Saving Edited 24bit Files in 32bit (float) DAW - Workflow?

[Eric Holtz]

Hello,

I am hoping to find clarification regarding what happens to a 24bit file each time it is saved and re-saved after editing using a 32bit (float) DAW. I am not mixing in a multitrack environment, but editing the actual waveform. The sources I work with are primarily LP (or 78) records, cassettes, and reel-to-reel.

My typical workflow for an LP record for example:

1. Record source at 24bit/88.2kHz/-15dB-12dB dBFS as one file per side of source; save copies as-is across multiple external hard drives.

2. Apply amplitude (clip gain) globally to both files, ensuring level consistency between both sides, and with that of the original source.

Save each as "...24-88_EDITS_01".

3. Open each "EDITS_01" and manually remove obvious clicks and pops, then apply light DeClicker - Save.

Note: I try to avoid adding EQ to a record, as I am not trying to remaster it, only clean it up.

4. Apply track markers - Save. Export track markers as 24/88.2 files - Save.

5. Export track markers as 16/44.1 with triangular dither for CD.

As you can see, there is a great deal of saving going on. Since Audition CS6 does all of its internal processing at 32bit (float), what happens to the 24bit file at each stage when I repeatedly edit and save it? Remember, I am working directly on the files themselves (Waveform editor) and not working in a multitrack, mixing situation.

Also, I thought I should only dither once, when going to 16bit. I've read elsewhere that dither should be applied at each step when saving the 24bit file, since it is going from the internal processing of 32bit float to 24bit integer. I have experimented with simply recording the original sources as 32bit (float)/88.2 files. Would using them throughout the editing process alleviate any conversion at each save? I've also read that Audition "may" apply dither to the 32 bit float file "behind the scenes" before being sent to the 24bit converter.

Any thoughts, tips, and suggestions would be greatly appreciated.

Eric

[SteveG_AudioMasters_]

Answered this in your other thread... but:

"I've read elsewhere that dither should be applied at each step when saving the 24bit file, since it is going from the internal processing of 32bit float to 24bit integer."

This isn't true at all - reason is in the other thread. Ideally, if you are going to process files in 32-bit mode, you should stick with it until the final stage, and only then do any dithering at the point of converting to a 16-bit int file. Incidentally, I think there is a good argument for making any MP3 copies direct from the 32-bit master without dithering - the main processing taking place in an MP3 conversion involves masking and reducing the bit depth of the 'masked' signals. If there's additional noise added to the least significant bit then potentially this could affect the coding algorithm. Or to put it another way - I can't see a single good reason for doing it!

[SteveG_AudioMasters_]

The other thing I should mention is that iZotope do a very good guide to dithering that you can download free of charge:

Dithering with Ozone

You don't need Ozone to use this - they explain the general principles, and they've got it pretty much spot-on.

[Eric Holtz]

Okay, say I open a 24bit file, apply noise reduction to it, click Save and close it. I then open it and apply Auto Heal Selection, and click Save. Are you saying that nothing is lost when saving?

"Ideally, if you are going to process files in 32-bit mode, you should stick with it until the final stage, and only then do any dithering at the point of converting to a 16-bit int file."

Are you referring to recording the original sources as 32-bit (float) files - the masters - and using them throughout the destructive editing process until I am finished?

I guess this is exactly where I am confused. Do you mean I should not be recording at 24-bit initially, but at 32-bit (float)? I thought I had read in some of your other posts that you recorded your initial tracks as 24-bit. If I have that correct, would that be for tracking instruments only, where all of the editing would be done in a non-destructive multitrack session? I've noticed there are some who record at 24-bit, and some who record in 32-bit (float). I didn't think it mattered since Audition will convert the file on the fly for 32-bit (float) processing.

Edit update: I was writing this post when you sent the answer to my other thread, so I missed it originally.

In a nutshell, if I open, change, save, and close a 24-bit file, then open, change, and save it again, I will be losing resolution. If I do the same with a 32-bit (float) file, nothing will be lost.

[SteveG_AudioMasters_]

Okay, I can clarify this a bit. Your sound device is only capable of outputting signals in an integer format, so as such it doesn't matter in the slightest if you store them as 24 bit files - indeed, they'll take up less room (if that's significant now - less likely in these days of massive storage capability). It's only when you come to operate on them, and even then only if you are likely to make large amplitude changes that you store, that 32-bit FP files come into their own. But when you use them, you really can mess around, pretty much with impunity. You can add 200dB to a FP file so that it's a solid mass of green, and when you re-open it and normalize it back to 0dB, it will be perfect. If you want to stop people peeking at your files without your permission, this is quite a good way of doing it, in fact!

