Will 32-bit converters take over soon?

[skunky_funk]

I have always been a perennialist when it comes to recording. I've always thought that it is not the recording medium that really matters, but the source, mics, preamps and monitors. Just take the past 10 years, and you'll see a very vast and quick change in trends in recording consoles.

 

Anyway, since 64-bit OS's are on the rise, I think 32-bits are probably next in line. But will you need them?

[Phil O'Keefe]

Is anyone even making a 32 bit A/D - D/A converter? And if they did, would they get anywhere near 32 bit performance out of it? Most 24 bit converters only have a 110 - 115 dB S/N ratio at best, out of a possible 144 dB range. 32 bit would be about 192 dB - far greater than a human can deal with... and then there are issues of thermal noise - molecular noise is going to pose significant issues for getting anything past 130 dB or so...

 

No, I don't think 32 bit converters wil be hitting the market any time soon, and even if they do, it's not going to be an easy "sell" to convince me of their merits. But I'm willing to listen.

[kptkarl]

I heard a terra-bit converter the other day at the local ID 10 T conference...WOW! Blew me right out the back door.

[UstadKhanAli]

So what does it mean when someone says that they are recording in 32 bits if no one is making a 32-bit converter?

[Zooey]

 

Originally posted by UstadKhanAli So what does it mean when someone says that they are recording in 32 bits if no one is making a 32-bit converter?

 

 

Good question. When you get into processing and mixing, there may be some benefit (for example, saving a pre-master mix as a 32 bit file). But for tracking? I think it just means you're wasting disk space.

[kylen]

True - if I put a 32-bit bitviewer in Sonar5 only 24-bits of the 32-bit word are used untill I either insert a 32/64 bit plugin or change one of the track sliders so a summing calculation is made. Assuming a native 24-bit source file of course...

 

In Sonar5 the busses operate at 64-bit float but I don't have a bitviewer that deep so I haven't "seen" it yet!

[blue2blue]

You might want to take a look at this thread at another site:

 

http://gearslutz.com/board/showthread.php3?t=47234&page=3

 

I linked to the last page in the thread because, while all the pages are probably pertinent, the discussion tightens up around issues of just how much input converter resolution is necessary as it goes along. (You might want to start scanning at the top and focus in as Dan Lavry and a couple of other knowledgeable types kick in.)

 

Lavry (yes, that Dan Lavry, the high end converter maker) makes a pretty convincing case that 24 bit converters will offer all the rez you'll need in anything approaching the forseeable future (while stipulating that, for processing, mixing, storage, etc, enhanced resolution generally helps and can't hurt). In fact, according to Lavry, 20 bit A/D is probably sufficient.

 

(I have to admit that Lavry's discourse challenged my ability to keep up, so I've got to do some more homework if I'm going to work through all the details. But it mostly sounded reasonable to me. And, you know, the guys is somebody. )

[blue2blue]

 

Originally posted by UstadKhanAli So what does it mean when someone says that they are recording in 32 bits if no one is making a 32-bit converter?

 

 

 

I would tend to think that what they're actually talking about is saving their DAW's files in 32 bit float mode.

 

There seems to be some confusion or disagreement as to the benefits, but as I understand it, saving a DAW project to 32 bit float allows individual tracks to keep longer word lengths that may have resulted from internal (destructive) processing so that, on reloading, you haven't lost that extra resolution as you might when saving project tracks in a standard fixed 24 bit format. At least I think that's the reasoning.

[Phil O'Keefe]

but as I understand it, saving a DAW project to 32 bit float allows individual tracks to keep longer word lengths that may have resulted from internal (destructive) processing so that, on reloading, you haven't lost that extra resolution as you might when saving project tracks in a standard fixed 24 bit format. At least I think that's the reasoning.

 

Just a minor correction... you can't save a file as "32 bit float"... you can save it as a 24 bit file, or a 32 bit file, but not as a "float" file. Floating point pertains to the mathematical / computational processing used on the file, not the actual file itself. And as Kylen mentioned, nothing is "in" those last 8 bits of a file recorded at 32 bits (software setting for file) when it was recorded through a 24 bit converter except zeros... but once you process it, then those last 8 bits start getting used.

[Zooey]

 

Originally posted by Phil O'Keefe nothing is "in" those last 8 bits of a file recorded at 32 bits (software setting for file) when it was recorded through a 24 bit converter except zeros... but once you process it, then those last 8 bits start getting used.

 

 

True, but you don't have to start with a 32 bit file to reap the benefits of 32 bit floating point processing. The last 8 bits only really get "used" when you save the result as a 32 bit file, not when you process it.

 

If you use real-time effects, you get the benefit of those last 8 bits by virtue of the mix engine being 32 bit floating point, even if your session is composed of entirely of 24 bit tracks.

[kylen]

New in Sonar5...

You can now play files of different bit depths in the same project. You can choose bit depths for playback, recording, importing, exporting, and processing (bouncing, freezing, applying effects). 32-bit support for these functions gives you full floating-point resolution for greater dynamic range and more precise editing.

They're refering to 32-bit float files. Both Adobe Audition and Sound Forge already had 32-bit file support. I think Sound Forge does both 32-bit integer and 32-bit IEEE Float - I found out the hard way they're 2 different things and not compatible, hehe...

 

If you can run VST plugins it's a lot easier to see where the 24bit stuff changes to 32-bit float with this free bit scope:

http://www.kvraudio.com/get/1654.html

 

As far as converters go I would tend to believe that 24bit converters are pretty much adequate, for me at least. I kinda tested this (unscientific subjective kinda thing) by playing one of my favorite 16-bit reference CD's thru my monitoring system and it sounded spectacular - we all do this every day no big deal. Then I recorded a favorite vinyl thru my recording chain to hard disk and played it back. Even that fidelity kicked anything I've done so far. That implies to me that my recording/playback preamps, monitoring, and A/D/A aren't the weak link in my home studio... Somebody had to tell me - I guess it may as well be me!

[blue2blue]

Originally posted by Phil O'Keefe but as I understand it, saving a DAW project to 32 bit float allows individual tracks to keep longer word lengths that may have resulted from internal (destructive) processing so that, on reloading, you haven't lost that extra resolution as you might when saving project tracks in a standard fixed 24 bit format. At least I think that's the reasoning. Just a minor correction... you can't save a file as "32 bit float"... you can save it as a 24 bit file, or a 32 bit file, but not as a "float" file. Floating point pertains to the mathematical / computational processing used on the file, not the actual file itself. And as Kylen mentioned, nothing is "in" those last 8 bits of a file recorded at 32 bits (software setting for file) when it was recorded through a 24 bit converter except zeros... but once you process it, then those last 8 bits start getting used.

 

Funny you should say that, Phil.

 

That, too, was my thinking. And I had started to post something somewhere correcting someone talking about saving 32 bit float files...

 

But I thought, maybe I just oughta check that before I leave that post.

 

And doing some quick googling I found things like this:

 

SONAR 5 now supports the importing and exporting of 32 bit floating-point files. In short, 32-bit floating-point files can more accurately represent very large or very small numbers. This accuracy provides better quality sound reproduction, extended dynamic range and more precise audio processing.

 

And I thought to myself, yeah, I guess you could save a 32 bit FP number just the same as you could save 32 bit absolute and it would make plenty of sense if you were processing in 32 bit FP.

 

So... not saying you're wrong but I've been dissuaded from my former thinking on this.