User Manual

40 User Manual v 1.0.0
UNFILTER
NOISE
Another important parameter in the LEARN process is NOISE. NOISE removes foreground
information from the input sounds before processing, and allows the algorithm to only focus on
the background information within the signal. To hear what this sounds like with your recording
turn on NOISE MONITORING. If this is on, you will hear exactly what the algorithm will hear
when NOISE is enabled. Why is this useful? For instance, if you are processing a piano
recording, you will generally have a lot of stationary harmonics and large frequency gaps
between them, which are devoid of information that the algorithm could use to estimate a
smooth frequency response. By removing the foreground, i.e. the harmonics, the algorithm
uses only the frequency response derived from the noise that was present in the original
signal. Attack transients, background or environmental noise, stochastic components in the
signal will all be used to derive the frequency response in this case. Since this process works
independently of strong harmonics you can derive a smooth frequency response in these cases
as well.
Note: As resonances are typically narrow, and depending on the acoustic situation may persist
over a longer period of time, some of them will also be ignored when using NOISE. In a
nutshell: NOISE mode may help with some signal types, but at the expense of precision under
some circumstances.