[ Ref. Piotr Barcz in 220321 MMD ]

I attempted this method back in the late 1990s. Scanning this way adds
in a level of complexity. Rolls are many meters long. Multi-tune rolls
can be as long as a football pitch: 80 to 100 meters.

There is a document online called 'Music from Photoshop.' This works
with a program called TIFF to CIS which converts the stitched-together
TIFF file images. The stitching of the pages gets real tedious after
a few pages. I still have 600 book stencils to scan this way as they
are mimeographs copies of books, which do not have perforated holes.

Some sort of guide is needed that moves along with the roll to
accurately measure how large each frame is. If the roll can be
sacrificed, direct marks or tape on the roll works. A wheel driven
odometer could also work. The displayed numbers would have to be in
each frame.

An alternative I have been experimenting with is to use an old junk
iPhone. I have an iPhone 6 which I saved for this purpose. The camera
is many megapixels, more resolution than a scanner. The downside is
that the camera does have some non-linear distortion.

It is best to mount the iPhone with a rigid support. I have an old
copy stand, an old photographic enlarger would also work. Diffuse,
even illumination also helps.

The ear buds that came with the iPhone have a switch for controlling
volume. When in camera mode this will take a photograph. (The volume
controls on the phone can also work as a shutter release.) They even
emulate the sound of the shutter motor.

There are two ways to capture: frame by frame or to record the roll
as a video. Apple QuickTime can then be used to save the 'movie' as a
stack of frames. They still need to be stitched together. Most of the
frames will be thrown out as they have redundant information. There
are programs that can convert YouTube videos to MIDI when the tracker
bar can be photographed clearly. Mike Ames converted a large number
of music rolls with a simple video camera.

My own scanner uses a pinch roller and CIS array, based on a design
by Spencer Chase for the IAMMP back in the early 2000s. I changed the
controller electronics from the Mark3 design to Arduino (based on the
AVR ATmega328 RISC-based microcontroller. The output is USB.

The Arduino is fast enough to clock the CIS at the 1-Mhz shift clock
as recommended. Each line does have to be buffered as the USB is a bit
slower. This does have the advantage that the CIS run-length encoding
can be done in the Arduino itself.

My method is to clock the data into a ring buffer. The host computer
sends a string command to scan one line. This can be combined with a
command to advance the paper one step.

If I was re doing this, I would probably now use Ethernet. That would
give between 10Megabits and 100 for a basic controller. A Raspberry-pi
controller could be used to clock the GPIO of the CIS using a python
script. Warren Trachtman was working on such a system at the time of
his passing.

Spool acceleration is not that much of a problem. Some rolls types
like the APP rolls I am scanning have acceleration encoded into the
roll. Robbie the editor of this digest has done a lot of work in this
area. Shown the acceleration to be linear dependent on the take up
spool diameter.

Julie Porter
Martinez, California