I ran across this e-mail discourse of January, 2003, which I wrote to
Peter Phillips of Australia in answer to a query about my roll-making
techniques of circa 1978.  It might be of interest to some.

 - - -

Peter, Thanks for your continued interest.

The Liberace rolls:

As I remember, Frank Adams (AMR, Seattle) got the QRS rolls and sent
them to me for Ampico coding, whereupon John Malone at Play-Rite recut
20 at a time, for distribution by Adams.  QRS also sent a tape cassette
of the Liberace recording session, but it really was of little use to me.
I simply added dynamic coding that seemed musically appropriate to me
at the time.

In each case, the first task was to edit the sustain pedal track.  In
every case it was so butchered by the QRS editor that the roll would be
useless to the knowledgeable listener, even with Ampico or Duo-Art
coding.

Due to the sluggish nature of pneumatic pedalling (much slower than
real pianists), I had to extend many notes at the beginning of a "pedal
down" event, so that those notes could be captured by the sustain
action when the dampers actually lifted.  This sort of thing is quite
apparent in almost any original hand-played roll.

Further, enough "land" had to be left in each "pedal up" event to allow
even very sluggish pianos' dampers to drop, before rising again at the
following "pedal down" event -- typically about 3/4 inch, depending on
paper travel speed.  This requirement further required extension of
many bass notes, since often a pianist will barely touch a low bass
note, catching it quickly with the pedal.  I found this to be the case
in my own recording sessions, using the crude (by today's standards)
recording apparatus that I developed in the early 70s.

Mechanics of Roll Editing:

The actual "cutting and pasting" was done on an editing frame that I
constructed for the purpose.  I will send you a picture of it with my
next installment.  It allowed about 12 inches of a roll to be examined,
taped, punched, repaired, whatever, at a time.  For convenience the
working area was tilted at about 30 degrees from the horizontal.  In
the work area was a grid pattern (obtained from Player Piano Company,
Wichita, Kansas, but now out of stock), representing all 83 playing
notes plus control channels at the bass and treble ends.  I still use
this frame for minor repairs of damaged rolls.

Back to the Liberace rolls:  My guess is that the Disklavier version
is simply a copy of the production roll that Frank Adams marketed in
those days, hence the similarity, and possibly with editing by others,
in-between.

There's a lot more to this story, which I'm happy to relate to you
for whatever use you can find.  Untold so far (and maybe unknown to
many) is that I designed and built the electronic control apparatus
that has been in use at Play-Rite (John Malone's firm) for all these
years, beginning in about 1976.  About every two years I typically
spend about a week at Turlock, updating and debugging that equipment.
It's quite archaic by today's computer standards, but as John says,
"It ain't broke", so he doesn't want to "fix it".  He and I have spent
much time chasing elusive "bugs" that would have rendered a batch of
recuttings useless, and it has been quite reliable for many years now.

I've considered writing up a summary of that equipment's functions
for those techies in MMD who are interested, with John's approval, of
course.

Until next installment...

Bill Flynt
Dallas, Texas
17 January 2003

 [ The reproducing piano rolls coded by Bill Flynt are listed
 [ at http://mmd.foxtail.com/MMMedia/PostWW2.html 
 [ -- Robbie