Logic support for 32-bit float files

I tend to agree with Edgar on the Zoom F6, although it is a clever device.

The Zoom has 2 converters on each input, one to handle low level signal, the other to handle high level signal. Both of these converters are 24 bit but are dealing with their own range. The resulting combination is then stored as a 32 bit float file on the Zoom. It's a clever and highly useful design but their marketing speak doesn't give the whole picture of what's going on.

However you look at it, any overs that are not clipped by storing as a 32 bit float file will need to be brought back to fixed point legal values before final output to your detination format - no one is going to be playing your audio back through an F6 or DAW. I'm not knocking the Zoom F6, it's a clever and useful device again for unexpectedness. The zoom doesn't store everything as 32 bit float though and there is good reason for this. If you have a perfectly legal 24bit audio file which never goes into overload and clipping, saving it as a 32 bit float file will give you absolutely no advantage, in fact, it willproduce a vey slight loss. However, if you have an audio file that is within the floating point DAW (i.e. it has never been fixed point clipped), where you have no idea of the max levels within that file - then storing it away as a floating point file will at least give you the possibility of avoiding clipping when the file is restored again. The Zoom F6 accomodates this nicely by not forcing everything to 32 bit loat, i.e. what I think they call the 'Dual' mode.

Going back to Logic's inability to import 32 bit float files natively, I share a lot of people's frustrations at this. After all, it's processing them internally all in float anyway. It is bewildering.

This is a cool discussion.

First of all, Logic is WAY OVERDUE with the implementation of importing 32bit float files without converting them to 24bit. I begged them for years.

Another thing is the confusion whenever the topic 32bit float is discussed:

• • Any AD or DA converter always has a fixed bit depth. 
  • There is no floating point converter.
  • There are some 32bit FIXED converters now available that people get confused with 32bit float
  • Why the need for 32bit fixed, I don't understand (beyond bragging rights and marketing), but that is a different discussion.
  • 32bit float is used for internal signal processing and as a file format (especially when delivering to Master Engineer)
  • 32bit float (Single Precision) uses technically only 23bit for the encoding (Mantissa), 8 bits are for the exponent and 1bit for the sign bit
  • Going to 64bit float (Double Precision) provides a 52bit Mantissa with 11bit for the exponent
  • Virtually all DAWs now use 32bit float and/or 64bit float for processing and/or summing
  • Logic lets you choose to use 32 or 64 for summing in the Preferences > Audio > Devices

About the Zoom F6. 

I think it is really irresponsible from their marketing department to claim "The Zoom F6 is the first professional field recorder to feature both 32-bit float recording". The tricky part is the word "recording". What I guess they mean is, they save the recorded signal as a 32bit floating file (which is great). What they falsely imply (intentionally or unintentionally) is that their converters are 32-bit float. However, if you look at their spec sheet, it says 16bit or 24bit.

So if you clip the converter (above 0dBFS at 24bit fixed), then you have a clipped signal and saving it as a 32bit float file will not magically remove the clipping distortion (unless I'm missing something here).

Hope that helps,

Edgar Rothermich

Graphically Enhanced Manuals (GEM)

www.LogicProGEM.com

https://www.youtube.com/c/MusicTechExplained/

Well...

I see this as more a comment on the features and quality of modern ADDAC recorders than about bit depth.

The 32 bits "debate" is only about the interface between the physical and the digital worlds. To understand how futile it is, just try to translate a 32 bits depth in electromagnetic physics. Have fun...

WD

The weird things are:

1. Logic's Freeze-Files are 32-bit.

2. In Logic, you are able to export tracks as 32-bit files.

And just to consider, more and more hardware is able to handle 32-bit audio files. Many mastering studios ask for 32-bit audio these days.

It's an obvious question: why can't Logic play 32-bit files? Even QuickTime Player is able to play them.

I understand that. Inside the DAW the bit depth is important.

My feeling is that 32 bits thing is more marketing for now. Why not 64 or 128 bits? Most of current 32 bits files are just 16 bits or 24 bits padded files.

In the real physical world this is another story. Anyways...

I believe Logic will be able to work with 32 bits files at some point in the future.

WD

Interesting.

I tend to agree about the dynamic range. It's just there are so many other factors involved. Characteristics of the sound wave, electromagnetic constraints...

Don't get me wrong, I see a use for higher bit depth. But as most people listen to MP3 and see vinyls as Hi-FI... 😉.

WD

It's always worth baring in mind that 32 bit float is not, and was never meant to be, a destination or output format.

As you say, it's really good for accidents and unexpectedness in location situations and transfering audio unscathed from one platform to another but any peaks exceeding 0dbfs need to be brought back under control before outputing to a general playable format (i.e. fixed bit depth).

I also agree it's a useful format and it would be massively useful if Logic could import 32 bit float files natively. Accidents and unexpectedness aside, it shouldn't be an excuse to forget about proper gain staging which is still a big part of getting a fantastic recording and mix. All the great recording and mix engineers talk a lot about good gain staging.

Interesting. I didn't expect this discussion going forward.

It is all about quantizing the real world and keep fidelity in digital transformations. This reminds me of the field of Information Theory I studied decades ago.

WD

Great post Edgar!

I think I was too lazy to get in those details. Thanks for doing it.

WD

Thanks for the discussion, educational and even a bit entertaining.

Cubase has had 32-bit FP support since at least 2005 but Logic's devs never seemed intrested in 32-bitFP support.