[net.audio] noise reduction: no free lunch

ark@rabbit.UUCP (Andrew Koenig) (01/15/84)

Consider a tape recorder.  You put a signal in and later get a similar
signal out.  No matter how much information is in the input signal,
the amount in the output is limited by the tape recorder.  A noise
reduction system such as Dolby or DBX works by redistributing the information
in the input signal so that the losses will come in less objectionable
places.  However, these losses CANNOT be eliminated, because you can't
create information that the tape recorder has thrown away.

Noise reduction systems that work by compressing the dynamic range of
the signal on tape pay for that in magnifying frequency response
irregularities in the tape machine.  No way to avoid it.  The more compression
you do, the more the irregularities will be magnified.  Dolby B does
the least compression, so it is least sensitive.  Dolby C is somewhat
more sensitive.  DBX is the most sensitive.

To fuel the flames a bit, I will point out that Dolby Labs makes the
claim that while output noise from DBX-processed tape will be less
than for Dolby C in the absence of any signal, Dolby C will have
less output noise than DBX in the presence of, say, a single low note
on a piano.  Given the difference in their approach, this makes sense.