Much has been written about preserving music rolls, all of which has
merit.  The more original rolls that I manipulate, the more convinced
I am that some sort of preservation is necessary if these performances
are to be preserved for anyone in the future who is interested in them.

The first machine I built to _read_ music rolls (there was no optical
scanning involved) consisted of pneumatically operated switches which
were read by a UM-1 board produced by Octet Designs.  The software for
this operation was written by Richard Brandle who has successfully
maintained updated versions of "Wind", his excellent roll editing
program.  The results were good, not excellent,  mainly because I was
not precise enough in the construction of the hardware.  The software
was great and offered options that no other software out there I have
tried is able to do.

When optical scanning started, I built an optical scanner with the
help of Terry Smythe using the electronic components he has described
in previous posts to this forum.  The results were excellent and many
rolls passed under the bar salvaged from a flatbed scanner.  My reason
for optical scanning was the reproduction of the original rolls on a
parallel punch perforator.  Larry Doe's scanners (I think he built 12
of them) are works of art; they also do an precision job of scanning
music rolls!

The purchase of the Tonnesen's Custom Music Rolls operation provided
me with Richard Tonnesen's excellent roll _reader_.  It was never
called a "scanner" and indeed it reads the rolls pneumatically using
a tracker bar and 100 electrical switches.  Nothing too exotic about
that.

A true measure of Richard Tonnesen's genius comes with the software
that runs the above reader.  Richard spent what must have been very
long hours designing a software suite that would handle just about any
situation one would find in reading original rolls.  Here is a very
good example.

The reader does not have an automatic tracking device.  As we all know,
old rolls have a tendency to wander back and forth for whatever reason.
Built into the software is routine to determine just how far each hole
in the tracker bar is uncovered.  The hole that has the largest area
uncovered "wins" and a message appears on the monitor of the computer
running the program that note number such and such has been deleted.
If the area uncovered for both holes is the same, both notes are read
and appear in the editor.

It is no problem to remove the incorrect "on" (determined by listening
to the roll) when the roll is examined after the read.  Most of the
errors come with holes near the edges -- the expression tracks in
reproducing rolls are a good example -- Duo-Art rolls being the biggest
"offender."

For each type of roll, a header is created containing all information
about the roll.  The same principle is used in Wayne Stahnke's software
suite which uses the annotation file along with the punch file to tell
the perforator just what it is supposed to do.  There is a template in
the optical roll scanning software for this roll information, but a
number of roll scanners simply insert the title and nothing more.

The most important piece of information in the Tonnesen file header is
the roll type, i.e., 88-note, Ampico A, Ampico B, Duo-Art, and all of
the Welte types.  That "roll type" line is later read by the perforator
and the punch software is programmed to reproduce as closely as possible
the chain bridging in the original roll.  The chains for the expression
tracks are not the same as those for the playing notes.  Such bridges
are inserted automatically by the punch computer and no extra steps are
necessary to insert chain bridging.

Through the years, Janet and Richard used scans or information provided
by others.  The chain bridging was already present in those files.  To
punch these as they were laid out, the letters "IM" appear on the first
line of the punch file header and the punch will faithfully reproduce
just what is in the roll file -- in other words it punches the "image"
it sees in the file.  Some original roll manufacturers did not use
chain bridging.  The punch can reproduce those rolls exactly using the
"image" feature.

Optical roll scanners can be built that are extremely accurate, however,
I am unaware of one that will produce a file that can go directly to
the computer that drives the punch with no intervening computer steps.
The Tonnesen system does just that.  There are no intermediate steps
top go from a CIS file to a eRoll MIDI file to whatever.  The Tonnesen
reader produces a .prf file and that is what drives the punch computer.

I do look at the scan or read in "Wind" just to see if any "clinkers"
have crept in or if the original roll I just read has tears in the
edges of the paper.  It is quite easy to remove any stray punches, or
in some cases when I am not paying attention and the roll being scanned
tears off its spool, a sea of black dots on my editing screen.

Optical scanners can handle rolls that the Tonnesen reader would have
difficulty with, i.e., warped paper, tears in chaining, etc.  The
optical software also handles tracking or mis-tracking quite well.

I find the optical scan a bit more labor intensive than the "read"
I get from Tonnesen's machine as more steps are required to get the
scanned information into a form that a punch can "understand".  The
roll which is produced by the reader file is extremely accurate,
however, I do punch a test copy and compare it with the original making
minute changes where necessary.

Each "step" of the paper through the punch head is 0.022 inches, a very
small distance.  Richard did this on purpose and his punch does not
"step" the paper.  It is pulled continuously through the punch head.
You can get away with that using such a small step distance and a
relatively high revolution rate on the punch crankshaft.  It helps to
understand the genius involved in all of this if you consider that it
was all built prior to 1979 when electronics were in a totally different
phase than they are now.

An accident in connecting an ungrounded computer with the on-board
punch computer resulted in full line or mains voltage hitting the
onboard computer.  Chips were literally blown off their respective
boards!

The main processor board had not been manufactured in years and it
took a lot of talking and even more money to get the firm to retool
and fabricate two new boards for me.  Programming the board was done
by Richard Brandle using the original source code written by Tonnesen.

A number of the electronic chips Richard Tonnesen used are no
longer made and a search of surplus electronic supply houses began.
Fortunately all the chips were found and the punch communications
cable is now optically isolated from the host computer.

The punch software is also a marvel.  Have two repeated notes where
the end of one is too close to the beginning of the other?  Child's
play for Tonnesen's software, as it looks ahead in the punch file,
determines the space between repeated notes and truncates the offending
perforation so that there is enough "land" or space between the two
notes in order for it to repeat on the piano.  Scan away and preserve
all you can.

Not everyone has the tools to make music holes in paper, but electronic
files can be played on solenoid pianos or sound equipment and the
listener can hear what the music sounded like when recorded.  If MIDI
goes away, and I do not think it will, someone out there will write the
software to interpret the information in a MIDI file and make it usable
for decades to come.

By the same token, there will always be pneumatically operated automatic
musical instruments.  Their number may diminish, but someone, somewhere
will always need rolls to play them.

Ed Gaida - Preserving music by punching holes in paper
San Antonio, Texas