[ Editor's note:
[ MMD member Wayne Stahnke developed the electric recording and repro-
[ ducing system for the Boesendorfer SE piano, and he holds a number
[ of patents related to modern solenoid piano systems. He is now a
[ consultant to Yamaha Corporation of Japan, and also owns his own
[ company, Live Performance, which creates music for the Disklavier,
[ some of which is derived from transcriptions of music rolls for
[ reproducing pianos.
[ In the 1970s Wayne built a precision pneumatic transcriber for 9/inch
[ rolls, which was used until he replaced the tracker bar and pouch
[ switches with an optical line-scanner camera, in order to transcribe
[ other formats such as Red Welte rolls. He notes that both pneumatic
[ and optical methods can produce excellent results if the transport is
[ well-built. The critical element of the process, he says, is the
[ measuring roller, which registers the distance along the paper with
[ 0.01 mm resolution (0.0004 inch).
[ Robbie Rhodes
In recent postings to the Digest, the term "master roll" appears several
times. This term seems to mean several different things, depending both
on the context and on the frame of reference of the contributor. This
note is an attempt to clarify the definition a little, and (I hope) to
shed some light on the subject of roll replication.
Master rolls are used to drive perforators. They contain the information
which ultimately finds its way into a performance, when a production roll
is used to operate an instrument. There are two quite different types of
master rolls, which are used in two vastly different ways. For the sake
of clarity, I will call these rolls "prototype rolls" and "pattern rolls"
in this note.
Prototype rolls were intended to be used, and were used, with perforators
that re-sampled the data when a production roll was made. As far as we
can tell, prototype rolls were in fact playable rolls; they could be
played on the same instruments that accepted production rolls. That is
the reason for the term "prototype rolls", because they are the
prototypes for the rolls which customers bought and played.
Production rolls made using the prototype-roll process differ from batch
to batch. On each production run there are inaccuracies in reading due to
bleed size, pouch porosity, and the like. However, the most obvious (and
by far the most egregious) difference from one batch of production rolls
to the next is caused by re-sampling. The production rolls are moved
intermittently in synchronism with the perforator ram, but asynchronously
with respect to the motion of the prototype roll. As a consequence,
there is an inherent inaccuracy of one punch advance with every leading
and every trailing edge of each perforation.
Thus single holes in the prototype roll will appear as either single or
double punches in the production roll; this is the best possible case.
Other inaccuracies can cause single holes to be missed altogether, or
stretched into triple punches. For an example of this phenomenon,
examine a Red Welte roll that has a test pattern beyond the rewind
perforation (these are not common, but every collection of moderate size
contains at least one).
"Pattern rolls" are very different animals from prototype rolls.
The purpose of a pattern roll is to trigger the interposers in the
perforator, row by row, as the paper advances. That is the reason
for the name "pattern roll."
One row in a pattern roll corresponds to one and only one roll in the
corresponding production roll. The total number of rows in the pattern
roll is exactly equal to the total number of rows in the production
Clearly, pattern rolls must move in lock-step with the perforator
crankshaft. The pattern roll may move continuously or intermittently,
but its motion must be such that, just at the instant the interposers are
to be triggered, the pattern roll has its corresponding row in place over
the tracker bar. This requires the use of sprocket holes, a
characteristic feature of pattern rolls.
In the interest of simplicity and durability, pattern rolls are usually
several times longer than production rolls. This allows a discrete row
of "punch" or "no punch" rows for each row in the production roll. They
are also usually wider, since sprocket holes on either side of the roll
are required. Thus pattern rolls cannot be played on the instruments
designed to accept production rolls. Perhaps for this reason, very few
pattern rolls have come down to us. The largest single collection of
pattern rolls is the Ampico master rolls, now in the possession of
Keystone Music Rolls of Bethlehem, Pennsylvania.
Production rolls made from pattern rolls are identical from batch to
batch. As an example of this, one can take an Ampico roll manufactured
in the 1920s and overlay it on a 1990s copy made by Keystone. The two
rolls match exactly; the scalloping, bridging, and punch pattern are
exactly the same.
Most production rolls produced in the United States were produced using
pattern rolls, with the corresponding synchronous perforators. Since
both the pattern rolls and perforators have now disappeared (with the
exception of the Ampico rolls and perforators mentioned above), we must
choose between two different methods of replicating the production rolls
One roll replication method simply uses the surviving production rolls as
prototype rolls. This is a natural approach, and the one that has been
used successfully to the present date. The other approach uses existing
production rolls as templates, from which the (now lost) pattern rolls
are recreated. Once the pattern roll is restored (usually in a computer,
not as a paper roll), we can once again make synchronous copies, exactly
as was done from the original pattern rolls in the 1920s.
For purposes of comparison, consider the sources of inaccuracies that are
associated with these alternate methods of roll replication. (We can, if
we like, quantify the contribution of each of these error sources; such a
discussion is beyond the scope of this note.)
First consider the sources of inaccuracies in using a production roll as
a prototype roll. In reverse chronological order, these are as follows:
(1) Inaccuracies in the (new) perforator used to manufacture the
(2) Errors due to re-sampling the (previously sampled) data; these
errors are exhibited as moire' effects, familiar to anyone who
has scanned a half-tone image on a computer scanner,
(3) Computational error (including roundoff error) in the computer
processing used to drive the perforator,
(4) Inaccuracies in the scanner (the number of error sources and
the error magnitudes are different in pneumatic readers and
the different types of optical scanners, but all have multiple
error sources that cannot be eliminated),
(5) Aberrations in the roll due to age and improper storage,
(6) Inaccuracies in the (original) perforator used to make the
production roll from the pattern roll.
In contrast, there is only one source of inaccuracies when replicating
a roll from the (restored) pattern roll. That is the entry given as (1)
above. There are therefore five more sources of error that contribute to
inaccuracies when using a production roll as a prototype roll.
Note that a (new) perforator can be built to any degree of precision we
want. Thus, we can equal or exceed the accuracy of the original rolls
manufactured during the 1920s, if we choose.
When we make copies of production rolls, we have fundamentally two
different approaches at our disposal. One approach uses the surviving
production roll as a prototype roll. A copy made using this approach
must of necessity be less accurate than the original; there are six
sources of error with this approach, each of which must be very small
if an acceptable copy is to result.
The other approach restores the pattern roll from which the production
roll was made. A copy made using this approach equals or exceeds the
original production roll in accuracy.