The monograph by H. W. Stephenson appears to be something that deserves
wider distribution.  I believe he is discussing a single valve system if
his main concern has been the Duo-Art.  (Was he looking at cross valves,
later types, or both?)  To offer any final and objective comments on any
of Mr. Stephenson's conclusions, we really need to examine his data in
its proper context.  In formulating his conclusions about valve travel,
for example, did his tests consider recovery time -- the time required
for the hammer to come back to a state of readiness for the next note to
play? This is important for rapidly repeating notes.

Repetition can and does vary according to the placement of a note in
the piano scale; the moment of inertia of a shank with hammer attached,
rotating about its pivot point, is greatest at the bass extreme of the
scale.  The effective rotational mass decreases markedly toward the
treble end of the piano.  That's why faster repetition is generally
achievable in the higher regions of the scale.

Consider also that when a note is struck lightly, it is not caught by the
back check as it would be with a more intense strike.  Aside from mass
considerations, the checked/not checked possibility requires that two
distinct conditions be dealt with during the release.  This is, of
course, influenced by stack pressure.  These and other factors (e.g.,
energy stored and later released by a grand's repetition lever spring)
could have a profound effect on ability of a pneumatically activated note
to achieve good repetition.

A valve gap on the small side not only constricts air being exhausted
from a pneumatic during activation, but also flowing INTO it after the
note is released.  The latter, of course, is not influenced at all by
stack pressure (the pneumatic is open to atmosphere over the top of the
valve), but rather by the dynamics of the action parts trying to achieve
a rest state after a note has been played.  My gut feeling is that a
slightly larger valve gap may be needed for release than for the strike
that precedes it.  I would give it at least .030 in.

Excessively large valve gaps need to be avoided because  of noise, if
for no other reason!  Valves with large gaps tend to add annoying clatter
to any music produced.  Excessively large gaps may also tend to cause
harmful pressure transients in the stack as a result of air leakage while
the valve element is in motion, traveling from one seat to the other.
While a single playing note might work OK, large numbers of notes being
played simultaneously or nearly so might suffer from the regulator's
inability to overcome the effect of the transients.

To keep on the safe side, I'll keep on setting Duo-Art valves in the .032
to .035 range.

Dave Saul