Do I really need to record at 24 bit? Why?

[Dr Huge Longjohns]

So I've been recording at 24bit/44.1 in Logic as long as I've worked in the box. But do I really need to use all that extra storage? Why shouldn't I work in 16bit as it's simply a question of how much dynamic range I'm working with and I don't record anything unusually dynamic. (And I dither down to 16bit at the end in any case.)

Or am I missing something here???

[Hugh Robjohns]

No, you don't have to work at 24 bit, and plenty of hit CDs were made with 16 bit tracking back in the 80s...

But tracking at 24 bit means the source files are only about 50% bigger than 16bit files, which is trivial with today's massive hard drive capacities, and the DAW is going to store everything as 32-bit float anyway.

The major benefit with 24 bit tracking is that you can afford to leave a generous headroom margin when tracking without compromising the signal-noise ratio, which makes tracking a whole load more stress-free. Tracking at 16 bit can be a bit tight on headroom if you're working in a quite environment. Not going to make a huge difference if your environmental noise floor is at -50dBFS, but anything quieter than that and I'd go 24-bit and think no more about it!

Personally, I'd say 24 bit tracking is a complete no-brainer. No practical reason not to, and significant benefits to working that way.

H

[Richard Graham]

A 24-bit file is 50% larger than a 16-bit file, not 30% - right?

https://picta.xyz/i/708db0512711/

Because I always record in a noisy environment with guitar amps and so on kicking out all kinds, I can't see any advantage using 24-bit, and plenty of advantage using 16-bit (smaller files means faster transfers from SD card, quicker backups, having to upgrade my storage 1/3rd less often...)

TBH I don't care about noise now. I used to try to get the meters all the way up without hitting the red. But for band recordings it makes no practical difference to me if I don't ever get into the orange.

And it was SoS that taught me about digital headroom!

[Dr Huge Longjohns]

The major benefit with 24 bit tracking is that you can afford to leave a generous headroom margin when tracking without compromising the signal-noise ratio,

My understanding is that 16bit gives me 96dB to play with? Am I really going to use that up even in a very quiet room with perfect equipment unless I'm recording a jet engine and mouse squeak in rapid succession on the same track? (My new concept album: Rolls Royce versus the Rodents.)

[Folderol]

I always record at 24 bit these day. I don't actually need to, but it means I can forget about setting anything up until I've got it all in the box and want to tart it up. I've still got acres of space on my hard drives.

[Hugh Robjohns]

A 24-bit file is 50% larger than a 16-bit file, not 30% - right?

Oops! Sorry... been a long day! I've corrected my original.

H

[ConcertinaChap]

The point (at least for me working in a reasonably quiet environment with acoustic instruments) is that I simply don't worry about getting a signal as close as possible to 0 without going into the red. The Ones Who Know say I don't need to worry so I don't and it all seems to work out. As Hugh says, it makes life just that bit easier.

CC

[Hugh Robjohns]

My understanding is that 16bit gives me 96dB to play with?

Technically it's closer to 93 because of the dither noise, but yes, broadly speaking. So if your average levels are around -20dBFS when you're tracking the digital system noise floor is going to be around 73dB lower... which is probably fine for recording guitar amps in a home studio... but it might potentially start to become an issue if you record heavily compressed vocals. Unlikely... but possible.

H

[Dr Huge Longjohns]

Out of interest, what was the theoretical 'bit depth' of 24 track tape ie what was the workable dynamic range before you got unusable noise issues? And of typical two track tape masters?

[Hugh Robjohns]

Out of interest, what was the theoretical 'bit depth' of 24 track tape ie what was the workable dynamic range before you got unusable noise issues? And of typical two track tape masters?

It's actually not that easy to make direct comparisons because the standard alignment for the reference level in an analogue tape machine is when there is 3% harmonic distortion -- something that wouldn't be acceptable at all with current digital systems! And... transient overload tends to be largely inaudible (or even desirable) on analogue tape -- meaning levels can be pushed harder to tape to help mitigate tape noise in a way that they can't with digital systems (to mitigate dither noise).

