Nice to see another subscriber interested in code and roll records.
While recovering "trapped" performances admirable, I would suggest that
the Holy Grail #2 of many MMDers is to create a practical algorithm to
convert between reproducing roll types.

The difficulties once again are the stack being split differently, the
heavy interaction between coding and quantity of notes on the Duo-Art,
there being several Ampicos (A, B, and with very early Stoddard-Ampico
different as well), and Welte being very sensitive to timing.

Perhaps creating a graded test roll with carefully planned notes,
accents, successful catches and those that just work would help
characterize an agreeable piano to start. This is more of a task for
someone familiar with the science of mechanics (no, put those spark
plugs away), than the adept C-coder, but the task will require
considerable code to make useful.

In the meanwhile, Bob, do you know anyone trying to acquire a pair of
IBM Selectric Composers (fully mechanical), manual, literature and
hundreds of fonts?

Karl¶
kap@firedragon.com