As an electrical engineer student fresh out of school in 1985, I went
to work for Hamilton Standard. I was first assigned to a programming
project for McDonnell Douglas's F/A-18 Fighter jet, in charge of
writing the test routines that it's "black Box" would perform upon
power-up and while it ran. When testing raw computer memory, I used a
few forms of tests -  cyclical redundancy tests, checksum tests, and
perhaps more intuitive to a piano test roll, pattern tests. The tests
were made in such a way to determine if the computer's memory retained
what was written to it, and to see if any memory bits were electrically
shorted together, and they the responded in the necessary timeframe.

The patterns that were used were specific for the types of problems we
would be able to detect, and were determined by circuit reliability
analysis. The simplest tests involved setting all the memory to "0"
and verify that is was all "0", set them to all "1" and verify they
were all "1", setting every other bit to "1"  and make sure there was
a "101010..." read, then set them to "010101..." and verify that
pattern, etc.

Applying this theory to a piano test roll is obvious - we listen to
see if a note does or doesn't play, or if unwanted ciphers occur. But
I was wondering how more complex tests might apply to a test roll.
Could specific patterns be utilized to detect if adjacent notes and/or
other pneumatic systems found in the Duo-Art were "shorted" together?
Perhaps the possibility of playing note #4 could effect note #60 under
some condition, or even effect another pneumatic system.

Have current test-rolls exhausted all of these kinds of test
possibilities, or are there still areas where more analysis might lead
to better system coverage?

Karl Ellison
Salem, Massachusetts