The first stage
in
our evolutionary tale of noise removal
is
the simple treble filter (or
'low-pass filter'). Less damaging than the volume control which removes the signal
a!together, the treble filter only removes a proportion of any signal present above a
g~ven
frequency (known
as
the shelf frequency of the filter). Unfortunately,
if,
at the
given frequency, you reduce the amplitude of the noise content of your recording
by
6d8
(thus making the noise half
as
loud) you will also reduce the genuine signal
at
this frequency by the same amount. This will
be
fine if your recording has little or
no
high frequency content, but natural sounds and modern electronic instruments
have frequency responses up to and beyond the limits of human hearing.
Consequently, the treble filter will only be successful
in
processing your antique
collection of '78's, and even then only
at
a cost.
But this gives us a hint as to how a more effective single-ended noise reduction
system could
be
designed: perhaps a device could
be
built which removes the high
frequencies when there
is
no
signal present, but leaves them untouched when the
noise is being masked by genuine high frequencies? Of course it can. It's a
Dynamic Filter (so called because the shelf frequency of the filter moves
dynamically up and down the frequency spectrum according to information
contained
in
the signal). But such devices are limited: for one thing, they can only
remove the noise which exists above the highest frequency of the music present at
any given moment. Secondly, they are based
on
filters with roll-offs typically of the
order -12d8/0ctave or -6dB/octave,
so
they always allow some high frequencies
through, even when they think that they're removing
it.
And thirdly, even though the
filters are designed to track the signal very quickly, they cannot respond
instantaneously, so they tend to round off fast transients such as snare drums and
samples (which have a habit of introducing high frequencies very rapidly into the
signal). And, because their raison d'etre is to reduce the signal bandwidth they also
tend to dull the genuine signal quite perceptibly. So
to
summarise dynamic filters: if
you're not compromising the signal you may not
be
removing as much noise as
you wish, and if you're removing all the noise you're probably damaging the
genuine signal.
Perhaps an alternative approach could give better results? Instead of altering the
frequency response of the signal to reduce the noise content, how about changing
the volume (amplitude response) of the signal
in
some way? This isn't such a
strange idea. Consider: if noise is always present
in
a signal then, if the total signal
amplitude drops down to the noise level, surely all the genuine musical signal has
disappeared? While there are many flaws
in
this argument (largely to do with the
statistical nature of broadband noise)
it
is, as a generalisation, nearly true. This
then suggests a device which will eliminate some of the noise: a Noise Gate. This
simple device detects when the signal drops below a certain level (a 'threshold' set
by the user) and then cuts off the signal entirely. It's just like turning the volume
control of your hi-fi to zero between tracks, and then turning
it
back up
at
exactly the
moment the music starts again. There are many enhancements to the Gate idea,
such
as
variable attack and release times, and hysteresis (all added to limit the
occurrence of damaging side-effects) but the principle always remains the same: if
the device decides that there is only noise present at its input,
it
totally shuts off the
signal. There
is
an
adage that says that a multi-track studio can never have too
many noise gates because, while they are pretty useless at stereo mastering, they
are invaluable for shutting off the intrusive hisses, hums, and buzzes of temporarily
unused synths and guitar channels.
DH-
1:
Rev.02 Ver.
1.06
Page - 5
July
12.1996
