This is a correction and amplification upon the last section of my
previous article -- it seems I changed units suddenly during editing
and I didn't see the error until now.  My apologies...

** "Will a recut of an Ampico roll, originally perforated at 30 steps
per inch, and re-sampled at 60 steps per inch, sound as good
as the original?"

   No, unless it is synchronously sampled to preserve the timing, in
which case you should have a perfect replication.  An Ampico roll
playing at Tempo 80 is traveling at 1.6 inches per second, therefore
48 perforations per second are moving past the tracker bar.  Stahnke
was sampling the position of his primary valve pouches at 100 samples
per second for the Cassette Converter system; this is equivalent to
perforating, and then playing back, a recut punched at 62.5 steps per
inch.

   If the re-sampled data is displayed on a "piano roll image" screen
on the computer one can see that chords, originally struck simul-
taneously, will "ripple apart" about every two seconds, corresponding
to the beat frequency between the 50 Hz sub-harmonic of the
re-sampling versus the 48 Hz sample frequency of the roll.

   The other major contributor to timing errors is the variation of
pouch switch delays from channel to channel.  In the digest of
95.04.21 Richard Tonnesen says that the pouch switches in his reader
have, typically, a 5-millisecond turn-on delay and 25 ms turn-off
delay.  Both durations are affected by the force of the pouch
reference spring (a short length of music wire) which "tends to get
out of adjustment," Richard reports.  Fortunately, his springs can
be adjusted; the pouch microswitches used by Wayne Stahnke and
John Malone/PlayRite also contain springs but they're not adjustable,
thus careful calibration and correction of the reader data is
necessary.

   Notice that the high resolution re-cut roll is playing at about
100 steps per second, or 10 milliseconds per step.  This "sampling
period" is so close to the delay times of the pouch switches that any
little variation in their time delays (caused by spring varience or a
clogged bleed) will affect the relative timing of events within a
chord strike, or worse yet, cause an expression command to arrive too
late or become lengthend or shortend.  Tonnesen and Stahkne both
operate their readers at about two feet per minute (0.4 inch/second)
so that the pouch switch delays are much less than the time between
perforations of the original roll; this helps prevent elongation of
the notes in the data file (and the copy roll).

   In anticipation of a howl of protest from owners of reproducing
pianos, let me acknowledge that the high quality re-cut music rolls
produced first by Larry Givens (and continuing to be produced by
people such as PlayRite, Keystone and Custom Music Rolls) are
entirely satisfactory and satisfying.  It is the engineer's challenge
to make the perfect replica; if it matches the original hole-for-hole
and scallop-for-scallop then he has achieved his goal.

   And if he has a supply of paper from the 1920's (of good vintage,
well aged) he could make a replica of the original roll that would
confound the experts.  See what I mean about counterfeiting?

-- Robbie Rhodes, 2 July 1995