Audacity Forum

stevethefiddle wrote:On the other hand, how much difference does it actually make in those "special cases"? Extremely little (insignificant?), so is it worth the time and effort? Probably not.

Steve:
I agree. Given the noise that's already in those recordings (phono pre-amp, vinyl surface noise, cassette hiss, mic pre-amp hiss, room background noise, cr*p on the YouTube audio, etc. etc.) the effect of adding dither noise at -90dB is surely insignificant.
-- Bill

stevethefiddle wrote:

Gale Andrews wrote:It also means you can't use "Detach at Silences" to get your white space back after running a Nyquist effect.

Not necessarily.
a) the silences could be marked before passing the audio to Nyquist...
b) the silence could be detected during processing ...and before converting back to 16/24 bit for writing to disk.
c) "silence" could allow for dither noise ...

Sure, or some other solution. All I meant was "the user" cannot get their white space back by detach at silences in any reasonable way.

stevethefiddle wrote:

Gale Andrews wrote:I think the level of this page should be kept understandable as far as possible without having to leave it.

Yes I fully agree and agree that it needs improvement. But it also needs to be accurate, and it is the "default" bit that I, like Bill, am having trouble with.

I'm not arguing with that - if processing 16-bit audio with 16-bit quality setting is done in 32-bit, the Manual is seriously misleading (and I've been misled by that and by other similar implications for years).

stevethefiddle wrote:A definitive answer about what Audacity does internally may help us come up with some wording that is both easy to understand and accurate. No easy task for such a complex thing.

Sure. Let me find out about this.

billw58 wrote: Adding dither when exporting a 16-bit project to a 16-bit file is not a bug, it's a feature!

I'm not quite convinced. I know quite a few users who stick with 16-bit quality for 16-bit output on the assumption there will be no dithering, and switch to that quality setting for that reason on the basis of our documentation. I know they could turn dither off, and are probably better off with 32-bit anyway, but we're misleading them if this is deliberately applied dither. For example on our Quality Preferences page: "Sample format conversion would be required upon export if you used the default 32 bit float sample format but exported a 16 bit audio file." That is a clear implication of no dither on 16-bit export of 16-bit audio. And if we are as clever as all this, let's figure out when we have a 16-bit audio track for export with no 32-bit processing required and keep the dither out of the export.

Anyway, let's not speculate further, but get some facts... Thanks for all the very high standard contributions.



Gale

Gale Andrews wrote: Anyway, let's not speculate further, but get some facts.

Agreed. As much as I would love to continue this discussion, we don't need to treat Audacity as a black box when there are people who know what's in the box.

-- Bill

Gale Andrews wrote: Anyway, let's not speculate further, but get some facts.

I'm still trying to get some solid answers, especially to if "all internal processing is 32-bit" but haven't forgotten. The source code claims quite clearly:

" Dithering is only done if it really is necessary. Otherwise (e.g. when the source and destination format of the samples is the same), the samples are only copied or converted...
There are only 3 cases where we must dither, in all other cases, no dithering is necessary.

* _INT24_TO_INT16
* _FLOAT_TO_INT16
* _FLOAT_TO_INT24 "

So I'm still assuming that applying dither at the point of exporting 16-bit audio to a 16-bit file is a bug (in the sense we don't actually intend this).

One other thing has come to light that I didn't know. Exported OGGs never have dither noise added (at least to silence). If you export 32-bit silent audio as OGG (which can only be 16-bit), the exported file is still silence.


Gale

So in the following workflow:
1. Preferences, Quality Tab, Default Sample Format = 16 bit.
2. Preferences, File Formats, Uncompressed Export Format is either "WAV Microsoft 16 bit PCM" or "RAW signed 16 bit"
2. Record a track from an ADC that outputs 16 bit.
3. Don't save the recording, but instead choose File / Export as WAV (or Export as RAW, depending on step 2)

I'm assuming that at no point was the data converted to 32 bit and back to 16, and that the contents of the file as saved after the export
has retained 16 bit accuracy.
Is this correct..?

thanks!
Tibor

tiborkerbosch wrote:So in the following workflow:
1. Preferences, Quality Tab, Default Sample Format = 16 bit.
2. Preferences, File Formats, Uncompressed Export Format is either "WAV Microsoft 16 bit PCM" or "RAW signed 16 bit"
2. Record a track from an ADC that outputs 16 bit.
3. Don't save the recording, but instead choose File / Export as WAV (or Export as RAW, depending on step 2)

I'm assuming that at no point was the data converted to 32 bit and back to 16, and that the contents of the file as saved after the export
has retained 16 bit accuracy.
Is this correct..?

thanks!
Tibor

The answer is: we don't know, exactly.

Some of us assume that all internal processing is done at 32-bit float. If this is true, then in your workflow the data would be converted to 32-bit float then back to 16-bit. Theoretically there should be no loss of "accuracy" during these conversions, except that Audacity might (according to preferences set in the Quality tab) apply dither when converting from 32-bit to 16-bit.

Quoting from an earlier post:

stevethefiddle wrote:1) Generate 10 seconds of silence and export as a 16 bit WAV file called "dither.wav"
2) Switch off dither in Preferences.
3) Export the same 10 seconds of silence as a 16 bit WAV file called "no-dither.wav"
4) Turn dither back on in Preferences
5) Import BOTH tracks back into Audacity.
6) Test.
Result - dither is applied on Export (no-dither.wav is silent).

-- Bill