dither from 16 to 16

Is it advisable to use dither when my source file is already 16 bits when burning to CD?

Thanks

read the Izotope Mastering guide. It will answer this question… but long story short… dither is used to mask quantization noise created by converting between bit depths. Adding dither to a file that is not being converted to another bit depth is just adding noise.

Thank’s bubba for the quick note. Be checking the izotope.
Thanks Man!

What Bubba says is correct but I bet you interpret it to mean that if you make a mix of 16 bit files, there’s no need to dither, and that’s not the case.

If you mix two 16 bit files, you get at least 17 bits of significance – and that’s assuming no FX.

Generally, if you’re doing ANYTHING to a 16-bit file in n-Track, it’s best to assume that the result is a 32-bit floating point file, which has 25 significant bits (plus exponent bits to make up the other 7). Therefore, dithering is appropriate.

And actually, Bubba’s post is correct but not complete.

Dithering when reducing bit depth is important and is the most common use of dithering. But it’s not the only one.

Dithering should also be used when converting to higher bit depths, to help mask the quantization noise in the original recording. There are two schools of thought here. The minimalist approach is to dither only the new low order bits, to put white noise where there would be nothing but quantization noise. (Yes, it’s funny to think of all zeros as being part of noise.) However, the minimal approach ignores the psychological aspects of quantization noise versus white noise. Q noise is mechanical sounding and irritating. White noise is natural sounding and our brains are very good at filtering it out. Therefore, many think the best approach is to dither considerably more than the original low order bit in order to mask out the nasty Q noise with nice white noise. This is what I would do if presented with an old 12-bit recording to master for general listening purposes. (No, 12 bits was never an audio industry standard, but it was an A/D converter standard, and it’s possible that there are audio recordings made that way … but yes this is an academic point rather than a practical one in most cases.)

We “dither down” at the end of any mix, as I mentioned above. Now, when would we ever “dither up”?

Well, let’s say we made a recording that peaked at -12dB and pushed the fader up so that it peaks near zero – or even more with compression. That’s essentially “dithering up”. For 24-bit tracks, there’s already plenty of low order info, and Q noise is way in the weeds compared to the noise we recorded in the signal (ambient noise, mike noise, mike preamp noise, etc.) But if someone made a 16-bit recording using an analog signal chain with noise below -90 dB or so, in an extremely quiet studio, well, then in this case it might make sense to “dither up”.

In general practice, though, there’s rarely a need to dither up. Just something to keep in mind if you ever find yourself pulling a signal out of the weeds, one that had been recorded at way too low a level and can’t be done over.

Quote (Bubbagump @ Nov. 15 2005,23:13)
dither is used to mask quantization noise created by converting between bit depths.

Err, I done said it… just not very verbosly. Didn’t specify converting up or down. :;): Here’s a link to Isotope’s guide to dither BTW:




Link

Eyup!

<!–QuoteBegin>

Quote
Dithering should also be used when converting to higher bit depths,


I had always thought that dithering should be the last process before burning to cd.
I record at 4800 and dither when converting down to 44.1 before burning.
In the meantime, if I want to do any processing in cool edit, I always convert to 32 bit float, but without the dither.
I am unsure now as to if I should dither when converting to 32bit float ????
Beefy is confused ???

Steve

Well, it depends… If you are working on source tracks in CoolEdit, I don’t see much reason to convert the bit depth at all. If you are mixing down… then the mix down to 32 bit has no conversion. The conversion happens in playback when the software has to truncate from 32bit float to whatever the soundcard can handle. So my advice… leave the source tracks alone, mix down to 32 bit. Do all your mastering type stuff (compress, EQ, whatever…) Convert sample rate to 44.1, dither/truncate to 16 bit.

The thing to remember is that N track takes a 16 or 24 bit source file, sticks a bunch of zeros on the bottom of it as it slurps it into the mix buss to make it 32 bit for processing, adn then chops off the zeros it added to accomodate the sound card.

so even when changing sampling freq. we must dither?

<!–QuoteBegin>

Quote
I am unsure now as to if I should dither when converting to 32bit float ?
Beefy is confused


Welcome to my world Steve. :(

Personally I don’t believe in extramarital dithering, but I am from the midwest.

Quote (dean russell @ Nov. 16 2005,20:57)
so even when changing sampling freq. we must dither?

No. Read what I said again…



<!–QuoteBegin>
Quote
dither is used to mask quantization noise created by converting between bit depths.


That is all there is to it. Sample rate is not in that sentence. :D And there is no must in dither. Please read that guide I sent you. It is very complete and will make this very clear.

Sample rate conversion has a comparable process called anti-aliasing… but it ain’t dither and works on a different principle.

Hey Dean,

I was searching the forum and I saw that you asked a similar question about dithering about a year ago:

http://www.ntrack.com/cgi-bin…=dither.

The theory can be a bit daunting, but if I were you I wouldn’t worry about what you should do “in theory” but trust what your ears tell you. It’s cool to understand all this stuff, and to try to get the best from what we have, but at the end of the day it’s all “just noise”. Sometimes people like it, sometimes they don’t. :)


Mark

Quote (Beefy Steve @ Nov. 16 2005,16:27)
Eyup!

Dithering should also be used when converting to higher bit depths,


I had always thought that dithering should be the last process before burning to cd.
I record at 4800 and dither when converting down to 44.1 before burning.
In the meantime, if I want to do any processing in cool edit, I always convert to 32 bit float, but without the dither.
I am unsure now as to if I should dither when converting to 32bit float ????
Beefy is confused ???

Steve
You're doing the right thing, Steve, if you're starting with 24-bit tracks. The difference in precision between 24 bit FIXED and 32-bit FLOAT is between 1 and 12 bits, depending on the signal value ... blah blah blah but bottom line is fuggedaboudit.

For processing 16 bit tracks in any nontrivial way, an argument can be made that dithering the zero bits is best. However, I'd have to do some careful studying to figure out what value to plug into N's "bits" box to get the ideal behavior.

BTW, I should have said " ... dithering CAN be used ...". The situations where dithering UP is important are relatively rare and definitely technical. Things to pay attention to when your mixes are already clear and open and beautiful sounding, but you're doing some unusual step for some reason on certain tracks or submixes.

Just to be terribly nit-picky, Bubba, you did mention noise "created" by changing bit depth. The only time noise is created by changing bit depth is when dithering down. So, strike the word "created" and we're good. (OK, flimsy excuse, and I stand corrected!)

Bottom line here is that Bubba's post is the rule of thumb to remember, and I was just covering some unusual cases, and causing more confusion than help in the process.

Good point, Mark. I like to fuss with the theory, but in practice I almost never dither except as the final step.

I have an untested hypothesis that a mix of all 16-bit tracks would sound better if all were dithered to 24 or 32 bits before mixing. But frankly, I doubt my engineering is good enough where that kind of subtle improvement would be noticeable.

When experts talk about the difference in quality between mixes of 16 and 24 bit tracks, they’re talking about pro quality mixes. On most amateur-level mixes, I doubt it matters much, and so the biggest benefit to recording in 24 bits is more headrooma and far less worry about optimizing the digital media (e.g., being sure to peak above -3dB when recording, carefully calibrating the analog chain to match the digital range, etc.)

Learjeff, you know I am just being pedantic. :)