If you have original note records from the three big players --
Welte-Mignon, Ampico and Duo-Art -- you have done much better than
I would have expected.  Those note records give quite different classes
of data.  But I am assuming you are planting the seed for a fascinating
technical discussion for which it is not absolutely necessary to have
a vault of original note records.

I believe the Welte-Mignon recorder had graphite inked wheels just
touching the paper and they were pressed harder depending on hammer
position.  With some curve-fitting this would give essentially the data
Wayne Stahnke decodes from his hammer position recordings.

(Ingenious resistance measurement of Welte-Mignon recording trace width
and almost any amplifier can recreate the original hammer positions
against time, and the sound would thus be recreated.  This was said to
have been done at Welte with a unique cabinet piano.)

(If the Welte-Mignon merely recorded key position, the data would
have been several degrees of separation from the sound of the string.
The data would not match the musical output.  Without simultaneously
measuring the sound or hammer position, there is no way back up the
data pipe and you are doomed to being a crafty editor to simulate the
pianists performance. )

(In late Welte pipe organ recording, and in other contemporary pipe
organ recording, especially with pneumatic or electro-pneumatic action,
the note record can be virtually the same as the pallet valve opening.
In that case, the simple recording of the note events from the key &
valve & pallet action is indistinguishable from live playing.  But
I digress.)

The best Duo-Art system, if I remember correctly, had a pencil
recording of the keys and required a technician to try to follow the
dynamics in real time, his knob inputs automatically marked along side
the note record.  Since erratic manipulation of piano keys can produce
uncorrelated and irregular sounds, a key contact is unlikely to give
enough information to accurately record timing, much less dynamics.
Editing necessarily followed.

I am glad I did not have to be there, it would have been maddening to
try to keep the original performance in memory and spend long hours
trying to bring it back to life.  Early Ampico and other systems would
have done little better with pencil recorders following the keyboard.

Ampico with Hickman's spark recorder gave loft and return events which
were transcribed into note events and intensities, cumbersome but
giving a reliable record.

Turning that old data into new performances would be close to the
original performance with Welte-Mignon and spark-recorder Ampico, but
you will find it difficult to have the performers give their opinion of
the results.

Karl Petersen