I have been recording and playing piano rolls for years, using various
techniques.  The first time I saw a roll I hadn't yet a player piano
but a pipe organ, and didn't know anything about pneumatic techniques.
So I made an electric reading device, which had 88 wires spaced as
required, making contact with a cylindrical brass bar located under
the roll paper.  The bar rotated so the wires kept it clean to ensure
a good contact.

I could read the rolls which have the tracks made of a row of holes
thanks to an electronic circuit generating a short, adjustable, delay
to avoid the interruption caused by the gaps between the roll holes.
So I could play the pipe organ, and other electronic musical
instruments.

When I had a player piano, I learned pneumatic techniques and made a
recording machine using a pneumatic stack operating contacts connected
to a MIDI acquisition device, with a software I made myself, using a
Commodore 64 computer (many years ago...!).

I discovered that the contacts where very critical because they tend
to bounce causing fast (4-5 msec) repeated on-offs of the signal.

This problem is important for anyone who is interested in MIDI
recording using such home-made devices.

I solved the problem with a software which validated the signal only
after a short delay, i.e., when the contacts had stabilized.

This delay depends on the physical construction of the contacts
themselves, and may be measured by an oscilloscope.

By this technique I recorded many rolls, including Duo-Art rolls with
their expression signals and could play them perfectly on a Duo-Art
reproducing piano equipped with pneumatic valves on its tracker bar.

I heard of Terry Smythe in Canada who made many roll recordings, but I
don't know what device he used: http://www.terrysmythe.ca/smythe.htm 

Leonardo Forti
Italy