Details of my Piano Roll Scanner

With a background in government administration and management, I was not at all prepared to undertake this endeavor. I could not have done it without the help (in random order) of Richard Stibbons, Jim Cullen, Bob Pinsker, Albert deBoer, Warren Trachtman, Gene Gerety, Spencer Chase, Kevin Keymer, Jack Breen, Ross Chapman, Bill Earnest, Peter Phillips, Wes Melander, Milton Schenk, James Stoyka, John Klassen, Henry Bohncke, Bill Flynt, to name but a few. My apologies to those I may have overlooked. The essential issue is that if I can successfully build and use a piano roll scanner, then most anybody can.

Basically, my scanner started out in life as a rusty old AutoTypist roll frame, an early automatic typewriter built around pneumatic player piano technology. After cleaning it up, painting, and substituting some conventional player piano spool box parts, I was ready to go, or so I thought. It has been an interesting adventure. Caution..... My scanner, while adequate from my viewpoint, is modest compared to what others have achieved.

Those interested in following in my footsteps are encouraged to participate in the Rollscanners discussion group. See:

And visit the following web site for detailed technical information.
Click on "Scanner Design".

Overview of my scanner, with the CIS sensor in place, roll attached, and computer ready to go. Note the piece of corner moulding just above the keyboard, serves as a handy parking lot for the CIS during roll changes. Of the 2 switches on the face, the left one is a start/stop toggle switch for the motor. The right one is a 10-turn pot to control the quality of the roll image. Richard Stibbons' MK3 circuit board, modified by Kevin Keymer, is mounted in the right side. The optical rotary encoder riding on top reports paper movement to the software.

Close-up of the right hand side of the CIS censor. Note the common headless finishing nail to provide a rigid constant positioning point for each time I return the CIS to its operational position. The 2 common brass rods provide a means to ensure the paper flows in as flat and stable manner as possible past the CIS sensors. BTW, the positioning nail could be dispensed with, for varying roll widths, to facilitate simplicity of placing the CIS as centered as possible over the roll, regardless of its width.

As the forward/rewind pneumatics of the original AutoTypist were removed, I had to find another way to facilitate forward/rewind. I did so with a super simple clutch made from a simple hex head bolt. Originally, the 2 wheels were pneumatically clamped together for rewind under friction. The bolt serves the same purpose, I just drilled a hole through both wheels. The bolt is inserted during forward scan operation, and the bolt is partially withdrawn for manual rewind.

Note a common friction brake from a conventional player piano spool box, to maintain a stable flow of roll paper under the CIS sensor.

As many of my rolls are fragile, I choose to rewind by hand. If anything goes wrong I can sense it immediately and deal with it. Very easy to damage rolls on motor driven automatic rewind. What you see here is a common gear from a typical player piano spool box mechanism. I just drilled a hole through its perimeter and screwed a piece of doweling to provide a simple thumb/finger hold.

This is a bird's eye view of my motor drive. After trying numerous different motors, I settled on this pancake 12vdc brushless gear-head motor, salvaged out of a piece of obsolete medical equipment. My experiments with a variaty of motors, including windshield wiper motors, eventually led me to a requirement for high torque and low shaft rotation. That pretty well narrows it down to a gear head motor. This particular motor is a Model GM15009005 PMI motor by KollMorgen Corp, which I could find no reference for on the internet, other than a series of hi-end brushless motors by this company. Shaft rotation is exceedingly slow due to the gear-head, but capable of huge amounts of torque even at low-voltage.

I only had 2 gears to play with, and 6 voltage selections off my multi-voltage 12vdc/800mAmp power supply. The target paper flow under the CIS sensor is now up to 8'/minute. To achieve this, I tinkered with my 2 gears and the voltages and arrived at an unconventional arrangement. I have the large gear attached to the gear-head motor, and the small gear attached to the scanner drive shaft. Then I set the voltage to 9vdc! The combination works like a champ! Loads of power and torque, seemingly rock steady, super quiet, and no overheating.

Note also the idler wheel from a conventional player piano spool box assembly. It serves only to take up the slack in the chain drive and minimize the risk of the chain slipping off the small gear. The 2 gears are from the original AutoTypist assembly, while the chain is a conventional player piano chain. Fortunately, the old AutoTypist, designed around pneumatic player piano technology, used the same chain drive specifications.

