Several people have asked me over the past few weeks how the tempos
were chosen for the new Telarc CD of Rachmaninoff's Ampico rolls.
Tempo is extremely important, of course, so I thought it would be
worthwhile to write a short article to be included with the liner notes
in the upcoming Telarc CD explaining how the choices were made. I have
attached a copy of that article, shortened slightly for the sake of
brevity. With best regards,
[ From the liner notes accompanying CD-80491 ]
TEMPO AND ACCELERATION OF MUSIC ROLLS
by Wayne Stahnke
The performances on this Compact Disc were recreated using a novel
three-step method for playing music rolls. In this process, a music
roll is first read by a specially designed scanner. By means of phase-
locked-loop technology, the resulting scan is then used to reconstruct
the master roll from which the music roll was perforated. Finally, the
restored master roll is translated by a computer program that contains
a mathematical model of the pneumatic mechanism originally used to play
the roll, yielding a disk file suitable for use with modern electronic
The first two steps are straightforward, and require neither interpre-
tation nor modification of the music roll data in any way. Since the
information embodied in music rolls is discrete, the process outlined
above yields an exact reconstruction of the master roll. Any diffi-
culties that arise from damaged rolls can be overcome by special
In contrast, the third step (translating the restored master roll)
necessarily involves judgments, and these can have a profound effect
on the ultimate presentation of the musical material. I am speaking
here of the individual adjustments that must be made for each roll.
The most important of these are tempo and acceleration.
In the case of Rachmaninoff, we are in a particularly favorable
position. The great pianist recorded a total of 35 selections on
music rolls, and for 29 of these he also made audio recordings.
Furthermore, Rachmaninoff was famous for playing a composition at the
same tempo from one performance to the next, sometimes for many years.
This reflected his conviction that there is a "best" way to play any
given piece. Thus, we have a reference that allows us to check the
tempo markings on most of Rachmaninoff's music rolls.
To be sure, the rotational speed of audio discs in the 1920s was not
yet completely standardized. There were variations among manufactur-
ers, and even for a single manufacturer the speed would vary from one
period of time to another. However, during the era in which these re-
cordings were made, the concert pitch of 440 Hz for A above middle C
was absolutely established. Thus, by adjusting turntable speed to play
a recording at an exact pitch, we can arrive at an exact playing time.
There has been much discussion about the relatively fast tempos on
many disc recordings of the 1920s. Some audiophiles have suggested
that pianists were forced to play pieces faster than they wished in
order to fit them on a single side. (The playing time of 78-rpm
records is limited to about four and a half minutes per side.) This
may or may not have been the case with other pianists, but in my
thinking Rachmaninoff was of such resolute character, and took his
art so seriously, that he would never have allowed himself to be
coerced by a recording technician on a matter of such musical import.
Far better to record another, shorter piece at the right tempo.
With all this in mind, I decided to match the playing time of the
translated music rolls to the playing time of Rachmaninoff's disk
recordings, regardless of the tempo indications on the rolls them-
selves. The difference in tempo between the two was usually small,
but in a few cases this procedure led to some startling changes. For
example, the music roll of the Prelude in G Minor (recorded on March
17, 1919, at Rachmaninoff's first recording session for Ampico) is
marked "Tempo 90," denoting a paper speed of nine feet per minute.
The corresponding Victor disc was recorded just a year and two months
later, on May 17, 1920. In order to match the playing times, I had to
effectively move the roll at 11.3 feet per minute--more than 25 percent
faster than indicated! This is a huge difference, and one that marked-
ly changes the character of the performance. It is supported by
compelling evidence, however.
To resolve the issue of tempo completely, I also turned my attention
to "acceleration" (I mean this in mechanical, not musical, terms).
The tempo indication actually denotes the paper speed only at the start
of a music roll, as it begins to unwind from its feed spool onto a
rotating take-up spool. In fact, as the diameter of the take-up spool
increases due to the continuously increasing paper load, the music roll
is pulled faster and faster--it is accelerated.
I first noticed that something was amiss with acceleration many years
ago while playing one of Rachmaninoff's early music roll recordings.
Near the end of the selection, where the score is marked "tempo primo,"
the pace seemed to be noticeably faster than at the beginning. Some
easy measurements confirmed the impression. Now, I had read that
Rachmaninoff was very good at returning to an original tempo when he
chose to do so. Thus my suspicions were aroused. What unexplained
phenomenon was causing the acceleration?
There could be only one culprit: a mismatch between the paper-handling
geometry of the recording and playback instruments. Accordingly,
I undertook a patent search and uncovered a surprising fact. The
recording machine at the Ampico studios in the early days pulled the
paper using a capstan, or constant-speed roller, with only a single
layer of paper on its surface. This arrangement caused the paper to
feed at a constant linear velocity. By contrast, the finished
production music rolls were accelerated on playback.
Unfortunately, this original Ampico recording machine was replaced
around 1925 by a new one that used a take-up spool, exactly as in the
playback instruments. Thus there are two types of Ampico music rolls:
those that should be played at a constant speed, and those that should
be accelerated. The dividing line between the two seems to date from
The new recorder was designed by Dr. Clarence Hickman, who was hired
away from the National Bureau of Standards by the Ampico company in
1924 to bring scientific methodology to reproducing pianos, which had
previously been designed almost entirely by cut and try. When Dr.
Hickman (who had earned his Ph.D. in 1922 at Clark University) began
to work on improving the recording machine, he apparently perceived
acceleration as a problem to be solved.
I interviewed Dr. Hickman at his home in New York City on May 19,
1979, after several years of attempting to locate him. I had a lengthy
list of questions relating to the Ampico mechanism that I hoped he
could answer. Alas, when I asked him specifically about acceleration,
Dr. Hickman replied that he simply could not remember the details
after so many years. However, upon examining the patents (which I
took along for the purpose), Dr. Hickman agreed with my conclusion and
advised me to play the earlier rolls at a constant paper speed and
accelerate the later ones.
Thus, for most music rolls the decision either to play at constant
speed or to accelerate could be based simply on recording dates. For
those rolls for which acceleration was uncertain, I could only make the
decision by ear. Once I learned just what to listen for, however,
I found it surprisingly easy to tell accelerated rolls from unaccelera-
ted ones. With the decision made in each instance I adjusted playing
time, as described above. (Acceleration has to be determined first,
since it affects playing time.)
It is therefore fair to say that the judgments regarding tempo and
acceleration have been made on the basis of purely objective measure-
ments. Furthermore, these measurements derive directly from compelling
historical evidence. This is deeply satisfying, because it allows us
to set aside any concerns we would otherwise have in this regard.