Hum is a low
frequency continuous tone, usually caused by pickup from the mains and
so is at 50 or 60Hz, or the first harmonic: 100 or 120Hz. It can be
dealt with to an extent by simply filtering off the low
frequencies. Often the hum is below much of the bass on a recording,
so careful filtering can produce good results. It can be
possible to dynamically filter the hum: turning its level down
only
in the quiet matches of music, on the assumption that in louder patches
the music will mask the hum. A better solution is a 'notch
filter', removing a narrow band of frequencies, as with whistles
(below).
Whistles are more of a problem because they are often in the audible
audio band. Very high-pitched whistles can often be
removed by a simple filter (a 'low-pass' filter because it passes lower
frequencies and blocks higher ones): but if whistles occur in the main
frequency range then anything you do to the whistle affects the
recording as well. In the days of analogue equipment even a very narrow
notch filter, designed to remove the whistle but leave the frequencies
each side of it, could have an adverse effect on the sound. With
digital methods it's possible to construct a much more effective
filter. In the illustration below - a frequency spectrum of a quiet
section of a recording made off AM radio - the whistles can be clearly
seen.
In the following illustration very narrow filters have been constructed
(the blue line), and the whistles have been much reduced.
This brief extract from the same recording demonstrates the sound
before and after the application of hiss reduction and whistle
filtering: the change comes with the change of voice and may not be all
that obvious on small loudspeakers; however the improvement is clearly
visible in
the spectrum above. The processing was done in
SonicWORX Studio (no longer available).