Understanding the effect of dither

Hey guys,

Dither seems to be a very complicated subject. Recently there was a thread about which to use and when so forth. I'm taking a Berklee mastering class and I finally understand what dither does and how it sounds. 

Unfortunately, I can't copy the Berklee materials, but they did have this site in one of the projects and I thought I'd share it:

http://www.24-96.net/dither/

The shootout examples are quite interesting.

Cheers, 

I found that 3 sounds the best to me with 7 a close second. 9 Was pretty transparent, but it sounded like an 8k bump.

Interesting, they're both from the same company MBM.

I found that 3 sounds the best to me with 7 a close second. 9 Was pretty transparent, but it sounded like an 8k bump.

Those examples on that page are a perfect example of how not to test dither. The whole point of the various types of dither is to move the dither noise to frequency ranges where your ear is less sensitive. And due to how the ear has varying sensitivity depending on loudness, listening to dither at any level higher than its "nominal" level (which would be in the ballpark of 10 dB SPL--according to my rough guesstimate) is utterly useless in determining its 'sound'.

Those tests are completely bogus.

From the site:

Dithering happens at very low levels and that our files display dithering "under a microscope". (54 dB of gain have been added AFTER dithering in order to eleiminate playback equipment distortion on your end and take the noise characteristics of your system out of the test equation to some extent.)

Almost all dithering type tests are done this way, and I would argue it's just wrong. In the real world background noise will mask the differences.

Keep in mind that +54dB = 500 times the amplitude. 

Why aren't these tests ever done without isolating and amplifying the noise several hundred times over?

awesome topic......Dithering and DC offset. Thanks for the info.

I would say that if you have to isolate and amplify something more that +20dB to make it audible (especially if it's a trained listener in a good environment), then it's just not ever going to be audible in the real world.

Exactly. Particularly when the item under test is specifically designed according to the equal loudness contours right down at the threshold of hearing.

Considering that I've never seen a double-blinded study of the audibility or otherwise of the various types of dither (under real-world conditions, not cranking it up), and that the only times people claim to hear a difference in the musical signal (not the dither itself) is when they turn it up significantly to hear how it fades out, I'll go out on a limb and say that no one can actually hear the difference between those dithers when the musical signal is 80 dB or more above the level of the dither. Which is a typical real world scenario. And then I seriously doubt they could hear the difference on fade outs without turning up the signal.

But please prove to me otherwise, I'd love to find someone who can hear the difference.  But until then, I just stick to triangular PDF, non-noise shaped dither when I go to 16 bits.

Izotope wrote this great dithering guide

+1

Awesome vid. Thanks for all the input!