Skip to main content

Hi everybody,

If, while tracking, you dither down to 16 bit from an external converter to an MDM, or any digital multi for that matter, what are the cumulative affects of the dither at mixdown? Particularly UV22-HR.

Thanks,

Tim L

Comments

Greg Malcangi Tue, 06/12/2001 - 01:28

Hi Tim,

The basic idea of dither is the addition of noise to hide quantisation errors. Modern dither moves the bulk of that noise way up into the high frequencies so that is *almost* unoticeable.

As a general rule of thumb therefore the use of dither is a balancing act between trying to use it as little as possible so as not to add too many artifacts while at the same time trying to avoid quantisation errors.

In short, I personally would try to avoid the situation you describe and track at the highest resolution your converters support. If you need to use an external digital multi-track I would try to get hold of the 24bit Tascam or even the Mackie.

Greg

Guest Tue, 06/12/2001 - 04:33

Greg, is the purpose really to try to mask quantization error, or to add a bit of noise for the reverb trails to fall off into?

My understanding [very possibly an incorrect understanding] was that it was more musical than straight 'truncation' in mix applications, but will "build up into a noisy ball of shit" if used when tracking.

When tracking, I'm from the school of 'no dither', but when mixing from 24 bit to 16...I've found a little added 'dither noise' from the Crane Song LTD. "dither CD" to be extremely helpful in making the stuff sound musical.

As always, YMMV.

bluebass Tue, 06/12/2001 - 08:57

I had a feeling summing all that dither shit together was not going to be a good scenario. I guess it's time to wake up, smell the coffee, and dump these damn Adats...what can I tell ya', I'm 'stubborn Irish'... sometimes I find myself trying to peel a potatoe 6 different ways with an old peeler rather than getting a new one.... :D

Thanks guys,

Tim L

anonymous Tue, 06/12/2001 - 11:07

Greg/Fletcher,

Greg, is the purpose really to try to mask quantization error, or to add a bit of noise for the reverb trails to
fall off into?

That and to prevent one from hearing the LSB flip (really crank up an *extremely* quiet passage of undithered music..... it's frightening).
Additionally, one can hear a bit into the noise floor as well. As a rule, I dither once and it's the last thing I do. There are many flavors of
dither out there. The Meridian sounds nice.

My understanding [very possibly an incorrect understanding] was that it was more musical than straight
'truncation' in mix applications, but will "build up into a noisy ball of shit" if used when tracking.

If you do it while tracking (which I have done with UV22) then it's best to not add further dither. As you say, it's much better to do it at the
end...*especially* if you are gonna do a lot of DSP stuff. Keep the errors down low and then whack 'em when you dither.

When tracking, I'm from the school of 'no dither', but when mixing from 24 bit to 16...I've found a little
added 'dither noise' from the Crane Song LTD. "dither CD" to be extremely helpful in making the stuff sound
musical.

Greg Malcangi Thu, 06/14/2001 - 07:46

Hi Fletcher,

<< Greg, is the purpose really to try to mask quantization error, or to add a bit of noise for the reverb trails to fall off into? >>

In effect, it's a bit of both. Ideally you want a reverb to tail off into silence. Silence is very easy to represent digitally, it's that little tail off into silence that's the problem. Even with a 24bit system the difference between silence and extremely quiet can only be represented by a 0 or a 1. There is nothing in between to smooth out the tail off. Anything in between is rounded up or down and hence the term "quantisation error". Dither in effect raises the noise floor and makes it virtually impossible with normal hearing to distinguish between a 0 and a 1, even at 16bit resolution.

Realistically though dither is in essence a bad thing. Who wants to add noise to thier recording? However, it ultimately comes down to the lesser of two evils. Truncating audio files, say reducing 24bit to 16bit for mastering, can produce some quite nasty quantisation errors especially in music that contains a lot of dynamic material. Under these circumstances I find the effects of dither to be a lesser evil than the effects of truncation.

Back in the early days of dither, which really did sound like just the addition of white noise, I sometimes prefered to live with the effects of truncation without adding dither. However, things have moved on a bit since then with various flavours of noise shaping. Noise shaping theoretically gives all the benefits of standard dither but moves much of the added noise up into frequencies which can't be heard. This results in a cleaner sounding and smoother mix in comparison to normal dither or truncation. These days I always use noise shaped dither when mastering down to 16bit from 24bit.

It stands to reason that you should try to avoid using dither unless you have to. I've never actually tried dithering every channel but logic would suggest that there must come a point where the cumulative effects become seriously noticeable.

Greg

anonymous Fri, 06/15/2001 - 20:44

hi brad, thanks for posting your opinions...
have you tried the benchmark converters? i tracked with their 24/96 a/d converters once, and faintly remember that their gear boasted about its dithering abilities (ns3 algorithm?), although i didn't have to bother with it at the time, so i didn't really check that out.
btw, i'm realizing now that the original poster was concerned w/ dithering while tracking to a 16bit machine, and i'm not needing to do that, but it's got me thinking about dithering options when burning to CD or at mastering time :)

-paul

Bob Olhsson Sat, 06/16/2001 - 14:30

The best explanation of dither I've come across is to think about it like this:

In digital audio there is always going to be some level below which the signal simply cuts out. Not dithering sounds exactly like an improperly set noise gate that is chattering and sounds like distortion as the signal crosses back and forth across the threshold setting.

What dithering does is to add just enough noise to "tune" the chatter of the gate in such a way that it sounds like innocuous hiss rather than distortion riding on the signal. There are many opinions about what flavor of hiss sounds innocuous and some "shaped" dithers build up in ways that severely color the sound although they may sound less "hissy" on first listen.

It is a misconception to think dither is covering anything up, the distortion IS being reduced and reduced by quite a bit. The trick is picking the right dither for the job which is probably more like picking a type of analog tape or a certain bias point than anything else.

bluebass Sat, 06/16/2001 - 17:23

Originally posted by Bob Olhsson:
The trick is picking the right dither for the job which is probably more like picking a type of analog tape or a certain bias point than anything else.

Very intriguing... could you elaborate a little on your preferences as far as choice of dither for different situations?

Regards,

Tim L

Bob Olhsson Sun, 06/17/2001 - 19:09

The "wrong" dither will sound cloudy or opaque while the "right" dither will sound fatter, richer and deeper with better reverb tails.

What will be "right" for one application can easily be "wrong" for another. You just have to use your ears and make sure that you are monitoring at the end bit-depth and not something higher.