Dithering Question

I think this may start a huge thread, so I feel kinda special to be the first to answer this. Here's the process I've learned which is the workflow I use every time without exception:

Dither only

1. *when changing bit-length*. For example, if I record at 24 bits, and I want to send someone a rough mix on CD, I first select the 16 bit resolution box, (CD audio is 16 bit.) then I'll usually select POWr3 dithering. POWr 1,2, and 3 are all basically a kind of "noise" which needs to be added in order to dither something.

2. *at that last step*. For example, sometimes I have a chain of plug-ins that each have the ability to dither, but I only use the last plug-in on the main outs to dither. This is because all of the plug-ins up to that point don't mess with the bit-length, so there is no reason to dither. (And again, because dithering adds a little bit of noise, too much dithering will raise your noise floor too much.)

There seems to be a lot of arguments about how and when to dither for the best results, but I was "sold" into the method I described here by my ears and by things I've learned about dithering, and I feel I have been able to get good results. Good luck to you!

I would personally only dither at the final bounce.

If you are rendering individual audio files back to 16 bit files I would consider it (say you are outputting stems for an OMF project) and of the three algorithms I would use Pow-R #2 if I felt I had to do so based on these graphics on Pow-R

http://www.weiss.ch/pow-r/documents/pow-r_brochure.pdf

If you are working at 24 bit mode I think these issues are less critical as dither works way down in the audio amplitude range and so the noise is much lower into the noise floor . The graphs show noise between 100 and 110 dB down from full amplitude.

POW-R is specifically designed to decimate from 24 bit down to 16 bit so if you felt the need to do a lot of individual dithering (like in the OMF post example) but were bouncing back to 24 bit I would look at another algorithm like Waves IDR or something similar. Really even TPD is not bad if it is 120 db down in the mix...

Izotope MBIT is the best dither. You can get it through the Izotope Ozone plug, or for cheaper, you can use Sample Manager 3 (not a plug-in) which also has the MBIT dither as well as the Izotope sample rate converter, which is also great.

Dithering adding noise

it is only required when you need to burn an Audio CD...

Never use dithering in the Bounce if your Mix require Mastering... or just nned to be added on a CD Compilation.

I suggest to Bounce at 24 bit everithing without use dither (you can also make a Mp3 for sending to friends by Mail...

BUT... if you need to bounce and convert from 48/24 or 96/24 to 44.1... Maybe a powr1 should be required...
in any case dithering is the very final step of your MIX... Don't use Dither is always a good idea in order to avoid a DOUBLE OR TRIPLE DITHERING processing...

more Dither = more Noise!

Remember All modern CD Audio Burning softwares make dither processing by defaul setting (see Toast.. or NERO (windows vista) Audio Preferences)

The possibility of having a double dither processing is really high and it is very dangerous for sound quality!

(my personal opinion is... I work on Mac with logic... I always play my files at 24 bit without Dither... sounds better)
I think is better to don't use Dither until your song will be burned on the final CD... (I'm not sure but also when you send CDs for duplications... they do dither again!!!)

fermusic

Fermusic is entirely correct. Dithering is necessary only for going from 24 bits to 16 bits. From 32 floating point to 24 bit dithering is not required. Only for going to 16 bit.

So, if you are going to have your product professionally mastered, give them 24 bit audio files. Let them dither, if even then. If your product does not leave the realm of a computer, keep it at 24 bit. If it goes to SACD or another 24 bit audio device, Blu-Ray or whatnot, keep it at 24 bit. Keep it at 24 bit until you absolutely have to go to 16 bit.

Basically, 16 bit is the technology of the late 70's still hanging on. The popularity of the CD has kept it alive longer than it probably should have been, but then, it is a pretty good format. And consumers are showing with the popularity of Mp3's that they don't care about sound quality very much at all.

But you still should. Only dither once, and only when you absolutely have to. What dithering does is add randomly generated noise that covers up the rounding errors from having to truncate from 24 bits of depth to 16 bits. 24 bits of measurement is much, much more that 16 bit- that isn't necessarily obvious. And when you have to fit all that measurement into a smaller grouping of numbers then error occurs.

Think of it like this: the number 1.40719, well, if you had to round it, what number is it? You can call it 1.5 if you wanted, but it isn't, you can round it down to 1, but it isn't that either. And that's a very simple way of looking at what truncation from going from 24 bits to 16 bits is like- this is an off the cuff example for illustration purposes only, folks, the math isn't meant to illustrate anything that happens in the real world.

So when a computer tries to make all the numbers rounded into what you are now asking them to be, it has a lot of numbers that don't easily become one thing or another, so you dither, which is a randomization of those errors which can be less easily heard.
And the kinds of dither such as Pow-R take advantage or your ears selective hearing, you hear better at some frequencies than others, so it concentrates the dithering noise where the ear is naturally less sensitive. I haven't ever found a preference for one dithering over another, you might, so try a few different ones.

Anyway, hope this helps, I will admit to being no expert on dithering so this is gleaned from what I've put together about it, there is always a chance I could be wrong about it and I apologize if this is so.