Dithering!

[The Philosopher King]

Quick Question:

I am Rendering/mixing-down parts as I build a song using export audio.

When all of the individual parts for the song have been mixed down, I will then mix them together to create a pre-master.

Should I be turning the Dithering Option to OFF (No Dither) while exporting the part files, since I will (as a final step) be Dithering all of the part files, after they have been mixed together (in the form of a song)?

Dither once seems to be the golden rule!

[solacerodgers]

Since this post got so long and out of control im going to post exactly what live says about dither and such right here.

Whenever rendering audio to a lower bit depth, it is a good idea to apply dithering in order
to minimize artifacts. Dithering (a kind of very low-level noise) is inherently a non-neutral
procedure, but it is a necessary evil when lowering the bit resolution.
Please note that Live’s internal signal processing is all 32-bit, so applying even a single gain
change makes the resulting audio 32-bit as well even if the original audio is 16- or 24-bit.
Dither should never be applied more than once to any given audio le, so unless you are
mastering and nalizing in Live, it is best to always render at 32-bit and avoid dithering
altogether.

[Tone Deft]

dither when you between awkward sample rates is the golden rule, such as going to 44.1k from 48k or 96k.

if you can't hear a difference, don't mess with it.

[Superchibisan]

you should really just dither once, but if you are doing things that way, it all depends on bit rate. if you reduce from 32 to 24, or to 16, you need to dither. otherwise you will hear truncation in the bass and treble.

i'd really just stay inside the project and freeze and flatten everything. that way you can stay in a 32 enviroment.

[ethios4]

dither when you between awkward sample rates is the golden rule, such as going to 44.1k from 48k or 96k.

if you can't hear a difference, don't mess with it.

What?? Dithering is applied when changing bit-depth, not sample-rate....you know that.

OP - you should maintain the highest bit-rate you can throughout your workflow. If you are exporting parts to be mixed later, you will benefit from exporting them as 32-bit. Only apply dither once, and at the very end of your process, when you go to 16-bit.

[Tone Deft]

it all depends on bit rate. if you reduce from 32 to 24, or to 16, you need to dither. otherwise you will hear truncation in the bass and treble.

32, 24 and 16 bit do NOT refer to the bit rate. "hear truncation in the bass and treble"? that's incorrect, in effect that's a filter and dithering is not a process that resembles a filter in any way shape or form. what does hearing truncation in the treble and bass mean? truncate the top bits you get massive distortion, truncate the lower bits and you lose dynamic range and get more subtle distortion. but none of that matters.

i'd really just stay inside the project and freeze and flatten everything. that way you can stay in a 32 enviroment.

apples and oranges. you can only dither in Live when you render the project, it has nothing to do with staying inside Live.

[solacerodgers]

it all depends on bit rate. if you reduce from 32 to 24, or to 16, you need to dither. otherwise you will hear truncation in the bass and treble.

32, 24 and 16 bit do NOT refer to the bit rate. "hear truncation in the bass and treble"? that's incorrect, in effect that's a filter and dithering is not a process that resembles a filter in any way shape or form. what does hearing truncation in the treble and bass mean? truncate the top bits you get massive distortion, truncate the lower bits and you lose dynamic range and get more subtle distortion. but none of that matters.

i'd really just stay inside the project and freeze and flatten everything. that way you can stay in a 32 enviroment.

apples and oranges. you can only dither in Live when you render the project, it has nothing to do with staying inside Live.

I was just about to comment on the "sound" of truncation. Wordlength is what dithering affects and what you would hear would be quantization errors with added harmonics or interpolation depending on the noise shape you used.

[Superchibisan]

wait wait wait. so 16 bit. and 24 bit. do not refer to bit rate? what do they refer to then?

[Tone Deft]

I was just about to comment on the "sound" of truncation. Wordlength is what dithering affects and what you would hear would be quantization errors with added harmonics or interpolation depending on the noise shape you used.

nope. wordlength does not change.

dithering twiddles the lowest order bits to help hide artifacts from other processes, that's all it is. how it's done and which method to use is a pretty deep concept.

[Tone Deft]

wait wait wait. so 16 bit. and 24 bit. do not refer to bit rate? what do they refer to then?

when the audio is sampled, 16, 24 and 32 bit refer to how large the binary number is that's used to represent the sample. this affects the dynamic range of the sampling process.

bitrate is sample depth times the sample rate, but most all digital audio systems are 32 bits, even if you only record with 16 or 24 bits. the data word is 64 bits (32 left, 32 right.)

[Superchibisan]

i can't believe i've been wrong all these years! my school must have lied to me. i paid good money for that! its all LIES.

[Tone Deft]

they directly affect the bit rate but there's two parts to the bit rate, so in short hand you could make that connection if you assume a sampling rate.

[solacerodgers]

I was just about to comment on the "sound" of truncation. Wordlength is what dithering affects and what you would hear would be quantization errors with added harmonics or interpolation depending on the noise shape you used.

nope. wordlength does not change.

dithering twiddles the lowest order bits to help hide artifacts from other processes, that's all it is. how it's done and which method to use is a pretty deep concept.

Wait what, wordlengths do not change? In digital audio (DSP) the wordlength is ever expanding now depending on your bit resolution and a fixed or floating point this will either be pre truncated or ( floating point ) be allowed to overflow into the wordlength. When you dither you loose wordlength and its usually the last steps that are truncated to fit into the new bits that are then requantized to fit into the smaller bits. All that said to say this all processing in the digital domain results in a change in wordlength.

[Superchibisan]

sorry, i used the wrong word. depth, not rate.

[Tone Deft]

I was just about to comment on the "sound" of truncation. Wordlength is what dithering affects and what you would hear would be quantization errors with added harmonics or interpolation depending on the noise shape you used.

nope. wordlength does not change.

dithering twiddles the lowest order bits to help hide artifacts from other processes, that's all it is. how it's done and which method to use is a pretty deep concept.

Wait what, wordlengths do not change? In digital audio (DSP) the wordlength is ever expanding now depending on your bit resolution and a fixed or floating point this will either be pre truncated or ( floating point ) be allowed to overflow into the wordlength. When you dither you loose wordlength and its usually the last steps that are truncated to fit into the new bits that are then requantized to fit into the smaller bits. All that said to say this all processing in the digital domain results in a change in wordlength.

OK, now you're getting into floating point DSP processing. I'm just talking regular digital audio sampling, dithering, and transmission.

where you're headed you can say all kinds of things about the audio, time domain vs frequency domain, floating point, normalization, all kinds of tricks.

dithering is taking the lowest order bits and using a pseudo-random process to 'blur' the data to cover up uglies from some other process. no need to convert to floating point, I guess you could and there might be dithering processes out there that do that.