Thanks, Craig Brougher, for stirring the memory of James D. Crank on
the subject of Dick Simonton and the Welte recording technology, and
for encouraging the input of Mike Knudsen.  Mr. Crank writes:

> The carbon rod under each key was the way it worked, all right.
> The logic for this system -- now, remember we are talking about the
> 1900-1905 era of technology -- was that the little carbon rod dipped
> into the mercury, which had a film of oil to prevent oxidation, and
> the lower it dipped, the less resistance that circuit had.  Over the
> period of time that the key was being depressed, and how far it was
> depressed, you actually got a curve if you plotted voltage against
> time.

The oil on the top of the mercury would give a time delay ramp if it
was actually thick enough to protect the mercury from air, as the oil
would be a dielectric, an insulator.  I suspect there was no oil in
this particular application, however other mercury switches may well
have used oil.

> As the musician played, the little wheels would depress against the
> moving paper roll.

As to the wheels placing varying widths of graphite ink: the feed
system for the ink could easily be that for marking colored lines on
resistors, pencils, etc., which was well developed in Germany and
elsewhere at that time.  The simplest method was to have the wheels
rotating at print speed as noted, the bottom of the wheel dipping
through a bath of ink, the tops of the wheels out of the ink.  This is
not readily seen in patent drawings as it was old art long before the
patent offices were set up.

For printing, either the wheel could be lifted against the paper, or,
if the paper were very wide and flexible, a pinch roller could press
the paper down against the wheel.  With a real knife edge as the zero
point, all lines would print all the time, and the width would vary as
the pinch rollers added force.  Very little movement would be
necessary.

The technology of using graphite was also well developed.  Colloidal
graphite is merely very finely ground graphite with material added to
keep it in suspension.  As an ink it would also have a vehicle to keep
it stuck to the paper, flexibly, while not acting as an insulator.
There is a wide selection of materials.  Try
http://www.achesonindustries.com/ADAG-ADD.html

> Then, some smart technician at the Welte factory could interpret the
> loudness of the note by not only the width of the line, _but also_
> the rate over distance of the line expanding.

While the change in width of the line clearly shows the key dip, and
the rate of change the rate of depression, it may not be necessary for
a technician to laboriously interpret the striking force of each note.
With a direct playback system, the intensities of each note can be
marked by some device which provides the value of the suction when the
note is being struck.  Simple pneumatic logic (a bank of or-gate
check valves into a spring loaded pneumatic attached to a recording pen)
can create an intensity mark for each note or for the bass and treble
groups.  Then this can be interpreted by a technician in much shorter
time for coding production masters.

> At one time there was a story that Welte could take this master roll
> and instantly make a punched roll from it, but that the dynamics
> would be missing.

This fits.  The punched roll would need the technician's interpretation
of the dynamics noted on the graphite inked original.

> The electrical technology of the day could have accomplished this, if
> the designer was really clever.  And, the Welte people certainly were
> clever.

Mr. Crank is one of the most clever people I have had the pleasure of
knowing, although he claims ignorance of electricity.  Since I have
spent most of this Sunday enjoying this correspondence, I must wait
until next Sunday to pay him a visit to discuss these topics.

Mr. Knudsen writes:

> I think Welte might have been willing to get instant playback through
> photocells, each cell reading one note track on the master roll and
> driving one vacuum tube whose output then modulated a magnetic
> amplifier, one of all these per note. {snip} The playback piano would
> have used large, AC solenoids, rather bulky and noisy, but perhaps in
> the form of a Vorsetzer.

Edwin Welte, or probably Karl Bockisch, would probably have been
delighted to use photocells, but they had not been invented, nor would
they have needed graphite ink to function.  What was available for
sensing was multi-leaf brush contacts, like a pair of little photo
retouching brushes, but made with conductive hairs.  Separating the
pair with an insulating sheet and allowing the brushes to lightly read
the traces would give a variable resistance reading.  The local
cross-sectional resistance of the local would be read, and the shape of
the trace, so hard to interpret visually, would not be relevant.

Passing a current through these brushes and the trace would give enough
signal to work a small electromagnetic actuator, perhaps a galvanometer
movement or a speaker voice coil on a pot magnet yoke.  This was all
well established at the time, and scientific instruments were of a very
high grade and sophistication in Germany then.  Siemens waited until
1885 to patent the voice coil design still used in 99 percent of todays
loudspeakers.

Amplifying that small movement would require no more than a balanced
valve, perhaps part of the voice coil actuator, to give a variable
opening to a tracker tube.  The simplest way would be to have the voice
coil sleeve vary the opening of a cross-hole through the central rod of
the magnet yoke.  An axial hole would connect the cross-hole to a
tracker tube.  Naturally, you would not feed an on-off valve and
pneumatic, but an amplifier valve and its pneumatic.  With the great
power available from the Welte Vorsetzer, and all amplification
accomplished pneumatically, there should be no problem in creating a
stunning performance.

Obtaining a suitable transfer curve might be done by shaping the hole
and/or having a variable rate spring opposing the voice coil force.

I would not be surprised if Mr. Brougher could take these ideas and
have a working model in a few hours!

All the best,

Karl Petersen
Washington, Illinois