44.1 kHz or 48 kHz?

[roman.i]

Try to listen to sounds with deep reverbs.

[BertKoor]

i imagine a quick blind test would let you know you can't hear a difference between 44.1 and 48k.

There are circumstances where....

For example the sound card only works at a fixed rate of 48kHz and support of other frequencies is done with shitty SRC by the driver.

You never really know what's inside the box.

[vurt]

i imagine a quick blind test would let you know you can't hear a difference between 44.1 and 48k.

There are circumstances where....

For example the sound card only works at a fixed rate of 48kHz and support of other frequencies is done with shitty SRC by the driver.

You never really know what's inside the box.

yup, but that would be an obvious reason, if there was ie ir reverbs being used with the irs being 48k might cause artefacts at other rates, is another obvious reason.

but all things being equal, i dont think a person could hear a difference between the two. on a system outputting both, with plug ins that work equally well at both settings, listening to the output of this, is what i mean. as in, there is no noticeable difference to our ears, only to how the programs we use interpret said data, which is why we use one or the other, is it for this or is it for that? we dont use either, because they sound betterer!!!

[mystran]

i imagine a quick blind test would let you know you can't hear a difference between 44.1 and 48k.

There are circumstances where....

There is a lot of individual variation in hearing. The general trend is that we all lose the high frequencies as we age, but where we start and how fast the loss happens varies a lot.

I have always been quite deep into the bell curve when it comes to high frequencies and even back in my mid-20s when I started doing audio development I could still hear (barely) up to 22-23kHz. I don't hear nowhere near that high anymore, but from personal experience I know it's a real thing.

Do I believe it is possible for someone to hear the difference in blind ABX on good headphones between 44.1kHz and 48kHz if there is frequency content up to Nyquist? Definitely. With speakers I doubt it, because tiny movements of the head can cause too much variation in high frequency response, but .. like.. yeah, it's plausible on headphones. Might depend a bit on the SRC filter too though, as a very steep filter might cause ringing that's more audible than the actual difference in frequency content and a very sloppy filter.. well see next point. A filter that's good enough to be transparent up to 20kHz, but not too steep is probably going to make such an ABX the most difficult.

Do I believe there is any practical difference whatsoever between 44.1kHz and 48kHz even if you can identify the difference in blind ABX? Well... if you have plugins with "sloppy" oversampling filters (eg. to minimize latency) then 44.1kHz might sound a bit duller, because such sloppy filters can cause a tiny drop in response down to 18kHz or something.. but this is not so much because of the sampling rate as such, rather it's because theoretically you need twice as steep filters at 44.1kHz compared to 48kHz... and if you don't use sloppy filters then whatever allows you to succeed in the ABX is so ridiculously subtle that it's entirely meaningless in practice.

ps. I used to use 44.1kHz even back when my hearing was still good enough that I might have been able to hear the difference in a carefully controlled test.

[mystran]

yup, but that would be an obvious reason, if there was ie ir reverbs being used with the irs being 48k might cause artefacts at other rates, is another obvious reason.

If your convolution reverb does not resample when loading an impulse response with samplerate different from that used for processing... then really the convolution reverb is broken. Broken plugins in general can behave differently at different sampling rates. Wanting to use (old?) plugins that only work properly at 44.1kHz (or 48kHz, but 44.1kHz might be more common) is certainly a valid reason to use a specific sampling rate... but it's still broken plugins.

[edit: although from experience.. it's really hard to make frequency dependent damping of algorithmic reverbs truly consistent without internally resampling to a specific known rate or a sufficiently high oversampling rate, so that's kinda where I'd expect the most variation.. but even that probably mostly doesn't matter in practice that much, because it's the kind of subtle that you start to notice when you spend half a year fine tuning the damping of a reverb algorithm]

but all things being equal, i dont think a person could hear a difference between the two. on a system outputting both, with plug ins that work equally well at both settings, listening to the output of this, is what i mean. as in, there is no noticeable difference to our ears, only to how the programs we use interpret said data, which is why we use one or the other, is it for this or is it for that? we dont use either, because they sound betterer!!!

