Jump to content

dithering... what exactly does it do??!


Recommended Posts

Hi nick,

Dithering is the multiplication of low level signals to avoid the distortion created by the truncation of those low level signals when their lower bits are removed.

 

OK, with that said. It's easy to think about this with a single tom sound. At 24 bit, we can capture the entire decay of the tom. If we lop off bits 17-24, at one point the decay will reach a level where the only valid data left is the peaks of those waveforms. The lower level information for the waveform is gone. This is truncation and creates a pretty nasty distortion that is in-harmonically related to the original signal.

 

By first injecting noise into the signal before lopping off those bits, the signal level down there gets multiplied so that it can be high enough to remain in the 16 bit version.

 

Because the distortion created by the truncation is noticeable only in the frequency ranges that our ears are most sensitive to at low levels, the noise is shaped, (POW, MBit, triangular, etc,) to deal specifically with those bands.

 

There is still much debate in the professional audio community as to the necessity of dithering masters, particularly for certain material. If you're mastering hot dance tracks, it's likely you won't notice. But string ensembles in a quiet recital hall, you probably will. But don't even think about auditioning different dithering algorithms unless the noise floor of your studio is extremely low, or if you have superb headphones. There's tons of articles, and debate, out there on the web about dithering.

 

Another use, an important one, for dither is in a signal processing chain. When using a plug-in that uses fixed point math which then passes off to a 32 bit floating point host, dithering the output of the plug has its advantages. Most of the better mastering plugs like this have a dither option on their output.

 

Hope that helps!

Link to comment
Share on other sites

  • 2 weeks later...

wow..

 

that's... involved. i think i get it. i should probably do a bit of research on it!

 

but basically, dithering is the compensation of the loss of lower level signal by overcompensating it to the point that it cannot be left out in the final bounce..

 

i think!

 

muchly appreciated fader8 :)

Link to comment
Share on other sites

Ok, so very basically stated, dithering audio sounds to me like the same thing as graphical dithering. just a way to more randomly blend transitions generated by reducing the depth of a sound to make a sound seem smoother/less distorted (but introducing noise as the randomization of rounded points on a wave form will not be consistently rounded in a single direction?) For example (imagine that this line is actually incrementing on one side of a wave with 90% of the wave not illustrated):

 

a 6 bits of the increment

000000001111111122222222333333334444444455555555

 

truncated to 3 bits (transitions are sharp and grouped)

000000000000000022222222222222224444444444444444

 

dithered to 3 bits

000000020020020202202222224224242424424424444444

 

The noise comes from the random up/down, but makes the line sound smoother because transitions are blended.

 

Is this on the right track?

Link to comment
Share on other sites

a 6 bits of the increment

000000001111111122222222333333334444444455555555

 

truncated to 3 bits (transitions are sharp and grouped)

000000000000000022222222222222224444444444444444

 

dithered to 3 bits

000000020020020202202222224224242424424424444444

 

:shock:

 

Neo? Is that you?

Link to comment
Share on other sites

Well I remember that THE one explanation that finally made it clear for me was from the Dithering Guide that the guys at Izotope published. It's a brillant (and FREE) attempt at clarifying one of the most elusive subjects in audio (there's still a bunch of others, though).

 

Here's a link to download it:

http://izotope.com/products/audio/ozone/guides.html

 

Then when you're done with it, there's more (there's always more) here:

http://www.digido.com/bob-katz/dither.html

Yes, it's the gospel by Saint Bob Katz. Buy his book, the Audio Truth lays here (along with a few other things, too).

Link to comment
Share on other sites

This thread seems like a reasonable place for this question. I currently have an Apogee Duet, but I plan to get the Ensemble, which has Apogee's proprietary UV22 dithering. Does this mean that when UV22 is enabled, the output of the Ensemble is captured by the DAW at 16 bits? Or would the entire mix be sent back to the Ensemble for dithering during the bounce process? If the former is true, wouldn't that be a problem in terms of signal degradation if you plan to process the signal after recording it (as most tracks do receive some treatment after tracking)?
Link to comment
Share on other sites

I don't work with the Ensemble, but I'm pretty sure you are being confused as to what is dithering and what's the converter itself. Dithering is only a PROCESS used to compensate for reduced wordlength when downconverting, for example from 24 to 16bit, but also from - say- 32 to 8 bit. Dithering is the injection into your original signal of a certain amount of noise (shaped according to a prioprietary algo), which - to make things VERY short - helps smoothing fade outs and reverb tails, and generally prevents blurring and data loss at very low levels.

 

Therefore, you can CONVERT a 24bit signal to a 16bit signal WITH or WITHOUT dithering, and you can bounce a 24bit mix to a 24bit stereo file WITH dithering if you so wish (however that would make little sense, to say the least).

 

So, within the limits of your audio interface, you will be able to record at any wordlength, any sample rate (dithering has nothing to do with recording), then use whatever files you got and bounce them at any wordlength and any sample rate, with or without dithering. Of course, you hardly would record at 16 bit to mixdown at 24, but you still can.

 

I'm sure the apogee Ensemble works like any other interface. Dither and wordlength conversion are completely different and discrete processes, even if they are usually handled in one operation. And whatever the fancy name of the noise shaping.

 

Hope this helps!!

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...