Mark Fontana wrote:

> Are the rolls generally encoded such that when played on a standard-sized
> takeup spool, the tempo will remain as intended despite the effect of
> othe increasing layers of paper accelerating the transport speed?

I believe the answer is no.  I encounted this question while working on a
project to convert live MIDI files to piano roll.  I asked QRS and they
do not currently compensate for tempo variation when encoding a single
roll.  They may sometimes use a different rate for the second song in a
medley roll.

> It occurred to me that I could include a function in my scanning software
> to adjust for the speed-up effect mathematically... Perhaps I could
> measure the diameter of the takeup spool (with music) at the end of a roll,
> then have the software correct the timing in the scanned data based on
> that measurement, the duration of the piece, and the original circumference
> of the takeup spool.

I have performed this experiment with the intention of calculating
"on-the-fly" a variable tempo adjustment when converting live-played MIDI
files to piano rolls.  I couldn't belive the results.  They indicated a
25 foot roll that started at a tempo of 70 was actually travelling at a
tempo of 85 at the end of the roll.   This is what the math said, but it
did not intuitively seem correct.  I was perplexed until  I read Zoltan
Janosy's thesis. He noted that this increase of tempo is countered
somewhat, "since the increasing mass (of paper on the take-up spool)
slows down the low torque pneumatic winding motor." Although he noted
that these forces won't exactly cancel each other out.

--Bill Jelen