|
|
Post by drumsound on Feb 14, 2024 22:16:45 GMT -6
Does anyone have thoughts on using 32 bit float compared to 24 bit. I'm starting a record tomorrow where the client has mentioned it because the possible mix engineer uses it. I've heard from some that it's snake oil, by others that it meshes better with 32 bit plugins.
Does anyone have any experiences, test, thoughts, anecdotes, or conjecture on the subject?
|
|
|
|
Post by audiospecific on Feb 14, 2024 22:27:41 GMT -6
Does anyone have thoughts on using 32 bit float compared to 24 bit. I'm starting a record tomorrow where the client has mentioned it because the possible mix engineer uses it. I've heard from some that it's snake oil, by others that it meshes better with 32 bit plugins. Does anyone have any experiences, test, thoughts, anecdotes, or conjecture on the subject? If you were not using plugins and bringing the track in at -4dbfs and mixing it I can see it being beneficial having 770db of digital headroom.
Otherwise, if you are a normal person who gain stages for the plugins, its pretty useless.
|
|
|
|
Post by Blackdawg on Feb 14, 2024 23:40:05 GMT -6
Sound wise there is no difference.
Most DAWs do operate in 32 but already anyways.
I have had it save my ass, I do a ton of live concert recordings and sometimes we get surprised by something. So we do 32bit even though the converters are 24bit. It still has saved my ass.
That said if you do your stuff right you don't need it. And sometimes I've had it make it more of a pain. But that's usually with trying to work with 32 but files from a sound devices in protools. Never have issues if it's made in protools itself.
It also makes your files a lot bigger.
So I doubt you'd see any gains from it really.
|
|
|
|
Post by lee on Feb 14, 2024 23:52:17 GMT -6
Where it is nice is that you can have a mix clip your DACs and mix bus, but the 32bit file preserves the headroom in the file itself, so even if your printed mix clips somewhere (even just inter-sample peaks), you can clip gain it down and there is no actual damage to the audio. And yes, if it was recorded low, you have a better chance of gaining it up without noise.
|
|
|
|
Post by sean on Feb 15, 2024 0:01:03 GMT -6
Where it is nice is that you can have a mix clip your DACs and mix bus, but the 32bit file preserves the headroom in the file itself, so even if your printed mix clips somewhere (even just inter-sample peaks), you can clip gain it down and there is no actual damage to the audio. And yes, if it was recorded low, you have a better chance of gaining it up without noise. Yes, this. |
|
|
|
Post by audiospecific on Feb 15, 2024 2:36:05 GMT -6
Where it is nice is that you can have a mix clip your DACs and mix bus, but the 32bit file preserves the headroom in the file itself, so even if your printed mix clips somewhere (even just inter-sample peaks), you can clip gain it down and there is no actual damage to the audio. And yes, if it was recorded low, you have a better chance of gaining it up without noise.
But there is diminishing returns recording low then gaining it up. Because A/D converter's signal to noise is not constant with signal level. So, if recorded low its going to have more noise added to it compared with a high level signal and its dictated by the A/D chip's signal to noise at the level recorded.
|
|
|
|
Post by the other mark williams on Feb 15, 2024 9:53:03 GMT -6
Where it is nice is that you can have a mix clip your DACs and mix bus, but the 32bit file preserves the headroom in the file itself, so even if your printed mix clips somewhere (even just inter-sample peaks), you can clip gain it down and there is no actual damage to the audio. And yes, if it was recorded low, you have a better chance of gaining it up without noise.
But there is diminishing returns recording low then gaining it up. Because A/D converter's signal to noise is not constant with signal level. So, if recorded low its going to have more noise added to it compared with a high level signal and its dictated by the A/D chip's signal to noise at the level recorded.
This is why some of the contemporary 32bit float recorders have two sets of A/D chips, each set optimized for a gain range. Literally just last week I had to film an orchestral event where I needed to take an audio feed off of a mixing board that was backstage (while I was filming out front), and the rehearsal was over before I could get to the event, so I had absolutely no way of knowing what the levels would be before hitting record. That 32bit float recording feature saved my ass. When I got it into post at the studio the next day, it was recorded so hot that it was almost pure distortion. I pulled it into Logic (which finally supports 32bit float in its audio engine as of a few months ago), gained it down by something like -36dB, and it sounds great. No distortion at all. I've had situations where the opposite happened, as well: where the signal was waaaaaaaay too low, but I was able to gain it up in post with very few (if any) deleterious effects. It is nothing short of remarkable what these recorders can pull off. Yes, certainly there are still "best practices" when tracking, and we need to continue teaching those to younger generations of course, but it really is incredible what is possible with 32bit float recording, especially when the recorder itself is designed for this kind of use. |
|
|
|
Post by damoongo on Feb 15, 2024 10:55:14 GMT -6
No benefit to tracking at 32bit if you have 24bit converters. But do any internal DAW clip processing etc at 32, and render the files you will send to the mix engineer at 32.
|
|
|
|
Post by robo on Feb 15, 2024 11:08:54 GMT -6
My understanding is that even when you record at 24bit (I do), any digital processing will spit out 32, so if you’re using analog inserts after plugins, it is best to dither down to 24 first.
