48000 came into use because it's divisible by 24, 25 and 30. Which are frames per second in movie and TV standards, so it's easier to montage.

Nobody hears the difference in blind tests of 44100 vs 48000.

 

96k will avoid aliasing that happens in the audio bandwidth at lower sample rates, the highs will sound better, so less sibilance, snares will not be so crunchy, hats less metallic. Oversampling is sample rate conversion up and then down again, could be more CPU friendly to just run at 96k for the session however SRC's can sound different, most of them use open source code but which one, you can see the response of many SRC here: https://src.infinitewave.ca/ , DB Poweramp 17 is one of the best to see how it should look, compare what you have with this, Izotope RX is also very good, as are others.

 

1500db headroom... a nuclear blast pressure wave is only rated around 250. That's a lot of headroom.

 

What I mean is that "I record in 32bit float" and "I mix in 32bit float" are two different statements. When your DAW processes 32bit float, it's giving you that headroom during playback only. A 32bit float project can't make 24bit audio become 32bit float, since that is decided inside the interface before it reaches the DAW.

You can do a simple test to prove this: clip your converter and record it inside a 32bit float DAW. If you can't recover the audio, it's 24bit. If you can recover it, it's 32bit float.

 

Last edited: May 22, 2025 at 8:58 PM

I record all my music projects at 44.1 and 24bit, and music projects with video in 48, 24bit.
There's a reason for that, which I won't get into, but you can Google it.
More isn't always better. Just make sure to record with 24bit (or higher) and you'll be good.

 

I recorded for years at 44/24 because stuff was going to end up on a CD anyway and even though that's not generally the case anymore I see no reason to change it because everything sounds fine so I don't feel like wasting my hard drive space. If there are any plugin/processing issues going on I'm blissfully unaware.

 

32bit float session support is not the same as 32bit float input stream without truncation during recording. If your DAW is not capable of this, it makes no difference. I already gave a list of those which do not, and so from your response I can safely assume yours is among that list.

 

44.1 for music, 48 for media.

 

It's not, I'm on Studio One. I think we're having a communication issue, because it seems that you don't understand what I'm saying, and I can't understand what you're saying.

Once again, my initial point was: seems like there are more people than I expect recording at 32bit float (considering 32bit float interfaces are still, to this day, very niche). I think people are getting 32bit float sessions confused with 32bit float recordings. If your AD is not capable of recording at 32bit float, a 32bit float session will not magically make 24bit audio into 32bit float. If your AD can record 32bit float, a 32bit float session will have 32bit float audio in it.

In any case, 32bit float audio is completely pointless for recording purposes, unless you're recording extremely dynamic content like field recordings. It has close to zero application in the studio, as 24bit already has infinite headroom, and proper gain staging is the foundation of mixing.

 

Even if you record in 16bit, the moment you have more than one track or even if you just have that one track, but do any processing (even if it's just level) , 32bit float would (in theory) have less quantization error than 24 bit.
But I'm totally with you: 24bit is plenty.
//edit// folks, we are talking about noise here. Quantizationerrors add noise. That can be shaped with dithering to make it more pleasant, but the moment we switched to 24 bit that noise is so low it shouldn't be a concern any more.
32bit float does make sense when you don't do a classic microphone preamp, but digitize directly the very low mike signal unamplified. For that 32bit float is a gamechanger. Preamps added noise, that's gone now, we're in the digital domain right beind the xlr cable so to speak.

 

Last edited: May 23, 2025 at 12:27 AM

people are confusing 32bit floating point sessions with 32bit recording, at the interface. That is your point. Mine is that even if you are getting 32bit floating point at the interface, but the daw being used supports 32bit floating point sessions; that does not mean that the input stream from the interface into the DAW is 32bit floating point. It's not, unless it has been updated to support this. It's 24bit. That means, each word passed from the interface into the session is being truncated by up to 8 bits. There is no recovery for truncation of the difference. It is like dither up to restore the lost word legnth.

 

I used to use a mix of 44.1 kHz and 48 kHz but a couple of years ago I standardized on 48kHz / 24 bit, simply because I was doing a few projects with video. As stated in various posts above most DAWs work in floating point internally so it's really not a huge issue. By the time you've mixed a few sound sources together no-one is going to notice the format of those sources and you can mixdown to anything you want, or even to several formats.

 

For EDM is 44.1 Good and for Video Sounds and/or Music-Videos is 48 the better choice - HiRes like 88,2 or 96 is used if you include high-quality studio equipment externally, mastering signal chain (hardware based) or you undertake high-quality recordings from the hardware area, 48 Khz is sufficient here, but 96 or 88.2 is more accurate.