A thick green folder crammed with issues of New York newspaper
in German from the WW1 era came to us in a box of glassware at an
auction years ago.  The glassware was cleaned and sold in the shop
we created on the square in Rushville, Illinois, in 2001.  We sold
the store in nearly two years ago, but this one box seems to have
traveled with us after the move back to Washington, Illinois.  You
can understand that we are still unpacking and repacking as we have
done for years with each move!

We had not looked through the folder of newspapers much, since
they were full of the news of the Kaiser's war and from the German
nationalist point of view.  The old fashioned type make it rather
difficult, though I did learn the German language in that form in
the 'fifties.

What I had not noticed among the newspapers before was a grade school
student's tablet with a lithographed cover showing a lion jumping
through a flaming hoop and the price of 5 cents.  The crispy yellowed
pages of lined newsprint inside had been filled with a tiny hand in a
European script and are apparently English translations of letters from
the original German.  Unfortunately they start and stop irregularly
and are only the body of the letters, the dates, salutations and names
of what seems to be one sender and a single recipient are not recorded.
I include the full text herewith.  If there are typos, they may be
mine, but we did proof them for content.  They didn't seem to need
polishing, so here you are.

 - - -

The past week I have been looking through our past notes and letters
and have a view of our progress, thinking to see where we should be
going.  I write this knowing well enough you oppose strongly and
equally my putting any of this down on paper, as you did about the
business in America.  I put it down now so it will not be forever lost
should something happen to us past our knowledge.  Perhaps the lawyers
will need it someday.  If we don't keep ahead of our competition,
we may have to actually patent our secrets.  When we are both gone,
at least it will exist for another generation to discover.  In another
twenty years recording techniques will probably be managed in a dozen
different ways and they will laugh at our silences.

Now that our factory in America has been stolen from us, you realize
how correct I was to have insisted we keep the recording systems and
the mastering machine here under our control while Von Bismark's
advisors led him down the muddy slope.  We did perhaps train up too
many workers in proper skills of manufacture, and I hope after this
war they do not compete directly with us or in concert with Votey or
Farrand.

Still our competitors have not developed a suitable recording method.
They have not moved past attempting to improve the results of the
pencil recording pianos.  All are fooled to think that the keyboard
is the place to record from, so their results cannot ever sound better
at first pass than a second-touch organ keyboard playing on a piano.
After almost ten years, I think that they simply do not believe it is
a large enough market to invest in better recording, and simply blow
the horn to sell wagonloads of low cost and lower value music "in the
latest style".

When we started, the pencil recorder was nearly useless for the serious
pianist or organist  because the timing of the playing itself was not
registered.  Now, I have always agreed that it was a useful tool if the
keyboardist operated the keys with the pencil recording in mind.  Of
course the pencil recording was a start.  We realized the final result
would be completed in a rather mechanical fashion where the personality
of the player was not a concern.  For organ notation where there is
only timing to record, the inked wheels were a fine step and are yet
more than enough.  The artistic part of the organ performance is
captured through registration and timing.  This is not so on the piano
and certainly not for any solo instrument, however many automatic
violins may bring to life the imagination of the uncritical listener.

It even yet makes me hot to think how idiotic it was to have a fine
pianist play beautifully at the keyboard and only leave a nearly
useless sheet of note events.  Even with a musician to edit and finish,
the performance sounds at best like an undistinguished student
imitating the original pianist.  The popular music reconstructed this
way in has been adequate worldwide for the market, of course, and only
a few musicians notice the difference in the conversions from the T-100
recordings, for good reason.

