[Ardour-Users] more on parallel eq'ing

[Jörn Nettingsmeier]

On 12/12/2012 09:07 PM, Jörn Nettingsmeier wrote:

> i doubt that "parallel eqing" as a technique makes sense. if this way of
> working inspires you (i can see how the two faders provide a very direct
> means of shaping the sound) _and_ you find musical use for the heavy
> coloration that results from combining a zero phase signal path with a
> non-linear phase one (the eq), then fine, but this is a rather
> idiosyncratic way of working, and the results won't be entirely
> predictable. i would venture than most engineers wouldn't be happy with
> such an arrangement.

in order to see how bad the results actually are, i did some 
measurements of a "parallel" eq scheme.

while i maintain that it doesn't make much sense to do it this way, 
there is less harm done than i thought with standard parametric filters. 
for steeper filters however, the combination is really deadly.

for testing, i set up a single pink noise source (JAPA), fed to two 
ardour busses: bus1 had no plugins, bus2 contained three different 
filters, only one of which was active at any one time: a 4th order 
highpass at 85 Hz, a 4th order lowpass at 12 kHz, and a four-band 
mitra-regalia filter (written by fons, with a gui by nedko arnaudov) 
with a pretty drastic curve meant to illustrate a practical "worst case" eq.

The four-band parametric is quite well-behaved in the parallel 
configuration:
http://stackingdwarves.net/download/Parametric%20Mitra-Regalia.png

the blue graph in JAPA is just the EQ, while the red graph shows the 
combination of the unfiltered signal with the EQ'd one. as expected, the 
boosts and cuts are gentler in the latter, while the interference 
artefacts are pretty minor (they are audible, but not really visible in 
the graph). the window on the right shows ardour3's "plugin analysis", 
where you can see the phase response of the filter (red) - it's really 
gentle. green is the amplitude response, ignore the funkiness in the 
bass, it's an artefact. likewise, the little "teeth" in the green curve 
are artefacts of using pink noise for measuring.

now, my droogs, the old higher order filters (with knives in them) kick 
it real horrorshow:
http://stackingdwarves.net/download/4th-order%20highpass.png

yummy. the phase response of such a filter is something an engineer 
should be aware of. its fall-off is 24db per octave, and while this is 
not a typical mixer strip eq, you might find it behind a rumble filter 
switch. analog synth fans will have a couple of those on their machines. 
you can imagine what a 48db/oct filter would look like when recombined 
with the straight signal. totally unusable for any systematic approach 
to mixing, but certainly interesting from a synthesis point of view.

a fourth-order lowpass is pretty much the same:
http://stackingdwarves.net/download/4th-order%20lowpass.png

these are chebyshev filters, and you could argue that with their 
passband ripple, they're not exactly high-end anyways. but a 
fourth-order butterworth (as often found in loudspeaker crossovers) 
doesn't look any better:
http://stackingdwarves.net/download/4th-order%20butterworth%20highpass.png
(i've had to cascade four 1st-order bw filters here for lack of a better 
plugin.)

*.*

bottom line: i don't see anything interesting in parallel EQing, and 
many pitfalls. but if parallel EQing inspires you, go ahead. just make 
sure that if you saddle that horse, you know how to ride it.

now i guess some of these artefacts sound a lot nicer and fuzzier in 
all-analog systems, which have their ways of "rounding off the edges", 
so i could imagine some weathered old hands from the analog days 
obtaining great results with such techniques.
but a textbook-correct recombination in the digital domain can get 
really nasty, so it shouldn't be blindly emulated in a DAW.


my 10 ¢


jörn


--
Jörn Nettingsmeier
Lortzingstr. 11, 45128 Essen, Tel. +49 177 7937487

Meister für Veranstaltungstechnik (Bühne/Studio)
Tonmeister VDT

http://stackingdwarves.net