I also dither to 24 for monitoring despite my interface being ok with 32 - it’s subtle but sounds better to me. Airwindows has a handy plugin called Monitoring which does this, along with some other useful tricks (isolate subs, mono, etc)
|
|
|
|
Post by lee on Feb 15, 2024 13:03:37 GMT -6
Where it is nice is that you can have a mix clip your DACs and mix bus, but the 32bit file preserves the headroom in the file itself, so even if your printed mix clips somewhere (even just inter-sample peaks), you can clip gain it down and there is no actual damage to the audio. And yes, if it was recorded low, you have a better chance of gaining it up without noise.
But there is diminishing returns recording low then gaining it up. Because A/D converter's signal to noise is not constant with signal level. So, if recorded low its going to have more noise added to it compared with a high level signal and its dictated by the A/D chip's signal to noise at the level recorded.
Indeed. There's always diminishing returns with poor analog gain staging! |
|
|
|
Post by audiospecific on Feb 15, 2024 17:42:07 GMT -6
I've had situations where the opposite happened, as well: where the signal was waaaaaaaay too low, but I was able to gain it up in post with very few (if any) deleterious effects. What I find is on A/D converters when you boost like a -20 dbfs recorded signal to -10 dbfs, you start getting clock based noises from the recorded signal at somewhere -80 dbfs to -110 dbfs and this varies depending on the ic chip and line stage used in the converter. |
|
|
|
Post by Dan on Feb 15, 2024 21:19:18 GMT -6
But there is diminishing returns recording low then gaining it up. Because A/D converter's signal to noise is not constant with signal level. So, if recorded low its going to have more noise added to it compared with a high level signal and its dictated by the A/D chip's signal to noise at the level recorded.
This is why some of the contemporary 32bit float recorders have two sets of A/D chips, each set optimized for a gain range. Literally just last week I had to film an orchestral event where I needed to take an audio feed off of a mixing board that was backstage (while I was filming out front), and the rehearsal was over before I could get to the event, so I had absolutely no way of knowing what the levels would be before hitting record. That 32bit float recording feature saved my ass. When I got it into post at the studio the next day, it was recorded so hot that it was almost pure distortion. I pulled it into Logic (which finally supports 32bit float in its audio engine as of a few months ago), gained it down by something like -36dB, and it sounds great. No distortion at all. I've had situations where the opposite happened, as well: where the signal was waaaaaaaay too low, but I was able to gain it up in post with very few (if any) deleterious effects. It is nothing short of remarkable what these recorders can pull off. Yes, certainly there are still "best practices" when tracking, and we need to continue teaching those to younger generations of course, but it really is incredible what is possible with 32bit float recording, especially when the recorder itself is designed for this kind of use. this x100. There is no real technical benefit, only detriment but the lack of error stuff is huge. After trying to fix numerous clipped tracks and the only way is to true peak limit or edit them, those modern zoom and sound devices recorders are a god send I always freeze and save at 32-bit even though my converters re 24 just do I don’t clip it. They can still clip the inputs of those 32-bit float recorders if they’re super careless too but you cannot help those guys. They need an analog limiter with at least a 1 second release or something. |
|
|
|
Post by sirthought on Feb 16, 2024 18:59:00 GMT -6
Isn't it all a moot point if your interface is set up for 24 bit?
And as for interfaces with two A/D chipsets...I notice the Neumann MT48 has two A/D chips, which is part of how they get conversion numbers so much more impressive than other devices, yet in the website listing they don't even offer up what is typically listed for such devices. Sweetwater does let customers know it's 24 bit/192kHz, but Neumann doesn't really cover that on their page. This surprised me, as I thought Merging was going after the people who really want 32 bit.
Off the top of my head I can think of only select interfaces that promote 32 bit. For sure the devices targeting video/film, like Zoom and Sound Devices, are pushing this as when you are in the field it's so hard to gain stage. But if you work in your studio all the time and take the time to gain stage, it won't matter if someone down the line is mixing or mastering with a 32 bit interface.
|
|
|
|
Post by Dan on Feb 16, 2024 19:29:38 GMT -6
Isn't it all a moot point if your interface is set up for 24 bit? And as for interfaces with two A/D chipsets...I notice the Neumann MT48 has two A/D chips, which is part of how they get conversion numbers so much more impressive than other devices, yet in the website listing they don't even offer up what is typically listed for such devices. Sweetwater does let customers know it's 24 bit/192kHz, but Neumann doesn't really cover that on their page. This surprised me, as I thought Merging was going after the people who really want 32 bit. Off the top of my head I can think of only select interfaces that promote 32 bit. For sure the devices targeting video/film, like Zoom and Sound Devices, are pushing this as when you are in the field it's so hard to gain stage. But if you work in your studio all the time and take the time to gain stage, it won't matter if someone down the line is mixing or mastering with a 32 bit interface. yeah it's a moot point anyway because none of these interfaces are even 24-bit and the audio will have always been 24-bits at some point too usually, they're just saving it as 32-bit float in the internal mixer or your computer if the drivers support it. You can still clip the analog circuitry before the chip too.
|
|