Depends on the definition of noticeable difference. If you hand someone two reference files A and B produced at different sampling rates (and resampled to common rate with quality SRC so as not to spoil the whole thing) and ask them to identify whether a third file X is equal to A or B, I find it likely that some individuals with exceptionally good high frequency hearing will be able to succeed with statistical significance. That's kinda the scientific meaning of "noticeable difference" and some 20 years ago I might have found it fun to try as a challenge, but now I don't have any chance anymore, such is the nature of aging.

But if you give those same people large pile of A/B signal pairs where either A or B is randomly 48kHz and then you ask them to identify all the 48kHz signals.. then I'm a whole lot less confident that anyone is going to be able to do much better than chance... and really it's this test that would give a more meaningful answer to the question of whether or not there's a practical difference.

[roman.i]

There are circumstances where....

For example the sound card only works at a fixed rate of 48kHz and support of other frequencies is done with shitty SRC by the driver.

You never really know what's inside the box.

Don't think this is the case with my audio interface (prism lyra),
maybe the clock works better with 48khz? IDK.

[mystran]

There are circumstances where....

For example the sound card only works at a fixed rate of 48kHz and support of other frequencies is done with shitty SRC by the driver.

You never really know what's inside the box.

Don't think this is the case with my audio interface (prism lyra),
maybe the clock works better with 48khz? IDK.

Try bouncing into a file at 44.1kHz, convert to 48kHz manually using some quality SRC and see if you can still hear the difference. If that solves the problem, then it could be the interface or driver.

If not, then it's probably some plugin (or DAW builtin effect) that's somewhat samplerate dependent. Having written a few algorithmic reverbs, the internal damping filters of those for example are notoriously hard to make truly sampling rate independent. Generally I'd expect 44.1k vs. 48k differences to be less obvious than 48k vs. 96k differences, but you never know.

[sQeetz]

I usually set everything up @44.1k because I don't give a crap.... some update of Live changed it to 48k and I didn't notice anything for the longest time. Changed it back to 44.1k because there is no difference anyway

[jethrobull]

I usually set everything up @44.1k because I don't give a crap.... some update of Live changed it to 48k and I didn't notice anything for the longest time. Changed it back to 44.1k because there is no difference anyway

Yep!
Used 96khz for ages thinking it had some kind of magic fairy dust. Updated the driver's one day, with the interface reverting back to 44.1khz and decided to just leave it. It sounds great and now I don't care about it. Just oversampling where needed

[Freaqpeak]

This video gave me some new insight how oversampling works. I used to always run at 44k1 and use plugins own oversampling when needed. Not so sure anymore. Will have to do my own tests again..
Anyways heres the video:
"Oversampling isnt that easy."
https://m.youtube.com/watch?v=qh-ybrG0qIg&t=1s

I still run at 44k1 because back in the day CPU usage was an issue and it sounds good enough for my own music. THOUGH 96k definitly sounded better for VST plugins back in 2000, but made it near impossible to make a track without bouncing to wave all the time.

[tundra5]

once you bounce it out no one is really going to hear the difference tbh , 44 is more than adaquqte

[billcarroll]

More good info on oversampling and running at higher sample rates. Higher sample rates for the DAW session are discussed at around 31:30, but the entire discussion with a plugin developer is good.

[Jafo]

Dan Worral

We old-timers have learned not to argue with IIRS.

[jtikey]

I'm using 48 kHz for everything

[Logga]

According to Dan Lavry of Apogee and Lavey Engineering, the theoretical optimal samplerate is around 60KHz, which is greater than 44.1/48KHz. 88.2/96KHz are the closest samplerates available that meet the 60KHz samplerate minimum.

88.2k is the obvious choice then, isn't it.

Only if your primary target medium is CD.

You can always use a high quality SRC at the end, so I don't think it matters.