got more data to destroy, so there'll be more left over when you're done destroying it (come dithering time).
I don't know for sure. Just asking a bit of a different question; yeah, I read the samplerates thread.

 

[colynomial]

I have a friend who had some of his work mastered at Sterling Sound in NY a few years ago. He told me that they actually stepped everything down to 44.1 before the mastering process. More reason to believe it may not be worth the extra HD space and added CPU/RAM strain when processing.

Though, I suppose if you were to think of it like the pixels in a digital picture, there is technically more "information" in a 2 megapixel picture than there is in a 1 megapixel one- as there is with 96K vs 44.1K waveforms. So maybe when you're applying effects to the original waveform the processing might have a smoother quality even if you do step down to 44.1 afterwards for mastering. For instance with the picture, if you want to say "sharpen edges" or something in Photoshop, it's able to define the edges with a higher degree of precision because it can get down to a finer level when comparing the contrast of adjacent pixels. So even if you compress the picture after you've applied effects the effects were rendered when there was more information available so they should be more "accurate". Then again I don't use Photoshop and know nothing about digital imaging- I'm just speculating.

 

[joeritson]

You're venturing into the realm of anti-aliasing, man. That's exactly what I'm getting at.
I think I'll stick to 96/24 until I prove myself wrong (Or until someone is kind enough to do it for me). So long as I get the dithering right, I cannot for the life of me see how this is not beneficial. I've got fucking shitloads of HDD so that's just not a concern.

 

[Metaltastic]

Dither doesn't affect converting from sample rates though, only for converting from 24-bit to 16-bit. There's no way to obscure the artifacts in downsampling, but the best way to do it is to record at 88.2 (rather than 96k) cuz it's an easy 2:1 stepdown to 44.1! (and hence why 96k I assume was designed for DVD audio producers, as it's a direct 2:1 stepdown to 48k). And all a higher sample rate allows you to do is record higher frequencies, and since 44.1 can record up to 22.5khz (and humans can ideally hear only up to 20khz, and I know mine's a fair bit lower), I don't really see the point! And Andy works at 44.1

 

[joeritson]

Dither doesn't affect converting from sample rates though, only for converting from 24-bit to 16-bit. There's no way to obscure the artifacts in downsampling, but the best way to do it is to record at 88.2 (rather than 96k) cuz it's an easy 2:1 stepdown to 44.1! (and hence why 96k I assume was designed for DVD audio producers, as it's a direct 2:1 stepdown to 48k).

This is extremely true. I stand corrected.

And all a higher sample rate allows you to do is record higher frequencies, and since 44.1 can record up to 22.5khz (and humans can ideally hear only up to 20khz, and I know mine's a fair bit lower), I don't really see the point! And Andy works at 44.1

Are you absolutely sure that 88.2khz records to double the frequency, or adds more data within the frequencies it records?

 

[Metaltastic]

I'm absolutely sure that up to 22.5 khz, 44.1 and 88.2 record the exact same amount of data - the difference is 88.2 continues recording up to 44.1 khz (the approximate maximum frequency that can be recorded at a given sample rate is half of said sample rate)\

EDIT: Whoops, make that 22.05, not 22.5

 

[Metaltastic]

SOURCE: http://en.wikipedia.org/wiki/Nyquist_Theorem

 

[MetalJonesy]

I'm absolutely sure that up to 22.5 khz, 44.1 and 88.2 record the exact same amount of data - the difference is 88.2 continues recording up to 44.1 khz (the approximate maximum frequency that can be recorded at a given sample rate is half of said sample rate)

+1 88.2 is obviously sampling twice as fast as 44.1, so when converting down to 44.1 it just takes away every other sample, whereas 96-44.1 is uneven so the data being dropped can be all over the shop

 

[joeritson]

In turn rendering any sample rate above 44.1khz completely pointless for audio recording, until we grow bat ears. Well explained

 

[Ermz]

Are you absolutely sure that 88.2khz records to double the frequency, or adds more data within the frequencies it records?

If I understand you correctly it does both.

The issue with 44.1khz is NOT that we can hear above 22khz and lose 'air', but that the highest frequency waveforms are rendered very coarsely. Think about it, how many samples are essentially representing that 22khz wave? 2? Hah.

This is why the highs on CDs can be considered 'harsh' at times (apart from the smashing that goes on these days). There is fairly severe distortion happening in the super-high frequencies.

We should be striving to mix at the highest possible sample rate we can, but until the final medium is actually printed in that sample rate, it's self-defeating.

 

[Metaltastic]

Thanks, and that's how I see it anyway! I've heard talk about plugins working more efficiently on files with higher sample rates and other little concerns here and there, but since 88.2 is the only sample rate that wouldn't leave enough artifacts when converting to 44.1 to defeat the purpose, it's not worth taking up literally double the HDD space IMO! (unless you have the space to burn, I guess! But even then, I feel like it'd be more of a strain on the HDD and session)

 

[joeritson]

SOURCE: http://en.wikipedia.org/wiki/Nyquist_Theorem

ZOMG WIKIPEDIA ISNT RELIABLE MAN! ANYONE CAN EDIT THAT!!!


*runs*

 

[MetalJonesy]

half the systems we use anyway (regards mics, speakers etc) dont go much past 20k either, fair enough the extra range may add harmonics to the material under 20k, but i really dont see the need to go above 44.1

 

[Metaltastic]

ZOMG WIKIPEDIA ISNT RELIABLE MAN! ANYONE CAN EDIT THAT!!!


*runs*

Hahaha, well it's just a summary of what I was taught in class

And Ermz, I don't quite follow - the highest frequency waveforms aren't rendered coarsely because all AD converters operating at 44.1k have an LPF set to 22k or so to get rid of the aliasing, so it shouldn't matter (and honestly, how can anyone say that low-passing a metal/rock/pop mix even at 18k would make any audible difference?)

 

[joeritson]

Hahaha, well it's just a summary of what I was taught in class

And Ermz, I don't quite follow - the highest frequency waveforms aren't rendered coarsely because all AD converters operating at 44.1k have an LPF set to 22k or so to get rid of the aliasing, so it shouldn't matter (and honestly, how can anyone say that low-passing a metal/rock/pop mix even at 18k would make any audible difference?)

Haha. No, I trust Wikipedia with almost everything. In fact I can't think of anything I don't trust it with.
What class did you learn this in? The maths in that article are one hell of a headfuck from my POV. (Maths B in Year 11, Maths A in Year 12, aka I can only just begin to understand those equations)
I'd love to be able to look at this shit and be able to figure it out. I mean, the physics of compression, phase and equalisation are one thing, but this?

 

[Metaltastic]

Bleh, keep that crap away from me man, I'd imagine you can make more sense of it than I can! I just learned about the basics of the Nyquist theorem and how it all works in my Audio Arts Production I class, we didn't delve into any of the math behind it!

 

[Ermz]

Hahaha, well it's just a summary of what I was taught in class

And Ermz, I don't quite follow - the highest frequency waveforms aren't rendered coarsely because all AD converters operating at 44.1k have an LPF set to 22k or so to get rid of the aliasing, so it shouldn't matter (and honestly, how can anyone say that low-passing a metal/rock/pop mix even at 18k would make any audible difference?)

Try to visualize an analogue waveform getting sampled. The lower frequencies, which stretch further across the time domain to complete one cycle are rendered more accurately than higher ones, that take less time. As you start to approach the super high frequencies, very few samples are rendering those waveforms. It doesn't have anything to do with aliasing or LPFing.