Re: [Audacity-quality] GUI text in vocal removal effect

[Steve t. F. <ste...@gm...>]

On 29 June 2013 13:59, Gale Andrews <ga...@au...> wrote:
>
> | From Steve the Fiddle <ste...@gm...>
> | Fri, 28 Jun 2013 11:21:55 +0100
> | Subject: [Audacity-quality] GUI text in vocal removal effect
>> I ended up rewriting the vocal remover code rather than trying to
>> untangle what was there before.
>> The help screen is virtually unchanged
>>
>> The error checking should be more sane now and the code less
>> spaghetti-like and easier to maintain.
>
> Thanks. FWIW, I don't understand (as in the previous code) what
> removing or retaining a range of 0 to 0 Hz is supposed to do.

Retaining a range of 0 to 0 Hz should be the same as "Simple (entire
spectrum)" (because you are retaining nothing)
Removing a range of 0 to 0 Hz should be "no effect" (because you are
removing nothing)

>
> Removing a 0 to 0 Hz band appears to remove audio that was
> only in the right-hand channel (irrespective what the audio was).
> Retaining that band silences the right.
>
> I notice different behaviour between old code and new in the case
> of a full frequency range music track with a 0.1 amplitude 440 Hz
> sine tone mixed into both channels. "Simple" Vocal Removal does
> not reduce the vocals in old or new code, but removes the tone
> and reduces the amplitude of the track.
>
> Retaining 300 Hz to 600 Hz in the new code produces a stereo
> track where the right contains only the tone and a little of the
> music. The left sounds little affected.  The old code produces
> the same audio in both channels, little affected (which was the
> result I was expecting).
>
> Removing 300 Hz to 600 Hz seems correct in the new code.

Oops, there is a bug in the "Retain frequency band" option, which is
fixed in the attached.
"entire spectrum" and "Remove frequency band" are correct as far as I  can tell.


>
> If you want the audio, please ask me off-list.
>
> How does it treat a lower band value that is above the higher
> value?

The code is tolerant of entering upper and lower limits the wrong way
round. The lower value is always used as the lower limit. It seems
pointless to throw an error message for this being that the code can
deal with it easily.


>
> Should it throw a specific rather than generic error when you ask
> it to remove or retain a range where the values are above half
> the sample rate?  I agree it's better to error than do something
> unknown as the old code did.

This is only relevant for low sample rates (below 40 kHz) but I agree
that it can be improved.
So as not to unnecessarily throw errors if the effect is used on
multiple tracks or in a Chain (where the tracks/files may not all have
the same sample rate) I've just limited the range to a maximum of the
Nyquist frequency.

Values over 20 kHz will generate an error message that tells the user
that they can't do that.
For low sample rate tracks, values above the Nyquist frequency (but
less than 20 kHz) will be treated as "as high as possible" (the
Nyquist frequency).

Steve

>
>
>
>
> Gale