From: Patent Tactics, George Brock-Nannestad
Steven C. Barr wrote:
> Perhaps what we need is a research team which can work on one single
> digital file taken from an average-condition 78, and simply try
> different ways of redrawing the waveform and auditioning the result
> until we can establish exactly what needs to be done to extract
> the "music" portion of the signal. Once we have identified that, we
> can then establish how best to restore the information which was
> made unavailable by the sonic limitations inherent in various
> old recording practices...?!
----- a comparison of the commercial tools available would be a most
interesting endeavour, and it will require a computer interface version of
the tools, i.e. Cedar Cambridge compared to Algorithmix and other less costly
approaches. The reason is that the settings can be logged and repeated
accurately. Before honing in on a narrower range of settings, we could
specify 10 settings per parameter that may be influenced, and each result
must be logged as well. The input digital file should be one and the same
digital transfer from a 78, unless we also want to compare pickups. As a
reference perhaps a straight transfer from an unused vinyl pressing of the
same piece played on an ELP turntable might serve. The unending variability
of signal from a 78 must be reduced in order to get a manageable data set (it
is large enough, just due to the parameter setting combinations), but the
categories experimented with might be for instance solo violin, solo soprano
voice, trombone, some of the signals that have caused problems for
psychological data reduction.
The only person that I can think of who might have the tenacity to undertake
such a task voluntarily could be Eric Jacobs - his experiments on the ELP set
an example to us all.
Now, merely subtracting the treated signal from the straight signal will not
tell us much. Listening to the discarded material on its own does not tell us
what its removal from the complete signal does for our listening experience.
To complete the experiment we would need to take a subset of the parameter
combinations and make a listening test for many to partake in. Then the
respective preferences would slowly appear, and possibly narrower sets of
parameter combinations could be tried in a second run-through.
To me the whole thing sounds so labor-intensive that it might be a credit-
giving project for a group of university students.
> In other words, can we take a 19th century acoustic recording, and
> infer from that what a stereo, full-fidelity recording of that sonic
> event would sound like...?!
----- that is not unlikely, but the noise-reducing tools will not perform
that. It would require analysis tools that could tell us something about the
reverberation in the original room and then model a soundscape including a