On Tue, Nov 26, 2013 at 11:27:46AM +0100, Raphaël Mouneyres wrote:
> as of multiband compresion for mastering purpose, i use 4 bands most
Ha, now I remember what I told Julien - which was exactly to
Depending on filter type and order, to obtain a flat response
in the crossover region it is not always true that the filters
should cross at the -3 dB points (and hence probably set to
the same frequency). It is true for first order ones. One filter
has +45 degrees phase shift at the nominal frequency, the other
-45. Combining two signals at -3 dB and 90 degrees phase
difference result in 0 dB, as required.
This doesn't work for second order filters. They will have +/-
90 degrees phase shift at the -3dB frequency. If you set both
filters to the same frequency and add the outputs, the region
around that frequency will cancel. If you invert one of the
two, the result will be +3 dB. To get the flattest possible
response in the sum you have to set them to different
frequencies. The result will never be really flat, but you
can come close.
If you want crossovers that exactly add up to a flat response
you need filters like the ones used in zita-lrx.
To answer Julien's question: just preserve the _ratio_ of
the frequencies of the two filters that form a crossover.
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)
Linux-audio-user mailing list