It can be a bit misleading though. If you operate Audition's mixer really hot, then your mix will still normalize fine (as above), but because your sound device can only operate in the real world (ie, integer format) your hot mix will overload it and it will sound distorted, even though it really isn't. So you have to be a little careful...

[Eric Holtz]

"It's only when you come to operate on them, and even then only if you are likely to make large amplitude changes that you store, that 32-bit FP files come into their own."

The original 24-bit files were recorded around -15dBFS to -12dBFS. The first edit is to raise the amplitude to approx. -3dBFS and save. Is this the sort of large amplitude change which you were referring? When making multiple edits and saves on a 24-bit file, there will be no loss in quality? I've read some suggest saving the 24-bit file as 32-bit (float), and use that all the way through the editing process until the end. I would then have the original transfer master as 24-bit, and an edited master at 32-bit (float). Conversely, should the initial recordings be done as 32-bit (float) files, and kept there all the way through the editing process?

I realize that as soon as I change anything on a 24-bit file, Audition will convert it to 32-bit (float), where it remains until it is saved. That save should be in 24-bit.

Am I even remotely close to correct on this?

I try not to run Audition hot, since as you stated, I would simply be overloading the converters on my interface.

[SteveG_AudioMasters_]

Generally this isn't too much of an issue when dealing with 24-bit files, but if you reduced the level of a 0dB 16-bit file by 48dB and saved it, you've thrown away half of the bits (6.02dB/bit). And if you subsequently amplified this back to 0dB, then you'd have lost half of the dynamic range - and there would be nothing you could do about it. But, if you'd done this with a Floating Point file, this wouldn't have happened - all the dynamic range would still be there.

Straightforward editing without level changes won't have any particular effect on a file - generally it's only level changes that will do this. Ideally though, I'd make the original saves as 32-bit FP and stick with this. Whilst not strictly necessary, it still means that the file will open and save again without having to do any form of conversion, and the less of these you do, generally the better - even if it only saves a bit of time.

[ryclark]

Most applied effects such as EQ, compression, noise reduction etc. count as level changes. So generally once opened in Audition save the audio as 32bit float.

[Eric Holtz]

ryclark​

"Most applied effects such as EQ, compression, noise reduction etc. count as level changes. So generally once opened in Audition save the audio as 32bit float."

Okay, so I open a raw 24-bit file, apply the aforementioned effects, then Save As a 32-bit (float) file, keeping that bit depth all the way to the final master. I will then have a 24-bit original transfer master, and a 32-bit (float) edited master. Got it. I know I have to add dither when going to the 16-bit files for CD, but I should not apply dither when going from the 32-bit (float) file to 24-bit. I hope I have all of that correct now.

Side note: What should I do when I want to edit 16-bit/44.1kHz files that came from ripping CD-RWs made in my standalone recorder? Should I immediately save them as 32-bit (float) files, then continue editing as I would with the 24-bit files? Many of those RWs sound really solid, and I would hate to have to re-record all of the source material again.

[SteveG_AudioMasters_]

Eric Holtz wrote:

Side note: What should I do when I want to edit 16-bit/44.1kHz files that came from ripping CD-RWs made in my standalone recorder? Should I immediately save them as 32-bit (float) files, then continue editing as I would with the 24-bit files? Many of those RWs sound really solid, and I would hate to have to re-record all of the source material again.

Depends what you want to do with them, really. Audition is going to open them internally as 32-bit FP files, so it makes sense to store them as that if you are going to edit them. But if you are archiving them, then there's no point in doing so in anything other than their original format.

If you want a basic rule of thumb, which won't let you down whatever the circumstances, then you should archive material in its original format, but make 32-bit copies of it for editing. This is always safe, as you always have the original to go back to, and all your editing is on a copy, as it should be. Everything I record on location is originated in 24-bit, and that's what I archive it as. But the working copies are created as 32-bit files, just so that I don't have the bother of having to change the message in the format option - I can use 'save in original format' every time, and it all remains as 32-bit material.

[Eric Holtz]

SteveG(AudioMasters)​

Thanks a million Steve. I had concerns relating to the working copies, and now I know how to work properly with these intermediary files.

Recap:

• Record original analog source material as 32-bit (float) or 24-bit files.
• If recording to 24-bit, archive multiple copies as-is, then save a working copy as 32-bit (float) for the editing stage.
• If source files are derived from ripped 16/44.1 CD-RWs and CD-Rs, archive them as-is, then save each file as 32-bit (float) and continue editing.

What about the sample rate for records? I record a variety of LP records (folk, rock, classical, etc.) at 88.kHz, which seems to be the sweet spot for my interface when recording LPs. Am I gaining any substantial quality benefits with 88.2kHz over 44.1 or 48kHz?

