This article is from the Audio Professional FAQ, by with numerous contributions by Gabe M. Wiener others.
This whole question is based on the premise that "analog systems don't
filter out all the information above 20 kHz." Indeed there are mixers
and power amplifiers and other electronic systems that are capable of
stunningly wide bandwidth, often exceeding 100 kHz, the same cannot be
said for the entire analog reproduction chain. The mechanical
transducers, microphones, speaker and phono cartridges seldom have
real response far exceeding 20 kHz. In fact, some of the most highly
regarded large diaphragm condensor microphones often used in very high
quality recordings seldom exceed 18 kHz bandwidth. Analog tape
recorders rarely have bandwidths as wide as 25 kHz, and LP
reproduction systems have similar limitations in reality.
So while it may be possible to send very high frequency ultrasonic
signals through parts of both analog and digital reproduction chains,
there are, in both technologies, fundamental and insurmountable limits
to the bandwidth that, in reality, lead to very similar actual
reproducible bandwidths in each.
Thus, one of the basic premises of the question is flawed. Analog
systems DO filter out information above 20 kHz. Further, the frequency
response and phase errors of even the very best well-maintained analog
reproduction systems have response errors far exceeding those of even
middle of the line digital equipment. Whether one person may find
those errors tolerable or even likeable or not is a matter or personal
preference that is beyond the scope of this or any other technical
There are a variety of anecdotal tales that are advanced to "prove"
that the ear can hear far beyond what is conventionally accepted as
the 20 kHz upper limit (an upper limit that, for the most part,
applies to young people only: modern high SPL music and noise levels
has lead to a widespread deterioration in the hearing of the adult
population at large, and especially amongst young males).
For example, there is an apocryphal story about Rupert Neve that
tells of a console channel that sounded particularly "bad". It was
later discovered that it was oscillating at some ultrasonic frequency,
like 48 kHz. Rupert Neve is rumored to have seized upon this as
"proof" that the ear can hear well beyond 20 kHz. However, there exist
an entire range of perfectly plausible mechanisms that require NO
ultrasonic acuity to detect such a problem. For example, the existence
of ANY nonlinearity in the system would result in the production of
intermodulation tones that would fall well within the 20 kHz audio
band and certainly would make it sound awful. Even the problem that
was causing the oscillation itself could lead to massive artifacts at
much lower frequencies that would completely account for the alleged
sound of the mixer in the complete absence of a 48 kHz "whistle."
Whether 20 kHz is an adequate bandwidth is a debatable subject.
However, several important facts have to be remembered. First, BOTH
analog AND digital reproduction systems suffer from roughly the same
bandwidth limiting. Second, digital systems using properly implemented
oversampling techniques have far less severe phase and frequency
response errors within the audible band. No analog storage and
reproduction system can match the phase and response linearity of a
digital system, both at low and high frequencies. Once those
demonstrable facts are acknowledged, then the discussion about
supra-20 kHz aural detectability can continue, knowing that, if it is
demonstrated to be significant, both systems are provably deficient.