Since the Easter weekend I've had a chance to audition and re-evaluate
a number of Duo-Art systems, some rebuilt by me or under my aegis.

>From examination of, among others, an early 1914 US-made grand
expression box, a 1925 British grand box and a 1926 US-made grand box
I have absolutely no doubt that even though they utilised standard
linear rate regulator springs, they were intentionally designed to give
the upward concave exponential response curves that I referred to
previously, and its is therefore vital that these be adhered to for
proper reproduction.

In each case, during rebuilding, inspection of the spill-valve lever
geometry and wear marks confirmed that the accordion linkages had been
originally set to impose an accelerative motion on the knife valve
similar to that fixed on some British upright boxes, and this was
retained.  This slight accelerative motion moves the knife valve heel
least per linear accordion increment at power 0 and most at power 15.

In the rebuilding of these boxes great care was taken to ensure that
there was no leakage past the knife valve in any position.

Whenever the knife valve heel is displaced to open the orifice, the
regulator pneumatic moves the toe to re-close it.  The knife valve thus
describes a compound movement across its orifice.  Therefore not only
the relative position of the orifice but also the shape of its top edge
affect the linearity of the response.  The 1914 US and 1925 British
boxes have 3/4" diameter circular orifices, the 1926 US box a 1-1/4"
x 3/8" slot with rounded ends.  Yet in each case a successively greater
movement of the regulator and thus greater extension of its spring is
required to re-close the knife valve between power 0 and power 15.

As Mike Knudsen, Julian Dyer and I elsewhere have noted, the regulator
pneumatic develops inherently less force against its spring the more
it closes, so from power 0 to power 15 an increasing rate of vacuum
tension is required in the regulator (and thus in the stack) to
re-close the knife valve.

These factors together ensure that, providing the pump has sufficient
tension, a clear exponential curve is obtained on all three boxes,
the gaps between successive levels increasing markedly throughout,
as is amply confirmed by repeatable plots of vacuum measurements.
So while the regulator springs are indeed linear, the expression curves
definitely and intentionally are not.  Note also that these Duo-Arts
meet the chord build-up tests in the relevant test rolls.

The closing of the box spill also contributes to the progressive
increase in the steepness of the curve.  However, I cannot stress too
much its real function in the accurate reproduction of incontestable
nuances edited into the rolls.  The spill is not simply to unload the
pump during low demand: a conventional spring regulator/spill would
have been simpler and cheaper.  The Duo-Art spill allows the theme
dynamic stream to modulate the accompaniment below power 10, a clever
and intentional form of time-sharing which results in far more levels
and subtlety than Aeolian disclosed in its literature.

The 1914 US box is in a New York Steinway model 'O' with very large
graduated pneumatics and original cross-valves; the 1925 British box
a Hamburg 'O' with medium-sized graduated pneumatics and original
circular valves; and the 1926 US box a little Gaveau grand with tiny,
one-size pneumatics and the smaller cross valves.  The pianissimo
readings required were 2.5", 3" and 3.5" respectively, and the
fortissimo 16", 25" and 38".  It is clear that power 10 is not 20"
in every case, and that 20" in no way plays them all about the same
loudness.

British Duo-Arts never have modulators, and the main vacuum trunk
is usually connected to the box immediately between the knife valve
orifices, so there is little pressure drop in the system and a
crash-valve is unnecessary.  However, they most frequently have in the
main supply an unloader spill set to a relatively high vacuum level,
which while protecting the pump (whose belt should never slip) allows
instantaneous response to demand for fortissimos.  At power 15 the
theme regulator is nearly closed, opening very smartly and very wide on
big chords to produce formidable fortissimos before re-closing almost
instantly.

Patrick Handscombe
Wivenhoe, Essex, UK