32 bit and dither

[tonzilla]

does logic dither, or should i say, how does logic dither from it's internal 32 floating point bit depth back to the output of 24 or 16.

does anyone know and how would it differ from Pro tools, cubase ect????

[David Nahmani]

I've never used dithering from 32 bit floating point to 24 bit fixed - I don't understand why it should be necessary or useful given the huge dynamic of 24 bit files.

 

Now when bouncing, you can choose to dither using a few options:

[tonzilla]

i see your point about 24 bit. the reason i was asking is because i was reading in Bob Katz mastering book about the importance of dither and noise shaping anytime a word length is lowered from its internal DSP word length. he was saying if i followed, it's the difference in how one daw relates, sound quality wise to another. i've also heard different people talk about the difference of logic's sound compared to pro tools and say logic is superior because of it's higher internal headroom ie bit depth. but after reading katz book i started wondering if it may have more to do with the way it's dithered and possibly noise shaped.

interesting.

t

[David Nahmani]

Well the audio engines are completely different so of course they'll sum audio differently, etc.. and they will sound different. But this has little to do with dithering.

 

And again, I don't understand noise shaping when working with 144 dB of dynamic. Maybe there's something I'm not getting? But if my quantization noise is at -138 dB FS I don't see the point in trying to mask it.

 

Anyway try this: take your most complex mix, bounce to 24 bit PCM. Now try the same bounce with one of the dithering algos. Can you hear a difference?

[tonzilla]

yeah it makes sense to me what your saying -138 db probably not going to hear. i think he was talking about the automatic dither going on behind the scenes and there may not be any noise shaping. i may have added that erroneously. but he was definitely not talking about bouncing tracks that i'm familiar with. any thanks for your insight.

[Larry Mal]

Well, there's no automatic dither. It may be checked by default on some programs, but you should always know if you are dithering, because that is pretty important.

 

Basically, what happens is there are rounding errors when going from a higher bit depth to 16 bit. Dithering makes those errors random, and less noticeable than outright truncation (throwing away) the extra bits. Instead of a clean slice, which is noticeable, it's an obscured slice.

 

And as far as I know, the "rule" is, you only dither when going to 16 bits. So otherwise, don't dither, and don't sweat it. Unfortunately, 16 bit was the best they could do in 1982, and somehow that standard by way of the popular compact disc is still the standard. So it has to be accommodated by dithering when going to the 16 bit CD standard. If the world had embraced 24 bit media, we wouldn't be having this conversation today.

 

Hope this was of any help, L

[Bluemoa]

The UV22HR option is grayed out for me, im assuming this only works if u have apogee hardware? I remember that Cubase has a software VST version of the dither.

[Eric Cardenas]

The UV22HR option is grayed out for me, im assuming this only works if u have apogee hardware? I remember that Cubase has a software VST version of the dither.

 

Change the bit rate to 16!

[Holger Lagerfeldt]

The UV22HR option is grayed out for me, im assuming this only works if u have apogee hardware? I remember that Cubase has a software VST version of the dither.

Just leave it greyed out, you're not missing out on anything... POW-r#1 or #3 is a better choice IMO.

Edited September 29, 2009 by lagerfeldt

[Holger Lagerfeldt]

i see your point about 24 bit. the reason i was asking is because i was reading in Bob Katz mastering book about the importance of dither and noise shaping anytime a word length is lowered from its internal DSP word length.

Dither is especially important when reducing to 16 bit but could be applied at any time you're reducing word length. However, adding multiple instances of dither could veil the sound slightly, and especially if it's colored dither or noise shaping instead of TPDF (flat dither).

 

The benefits of adding dither from say 32 bit float to 24 bit fixed is very theoretical.

[fader8]

the reason i was asking is because i was reading in Bob Katz mastering book about the importance of dither and noise shaping anytime a word length is lowered from its internal DSP word length. he was saying if i followed, it's the difference in how one daw relates, sound quality wise to another.

 

With certainty, there is a small amount of dither, probably a flat TPDF, used by Logic whenever it truncates its 32 bit internal resolution to 24 bit. This is just good engineering practice and reduces rounding errors and other issues downstream. You may have been misinterpreting what Bob wrote regarding re-dithering and cumulative dithering to 16 bit, but Logic almost certainly disables this when bouncing to a 16 bit Pow-r dithered file, for example.

 

Note that Logic doesn't offer dither as a plug-in, but only during this bounce process. By doing this it avoids the problem of the audio stream leaving the dither plug at 24 or 16 bit, then having the wordlength expanded to 32 bit again, then re-dithering and truncating to 24 bit to talk to the hardware driver.

[tonzilla]

to clarify. this post is not really about bouncing files although related.. i was originally asking if any body knew whether Logic handled the internal DSP of 32 bit back to say 16 bit ( if you were working at that depth) with dither or truncation. as David pointed out earlier it's not an issue at 24bit because the quantization noise produced at 24 bit would be so low ( -139) I think Fader 8 answered my question. thanks! I would have to guess all daws will do this or there would be a lot of degradation working at 16 bit. god forbid

[tonzilla]

how do like the duende fader 8

[fader8]

to clarify. this post is not really about bouncing files although related.. i was originally asking if any body knew whether Logic handled the internal DSP of 32 bit back to say 16 bit ( if you were working at that depth) with dither or truncation. as David pointed out earlier it's not an issue at 24bit because the quantization noise produced at 24 bit would be so low ( -139) I think Fader 8 answered my question. thanks! I would have to guess all daws will do this or there would be a lot of degradation working at 16 bit. god forbid

 

But it's never "working" at that bit depth, so I'm not sure I understand your question.

 

As for Duende, I love the X-Comp with its "Bleed" controls for parallel comp.