For playing the organs from rolls, our simple and stable pneumatic
system reads with great repeatability from the simple punched rolls
giving the precise note events expected.  There is no regulation of the
mechanism needed.  It simply plays or does not.  When an arrangement
is hand played for the orchestrion of the piano accompaniment, it is
recorded from the organ keyboard also with the simple note events,
and they sound exactly as they did during the recording from the organ
console with almost no editing.  The organist learns to play the remote
piano with considerable artfulness and seems to be able to quickly
adapt, just as a pianist would adapt to a piano with even serious
technical defects such as missing strings and dampers.  There is much
teasing of the keyboard going on, but the notes do not play unless the
organist desires.  This technique of piano recording is an abstraction
and, though very useful for orchestrions and simple pianos, does not
intend to give artistic results.  It is perhaps quite useful for
popular piano pieces, however I have no reports of anyone doing it,
which seems a great oversight on their part.

Before we created our system, I was watching a technician regulate
a piano action, I realized that there were many more abstractions of
hardware and energy transfer which happen between the pianist and the
note sounding from the string.  The key and the action parts could
truly make many gyrations, and a vast range of inputs at the key will
give the same sound at the string.  The speed of the hammer when it
hits the string is the essence, nothing else.  You know that many notes
are sounded although the key is not depressed all the distance to the
key bed and notes can play softly or be repeated although the key is
not fully released.  The great mistake is to record the key position,
thinking it is mostly the key position which defines the note.

Just as we moved from pencil recording from the organ because the line
quality and wear of the pencils was impossible to control, we had then
to think about moving from the simple inked disks for piano recording.
We adopted inked disks shown by Keuffel for ruling paper and Eberhard
for marking rings on pencils, but it was not easy to intelligently
start and stop ruled lines on a recording roll, but it was a beginning.
Their inked disks were of sheet brass and made continuous lines, but
ours needed to show a change due to the note event.  We needed only to
develop the method to have them print on the paper when a note is being
played.

Heretofore the disks have all been on the same shaft with the lower
edges in a tray of ink.  Putting each wheel on a separate arm with an
inked pad wiping them worked well, then the signals from the instrument
to electromagnets print the line.  The ink traces are marked at the
start and end, punched, and you have the editing master.  Since the
ink trough did not move with the wheels and gave a strong variation
in the result, we changed to a saturated ink wiper developed.  Because
we could not clearly know how much free motion occurred with rollers
or pencils which lift free of the paper, the adjustment of each note
varied in timing and force gave erratic records on paper.

Having inking disks rolls with rubber faces in a sharp "v" allowed our
recorder to give continuous lines for position, then widen the line by
pressure of the rubber rollers when the electromagnet pulls in.  The
line reverted back to a fine ruled line when not actuated.  The key
idea was to maintain a base line reference and show a clear change.
Without the fine line for the datum, there was no true way to know when
the note begins and lets up.  Now we had our good organ recorder and it
has hardly needed to change.

I thought constantly of a way to simply and accurately record the
strength of the notes from the piano.  The question of deriving the
hammer speeds from an automatic record, recreating them on the piano,
then somehow being able to manufacture equally stable recordings
cheaply in volume and pianos to play them was all wrapped together
in a complete system requirement.  Any method would have to provide
all these and be automated without the need for a great staff of roll
editors, or it would be commercially worthless.  As much as we may
enjoy the music and the people who we meet through it, we always
realize that this is our business enterprise and sustains more than
our artistic pleasures.

This goal and its balance occupied my mind continually for months and
I am sure the burden did not make me a useful person to live or work
with.  Even going out with the family or to the baths did not distract
my mind from this question.  I looked at technologies from other parts
of art, craft and science.  I watched for clues from nature.

One Friday I spent the sunny afternoon skiing.  Remembering Uncle's
sound example, I kept an eye on the tracks already on the slopes.
The lines and defects in the tracks of earlier skiers showed clearly
that here they found rocks or there they had to correct to avoid hidden
objects.  A flash of recognition came for the note recording.  The
marks obviously showed where they went and how hard they loaded each
ski.  The depth showed their weight or their force in turning or
adjusting.  You could almost see their arms on the poles and their
weight shifting on the skis.