Having said that, a high-end 1/2-inch mastering machine running at 30ips might achieve something like 75dB SNR. A 1/4-inch 2-track machine at 15ips would achieve maybe 68dB on a good day with a favourably weighted measurement! With more conservative levels to reduce distortion, 65dB SNR is a more realistic figure.

A pro multitrack would be worse than that because of the much narrower track widths -- hence the standard requirement for Dolby A and then Dolby SR noise reduction, and a semi-pro machine with slower speeds and even narrower tracks would be worse still.

With the Dolbys bypassed, a really good pro 2-inch 24-track (like a Studer A827) would manage about 60dB with CCIR EQ (and a bit more with NAB EQ), and maybe 70dB SNR with Dolby A. Dolby SR would push that a bit further to maybe 80dB-ish on a good day.

All in very round numbers, of course.... (the specifics of the measurement bandwidth, weighting, rec-rep EQ, and bias level all play a significant role in the actual figures), and the SNR is a trade-off against bandwidth and distortion.

The important point to take away is that when working with digital gear, you have to leave generous headroom margins, whereas you didn't/don't quite so much with analogue tape.

That means a SNR that might be acceptable with tape simply isn't with digital. Peak clipping in the digital domain is so horrible and blatant with harmonic sources (you can often get away with clipping noise-like signals like snares and cymbals!) that it's simply not acceptable to most. Whereas it is not only acceptable, but often desirable with analogue tape recording.

Modern 24-bit converters are just about on a par with the dynamic range capability of a good analogue console. 16-bit converters fall somewhat short... But then being practical, the console noise floor isn't usually the limiting factor in most recording environments.

H

[Dr Huge Longjohns]

Very interesting, cheers!

[jimjazzdad]

I do all of my live (mostly orchestral) recordings in 24 bit now, but I always feed a split to a back up machine , my old Korg D888 - a 16 bit machine. I made a lot of good recordings on that old D888. I would be hard pressed to tell the difference between my 24 bit master and the 16 bit back up, if I didn't already know. But its a lot easier to deal with the occasional tutti in 24 bit.

[Tim Gillett]

My understanding is that 16bit gives me 96dB to play with?

Technically it's closer to 93 because of the dither noise, but yes, broadly speaking. So if your average levels are around -20dBFS when you're tracking the digital system noise floor is going to be around 73dB lower... which is probably fine for recording guitar amps in a home studio... but it might potentially start to become an issue if you record heavily compressed vocals. Unlikely... but possible.

H

Because its dynamic range is much smaller wont a heavily compressed vocal be easier to record?

[Hugh Robjohns]

oops -- sorry. Not writing what I'm thinking -- and not for the first time today!

I meant recording a vocal that would subsequently be heavily compressed. The processing would potentially pull up the noise floor unacceptably.

[Guest271017]

I record 32 bit float in Cubase, any other DAWS have that option?

[muzines]

I record 32 bit float in Cubase, any other DAWS have that option?

What are the reasons why you choose to record in that format?

[Guest271017]

I record 32 bit float in Cubase, any other DAWS have that option?

What are the reasons why you choose to record in that format?

Cubase does all internal processing at 32bit FP. I know my real recording bit depth is controlled by my DAC. Been a long while since I was involved in discussion on the topic so the reasons are a bit muddy. One has to do with how buffers get filled. There are plenty of discussions about it on Google, don't want to link to competition sites, don't know if it's frowned upon here. It's not a matter of headroom or file size, it's how everything is handled ITB.

[Hugh Robjohns]

I record 32 bit float in Cubase, any other DAWS have that option?

They all process audio with that kind of word-length (and some work with even longer word-lengths), and it makes sense to store work-in-progress projects in that format rather than continually convert to and from 24-bit fixed point formats.

But the raw source tracks will never be more than 24-bit, and the debate here is whether 16-bit source files are acceptable.

H

[ConcertinaChap]

I do all of my live (mostly orchestral) recordings in 24 bit now, but I always feed a split to a back up machine , my old Korg D888 - a 16 bit machine. I made a lot of good recordings on that old D888. I would be hard pressed to tell the difference between my 24 bit master and the 16 bit back up, if I didn't already know. But its a lot easier to deal with the occasional tutti in 24 bit.

