Re: [Audacity-devel] track mixing

[Anthony A. O. <Ant...@ep...>]

Whops! I'm still quietly trying to condense my comments and suggestions, when I
came across a classic topic.

On Sat, 1 Dec 2001 21:51:08 -0800, Joshua Haberman wrote:

>* Dominic Mazzoni (do...@mi...) wrote:
>> I totally agree - if it's possible to record levels above 1.0 then
>> there's no reason not to and we should then just automatically
>> de-amplify the audio afterwards, making it totally transparent
>> to the user.

>Don't do that! I don't think a tool should try to outsmart a person. The
>user should have that option, but Audacity shouldn't do this behind the
>user's back, IMHO.

I have to agree here, and I'll explain why in a minute.


>> Note that most sound cards send integer data, usually 16-bit, and
>> even the high-end ones often just send 20-bit or 24-bit integers,
>> and in each case -1.0 is defined to be the smallest negative
>> signed integer that can be represented in that many bits, i.e.
>> in 16-bit, -1.0 corresponds to -32768.  So they're not ever
>> actually sending raw data that's greater than 1.0.
>
>That occured to me after I thought about it some more. However I think
>this general idea could still be useful. For example, say that I amplify
>an existing section so much that it clips. It would be nice to be able to
>look at the waveform and see what part has clipped by observing what part
>exceeds the maximum amplitude line, knowing that no data has been lost
>(ie. I don't want Audacity to clamp for me).

>This particular example is contrived, but more realistic examples would
>be when quick mixing or applying an effect that happens to increase
>amplitude. It would be nice to have the flexibility to edit freely
>without worrying about clipping that results in lost data.

The casual user should get in to the habit of respecting the 0 dB Fullscale
barrier. First of all, all professional programs have a certain amount of
headroom internally, before they pass on the 'X' bit word of sound data to a
device.
A lot of plugins behave this way as well and they ALL have one thing in common.
They tell the user by how much the signal overshoots the 0dB line.

For example, the Waves Renaissance EQ, has an output fader(mono/dual
mono/linked) around which three groups of dB values are displayed :
1) beneath the faders the values of the actual signal are displayed, one for
each channel

2) this single value displays the difference to 0 dB Fullscale. The Waves plugin
actualy does this for -0.1dB. Clicking on this value will move the output faders
by that suggested amount. Should the signal peak at -3.0 dB, -2.9dB will show up
here. The value resets once the usr cilcks on it, or 1). Clicking on 1) does not
change the output fader.

3) output fader position - 0 dB is bypass. The Waves plugin allows for +12/-24
dB of amplification.Vegas for example allows +12/-inf dB settings.

Thus the user is left to make the descision with all the necessary information
he needs, readily at his fingertips.

The 'smart' quickmix or export of a final mix is a great idea. This final
amplification operation, just like all others during an export, should be of the
highest possible quality(float?48 bit mixer dithered to 32,24or16?). Therefore I
think having a mixing settings each for work and export mix operations might be
a good idea.

That's pretty much what I propose for Audacity in terms of clipping control.
Tell the user how much is clipping, if possible. It usualy isn't possible with
plugins, but a clipping check post-plugin(more than five 0dB FS samples
perhaps?) would help to issue a warning to the user.

I've got quite a few comments on the interface in the pipeline. I'm not shure I
see the point in having a multitrack display, when Audacity can only take one
audio region per track. Mind you, I've hardly seen such a great visualization of
volume automation in any other software. More on many things in a few days.

I included a screenshot of the Waves plugin(REQ) to illustrate the interface.

 Take care
  Tony