On 04/06/07, Steven C. Barr(x) wrote:
> ----- Original Message -----
> From: "Don Cox" <[log in to unmask]>
>> On 03/06/07, Steven C. Barr(x) wrote:
>>> The applicable question here is not too hard to define...though
>>> difficult to answer! "Is it possible, using a recording of
>>> admitted (but not accurately defined) sonic inaccuracy, to
>>> recreate using digital methods an accurate recreation of
>>> EXACTLY what was recorded using earlier inaccurate methods?!"
>> It depends on the nature of the inaccuracy. Equalisation problems can
>> be solved quite easily. Nowadays, many clicks and some crackle can be
>> removed digitally. Intermodulation distortion would be very hard to
>> correct. Resonances are hard, as they vary with level. Bad microphone
>> placement, especially close micing, would be very hard to correct
>> unless multi-track tapes exist with the outputs from each mic kept
>> separate. (Which is not usual in the 1950s). You can't correct a
>> problem until you have analysed it, identified it and studied it.
>> Recordings have many different and independent faults.
> Remember that these Gennett phonorecords were recorded acoustically...
> which means the frequency response was essentially defined by now-
> unknown details of the recording horn, as well as the minimal
> bandwidth inherent in the acoustic-recording process...
I think the cut off at the low end would depend on the size of the horn,
which can be seen in the photos.
However, there are two factors: the frequency range (which frequencies
are recorded at high enough levels to detect in the noise) and the
varying levels within that range. The latter will include peaks and
troughs caused by both physical resonances and by phase effects - and by
the room acoustics.
When you get to the electrical era, there is also deliberate
(un)equalisation to consider. How accurate were the circuits that were
I think an engineer who is a competent musician and is careful and
patient can improve the sound considerably, but it is an art rather than
I suppose one might get a live band to record the same music
and then adjust until the frequency response matched as nearly as
possible. In practice, the good engineer will rely on his memory of
hearing similar bands playing live.
There are some interesting algorithms now being used for image
processing which might be applied for digging more high frequency
information out of acoustics. Look at the examples
(An audio file is basically a 1-dimensional monochrome image.)
[log in to unmask]