What about the sample rate for cassettes? Most of my tapes are metal and high-position (CrO2) TDK, Maxell, and Fuji from the '80s and '90s, recorded using Dolby-C on a well maintained and serviced Pioneer CT-50R deck. Some of the cassettes are direct recordings of the first-play of (then new) vinyl records. Other tapes are mixdown masters of various 4-track cassette recordings of music groups.

Am I gaining any quality benefits with recording over 44.1kHz or 48kHz, even if the tape sources are records, and the playback machine is a Nakamichi MR-1?

If not, which is recommended, 44.1kHz or 48kHz?

[SteveG_AudioMasters_]

Eric Holtz wrote:

What about the sample rate for records? I record a variety of LP records (folk, rock, classical, etc.) at 88.kHz, which seems to be the sweet spot for my interface when recording LPs. Am I gaining any substantial quality benefits with 88.2kHz over 44.1 or 48kHz?

What about the sample rate for cassettes? Most of my tapes are metal and high-position (CrO2) TDK, Maxell, and Fuji from the '80s and '90s, recorded using Dolby-C on a well maintained and serviced Pioneer CT-50R deck. Some of the cassettes are direct recordings of the first-play of (then new) vinyl records. Other tapes are mixdown masters of various 4-track cassette recordings of music groups.

Am I gaining any quality benefits with recording over 44.1kHz or 48kHz, even if the tape sources are records, and the playback machine is a Nakamichi MR-1?

If not, which is recommended, 44.1kHz or 48kHz?

Personally I do everything at 44.1k, and have done for years. With vinyl, this is based on looking at the signals coming from the cartridge, rather than the noise. Most vinyl is mastered from open-reel tape, and to get the response of this over 20kHz requires little short of a miracle, so it never makes it to disks. The only exception to this is if you have material that used CD-4 encoding, but even then if you use an external decoder (which you'd have to do) you could record the multi-channel results at 44.1k without losing a thing. Cassettes are just the same; I use a Nak LX-5, and it's really difficult, even with good material, to get a response above about 18kHz, and even at that level it's pretty meaningless because it's got a lot of noise with it, inevitably.

The only thing that the sample rate relates to is the highest frequency you can record - it's always the Nyquist rate, which is half of the sample rate. So for 44.1kHz, that would be 22.05kHz, which is comfortably higher than the highest frequency you can hear, even as a child. Adults get nowhere close to this - that's because presbycusis gets us all in the end - gradually...

Back in the bad old days, there was something to be said for recording at a higher sample rate, because old A-D converters used brick-wall filters to stop aliasing (a folding-back of any frequencies above the Nyquist rate) and these filters caused phase errors which were sometimes audible. But all modern converters use an oversampling technique that eliminates the need for this filtering - so 44.1k sounds identical to anything higher - not least because your equipment can't reproduce it anyway!

48k tends to get used for video, rather than pure audio, although there's a DVD audio spec that uses it. Since all CD audio uses 44.1k, and it's been experimentally (and reproducibly) proven that even with really good equipment, nobody can tell the difference between 44.1k and any higher rate at all, there seems to be no point at all in using a higher one - unless you want to use up a lot of storage space storing noise, that is. But hey, it's your call...

[Eric Holtz]

SteveG(AudioMasters)​

My typical editing for an LP record involves taking a copy of the original 24-bit or 32-bit (float) file, then boosting amplitude by roughly +10dB to +13dB, removing clicks and pops, and using Auto Heal and light noise reduction if and where needed. Except for click removal, cassette transfers get the same treatment, using (sometimes) several saves of light noise reduction.

So in summary, it really doesn't matter whether I am using a 24-bit file or a 32-bit (float) file, saving, opening, and saving it again throughout the editing process, but it is preferred to use 32-bit (float) for recording the original source material, and then editing that file. I was worried that I was stuck with a bunch of 24-bit files that I couldn't heavily edit for fear of the possible conversions at each save. In your opinion, I don't need to save a 24-bit file as 32-bit (float) first, before the editing process?

I have a couple other questions while I am here:

1. Should I apply a LP filter on the files, or does it really matter? At 88.2kHz, there is a smooth slope all the way to about 30kHz before it's all noise. When prepping tracks for CD, should I add, say a 20kHz LP filter to achieve a smooth slope down to 22kHz, removing the typical brickwall?

2. Is it overkill to record cassettes at higher than 44.1kHz, even if they are metal and high bias masters being played back on a well maintained Nakamichi deck?

I didn't think I should since the deck's specs are rated to 20kHz.