I thought, as did Uncle, that if one could ski directly in their tracks
and pole at the same spots, one would reproduce not only their path but
the speed and dynamics of the movements of their arms, poles, hips and
shoulders.  Where there was a little jump and they left the ground for
several meters, my mind did not follow but could only guess where they
might have been once the record of the skis touched the ground again.
It was what we had learned with the note recording, that it is
important to see the continuous track to know what relative changes
actually happen.  My mind thought of skiing in wet mortar and running
down through it again to recreate the event, then I realized that this
is, of course, what happens in the phonograph on a very small scale.

So I thought how a record of dynamic notes on the piano might be like
this.  It is good that the paper can move at a firm rate, so
eliminating one variable in the record.  Instead of the inked rollers
quickly going on an off from the same contact closing that operates
pull-downs in the organ chest, a variation in force could give a
variable width line, the trace of the inked roll always visible and
increasing in width smoothly as the hammer flies.  When the hammer is
caught by the back check, the width will continue at about half, then
drop back to a fine line when released.  Since I had realized that the
hammer speed was what needed recording, I had to think how to capture
the speed.

I saw now that moving many steps away from the hammer toward the key
was clearly a mistake when recording, but making an measurement of the
hammer speed seemed impossible to me.  An abstraction into the hammer
position still would give a more basic and completely accurate
representation to reconstruct the speed.  If you map the hammer
position over time, the energy and the speed can be recreated.  So we
needed a way to push the inked recording wheel in proportion to the
hammer position.

The hammer was too much affected by anything we did to sense its
position directly because so little energy is in the hammer itself.
It obviously could not push the inked wheel directly, unless the wheel
was too small to be of any use.  An amplifier of some kind would be
needed.  The position of the hammer must be duplicated without any
forces affecting the hammer travel.  What ideas we tried!

A variable resistance from a card with a carbon track and a silver
contact, made small enough to avoid unnecessary friction, had too
little travel and was not repeatable.  Amplifying a phonograph also
has only fair results from such technique.

The proportional pneumatic amplifier was the obvious thing to try.  The
phonograph can be amplified by such means with astonishing power for
public performance, and with better justice to the quality of the sound
than I would expect.  I wish it could have been done with a simple vane
on the hammer shank passing over a slot to bring up the force on a
pouch loading the inked wheel.  A single amplification was added with a
spring loaded spool valve.  This reads hammer position and gives force
in proportion.  But the bleeds hissed continuously, even when reduced
with the amplifier valves.  The hissing from the bleeds on an entire
piano would have been too much distraction.

The idea of variable resistance directly powering the magnets was so
simple, but to put a variable resistor which would carry enough current
on the hammer shank did not serve.  The wiper created too much drag if
placed at a large enough radius to obtain a reliable and wide
variation, and the output was not well controlled if the radius was
reduced.

When our piano was being set up in the summer cottage, the technician
was refilling the mercury switch and was chasing a few bits of mercury
into the trough with a pencil tip.  The idea of the variable resistance
having a mercury contact against graphite came to me, since there would
be no real friction in the movement against the mercury, but only
a small spring effect.  The hammer shank rotating through a small arc
was not the place to install a carbon element and a pool of mercury.

Mercury is most stable when in the bottom of a vessel of small
dimensions, and the carbon element gives maximum controllable variation
when moving perpendicular to it.  We tried carbons with flat, cone
tipped and oblique truncated tips.  Only after having all the other
elements of the recording and playing systems completed did we settle
on a wedge tipped 5 millimeter carbon rod pushing into a few drops of
mercury in a Hahnemann pill bottle.  When the rod goes down and the
mercury rises slightly, of course.

The zero point adjustment resolved to a simple screw raising the bottom
of the pill bottle, easily bringing the zero point on the inked wheel
into a fine line.  With a light arm extending from the side of the
hammer shank, back over the pivot, a spring balanced sticker ending in
a hole in the action base at the keybed pushed down the tip of the
carbon rod which rose to the top of the keybed on a copper coil which
also balanced it.  These two very light springs and feathery sticker
have no perceptible effect on the playing of the instrument, and the
action and contact bank are both easily and independently serviced.

