On Sat, Mar 27, 2010 at 02:45:18PM +0000, Folderol
wrote:
What! Are you telling me that going round the
edge of a CD with a felt
tip pen *doesn't* improve the 'clarity'? :D
Depends on the price of the felt tip pen.
Regarding the 'feel' of filters, I've put up some examples
on <http://www.kokkinizita.net/linuxaudio/paramfilt.html>.
These are three examples of a parametric filter set to
1 kHz, fixed 'bandwidth' and gain between -15 and +15 dB
in steps of 5 dB. The curves for +15 dB are identical for
all three.
If you compare Type 1 and Type 2, it's easy to see that
the two would 'feel' quite different if you turn the gain
knob. Most filters you'll see in practice will be somewhere
in between these two. For example my Ladspa filter plugin
is such a compromise, visually somewhat closer to 1 than
to 2.
Type 3 looks quite odd, but it is in a sense the most
'natural' of all three: it corresponds to a variable amount
of a *fixed* bandwidth (i.e. independent of gain) 2nd order
resonance being added or subtracted. Again this 'feels'
different.
Now the DSP code used to produce these curves is identical
for all. It has three internal parameters, c1, c2, g, and
the only difference between the three forms is in the way
that the three 'user' paramters Frequency, Bandwidth and
Gain are mapped to the internal ones. Given a certain curve
(i.e. values for c1, c2, g) all three can produce it.
So this is in fact three time the *same* filter, just
'presented' differently.
None of this really addresses the difference in *sound* of these
supposedly-"magical" filters.
But Google turned this up, which I found interesting:
Has anyone tried emulating these Neve filters in software using LADSPA?
Anyone want to? (um, hi Fons!)
-ken