Alias due to lamps (Eckard Blumschein )

Subject: Alias due to lamps
From:    Eckard Blumschein  <Eckard.Blumschein(at)E-TECHNIK.UNI-MAGDEBURG.DE>
Date:    Mon, 18 Jun 2001 09:53:15 +0200

Dear list,

Several discussions nurtured my suspicion that a perhaps overlooked problem
might be a rather ubiquitous one. Some years ago, I investigated power
supplies for arc welding. They caused strong components of arc sound within
the frequency range above the limit of audibility, in particular at 50 kHz,
while the signal was sampled with 44 kHz. In any case, Shannons condition
for correct sampling was violated. Without anti-aliasing filter, an alias
was clearly dominating the recorded sound. Both Robert Masta and David
Isherwood confirmed that meanwhile the anti-alias filter does an excellent
job. I am not quite sure whether this only holds for typical sound, or they
also took into consideration that there is an increasing application of
power electronic switches. Those who are not familiar with the problems
might for instance have a look into US patent 5,828,185 concerning HID lamp
design above 19 kHz but below the lowest lamp resonant frequency. In
general, power electronic switches prefer to operate between typically just
20 and more rarely 100 to 200 kHz. Sources of strong inaudible sound are
not only magnetic components but also the plasma of lamps and in rare cases
even a filament of a lamp. Imagine a radiation of 90 dB SPL at 50 kHz.
Well, I may choose a steep filter and/or 96 kHz sample rate as to avoid the
alias. However, I am seeing problems:
1) I doubt that all those who are recording sampled sound are aware of
possible sources of inaudible sound.
Energy saving lamps are most widespead candidates of inaudible emission,
and they may irregularly change their behavior.
2) Is there already any device or method that alerts an inadmissible level
of inaudible sound prior to and during audio recording?
3) I cannot decide whether reported subtle differences between records with
48 kHz and 96 kHz are actually due to the higher sample rate, they just
reflect less aliasing, or they are due to different anti-alias filtering.
4) Don't have very steep filters dynamic limitations?
5) How to in future exclude mysterious artifacts if the experts are not
even able to easily agree on Martin Braun's handling of GIPOS data?
6) Even if the manufacturer/vendor of a source of inaudible sound is
responsible to obey the rather high pegel limits, is he also obliged to
point to the possible risk of aliasing influences?

Please respond directly to me. I will collect all responses.

Eckard Blumschein
http://iesk.et.uni-magdeburg.de/~blumsche/AuditoryFunction.html