This is, of course, just the first element in the recorder and just the
start of the entire system, but this concept came first and the others
were developed to accommodate it.

We start with a thin base line on paper which turns into an increasing
wedge, then a reducing wedge, then at about half width remains at a
constant width for the duration of the note, then is a reducing wedge
back to the thin base line.  The smaller the ink disk becomes and the
smaller the paper radius against it becomes, the smaller the mark from
the fastest accent note becomes.  This reduces the minimum paper speed
and makes the ink delivery easier.  In practice the ink disk cannot get
so small that it cannot be inked, but the paper can go over an even
smaller roller when it meets the inked disk than perhaps we have
chosen.

Reading the note record was the next challenge.  Reading manually using
templates or angle measurements of the first wedge for each note would
be out of the question from the amount of labor required, so automatic
reproduction from the note record was needed.  The shape, or actually
the rate of widening of the leading edge of each note record had to be
individually recognized and reproduced in the position of each hammer.

The only method which occurred to us was to use conductive ink, to read
the change electrically and amplify to drive the piano action.  Our
practical experience has been with pneumatic systems, and the pneumatic
amplifier, but reading ink would not be practical pneumatically, and an
electric analog was decided.  Again the conductivity and adaptability
of carbon was taken up.  Carbon ink was chosen using graphite pigments
which are plate-like particles easily suspended in liquid.  There are a
number of interesting industrial uses of this material, particularly in
plating, ceramics and small electrical resistance devices.

This suspended graphite was now our ink, and, when dry, was our
conductive element of the note record which required accurate reading
and application to reproducing the hammer stroke.  Since the ink is on
the top of the paper and may vary in thickness, the important factor to
read is the width.

We had a choice of several methods to read the width.  First, a small
array of individual contacts can be used which are gradually connected
together through the widening ink line.  These can be interpreted
through a resistor network so it is not critical that the exact lateral
position of the line be maintained.  The complexity of such a system
made us look to something more pragmatic.

A pair of contact brushes wide enough to always stay on the minimum
line from the narrow disk edge and wide enough to cover the full width
at maximum hammer travel, but narrow enough to avoid contacting the
next row of the record was considered.  If the brushes were narrow in
the direction of travel and spaced apart the length of the loudest
staccato note, then a short duration of output would occur when the
first brush was just leaving the note and the second brush was just
coming up on the beginning of the note.

It was obvious to shape the first brush in a vee like the trailing edge
of the note and the second brush in a vee like the leading edge of the
note, so this staccato note would simply give a very short pulse of
maximum current.  Anything running longer would gradually build up due
to the longer widening of the beginning of the note against the second
brush.  Then it was seen that the first brush should be about the width
of the sustained line produced when the hammer is on the back check.

Trials of these brushes and note recordings over a wide dynamic range
were very illuminating when we fed the signals into a recording
galvanometer.  The results indicated that the range of dynamics was at
least 100 to one and even soundless notes in which the hammer did not
hit the strings could be repeatedly recognized at a consistently low
level!

Now we needed two systems more.  One to play the recording back
immediately when the ink was dry and a complete and thorough system to
encode the values to a punched roll and play them back in a commercial
pneumatic piano.

For the immediate playing we needed to select an amplifier to drive
the piano with the output from the brushes.  The forms of electrical
amplifiers commercially available had changed rapidly during those
times and the universities and technical instrument makers and
electrical equipment companies all had their favorite patented systems.
None described their systems in terms which were compatible with one
another, almost as if they were trying to avoid being compared and
found wanting, and we found many to be wanting.

Combining established electromagnet designs with the most linear
response from our experience with pneumatic systems, we simply put an
open coil and pot magnet (of the kind Siemens patented in 1884) on the
stem of the pneumatic amplifier we had used earlier.  The coil was
powered directly with current flowing through the brushes.  It was
quickly evident that we could not stop the note record or the graphite
traces would overheat and burn up the paper.  We were able to reduce
the current.  Our approach wisely ignored the thickness of the graphite
traces on the paper.

