I've been working on an expression simulator so I can hear the results
of some scanned Triphonola rolls, and I finally got it to work last
night. Now I need to get the simulator tuned to work properly!
There's a piece of information I need that has proved rather elusive
so far. This is: how fast is the crescendo / decrescendo, both in its
normal setting and when the 'fast' ports are open?
For those unfamiliar with this system (i.e., almost everybody), which
is Hupfeld's second reproducing piano system (their first being the
DEA), some explanation may help set the context of the question.
The Triphonola is a Theme and Accompaniment system, with the familiar
'snakebite' Theme accents just like Duo-Art or Artrio-Angelus. Each
of these two playing levels is defined using a pair of crescendo
pneumatics mounted on a common board so their effect is additive
(rather like a Duo-Art with a two-segment accordion, both segments of
the same size). These pneumatics are either opening or closing all the
time, with either a slow or a fast rate of change, almost exactly like
a Welte-Mignon and complete with the fixed Mezzo Forte (MF) reference
point.
Despite the Mignon-like MF point, the Triphonola is much more like a
two-segment Duo-Art, and I've adapted my Duo-Art simulator to model it.
The critical difference between the two systems lies in the value of
the rate-of-change parameter. The reason for this is that although the
Duo-Art is often described as using 16 fixed levels, in reality it
takes a finite time to switch between them. Each accordion segment can
just as easily be thought of as a crescendo/decrescendo between fixed
limits.
Because the Duo-Art has such a fast ramp-rate it makes great use of the
fixed reference points, but there are notes that get played when it's
at intermediate levels. A Triphonola model uses just two accordion
segments and slows their movement down. Both systems are otherwise
very similar, particularly in having an additional, even-faster, ramp
rate, when switching between Accompaniment and Theme levels.
Actually, I have a theory that it would be possible to model _all_
reproducing piano systems in a single simulator by suitable setting of
parameters for each of the sub-components that all these systems are
built from! That's a diversion for now, something to consider another
time perhaps...
Triphonola instruments are vanishingly rare in the UK, but rolls do
turn up. Hupfeld had a phenomenally good catalogue of classical music,
and the highest production quality of any roll manufacturer, but
because of the rarity of the instruments Hupfeld has been broadly
ignored in favour of Welte-Mignon, Ampico and Duo-Art.
Hupfeld 88-note rolls are widespread, and much prized. Triphonola
expression coding lies wholly in the margins, so the 88-note and
Triphonola versions have identical performances on them. Scanned 88-note
rolls, even with a fixed-theme model, are distinctly flat, and a poor
way of sharing this hand-played repertoire. Simulating from the
Triphonola expression is an obvious improvement. Better still would be
to have a piano, but that's a lot harder.
I feel I'm getting close with the simulator -- the basic mechanics I'm
happy about, now it's a case of getting the details sorted, hence my
initial question. I'd appreciate any information that can be provided.
Julian Dyer
|