Print

Print


Norman wrote:
> Jon Noring wrote:

>> I found the following article which discussed the recording speeds
>> for "78rpm" records:
>>
>>  http://articles.findarticles.com/p/articles/mi_m0HEO/is_13_26/ai_94718185
>>
>> How accurate is this article? After the mid-20s, did recording
>> engineers religiously record at 78.26 rpm as the article implies, or
>> was there still a lot of small variation in recording speed?

> It is well-intentioned, but rather simplistic, as David Lennick has
> already commented.
>
> [snip]
>
> These records invariably include a piano, and so the only assumption
> we have to make is that it was tuned - more or less - to standard
> pitch for the above-mentioned countries, i.e. A=440Hz for the U.S.A.
> and A=439Hz for Britain. That relatively small difference is not a
> significant problem. (Much useful discussion on Standard Pitch was
> done on this list a couple of years ago, and we concluded that A=440
> became first established in the U.S.A.; it was adopted elsewhere in
> much of the western world in the later 1930s).
>
> [snip]
>
> Nevertheless, there remains the problem of pitch change during the
> cutting of a record. In the type of record mentioned in this posting,
> any change in pitch is usually downwards. This because a cutting
> turntable with insufficent torque will speed up as the drag from the
> cutter decreases as the groove velocity falls towards the centre of
> the disc. On replay, naturally, the pitch will drop.

Interesting. Has anyone done an analysis (probably based on pitch
sampling throughout a recording) if the pitch drop due to cutter drag
is essentially linear with respect to time (or some such predictable
curve)? Also, is this actually the norm to expect (variable recording
speed due to cutter drag) rather than the exception? What recording
companies used equipment which guaranteed, within practical
considerations, constant cutting lathe turntable speed?

Obviously, if one is to correct this pitch drop effect, it has to be
done during digital restoration since the playback (for producing
the raw transfer) must be kept rock-solid constant in speed.


> Some time ago Michael Kieffer generously sent me 78.26 rpm transfers
> of master pressings of the Louis Armstrong's Hot Five 1926 versions of
> 'Cornet Chop Suey' and 'Georgia Grind'. This means I can now compare
> all 6 titles from that session from masters, and complete an exercise
> I started some time ago. The preliminary results of this indicated
> that the masters (nominally intended to be 80 rpm, Okeh's 'standard'
> at the time) were actually cut starting at ~81.6 and ending at ~82.4
> rpm. As soon as I get time, I'll complete the rewrite of this.
>
> In short, if you want to be absolutely thorough on this topic, you
> need to check & if neccessary correct the pitch of each session.
> Fortunately, the lathe *usually* behaves the same throughout a
> session... but don't count on it!
>
> A year or so ago I wrote a couple of rather rambling articles on
> pitching American & British Jazz & Dance 78s, which can be seen on my
> website at:
>
> www.normanfield.fsnet.co.uk/pitch.htm


The original reason I brought up this article is for the purposes of
Project Gramophone discussion and planning.

As currently envisioned, Project Gramophone will concentrate on
producing very high quality *raw digital transfers* of original source
78 rpm discs (and other formats, too, but the primary source material
will be 78 rpm discs).

The digital raw transfers (at 96k/24bit resolution) will then be
archived and made available online in "library fashion", such as at
the Internet Archive, for use by those who wish to digitally restore
them.

Of course, it is also envisioned, as allowed by law and proper
arrangements with various rights holders (dependent upon country), to
create listenable, online versions (probably streaming audio) of these
raw transfers. Since most of the transfers probably will not, right
away, be properly digitally restored, some automated (and rudimentary)
restoration will be done on them to produce at least listenable
versions for the general public -- they will obviously not be pitch
corrected unless there are clever ways to autodetect the necessary
pitch adjustment (and get it right most of the time.)

It is clear that because of the large numbers of sides needing to be
raw transferred, there is no time to spend on adjusting the exact
speed of the original recording -- this requires proper pitching and
the like (not to mention the variable speed of recording due to cutter
drag as Norman mentions above!) Fortunately, changing the pitch during
the digital restoration step is trivially easy to do by simple
resampling (and the high digital resolution 96k/24bit assures this can
be done with effectively zero impact -- even variable cutting lathe
speed can be digitally corrected so long as we can determine the speed
variation curve.)

However, this still begs the question as to what speed(s) Project
Gramophone should standardize upon for the raw transfer stage since
it is important for several reasons to settle upon a standard speed
or set of speeds and religiously keep to them (the exact speed of raw
transfer will, of course, be recorded in the metadata associated with
the raw transfer, and will be available to the restoration engineer.)

For example, should PrGram simply set 78.26 rpm for all transfers of
78 rpm discs, regardless of era and recording company? Or should we
settle upon a set of speeds? One advantage of establishing one and
only one speed is that it is easier to calibrate and keep calibrated
the turntables. This is especially important when we set up multiple
transfer "stations" run by different groups. We have to keep things
as simple as possible, yet assure strict conformity to the same set
of standards so we get consistent results across the board, and those
who later digitally restore the raw transfers know the transfer speed,
the equalization, etc., etc.

Thoughts?

Jon Noring