In digest 960307 Wayne Stahnke wrote:

> To hear the performance, you need a program that models the
> behavior of Ampico instruments.

We have applied the term "Emulator" to any hardware and/or software
process which duplicates the essential (as in "essence") function of
another "real" process.  The internal functioning of the emulator is
immaterial -- the only requirements is that the output data from the
emulator is the same as the output data from the "real" process
which is emulated.

An emulator for a reproducing piano roll receives as input the same
binary data that exists at the tracker bar and valves of a pneumatic
piano.  The output data from the emulator is sound: acoustic sound
pressure, generated by an electronic synthesizer, amplified, and
converted to acoustic pressure by a loudspeaker.

In evaluating the faithfulness of the processes which Wayne Stahnke
and I are discussing it is appropriate to ask, "What is the Measure
of Success?"

The very practical, and obvious, Measure of Success in preserving
music roll binary data is to punch a roll from the recovered Master
Roll file, lay it against an original, and inspect it!

The measure of success of the reproducing piano emulator is in the
listening.  Horst Mohr built a _complete_ Ampico-A emulator.  The
input is an Ampico roll, the output is music from a loudspeaker.
Within Horst's emulator system is a software element which converts
Ampico binary data into MIDI data.  If Horst has a _really good_
piano synth (which converts MIDI data into electric signals), a
listener should be unable to discern the difference between his
emulator system and a real Ampico piano.

For simplicity we will call the "Ampico to MIDI" process the
"Ampico Emulator", while bearing in mind that this is simply one of
the elements of the overall Ampico-to-sound emulator system.

Emulators for "Ampico to MIDI", "Duo-Art to MIDI" and "Welte to
MIDI" have been written by Richard Brandle and Wayne Stahnke.  I
wrote a "Ampico-B to MIDI" emulator first, because that is the
simplest, and I am developing my "Ampico-A to MIDI" emulator.  The
latter seems to yield musically pleasant results, but I haven't yet
held "comparison listening tests" to determine its Measure of
Success!  I will prepare a MIDI file of the Rachmaninoff roll and
post it to Jody's ftp account for all to evaluate.  But be patient;
it will be several weeks.

How does one create an Ampico roll from an original DuoArt or Welte
performance?  Answer: by using a translation process.  We use this
term because of the analogy to translating text from one language to
another.  The content is the same in both languages, the difference
is in the vocabulary and grammar rules.  Most important, the data
is already edited for a split-stack piano.

Translation of music roll data between different reproducing piano
formats is much, much simpler than converting a MIDI performance
recording to play on a pneumatic reproducing piano.  This is because
a MIDI output device, like a synthesizer or solenoid piano, can play
notes simultaneously with differing velocities, whereas the
pneumatic reproducing piano (and its electric relative, the Marantz
PianoCorder), must strike simultaneous notes at the same velocity.
Converting MIDI files to rolls for these "split-stack" machines
requires the same tedious manual editing that was needed in the
1920's.  I believe that a computer process can only aid, but not
supplant, the work of a skilled, musically-trained editor.

The executives at Ampico in 1927 probably asked Mr. Stoddard and
Dr. Hickman if an automatic machine could be built to create music
rolls directly from the data recorded by the spark chronograph
recorded.  I wonder what the reply might have been...!

Translator boxes were built in past years by Wayne Stahnke and Will
Dahlgren which perform simple real-time conversion between serial
data formats for devices such as PianoCorder and IMI Cassette
Converter.  For the most part the users of these simple boxes are
satisfied if the music is pleasant-sounding; knowledgeable listeners
("piano critics") claim that the translation is deficient.  The
simple boxes do have a notable feature:  they seem immune from
claims of Copyright Infringement, because a permanent data record
(a disk file) is not created!

-- Robbie Rhodes