In digest 95.11.24 Walter Tenten hits upon one of the
least-talked about but, IMHO, most critical aspects of transcribing
roll data into faithful electronic copies: The effective pneumatic
"length" (both in their time and displacement senses) of the holes as
influenced by the non-uniform size and placement of tracker bar holes,
as they differ among formats.  (WHEW!)  I will assume (always
dangerous!) until someone tells me differently that most if not all of
the optical roll scanning techniques tried so far "look" at the roll
along a straight, essentially zero width line which is perpendicular
to the line of travel of the paper.  Reading even a simple
non-expression roll in such a manner cannot, absent careful data
manipulation, produce a sufficiently faithful electronic copy, either
for electronic playback or for producing new copies of the roll.
Typically, the optical sensor will trigger "on" as soon as the leading
edge of the hole passes the sensing line and will not go "off" until
the trailing edge has passed the same line.  Pneumatic sensing
requires that the hole be a bit more open (or closed) before a
transition is sensed and so will produce a shorter duration for the
hole than will optical sensing.  Consequently, the electronic image
will sound "smeared" as notes ending will overlap following notes
sounding by "half-a-hole" or so.

	Now, complicate this by such considerations as this:  On a
Duo-Art tracker bar, the holes for the eight intensity tracks are
approximately 0.26" in advance of the note holes and are approximately
three times as "long" as the note holes.  The resulting pneumatic
performance of this tracker bar MUST be duplicated.  On the Ampico "B"
tracker bar, some holes are effectively elongated by being routed out
and the intensity holes (2B, 4B, 6B, 6T, 4T, and 2T) are "delayed" by
several thousandths of an inch beyond the note holes (bet you never
noticed!)  This is presumedly what allows "B" coded rolls to reproduce
relatively accurately on "A" mechanisms.  When played on "B"
mechanisms, the faster actuating "B" expression mechanisms are delayed
by this distance so that the effective expression between the "A" and
the "B" is equalized.  (Anybody want to discourse on this as a
separate topic?)

	Now I'm quite sure that a data manipulation algorithm could be
produced that could compensate for these problems.  Remember though,
you'll need a different algorithm for EVERY different roll format and
EVERY specified tempo value!  But why bother with all this when a
*properly designed and used* pneumatic reader cleverly ignores (or
automatically compensates) for these considerations.  After all, the
goal should be to reproduce the (digital) pneumatic performance of the
tracker bar as it responds to a roll.

	And of what use is the finished electronic image?  In my
estimation, there are three: 1)To produce an image that can be used,
with properly designed equipment, to produce new copies of the roll.
Richard Tonnesen is probably the leader in this process.  2) To allow
existing *properly restored* pneumatic instruments to be operated
*faithfully* by means of a sensitively designed (and reversibly
installed) apparatus which would be, in effect, an electronic
replacement for the paper.  (Remember the CC-3 Cassette Converter of
the '70's?!)  3) The third use would be to allow existing roll
performances to be recreated, with proper expression, on modern day
synthesizers.  This, while technically interesting, does not seem to
me to be as much of a necessity as the others.  However, I loudly
applaud the efforts of Richard Brandle (and others) who are engaged in
efforts to this end.  The dual of this, i.e., manipulating modern
(MIDI) performances so that they can by used, perhaps with the help of
the equipment in 2) above, on vintage instruments, can also be
debated.

	I will have more to say on 2) above in a future post.

-John Grant