My final configuration required a 12vdc/800mAmp variable voltage power supply for the motor, a 12vdc/300mAmp variable voltage for the light source, and modest fixed dc voltage power supply for the MK3a board. For the latter I found a 6vdc power supply. The MK3a has its own voltage regulator to 5vdc, so an external power supply between 6 and 15vdc will do nicely.

My final light source is an unusual extremely slim ccfl flourescent tube that operates off a 12vdc inverter. I experimented with a variety of light sources such as an under-the-counter flourescent fixture with a 110vac 15" x 1" tube, a 110vac 12" x 1/2" flourescent tube, a 110 vac 12" x 1" incandescent tube, and found all will provide adequate light. What is critical is distance from the CIS and its placement slightly below an imaginary right angle line from the face of the CIS. I eventually settled on this super slim 1/8" x 12" floiurescent tube available from AllElectronics in the Los Angeles area. Thank you Albert for finding this nifty little light source.

Took about an hour of tinkering with the placement of the tube, which incidentally is imbedded within its own 3-sided shroud, so the splash of light is very narrowly focused. By chance the shroud for the tube is exactly the width of my roll transport, so it is simply held in place by friction and a single dab of plastic glue.

Operating it at its normal 12vdc, the light was fine for original rolls, but posed problems trying to strike a balance of adequate hole pattern image and specks when attempting to scan newer rolls perforated on light highly translucent paper. Again, I tinkered with the voltage and discovered that this 12vdc bulb/inverter combination will work just fine at 6 and 9 volts. Below 6 volts, the lamp did not have even light distribution end to end. I settled on 9vdc, coupled with 2 layers of neutral density filter, which provided a light source adequate for all types of music rolls.

The little 12vdc inverter carries on it a "DANGER - HIGH VOLTAGE" warning label, so I mounted mine in a hidden location where it is highly unlikely I will ever touch it. The side you see in the pic is against the wall of a wooden file cabinet.

The lamp connector did not fit the inverter connector, notwithstanding that what I bought was supposed to be a matched pair. So I cut off the lamp connector, and soldered in a pair of single pin connectors salvaged off an old motherboard.

Note the optical rotary encoder swung over backwards to facilitate roll change.

Here can be seen the all-important MK3a board designed and developed initially by Richard Stibbons, enhanced by him, Wes Melander and Kevin Keymer up into the current MK3a board. The simplicity of his design is quite deceptive. Nevertheless, the board works very nicely. Details of his board are available at

With zero knowledge about electronics, I was/am at a distinct disadvantage when it came to this portion of the adventure. I still have little or no knowledge about electronics and programming, but Richard has done such an outstanding job at documenting his design that anybody can build one.

This is what I see on my computer screen during a roll scan. Basically, my scanner is now running at 8'/minute, continuous flow 12vdc pancake brushless gear-head motor running at 9vdc, transmissive 12vdc ccfl 12" x 1/8" lamp running at 9vdc, 119 lpi, zero acceleration into a fat take-up spool, an optical rotary encoder, manual rewind, no end-of-roll sensor other than audible ripping of paper off the core, and a 10-turn Bit-Twiddling pot.

And this is I scanning an Ampico roll.

These are approximate dimensions, using 1" x 6" finished lumber that is clean and straight grained (suggest birch or maple). The angles are approximate, with the objective being to use gravity to hold the optical encoder in place. The only really critical measurement is the gap between the paper and the face of the CIS. Start with ~1.5mm, then if necessary, small shims can be used to move the CIS slightly further away from the paper. The small dots top and bottom indicate where to drill pilot holes (suggest 1/8") through both side frames while clamped together. This is important to ensure paper alignment top and bottom are perfectly parallel to one another. The placement of the top and bottom chuck pilot holes allows for future mounting of nickelodeon and orchestrion rolls. All 4 pilot holes will end up larger by the size needed for the actual player piano parts used. In my scanner, I mounted the 2 side panels 13 1/2" apart.

While this image has poor resolution on your computer screen, it has been shrunk here for uniform display purposes. The actual image is about 3 1/2" times larger, so "right-click, save image as" and save this image to your computer in a directory of your choice, then print the image (dimensions.jpg) on 14" paper in landscape mode. Details should be quite legible.

Back to Rebirth