Peter Schmalkoke, 2001-03-15, translated 2001-03-17
 

Perturbative fields:
--------------------

Within the range of houses, which are attached to electricity mains,
alternating electrical and magnetic fields are usually to be found,
with can easily cause substantial disturbances to recordings with
sensitive equipment.

These perturbative fields should be shielded as far as possible. That
means: Utmost distance from all lines, which lead 230V, from transformers
and all electrical devices, which are operated with alternating current.
Furthermore, all lines, that carry weak signals (for instance microphone
cables) must be structured well shielded. Symmetrical lines and amplifier
inputs are optimal. Metal housing of microphones must be grounded or con-
nected to the earth of their amplifier input. a further screen can possi-
bly be achieved by grounding all large metal parts, e.g. housings of
electrical appliances (even lamp screens), metal shelves and desks.

Unavoidable disturbances must subsequently be eliminated from the recor-
dings. The expenditure (of time) is not smaller and the signal quality
becomes likewise worsened thereby. Clear recordings under good conditions
can not be substituted by anything.
 

Removal of powergrid humming from a recording:
----------------------------------------------

The problem of perturbative fields also exists within my other research
object, the magnetic signals in the natural environment. In the meantime,
I have found a promising solution, as the strongest disturbances are
derived mainly from the electricity power grid and therefore consist
mainly of the harmonic wave spectrum of 50 cycles per second (in europe).
I.e. the disturbance is essentially periodic with a period duration
of 1/50 sec. = 20 ms. A (digital) filter for additional treatment of
the signal can be built, which operates like this:

The input signal is sent into a delay line, and echoes are taken from
there, which represent the signal with delays of exactly 20 ms and
multiples of that time. These artificial echo signals are then merged,
say averaged. By interference thereby only the amplitudes of those fre-
quencies remain unchanged, which have periodical oscillations at 20 ms,
thus only multiples of 50 cycles per second. Everything else is weakened.
What remains, represents the extracted disturbance. This mixture is then
to be subtracted from the input signal, which contains these disturban-
ces, and in the result they are then perfectly extinguished!

For an illustration I have simulated the behavior of such a filter. For
the purpose of better representability the fundamental of the spurious
frequencyone is in this example assumed to be 480 Hz and eight echoes
are used. The examined input signal consisted of a pure tune, wich rose
evenly from 0 to 5.5 kHz during 10 seconds.
 
It is to be seen clearly that all multiples of 480 Hz are faded out
selectively and the bandwidth of these "needles" is already very small.
The actual weakening with the accurate frequencies may be even stronger,
than the resolution represents here. Obviously a usefull result is
already to obtain with eight echoes. The ripple is reduced when usig
further echoes.

If only one 20 ms echo is subtracted the simplest case, then the result
is not yet particularly good, because the bandwidths of the weakened
areas are very large, 50% of the spectrum remains at least 6 dB below
the peak values, so that a substantial portion of the information in
the signal is lost. Additionally there is a strong doubling effect in
that case. With 20 ms however, it does not irritate the ear that much
to make the procedure perfectly useless for listening.
It would be better in any case, to operate with many echoes over a
larger period of time. Then the individual echoes are weaker in the
output and the suppressed proportions of the spectrum are smaller,
the whole signal is thus less degraded.

With two echoes the strength of each individual echo in the output signal
is already 6 dB weaker than the original, and it falls reciprocally with
the number of echoes. A test with 50 echoes and a signal with a strong
disturbing proportion proofed a good extinction and hardly perceptible
alteration when evaluating by hearing.

With many sequential echoes, then distributed over a larger period,
this arrangement exhibits a typical behavior: An abrupt change in the
strength of the spurious signal steps up immediately as a disturbance
at the output and evenly fades away to zero again in the course of the
total echo duration. Extinction is based on the regularity of the distur-
bance.

In order to make the procedure applicable, a program can be written.
As a simple solution can be used for additional treatment of existing
recordings. That could be done even with BASIC, would however be very
slow. A more advanved version could be designed for direct use while a
recording takes place, for instance as plugin for Cooledit or Winamp.