Some thoughts on Dithering.

[soundandform]

A few months back, I gave a short presentation about Dither at an AES meeting in Nashville. People seemed to enjoy it, so I've typed up an extended version here: http://www.smassey.com/blog.html#29_apr_2008

Steve
Massey Plugins Inc.

[Colin F.]

Great article. Cheers for that. I played around with many different dithering options and eventually realized that it made little to no noticeable difference myself.

[aitikin]

Very good read. I'll probably continue to dither the stuff I do at school, but in my own, probably not anymore. (Recording a percussion ensemble right before a 15 person woman's chorus is not an ideal situation)

[Jon~T]

Excellent article Steven

While I understand the theory behind dither, I never understood why it was sooo important to do it, one of the only Rules of recording , but no one would/could/should be able to tell if you actually dithered or not.

I still get a chuckle out of studio gear lists with dither plugs listed.

[GooberNumber9]

Wait a second:

85dB SPL is a standard peak calibration level for movies and is also the threshold of hearing damage, at long durations. Under this condition, 90dB below that is well under the generally accepted "threshold of human hearing". Expecting to hear the effect of dither is like expecting to hear the fluttering of a fly through the din of a jackhammer -- it's just not going to happen.

I thought 85 dB SPL was calibrated to -20 dB FS. That means that in your calibrated THX theater a peak at full scale would be at 105 dB SPL and your quantization noise would start creeping in around 15 dB SPL.

To me the real issue would seem to be whether the downstream analog audio devices have 90 dB dynamic ranges and whether their thermal or other noise contributions are covering up quantization noise. Also, our hypothetical THX theater certianly has an ambient noise floor higher than 15 dB SPL even if there aren't 500 breathing, shuffling, snacking, and inevitably talking, patrons in it.

Ok, so yes this post is just a pedantic call-out in a minor point and I'm not disagreeing on the larger issues (although I have thought I could hear the difference between a mix of mine with dither vs. without).

Todd Wilcox

[Brett Siler]

At least half of the sessions I have done (probably more) I never dithered. Nobody complained.....

[fossiltooth]

At least half of the sessions I have done (probably more) I never dithered. Nobody complained.....

I used to do separate bounce for Ref mixes, converting down to 16 bit in PT using the dither on the L2.

Recently, I've gotten into the habit of bouncing final mixes at 24 bit without dither and then just letting iTunes or toast do whatever they do in converting them 16 bit. I'm not sure if they're truncating or applying dither, or what the hell they're doing.

Regardless, whatever happens in the process isn't disastrous enough to be annoying. In fact, it might sound even better than when I bounce to 16 bit. Then again, I could just be mixing better these days... I should do a blind test sometime.

[SpencerBenjamin]

I could help you there. In getting blind, that is.

*cough*

[Brett Siler]

Recently, I've gotten into the habit of bouncing final mixes at 24 bit without dither and then just letting iTunes or toast do whatever they do in converting them 16 bit. I'm not sure if they're truncating or applying dither, or what the hell they're doing.

Thats what I usually did/do in Toast if I know they aren't gonna send it somewhere else to get it mastered. If they are gonna have someone else master it the I just like to put the 24 bit files on a data disc.

There have been some people that have said some stuff about iTunes doing weird things to their mixes if they haven't dithered to 16 bit. I personally haven't heard anything that jumped out at me and made me go "Jesus, what the hell happened".

[fossiltooth]

Thats what I usually did/do in Toast if I know they aren't gonna send it somewhere else to get it mastered. If they are gonna have someone else master it the I just like to put the 24 bit files on a data disc.

Yeah, but I usually do both... one 16 bit ref CD for client reference and one 24 bit master CD (or data transfer) for the ME.

They used to be completely different bounces. The client ref would be a 16 bit dithered bounce using L2 or the like to bring the peak level close to -1db without doing any real GR, and a 24 bit "Master" version with a lower peak level and no dither. Nowadays I just use the same bounce for both purposes. It saves tons of time. 24 bit, no dither, L2, Maxim or Massey 2007 taking off 1db at the most and bringing the peak level up to -1db or so. Works for the client reference, works for the Master copy. No problem. I'll let the ME worry about the subtleties of dither if he likes.

For those who didn't bother to read Steve's article, he's basically saying "For practical purposes of 16 bit CDs with a peak level close to 0 dbfs, dither is essentially meaningless".

I'd probably be inclined to agree. My noise floor is probably significantly higher than the dither level, especially if I tracked to tape, or had a huge multitrack band session with tons of open mics. Even if that wasn't the case, the noise floor of any listening environment is much higher than the quantization noise or dither. It's not effecting my mix enough for me to care.

Incidentally, Steve's article features some wise words, and it's brilliantly written. It's a worthy and easy read. The link is in the first post. Check it out!

[leigh]

Steve,

You raise good points in your article. Dither certainly deserves to be de-mystified, and understood for what it is.

I agree with this statement, for most types of program material:

"Expecting to hear the effect of dither is like expecting to hear the fluttering of a fly through the din of a jackhammer -- it's just not going to happen."

However, I wonder if this statement is a bit hasty:

"Maybe... just maybe, with a highly un-normalized Classical CD from the 80's listened to in an anechoic chamber with the lowest-noise-possible audio reproduction gear, you just might barely detect hints of quantization noise on the quietest passages."

...at least as regards to program material. There are some very quiet passages in pop music too - song fade-outs and fade-ins. And while it's a safe bet that, 99% of the time, the quantization noise over those passages will be masked by ambient noise in the listening environment, or circuitry noise in the playback equipment, there is also the "first do no harm" idea in audio mastering (well, not that all mastering engineers follow that these days, but it is written in the holy books of Good Engineering Practice).

To wit: if you are converting a 24 bit mix to a 16 bit file, you are inherently taking responsibility for making that conversion process as clean as possible. If you simply truncate the sample words, you are introducing a distortion to a signal that previously was clean. If you truncate and add dither, you introduce a low level noise to a signal that was previously clean - but the noise is *not* correlated to the signal in any way, and so if it sounds like anything, it sounds like white noise, instead of crunchy distortion.

Now, that last "if" is a big one, because you're right that most of the time, nobody hears those sounds at all. But you never know how that audio might be manipulated or transformed in the future, possibly in a way that would highlight the quantization noise present.

Another way to put it - dithering is kind of like the audio engineering version of Pascal's wager - there's nothing to lose, and potentially something to gain. (Well, I suppose there's money to lose, if you buy into expensive dithering plug-ins or something, but I'm just talking about using the garden-variety dithers included with just about every DAW under the sun.)

In conclusion: Dithering is bullshit. Dithering is not entirely bullshit.

Leigh

[MoreSpaceEcho]

^agree.

[fossiltooth]

Great points on all counts, Leigh. Still, when a discussion tilts too far in one direction, someone needs to push back a little bit.

Steve did that.

Sure, it was partly motivated by self-interest, as the Massey 2007 does not offer dither, but still, his thoughts on the subject are valuable and help bring some wisdom and balance back to the discussion over the essentially inaudible "black magic" that is dither.

Great F*ing thread.

[leigh]

Still, when a discussion tilts too far in one direction, someone needs to push back a little bit.

Steve did that.

For sure, I agree.

[sears]

The secret to killer tollhouse chocolate chip cookies is sea salt.