Handbook for Sound Engineers

Consoles 965

tapering off the values toward the ends of the set. The
taper that is applied varies according to the type of
window, and the differing types are best suited to
differing interests of compromise. Say a brick-wall filter
had been described as in the figures; one window may
optimize for stop-band rejection, Fig. 25-134, another
may trade that against sharpness of the filter cutoff rate,
Fig. 25-133, etc. Many thanks to Momentum Data
Systems’ software for the curves.

25.21.1.3 Symmetrical FIRs

There is an FIR implementation that has some quite
interesting properties and as such is probably the most
used, so much so that the majority of commercial design
packages assumes as a default that one wishes to design
symmetrical FIRs and that FIR has become almost
synonymous with the symmetrical filters they afford.

These allow the imposition of a frequency response
(in the case of a conventional-style EQ) without altering
the phase response, unlike ordinary EQ (and nature) in
which any frequency response change comes with a
corresponding shift in phase response for free. Although
this characteristic might at first blush seem ideal and a
major leap forward for audio technology, in practice
they are only rarely used; yes, Virginia, they do sound
different to conventional EQ with equivalent frequency
responses, but not necessarily better. (An odd effect is
that one seems to need more phaseless EQ cranked in
than conventional EQ for a similar subjective effect.)
Certainly, it’s not better enough to displace conventional
EQ, which can be readily and far more efficiently

created in either digital or analog form. The difference

alone, however, is sufficient reason for existence in

music production, and special-effects units and audio

workstation plug-in software specifically to do symmet-

rical FIRs are available.

Symmetry refers to the fact that the coefficient set is

arranged to be symmetrical about the center—the

midpoint—of the filter; identical coefficient set-lets tail

off toward the end of the set as tail back toward the

front. The midpoint of the filter is regarded as the time

center—in other words, a symmetrical FIR has an

intrinsic time delay of a passed signal of half the length

of time the filter takes to calculate; in our now-famous

50 Hz capable, 960 point filter, the effective time delay

is 10 ms, or half of the time it takes for any one data

sample to transit the entire filter, being 20 ms. This time

delay is another major downside to symmetrical FIRs;

in order to keep everything in a multisource console

time aligned, all other sources would have to be delayed

by the effective time delay of just one FIR’ed source.

Note that not only is half of the filtering done after

to the time center, but, and this is the head hurter, half of

it is done before the time center, leading up to it. The

filter only remains causal because of the intrinsic time

delay. That the ear can deal with filtering effects before

something has happened and integrate it all into an

acceptable sound is a true amazement.

25.21.2 Recursive Processing

This concept was approached in achieving spin echo

and reverberation; feeding an already manipulated input

sample back around in a loop to be reprocessed along

Figure 25-133. The same 33 point filter Hanning windowed.

0 0.15 0.3 0.45 0.6 0.75

0.583

0.292

0

0.292

0.593

Time–ms

Frequency–Hz

0 4800 9600 14,400 19,200 24,000























Figure 25-134. 33-point filter Harris windowed.

Frequency–Hz

0 4800 9600 14,400 19,200 24,000























0 0.1333 0.2667 0.4 0.5333 0.6667

0.583

0.292

0

0.292

0.583

Time–ms