Topic: Is 24-bit AIFF really 32-bit?

[mfisch]

So I've got this edirol fa-101, mac mini w/ core audio, and cubase se recording into AIFF files.

I guess I thought this was 24-bit audio since the edirol says 24-bit on it, and so does the cubase meta information.

Upon replay, running a VST plugin that counts data bits in an audio stream ... it really looks like 32-bit data. This makes sense really, to ensure even wordcounts when processing in a computer -- but whats really going on here?

Is all my audio really 32-bit? If I archive this to flac/24 is the compression I'm seeing only due to the fact the extra 8 bits have been truncated?

[nic]

its possible the app your using works in 32bit float files.

[mfisch]

I'm sure it does, and the data coming from my AD apparently does too (because theres actual data after the 24-bit mark).

Anyone else have experience here? I guess theres no chance of my flac'ing these without losing that extra data huh?

[mfisch]

Refreshing this thread with an update (since I got no real answers last time).

I guess I need a real hardware guru to answer this one. I'm not debating the utility of high-resolution audio recording here, this is strictly a detached technical discussion.

I record audio with my "24-bit" (at least thats whats painted on the side) Edirol FA-101 over firewire 400 to a mac running cubase se.

So:

Edirol FA-101 A/D (24-bit?) --> 1394 400 (32-bit) --> Mac OS X (32-bit) --> Cubase SE (32-bit)
Everything supports 32-bit float except the A/D (supposedly).

  Now I found this neato plugin (that escapes me at the moment). It simply decodes the digital values of each audio frame, and lights up 32 lights on a colored strip when they change inter-frame. The resulting effect is a visual representation of the bit-depth of the audio being forced through it.
  When I play my FA-101 recordings (32bit float files remember) I get 30 to 31 bits of detail on average. This is in contrast to a "24-bit flac" recording which indeed shows 24-bits, and a CD audio file which properly displays 16-bits of detail.
  This seems to indicate my files contain almost (if not completely) 32-bits of detail. so ... is my Edirol likely decoding at 32-bit and edirol is just advertising 24-bit for product differentiation purposes? Maybe I'm just seeing noise on the 1394 bus?

comments/questions?

[Steve J]

I can't answer this directly; but I seem to recall that a similar discussion came up regarding how Cool Edit Pro/Adobe Audition deals with files while editing, i.e., the program's "native" format is 32-bit float or some such. That's the best I can recollect.

[mfisch]

My question is really hardware specific to the Edirol FA-101, but I wondered if anyone had done similar research on other USB/1394 devices.

I'll try asking yamaha but something tells me:

A) I cant talk to anyone who has a clue.
B) People who have a clue's boss doesnt want me to know I have 32-bit hardware.

anyone here work for yamaha? =)

[Brian Skalinder]

comments/questions?

I think Cubase and the plug-in are simply operating in the 32-bit realm.  I suspect Cubase is like CEP/AA in that it does not have a native 24-bit file format.  Rather, it opens 24-bit files in a 32-bit float workspace.  If one saves the file with default settings, it saves as 32-bit float, even if it includes only 24-bits of audio.  In such a case, there are, in fact, 8 extra bits of information in this case - but it just so happens all the information = padded zeroes, i.e. no meaningful content.  CEP/AA will only output the actual 24-bits of audio content if one specifies the proper Save As settings before saving the file.  I think the plug-in cannot distinguish between the 24-bits of actual audio content and the 8 extra bits of filler, so it reads the file as 32-bit.

Then why don't your 24-bit FLAC-sourced files also read as 32-bit?  I'm betting because you simply opened these files in Cubase, but did not then save them in the 32-bit float workspace as you do your FA-101 sourced files.

Wanna test this?  Open the 24-bit FLAC-sourced files in Cubase and use the same Save As options with those files as you do for your files recorded from the FA-101. 32-bit files.  I bet the FLAC-sourced files then, too, appear as 32-bit files to the plug-in (even though they're 24-bits of audio + 8-bits of filler).

[mfisch]

comments/questions?

I think Cubase and the plug-in are simply operating in the 32-bit realm.  I suspect Cubase is like CEP/AA in that it does not have a native 24-bit file format.  Rather, it opens 24-bit files in a 32-bit float workspace.  If one saves the file with default settings, it saves as 32-bit float, even if it includes only 24-bits of audio.  In such a case, there are, in fact, 8 extra bits of information in this case - but it just so happens all the information = padded zeroes, i.e. no meaningful content.  CEP/AA will only output the actual 24-bits of audio content if one specifies the proper Save As settings before saving the file.  I think the plug-in cannot distinguish between the 24-bits of actual audio content and the 8 extra bits of filler, so it reads the file as 32-bit.

Then why don't your 24-bit FLAC-sourced files also read as 32-bit?  I'm betting because you simply opened these files in Cubase, but did not then save them in the 32-bit float workspace as you do your FA-101 sourced files.

Wanna test this?  Open the 24-bit FLAC-sourced files in Cubase and use the same Save As options with those files as you do for your files recorded from the FA-101. 32-bit files.  I bet the FLAC-sourced files then, too, appear as 32-bit files to the plug-in (even though they're 24-bits of audio + 8-bits of filler).

I couldnt really figure how to read the results of my test .... Maybe you can give it a go and tell me what you think its doing?
The plugin is available at http://www.tobybear.de/p_utilbag.html. Weird that Toby claims there is no mac port, ... Im using it. *shrug* as to where it came from =)