dither to 24 bit?

Jah Servant · Post by **Jah Servant** » Sun May 04, 2008 2:32 pm

I'm wondering when mixing down a track from cubase through scope, and recording in wavelab to 24 bit, should I add 24 bit dither in scope? I'm wondering because I was reading in Mastering Audio by Bob Katz that you should dither after any daw processing because the processing is done at a higher bit rate (he specifically says even when recording from 24 bit to 24 bit). I hadn't really thought of this before but it makes sense. Anybody do this? Should I do it? The dither noise level is so low at 24 bit it really wouldn't ever be a problem anyways.

Post by **garyb** » Sun May 04, 2008 2:36 pm

dither doesn't always sound better. you should only use dither if it sounds better. the ear is the only test.

alfonso · Post by **alfonso** » Sun May 04, 2008 2:56 pm

I imagine that if you are ending anyway at 16bits, doing it only then will cover all the necessities....if your final product is at 24bit, it makes sense.

firubbi · Post by **firubbi** » Sun May 04, 2008 9:51 pm

is DVD support 24bit audio @48khz?
thanks

arela · Post by **arela** » Sun May 04, 2008 10:22 pm

DVD-Video is 24/48. DVD-Audio is 24/96 i think.

Jah Servant · Post by **Jah Servant** » Mon May 05, 2008 3:12 am

Alfonso, actually, according to the dither chapter in the book, you can't correct it afterwards. Missing dithering can cause some distortion and you can't correct that by dithering afterwards. I guess Garyb is right, if it sounds best then do it.

Bob Katz is saying that if your processing is done at 32 bit, even though your files are 24 bit, you should dither to 24 for your new 24 bit file. It seems that wordlength often increases in daw processing. He says even if you are using a 16 bit file it will come out to be 24 bit afterwards and will have to be dithered again. Of course it doesn't improve the sound by going to 24 bit, but it just spits it out as 24 bit. Then you need to dither to avoid the truncation and distortion when going back to 16 bit for cd. Regardless of that I'm talking about 24 bit to 24 bit. If I understand correctly he's saying that if I have a 24 bit file and I am taking that to eq it then I should dither after the eq to get an untruncated 24 bit file again. Because the wordlength increases during the processing, say to 32 bit. So it got me thinking about mixing to stereo and the need for dither there.

alfonso · Post by **alfonso** » Mon May 05, 2008 5:28 am

Jah Servant wrote:Alfonso, actually, according to the dither chapter in the book, you can't correct it afterwards. Missing dithering can cause some distortion and you can't correct that by dithering afterwards. I guess Garyb is right, if it sounds best then do it.

Bob Katz is saying that if your processing is done at 32 bit, even though your files are 24 bit, you should dither to 24 for your new 24 bit file. It seems that wordlength often increases in daw processing. He says even if you are using a 16 bit file it will come out to be 24 bit afterwards and will have to be dithered again. Of course it doesn't improve the sound by going to 24 bit, but it just spits it out as 24 bit. Then you need to dither to avoid the truncation and distortion when going back to 16 bit for cd. Regardless of that I'm talking about 24 bit to 24 bit. If I understand correctly he's saying that if I have a 24 bit file and I am taking that to eq it then I should dither after the eq to get an untruncated 24 bit file again. Because the wordlength increases during the processing, say to 32 bit. So it got me thinking about mixing to stereo and the need for dither there.

My only doubt is if what's happening at the very low end of a 24 bit word is of any meaning when you go to 16 bits. I think the dithering noise applied in the last stage is much louder and that resolution doesn't exist any more. But well, Bob Katz is more right than me in any case. I'm not sure that what he said applies to something that will end only to 16 bits. Probably a good reason is that you never know if you will need the higher resolution one day to be delivered as such, in that case is better to do as he suggests.

Post by **garyb** » Mon May 05, 2008 9:15 am

technically, he's right, of course.

i just don't think a dithered file always sounds better than not. really, if the track is the typically undynamic stuff of pop music, the dither does nothing of value except add a slight amount of noise. the truncated bits are useless information anyway. for very dynamic music, dither might be essential, if the resources are available. to put dither on every effect, eq, mixer channel(the mixer operates at a VERY high bitrate!), on EVERY change of bitrate would be ridiculous. who's going to calculate the latecy/phase issues?

sometimes some of these guys go a little bit crazy.

your ear will tell you that sometimes, dither sounds worse than truncation. truncation is NOT bad, necessarily! dither is noise! noise is garbage(sometimes

).

use it if it sounds better. otherwise, don't even worry about it at all...

Jah Servant · Post by **Jah Servant** » Mon May 05, 2008 9:47 am

ok, I actually wrote Bob and he replied and says:

"The sonic difference between truncating a 32 bit float signal (as in Cubase) to 24 bit versus dithering it to 24 bit is almost inaudible! And for all practical purposes it is insignificant. Routinely I do dither 32 float to 24 and recommend it as a safety measure, but the last time I performed a blind listening test on it, it took 15 minutes of careful scrutiny to discover there MAY BE a very slight "warming or deepening" of the sound. But considering cumulative losses, it's also a good thing. But don't lose any sleep over it!"

So there is his answer. It's just what I read in the book made me start to think and although I know it'd be a small difference I was really curious as to what might be best in this situation. So it's possibly not making any difference to the sound at all and even if you do dither, the noise level is so low, -141db at 24 bit I think, that it's inaudible.

Post by **irrelevance** » Tue May 06, 2008 12:17 am

This is why I love this forum! So much good quality information. Thanks guys.
I have been thinking that rayjer than dithering from 24 bit to 16 bit. That I could record in real time to VDAT at 16 bit. I have not tried this yet but this has to be better than adding random noise to the mix!?

Post by **garyb** » Tue May 06, 2008 12:28 am

higher bit depth=headroom.

if you are recording live instruments or voices, use 24 or 32bit. you can then dither or not. it's not a big deal. the file sizes don't become stupid and there's really no penalty, compared to doing sample rate conversions

tgstgs · Post by **tgstgs** » Tue May 06, 2008 12:41 am

if the use of a dithering algo after a wavemanipulating algo makes sense,
you should have knowledge how that algo manipulates and what dithalgo is used to apply;

if not _ well _ i dont know what im doing but i dont dither of doing well you know _

dithering vibes

Post by **garyb** » Tue May 06, 2008 12:55 am

yes.

if some of the most well trained ears on the planet in the best listening environment need 15 minutes to be able to discern a difference, shakily, it's just not that critical. there are many more places to improve your production before worrying about this type of minutia, beginning with the monitoring environment....

Jah Servant · Post by **Jah Servant** » Tue May 06, 2008 3:21 am

you are right Gary, but it's always nice to have the info, you know the saying "the more you know is the more you know"
But for this particular matter it is not going to make a difference for anybody most likely, I am probably making enough other mistakes to mask over anything gained by this small practice.

Post by **garyb** » Tue May 06, 2008 10:59 am

Jah Servant wrote:you are right Gary, but it's always nice to have the info, you know the saying "the more you know is the more you know"
.