Lovely piece of kit, the D888, I thought very highly of mine and would have bought a 24 bit version like a shot for site recording if they'd made one, but your last sentence summarises precisely for me the advantage of 24 bit.

CC

[CS70]

So I've been recording at 24bit/44.1 in Logic as long as I've worked in the box. But do I really need to use all that extra storage? Why shouldn't I work in 16bit as it's simply a question of how much dynamic range I'm working with and I don't record anything unusually dynamic. (And I dither down to 16bit at the end in any case.)

Or am I missing something here???

No. On occasion I still use the BR800 as a practical recorder when I dont want to bring the laptop. It records at 16bit and I have had no problems with that at all.

You need to pay a little more attention to the gain structure so that you don't clip things or they dont' get too low - as Hugh says, it's just more forgiving to work at 24bit as the sweet band is larger.

[Drew Stephenson]

Another for the "It's just easier" camp here. I'm mostly recording myself so just being able to set things up with plenty of headroom means I don't then have to keep an eye on the faders whilst trying to do everything else. It also means I can jump straight to overdubs and supporting parts without needing to reset the gain.

[muzines]

Cubase does all internal processing at 32bit FP.

Yes, basically all DAWs do (a few run 64-bit precision).

I know my real recording bit depth is controlled by my DAC.

Yes, and so if you are recording to 32 bit files, there will not be any extra data or "quality" improvements over 24-bit, you're just writing bigger files with more zeroes. And consequently, when playing back these files, there is more overhead as the disk has to deliver larger files with extra data all set to zero and giving no benefit. So: bigger files, poorer track performance, and no audible benefits at all.

(For example, if you use 24-bit audio files in Logic, and you are using 75% of the disk performance playing them all back with a given buffer size, but then you "freeze" all those tracks to bake in the plugin processing to free up the CPU, the disk requirements are now higher, as Logic is playing back 32f files (to retain processing precision), instead of the 24-bit files - so your disk requirements now move up to 100% and you're in danger of not being able to play everything back.)

If the files are 32f files, then the audio resolution is still only 24-bit anyway (the other 8-bits are for scaling the 24-bit audio data).

Been a long while since I was involved in discussion on the topic so the reasons are a bit muddy. One has to do with how buffers get filled. There are plenty of discussions about it on Google, don't want to link to competition sites, don't know if it's frowned upon here. It's not a matter of headroom or file size, it's how everything is handled ITB.

No, there is basically no point in recording to 32-bit files, unless you have a specific reason to do so. Bigger numbers are not better, and in 99.9% of cases, 24-bit audio recording, done well, with good gear, is all you need. (Sample-rate arguments are a bit different but are not relevant in this discussion).

As Hugh says, it's worth retaining 32f files for intermediate processing, but for straight recording from acoustic sources with an audio interface, recording at 32-bit is a waste of resources for no benefit....

But I digress...

[ef37a]

I would respectfully suggest that anyone questioning the "easy life" 24 bits gives you has never done any live music to tape recording?

I recall recording a girls choir and grand piano to a stereo Ferrograph series 7 (DC heated triode pre amps) and that was a trial. The music (in Polish!) had a dynamic range well beyond the maybe 60dB of the recorder (no" nice" squash/tape sound wanted on choirs!)

Had I had my present rig, NI KA6 and perhaps the ZED10 I could have got a level at several FFs and then left it running and gone to the pub! Well, almost.

Dave.

[Normonster]

A 24-bit file is 50% larger than a 16-bit file, not 30% - right?

Oops! Sorry... been a long day! I've corrected my original.

H

This is a funny one, even if off topic...

In Pricing the two approaches yield Margin and Markup.

If A-B=C, then C/A=Markup and C/B=Margin. Percentage growth rates follow the Margin calculation, so 24-16=8, 8/16=50% (the Markup formula would have been 33.333%)

Cheers.

Since I can I think I'll try recording at the higher bit rate.

[Dr Huge Longjohns]

Eh?