dithering?

[Ben B]

There are people who say they can, but I don't know of anyone who has actually demonstrated this ability when the music is loud.

It's easy to hear the effects of using (or not using) dither in a contrived test. Here's an example:

1. Create a 24 bit session, and insert a signal generator on an aux fader.
2. Set the frequency to around 200 Hz. (The frequency is not particularly important, but using 200 Hz makes the test a little less irritating to listen to.)
3. Automate a fade-out to negative infinity over a period of several seconds.
4. Bounce this to disk twice while converting to 16 bit resolution, once with dither, and once without.
5. Close the session, and create a new 16 bit session.
6. Import both bounces into the 16 bit session.
7. Cue up to the very end of the fade out, and listen to approximately the last second of audio as it fades to nothing. (Do this over headphones for the maximum audible effect.)

In the dithered file, you should hear a smooth fade out, with a little hiss at the end. In the non-dithered file, you will hear a sort of zipper-like distortion as it fades out.

So, the distortion exists, but it is debatable whether or not it is audible when the music is loud.

-Ben B

[Ethan Winer]

Originally Posted by GRO

can anybody actually say they hear a difference between when dithering is applied and when its turned off?

Many people will claim they can hear a difference, but from the tone of your question it seems you already know the truth.

--Ethan

[GRO]

well in the studio where i work for a local label they swear by dithering so i brought it up last night and they were like listen....and i was like listen to what? they must of heard some real sonic tone changes only dogs could hear. i never could tell a difference and after lots of reading i am convinced the change is subtle

[chrisj]

Originally Posted by GRO

can anybody actually say they hear a difference between when dithering is applied and when its turned off?

Yes, especially with boutique dithers. I'm not at all sure dithering to 24 bit counts, though it is technically correct to do. Dithering to 16 bit, hell yes if you have good monitoring.

I've done mastering. I'm using a Lavry Black and Channel Islands monoblocks, very radical speakers and room treatment. I'm sure there are systems where you can't hear dithering vs. truncation no matter what you do, but I don't want systems like that.

Almost any EQ move or fader move will be a bigger difference MUSICALLY- if you're into the textures of things, you do get to a place where you have to pay attention to dithering because EQ and fader moves will no longer help you.

As a musician, you shouldn't have to care unless you're a massive tone snob, and then you should indeed care about dithering.

[Ben B]

If they were the ones switching it on and off while insisting that you "listen," then it was not a valid "test."

-Ben B

[Ben B]

Here's another thing: if they were merely switching a dither plug-in on and off, nothing is happening anyway apart from the addition of very low level noise. The distortion that dither corrects only manifests during a change of word length. Listening in a 24 bit session "with and without dither" reveals nothing at all.

-Ben B

[Joe Porto]

Originally Posted by GRO

can anybody actually say they hear a difference between when dithering is applied and when its turned off?

Depends on the bit length. At 8-bits, you can most definitely hear lack of dithering as an instrument or reverb tail fades out. You will get truncation effects at it's lowest levels. Adding dithering to an 8-bit signal will add noise that is well above the noise floor.

At 16-bits, not so much, although if you mix OTB through compression, you could potentially hear the same truncation during fades, but most likely, the self noise of your mics and pres would be above that level

At 24-bit, the dither noise is so low that you'd have to turn the volume up to deafening levels to hear the effects of dithering or truncation, and at that point, the noise floor of any analog stages would be louder than the added dithering noise.

[ark]

A friend of mine who is a serious signal-processing expert tells me that the way to hear the effect of dithering is to listen to a very quet passage with the volume turned all the way up. Listen particularly to how reverberation dies away after instruments stop playing.

I haven't tried this test myself, but I trust him.

[OceanMan]

it is generally minimal. sometimes i think it is possible that it affects you on an almost subconscious level, making you simply "enjoy" the particular song more......
maybe i just need to lay off the crack

[Ben B]

Originally Posted by ark

A friend of mine who is a serious signal-processing expert tells me that the way to hear the effect of dithering is to listen to a very quet passage with the volume turned all the way up. Listen particularly to how reverberation dies away after instruments stop playing.

I haven't tried this test myself, but I trust him.

You will have to be converting the word length to hear this phenomenon. Simply turning up the level with dither activated while maintaining the original word length will only allow you to hear the dither noise.

As I pointed out in my last post, you'd need to bounce the mix twice while converting the word length -- once with dither, and once without. You would then need to compare the bounces.

-Ben B

[Nordenstam]

Correctly implemented dither completely removes the truncation noise from word length reduction. So it follows that 16 bit dither noise will sound the same in both 24 bit and 16 bit containers. Anyhow, most all dither routines does the truncation to 16 bit as well. Use a bit scope and check the output of the dither routine if in doubt. I have not been able to find a dither plug that does not truncate as well.

Second time in ten minutes I post this link: 24 vs 16 bit not audible? with 10/10 blind test results in this post: 24 vs 16 bit not audible?

Another test: Science Versus Gearslutz

EDIT: more on the subject in this post!

[theblue1]

Point of correction: dither does not "completely remove" truncation-generated alias error -- but it will reduce it with mathematical significance, no question.

Think of it (somewhat crudely) this way: truncation imposes a 0 value at the truncation point. In essence, you're "rounding everything down" -- so a simple randomization of values of the new LSB (least significant bit) actually brings the results statistically closer to the results that might have resulted had we been able to do a proper round-up/round down rounding.

[Ethan Winer]

Originally Posted by ark

A friend of mine who is a serious signal-processing expert tells me that the way to hear the effect of dithering is to listen to a very quet passage with the volume turned all the way up. Listen particularly to how reverberation dies away after instruments stop playing.

Yes, that can sometimes reveal a difference between dithered and truncated. But if you have to turn up the volume so loud that you'd blow out your speakers and eardrums on normal parts of the music, how important is dither really?

--Ethan