Print

Print


Aaron wrote:

> I am surprised that "pitching" is being suggested ex post facto as the
> proper procedure for doing a 78 transfer. The reasons are as follows:
>
> 3. The final reason is that the algorithms which are used to achieve
> this pitch change simply
> go back and toss out enough samples to make the program the desired
> speed or length.

This I believe is not correct. Implicit in the assumption of samples
is that they represent points along a wave form which is the Fourier
sum of sine waves. Thus by doing the D-A conversion (but not really),
one can predict with high precision the value a sample should have
that's arbitrarily somewhere inbetween any two samples (which is what
resampling is intended to do, at least that's how I understand it.)

Thus, I am of the understanding, which may be wholly wrong, that the
published algorithms for digital resampling are very well developed
and universally implemented by the better sound processing tools. And
that one can even predict the distortion caused by these algorithms, and
that it is typically very small. But then, I've been surprised by what
these tools can and cannot do. Anyone?

(Btw, I just found the following online document goes into the general
resampling algorithm known as "Bandlimited Interpolation":

http://ccrma-www.stanford.edu/~jos/resample/resample.pdf

It states on page 2:

"...Since Shannon's sampling theorem says it is possible to restore an
audio signal *exactly* from its samples, it makes sense that the best
digital audio interpolators would be based on that theory. Such
'ideal' interpolation is called *bandlimited interpolation*."

Putting it another way, if we accept that, in principle, we can do a D-A
conversion, and then resample that using A-D, then we don't run into
the problems Aaron is mentioning. Thus, this resampling can be done
without ever going into the analog realm. Even if some audio processing
tools don't use bandlimited interpolation -- and maybe Aaron is
referring to some of these "cheap ones" -- one can certainly find a
specific tool, and probably free, which does proper resampling.)


There's a lot of factors to take into account when transfering, and
adding precise pitch control in the analog stage is just another added
burden if it is not necessary and there are other reasons not to do it
(like one has to do a large number of transfers in a limited time frame
-- digital restoration is done at a later time and at a more "leisurely"
pace.) Of course, if one has the time to do it in the analog stage,
then there's nothing wrong with that (although one may be able to do
spectral analysis or some such tool in the digital realm to very
precisely fine-tune the pitch.)


> 4. For these reasons I believe that (with the exception of the Weiss
> EQ-1) a really fine analog(Sontec)
> EQ with an excellent turntable turning at the exact required playback
> speed  will in general reveal far more musical results
> than attempting either eq or pitch changing in the digital domain. I
> think it is unlikely to change in the near future.

George Morrow did mention the need to try to get the equalization
right in the analog realm before digitizing. He said the equalization
can be done using a simple change in capacitance (or whatever) in the
phono input, to emulate what was done years ago. He also said that
someone has actually put together the table of values to use depending
on label and label number -- by and large, he said, the recording
companies kept their equalization settings the same for a period of
time (but not always), since that was tied in with the "marketing" of
their product, and they wanted uniformity in the equalization.

Again, comments from the experts on both of the above are wanted.

Jon