Does noise reduction work well if...
-
- KVRAF
- Topic Starter
- 7402 posts since 17 Feb, 2005
The sample in question is converted to a 32-bit float, scaled to a much lower amplitude, like -200dbFS (knowing floats DR), and reduction applied. Or is there a proper 'gain staging' for it? Is the FFT just as accurate that low?
-
- KVRAF
- 7893 posts since 12 Feb, 2006 from Helsinki, Finland
It seems a bit silly to gainstage like that, but as long as you don't go low enough to run into denormals, it shouldn't really matter.camsr wrote:The sample in question is converted to a 32-bit float, scaled to a much lower amplitude, like -200dbFS (knowing floats DR), and reduction applied. Or is there a proper 'gain staging' for it? Is the FFT just as accurate that low?
-
- KVRAF
- 3080 posts since 17 Apr, 2005 from S.E. TN
Tis been awhile since I used a spectral noise reduction. Don't offhand recall the settings params, and certainly haven't used every spectral NR implementation.
Maybe there could be practical problems with real world programs, such that the programmer never imagined that his process would be used on such low level signals? And therefore perhaps the user interface settings would have too limited a settings range to operate properly?
Ferinstance, recall a spectral plug from long ago that was written to be an "automatic" NR kinda like a spectral burwen. No noise print, just a threshold knob and a few other controls. Betcha that one wasn't set up properly to work on -200 dB signals.
Maybe there could be practical problems with real world programs, such that the programmer never imagined that his process would be used on such low level signals? And therefore perhaps the user interface settings would have too limited a settings range to operate properly?
Ferinstance, recall a spectral plug from long ago that was written to be an "automatic" NR kinda like a spectral burwen. No noise print, just a threshold knob and a few other controls. Betcha that one wasn't set up properly to work on -200 dB signals.
