Topic: DAW workflow question: 32 bit > 24 vs 32 all the way

[checht]

I'm considering which DAW to move to from Audacity. Down to Reaper and Logic Pro, and tending towards logic as I'm a lifetime apple person and my low vision is an issue learning new software. Haven’t found a visual accessibility skin for Reaper that works for someone needing high contrast, and I'm dialed in on that on logic.

But, logic cannot import 32 bit float files, so I'd need to convert my recordings to 24 bit.

My question is whether this would be a problem. I have gotten used to setting gain for recording so that levels go above zero a bit. Not every show, and not the whole show, but some recordings end up in the red somewhat.  Would converting to 24 bit lead to clipped files? I'll likely split the poly wav in Audacity anyway, so reducing bit depth could happen at that time. I could normalize first I guess…

Any other considerations I should be mindful of? I can certainly learn Reaper, and probably figure out how to customize the interface myself, just seems like logic will be easier/faster new platform for me. Suggested workflow?

Thanks!

[morst]

32 float only has 24 bits of meaningful audio data (the mantissa) anyhow.
As long as you don't clip during transfer to 24 bits then you should be fine?!

[aaronji]

^ And the actual dynamic range is very unlikely to exceed even 16-bits. I suppose it would be preferable to do all mastering prior to dither, but I doubt it makes much difference (especially since the DAW probably upsamples to 32-bit float for processing).

[checht]

1. Thank you both.
2. Yes, Logic say that the workspace is 32 bit float.
3. Should I normalize in Audacity prior to bit depth conversion?

[morst]

Mathematically speaking, to avoid rounding errors, it would probably be good to do the work on the original bits, but if the rounding error is really at the limit of resolution, I wonder how audible it would be if you didn't do it the optimal way?

^ And the actual dynamic range is very unlikely to exceed even 16-bits. I suppose it would be preferable to do all mastering prior to dither, but I doubt it makes much difference (especially since the DAW probably upsamples to 32-bit float for processing).

[live2496]

If your shows are recorded with some peak levels exceeding 0 dbfs, then dropping the level 1 db or whatever is needed is a good idea in Audacity first. Just ensure that peak levels are addressed and then export a 24-bit file. 

Gordon

[checht]

Makes good sense.

Many thanks for your help!

~Chris

[Gutbucket]

Sounds like you'll need to convert to 24 bit to import into Logic.  In whatever software you use to make that conversion (Audacity?) you'll want to select the desired 24bit-wide range from the 32bit float data to export. This shouldn't cause data loss as the actual audio data, once it has been recorded, will fit comfortably within 24 bits. To do that you'll normalize or otherwise specify the range to be used. 

Once your are manipulating the audio data in Logic its internal workspace will use a 32bit float calculation space, which is good as that's where its needed again for calculation accuracy.

Once that work is complete, you can then normalize and export the final result from Logic at 24 or 16bit, as the finished output should fit into the dynamic range of either format.

[detroit lightning]

Was gonna start a new thread, but looks like my question fits in here...

Recorded 32bit w/ a Zoom F3 last night, first time where levels "clipped" - visually only.

Is my first step (audacity) to "amplify" down to -1db (or anything below 0) while in the 32 bit realm? Or is there another way to do this...

[Gutbucket]

That's the essence of it, you want to normalize such that the highest peak of the file is comfortably below 0dBfs. -1dB is fine, I'd probably default to -3dB or so.  I've not used Audacity in a long time and can't recall the actual procedure to use to do this.  Could be "amplify" (entering a negative value in your case) after identification of the highest peak, or "normalize", which should locate the highest peak and shift it down to whatever level you specify.

[detroit lightning]

That's the essence of it, you want to normalize such that the highest peak of the file is comfortably below 0dBfs. -1dB is fine, I'd probably default to -3dB or so.  I've not used Audacity in a long time and can't recall the actual procedure to use to do this.  Could be "amplify" (entering a negative value in your case) after identification of the highest peak, or "normalize", which should locate the highest peak and shift it down to whatever level you specify.

thank you!

Yeah, I believe the "amplify" function is the way in audacity (seems like I should probably switch to another program more widely used these days...).

I have some normalization questions, but I'll poke around the forum first.

[Gutbucket]

Fuzzy memory of some commonly used DAW's normalization function (Audacity?) having two options- one being this kind of straight-forward level normalization to whatever peak value you specify (the desired outcome), the other option limiting peaks that exceed the value or something like that, which is not what you want.

[checht]

Fuzzy memory of some commonly used DAW's normalization function (Audacity?) having two options- one being this kind of straight-forward level normalization to whatever peak value you specify (the desired outcome), the other option limiting peaks that exceed the value or something like that, which is not what you want.

Normalize in Audacity does the former, changing the amplitude of the entire wave,

[morst]

Yeah, I believe the "amplify" function is the way in audacity (seems like I should probably switch to another program more widely used these days...).

Audacity is still widely used!

[Gutbucket]

Thanks checht.  In that case, Audacity's normalize function is likely the easiest way to accomplish what you wish to do, detroit.