It was important to favor the outside of the note records, so the
trailing brush elements were weighted through a resistor network which
gave more current through the outer elements and less through the
center elements.  Then the rate of the sleeve valve on the amplifier
was biased to output the same hammer strikes that had created the note
records.  This could have been done by using negative rate springs, but
they are notoriously unstable to keep in adjustment.

Developing the system on the one-key model, I thought first of driving
only the hammer since it was the hammer position which needed to be
reproduced.   Immediately the need for the back check to control the
hammer and the repetition action to do its office showed that driving
the key was best and did quite obviously reproduce through the action
the hammer energy desired.

The pneumatic amplifiers to give proportional force replaced the unit
valves on simple note pneumatics in a very ordinary looking stack.

A first complete recording contact system is housed in a relatively
ordinary normal wing piano.  Cables from the piano pass to the left
into a large cabinet with a roll of paper, the graphite ink recorder,
and a serpentine drying section.  The ink is stable enough to allow the
paper to go up and down over rolls with large solid faces supporting
the back side of the paper at the top and rolls of notched, knife edged
disks at the bottom so damp ink on the first few passes sees no more
than the tip of a blade and does not track or run.  The winding drum at
the end of large diameter assures a constant speed.  The paper is cut
at the winding drum and respooled on a small diameter roll with
flanges, much like used on the orchestrions, and the leader is trimmed
and the labels previously created for the session are applied.

All the panels are removable for service.  The rubber rolls have to be
started off completely clean.  Very soon, we adapted a trough inking
system much as taught by the paper ruling systems, since the wicks
would dry irregularly and notes would be missed.  An attempt to simply
seal the ink and roller portion was not adequate since the ink would
separate enough to give problems although it is supposed to stay in
suspension forever.  Excellent quality thin drafting vellum is used,
and it is fortunate that this is manufactured in quite long rolls.

In the optimum recording session, the artist plays several pieces in
succession without shutting down the recorder.  It takes nearly an hour
to prepare the recorder when the ink reservoir is dropped and the
wheels are cleaned.  The pianist and our staff usually go to lunch and
I restore the system for the afternoon session.  Nearly five meters of
paper are wasted each time, but this is an annoyance rather than a
concern for expense.  We are careful to put all the waste in the
furnace without involving other employees.

Once the dried note recording is spooled, and it is usually done for
the entire session rather than individual pieces, it is installed in
a keyless cabinet piano player (in the same studio) of high quality
adapted with the hidden amplifier playback system described.  Since it
has to run at maximum vacuum, the interior of the pneumatic section is
heavily padded to reduce pneumatic noise.

Neither the recording system or the player are displayed or described
to the artists, but kept as a proprietary secret with an intentional
air of mystery.  It is essential for the artist to hear their efforts
played again so soon after they are made, that they know that no simple
marking and editing could be accomplished in so short a time.  At the
beginning, we did not realize how important the "approval" of the
recording was in most artists.  minds.  It would simply not do for them
to record and go home without hearing the piece and acknowledging it as
their own.  They could also become confused later on as to the quality
of their playing or the recording.

So it became important for us, also, to have their imprimatur at the
outset when their enthusiasm was fresh.  This wonderment was a very
powerful potion to entice an enormous range of artists to record in the
first years, since they would engage the curiosity then the demands of
their associates and rivals to be inducted into this mystery, this
mystery which gave their playing to eternity and spread their names
throughout Europe.

That describes but half the system and, although it is the only part
about which anyone asks, is the simpler and, perhaps, the less
important half.  As I said, to be a commercial success, the system had
to go from the keyboard to the perforated roll on the shipping dock
without any great editorial effort.   We are now halfway to the goal,
and it is only a curious academic exercise without the second part.

