From: Patent Tactics, George Brock-Nannestad
This is in reply/comment to Jon Noring, Aaron Luis Stevenson, Goran Finnberg,
David Seubert, Mike Richter.
1) Pitching of 78s is an activity that requires you to play the same part of
a record over and over again, while adjusting the speed, and this causes
demonstrable wear, unless you use an ELP Laser Turntable. For this reason I
have proposed to IASA that we recommend that any pitching is made on a
secondary carrier obtained by a first transfer. Now, in the old days, a
cassette was ideal for this, because you would not use tonal fidelity, only
pitch fidelity as the goal. That is, if you are not working with acoustic
recordings - here we have interplay between pitching and EQ. The cassette
player I use has a very wide variation plus or minus 20% corresponding to 3
semitones either way. I have a calibration tape for the cassette player, and
when I play that at the speed decided to be good for correct pitching of the
cassette, I can measure the pitch of the calibration tone now emitted by the
calibration cassette, and I can then determine the percentage speed deviation
from the speed of the original used at transfer. I facilitate this procedure
considerably by using the special calibration disc SC-1 that I developed in
1983 and which I have described in this forum previously.
2) Equalisation means two things: it means a) the gentle slope changes in the
transfer function that serve to emulate the change from constant-velocity
recording to constant-amplitude recording that is either inherent or
deliberate in almost all analogue mechanical records. However, it also means
b) the inverse filtering that needs to be performed on an acoustic record in
order to counteract the influence of the horn-recording-soundbox combination
on the spectral balance of the sound. There is a 40 dB (worst case) dynamic
hidden here, and for this reason it is extremely important to have good
dynamic range housekeeping in a 16 bit system.
Now, pitching can only be performed reliably on an equalised signal in the
case of an acoustic recording where the instrumental cues that are most
practical for pitching are not readily apparent, and the adjustment of the
inverse filters is dependent on the frequencies that have been influenced by
the recording setup, but these frequencies are dependent on the replay speed.
So we have an iterative process, and that is slow going. Again, the
mechanical record suffers, if we have to try too many times.
Both 1) and 2) point to a "quick-and-dirty" transfer to digital, provided a)
the dynamic range is available, and b) we know precisely what the replay
conditions were. If b) is not fulfilled, we may be able to do a "perfect"
transfer for the particular record by digital work, but we will not be able
to relate it to the original, and we will not be able to accumulate
knowledge, only experience.
3) The reason that a high sampling rate and a high resolution (20 or 24 bit)
are required, is that it is not the bandlimited and dynamically limited
recorded signal that we are worried about, it is the NOISE that is both
broadband and with a large dynamic. In order to be able to distinguish
between the two, we need a good representation of the NOISE. I think that i
was the first to express this apparent contradiction in archival circles in
the IASA Conference in Bogensee, near Berlin, in 1990. The fact that anything
from 18 bits and up is bogus in the absolute sense is immaterial: it only
means that there is some more distortion, but it ensures that interpolation
algorithms will work, albeit slowly.
4) As to sampling-rate conversion, I would refer to the book by John
Watkinson: "The Art of Digital Audio", Focal Press (mine is 3rd edition
2001), pp. 136-150. Now, these pages as well as many others cannot be
understood without having read all of the book, and indeed, I consider it
imperative for anybody seriously working with audio in a modern, digital
world, to read and understand as much as possible in this book, all 739 pages
of it! If you can't, then obtain the services of an engineering student who
can help you. Get a grasp of it. It is frightening to have to span the
stretch of more than 100 years of development, but we cannot avoid it if we
want to transfer analogue content from 1900 to digital content in 2003. We
will never get back to the stable period from 1955 to 1985, in which it was
sufficient to be good at magnetic recording. All that is made on the basis of
too little knowledge is detrimental to the digital content that the future
will have to rely on.
Kind and thoughtful regards,