I have followed with interest the thread that has developed in
the wake of my MMD posting of 10 September, 2003, suggesting that
telephone technology might have played a role in the early Welte
recording process.  Opinions expressed have varied widely, and
I believe certain items stated as facts in some postings call for
further scrutiny and examination.

I, along with many others, firmly believe that Welte did have, from
the outset, an effective way to measure and record dynamics of piano
performances.  A credible body of supporting evidence exists in the
library of amazingly lifelike recordings that came from the Welte
studios in the pre-WW-I era.  Bear in mind, too, that no one up until
that time had any experience whatever in inserting human interpretations
of dynamic coding on piano rolls.

If dynamic coding had been 'faked' by human editors, as was done in
Duo-Art and early Ampico recordings, one would reasonably expect auditory
evidence of a 'learning curve' in the early recordings as editors built
up their skills from a starting level of zero.  Improvements over time
in early Duo-Art and Ampico recordings certainly suggest that editorial
skills underwent such a developmental phase.

Welte recordings, however, do not follow this pattern.  They were quite
good early on, with very lifelike dynamics, and remained so as the
system enjoyed its notable early success.

There are many folks, however, and I'm one of them, who find it
extremely difficult to accept the story of carbon rods suspended from
springs and plunging into a trough of mercury.  The technological
shortfalls of such a scheme are many, and have been well documented
elsewhere.

Dismissing carbon microphone technology as too primitive to be useful
in the subject time frame is not in accord with the historical record.
More than 25 years had passed between the telephone's invention and the
Welte-Mignon's debut.  The telephone was invented in 1876, and technology
improved enormously in the quarter century that followed.

In 1882, for example, magneto wall phones were in use, and by 1900, the
basic design of the carbon microphone element had become pretty much as
it remained throughout the 20th century.  The well known "Eiffel Tower"
handset, also known as the "skeleton" phone, highly sought after by
telephone collectors, was introduced in 1892 and continued in production
until 1929.  The handset was, of course, designed to operate in any
position.

Perhaps 100-year-old telephone components are cantankerous and
unreliable today.  Is that really surprising?  Can we expect today's
experience to be a reliable indicator of how ancient equipment worked
when new?  It seems unlikely.

In my earlier posting, I suggested that pressure sensors using carbon
microphone technology may have been employed, but did not mention
exactly where a pressure sensor might have been placed in the piano
action.  There are various possibilities, and I'll comment later on
one of them.

In a piano action, elements comprised therein behave in accordance
with basic laws of physics.  Every freshman physics student learns the
fundamental relationship that says force equals mass times acceleration,
or f=ma.

A pianist exerts a force on a key, and thereby causes a hammer to
accelerate.  Although this is an oversimplification of a really
complicated system, we can cut through to the crux of the matter by
looking at what's really important.  We see immediately that 'mass' as
a constant, and it follows that acceleration is directly proportional
to force.

When a pianist applies a certain amount of force to a key, a proportional
acceleration is imparted to an associated piano hammer.  While a given
amount of force applied to the key, the hammer continues to accelerate
until the jack fly disengages.  At that instant, a certain hammer
velocity has been reached, and the hammer continues its travel without
further acceleration.

If applied force can be measured, it should be possible to relate it
directly to acceleration, and ultimately to hammer velocity.  However,
things get a bit sticky when we realize that force may not be a constant
throughout a key's travel.  If force varies while being applied, it is
necessary to integrate force over a specific time interval to obtain
acceleration up to the moment of jack fly release.  That is why Ampico's
Dr. Clarence Hickman, in his great wisdom, avoided that problem (along
with some others) by direct measurement of hammer velocity after jack
fly disengagement.

There's no doubt that Hickman's approach was a superior and
scientifically "clean" way to make such a measurement.  But it is not
the _only_ way.  Suppose a pressure sensor were interposed between key
and balance rail, for example, in such a way as to measure force
(possibly time varying) applied to a key while it is being depressed.
Further suppose that measured pressure is transmitted to a mechanical
integrator of some sort (inked rubber wheel? ) whose output could be
recorded in real time.

In my opinion, such a device could be built whose measurements would
correlate fairly closely with Hickman-style hammer shank velocity
measurements.  It would likely be less accurate, and certainly subject
to a greater number of potential sources of error, but useful enough,
perhaps, for a strong 1904 beginning.

Dave Saul