When the original note record is created in the marking machine, it is
complete and stable.  Then it needs to be decoded into a basic note
record and a derived dynamic scheme which is simple enough to drive a
pneumatic piano with a divided stack.  The treble and bass sections are
treated separately but in parallel.  Each set of 44 notes is read with
the same kind of brushes used in the cabinet player.  At a reduced
speed, another paper roll of normal width is marked with a line for
each note position and duration, and also with a cross bar at the
beginning of each note, and at the end when the hammer drops off the
back check.  This marking allows adjustment to be made in note timing
without losing the original timing by punching it out or taping it
over.  The graphite recording is almost never consulted again since the
hammer timing and force values are placed on the marked master

While the note events are being marked on the sheet, amplifiers like
those on the playback piano create the same suction in the pneumatics
for each note.  Each individual pneumatic comes up to pressure, drops
off and returns to the back check value.  It is the nature of the
recording that the striking value only occurs at the beginning of a
note, and instantly reduces to a minimum until the note ends.  This is
very important and allows a very simple additive device to combine many
note intensities into split stack levels.  Tubes from each pneumatic
pass through check valves to a spring loaded combining pneumatic.
This does not sum the forces of all the playing notes, but moves in
proportion to the playing level of each note being initiated at any
moment, as a maximum, and drops off to the next lowest level quickly
as well.

The volume of this system is kept at the absolute minimum to accomplish
the result so that all possible lag due to volume and length of tubes
is eliminated.  Although we initially planned to be able to run this
at a reduced speed to avoid interaction between notes and allow the
crescendo elements time to work, it was found that there was no need to
reduce the speed since the results at full rate were indistinguishable
from those at reduced rate.

The expression in the playing piano system is a split stack.   Two
rates of change are called by simple lock and cancel perforations, and
a mezzo-forte lock is introduced as a locating device.  It is not the
intent to simply change the stack vacuum and hold it at exactly the
value needed for each note.  The interaction of the crescendo and
forzando to create a vacuum level are completely automatic and not
selected as fixed matches.

In creating the playing master, the note records and the result of the
level pneumatics are inked directly on the edge of the playing master
during the note marking.  The level can be consulted if there is need
to manually adjust the result of the automatic system.  While the level
pneumatic dances rapidly between a vast range of playing levels, four
small pallet valves damped with dashpots ride on the level pneumatic.
The change in direction of the level pneumatic signals lock and cancel
of the fast and slow crescendo and decrescendo necessary for the piano
to simulate those intensities.

While this system created even the original recordings with amazing
success, there has always been some editorial correction necessary, as
when two very different intensities are needed at the same instant.
This same system has been used since the outset of recording and has
been remarkably stable.  There was only one realignment of rates after
the first year.

So that the level changes from these small increases and decreases
not drift too high or too low, two techniques are employed.  During
passages when no note events are anticipated, the intensity is brought
to zero or to mid-range against the mezzo-forte hook.  or to maximum,
so that the next playing levels have a firm frame of reference.
Adjusting the individual piano so that the rates of change are balanced
is critical and makes a vast difference in the quality of the
reproduction.

The result is played back on a well regulated T-100 piano in the
studio, and those minor adjustments mentioned are marked and proved.
This is the first master roll and is duplicated in the normal way to
a production master and then to commercial copies.  Later changes in
physical scale of the rolls to the T98 and American scales are simple
mechanical accommodations for increased commercial acceptance.

The next steps needed to improve the process are to automatically
regulate the player systems.  The piano will always require some
attention of a technician to tune and regulate the piano action,
but the same technicians are often incapable of regulating our system
to provide even adequate performance.  The mechanism should be
self-regulating to variation from atmospheric changes and normal wear,
since the owner will always assume that the defect is in the design of
the system and not in lack of ability of the technician who has just
charged a premium for his expert service.  A simple mechanical routine
based on reiterating bleed adjustments to a time base can trim the
crescendo rates each time a roll is ....

 - - -

Now, there it ends abruptly since the cardboard back of the tablet and
who knows how many pages have been ripped off.

Karl Petersen
Washington, Illinois