Re: [Audacity-quality] Noise Removal Effect broken?

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

On Mon, Sep 13, 2010 at 3:29 AM, Bill Wharrie <bi...@go...> wrote:
>
> On 12-Sep-10, at 6:50 PM, Gale Andrews wrote:
>> [snip]
>
>> To me, we're largely revisiting the discussion of a few months ago. As
>> I recall, a test like you are doing needed about +18 dB  on the
>> Sensitivity
>> slider with then little or no wobble (I think I increased smoothing
>> a bit
>> above default and reduced attack/decay a bit below).
>
> I've quickly read through that thread on -devel and it seems to me
> that it was mostly (all?) about the Sensitivity slider, and fixing the
> problem where the Attack/Decay setting was non-functional.
>
> I still maintain that the effect is not doing what is described on the
> wiki. That is, a multiband noise gate, where the threshold for each
> frequency band is determined by the noise sample. It appears (as
> others have noted) that the effect is using the overall full-spectrum
> signal level to decide whether or not to gate (or downward expand) the
> frequency bands, rather than looking at the signal level in each band.

>From tests that I've done I agree that the wiki description does
not appear to match what the effect is actually doing.

Was the wiki description written based on what the code is supposed
to do, or on what the documenter thought it was doing? In other words,
is the code wrong or is the wiki description wrong?

>
> Raising the Sensitivity slider causes the effect to treat more of the
> desired signal as noise. A slight increase in Sensitivity can help in
> some cases, but +18 dB is way over the top.

What exactly does this dB measurement actually refer to?
If set to 18, is the noise sample being amplified by 18 dB?
"Sensitivity" would imply a change of the gate/expander threshold
level. is the threshold level raised by 18 dB? If so, which threshold
- a threshold based on the noise frequency band, or based on the
overall full-spectrum signal level?

>
> So, I don't think this is a re-hash of the -devel discussion from
> July.  I think it would be useful to look at the heart of the effect
> to see if it really is doing what is claimed.
>
> -- Bill
>