Jim Green in 1983 at AlphaType/Berthold in Tampa.
The terminal shown supported Bourne-Shell UNIX.
I wrote 4 big notebooks full of application software
in C shown on the shelf and collected applicable
textbooks and manufacturer's catalogs of components.
During my 2nd year of engineering design work at
Alphatype/Berthold I prepared
graphical servo simulations
in the C programming language
in a UNIX
and produced mathematical
modeling of servo performance
in C for a very high-resolution
servo system with micron positioning accuracy that was capable of fast motions to support
Berthold's High-Speed CRS typesetter. A
and built with
was responsible for
the resolution of the optical shaft angle encoder
in the system to achieve extremely
. It did this by
dividing the optical shaft angles
into subintervals 1/16 or 1/32 as wide as the resolution interval provided
by the manufacturer of the shaft angle encoder in the first place.
I designed a Multibus interface
for the XY servo positioning system
for the paper feed and lens position axes involving digital logic and parallel interfaces that connected to an
existing servo amplifier
card in which
was built with bipolar transistor technology
. Today integrated power op amps can often be used
to this job.
The twin servo amplifier
Typical DC Servo Motors,
with or without integral
DC tachometers and/or
shaft angle encoders.
A shaft angle encoder
typically outputs two
signals with a phase shift.
behaved like a pair of power op amps that could source
high currents for driving the DC servo motors. Also,
I redesigned the old Eurocard servo controllers into a form that was easily
analyzed mathematically as a second-order servo with a tachometer-controlled rate loop
inside a position loop, which provides excellent holding torque.
The behavior of the system was easy to simulate, as the 2nd order servo is mathematically described as
a 2nd order polynomial with two complex roots that is well-defined. There was a fast positioning mode
with different 2nd order polynomial description, but the intial conditions were well-specified and
the mode-switching problem was analytically straightforward.
This system demonstrated higher performance and superior characteristics. Two
cards were used, one to control film feed and another to control lens
movement, each separately tuned for its unique inertial load, which I had to
compute from calculus and the density of materials, because the inertial moments
appeared in the constants of the 2nd order equations describing the system.
Also, I designed a servo simulator test instrument and a phase
measurement instrument for making adjustments to the servo system.
The servo worked just as
predicted, since the servo bandwidth was under 200 Hz.
A mode-switched servo design was employed using a tachometer-stabilized
system that could be mode-switched to change the feedback characteristics
of the servo loops. Modeling used a time-domain continuation approach
using initial conditions in the
that were well-determined from previous phases of the motion.
An alternative simulation model
Finally, I developed a
-based servo simulator, controller, and
analog I/O multiplexing card for production testing of servo controllers,
and made proposals for phase-locked loop
. Most of
the hardware was built up and checked out by me personally. The servo
controller typeset the most demanding jobs with a positioning accuracy
of 1 micron, at higher speed than the previous system. I probably deserved a raise!
as a C programmer in a UNIX environment, I worked on character
generation algorithms for phototypesetters and conversion of vector-outline characters
to a form suitable for an interactive computer graphics
terminal, the spectacular Berthold Magic System. My software
contribution included character representation re-coding and smart character
fill routines and produced handsome output on a printer, spinning out
character sets as a batch system that converted alphabet after alphabet
from vector-outline form to run-length-encoded form without human
intervention. This went unsabotaged, probably because it seemed within
the reach of ordinary human intelligence.
Dr. Peter Krumhauer came up with some novel and ingenious methods for
applying the theory of cubic splines
to character generation
during my first year at AlphaType/Berthold. These were chains of curves described by
third-order polynomials featuring four constants per polynomial and enough degrees of freedom to specify one tangent
vector at each end of every cubic seqment. This was sufficient to fluidly describe virtually any
character outline with small set of functions. I wrote
programs for applying the spline theory to the generation of large
multipage images that generated a beautiful, curvaceous "W" character that could be
size-magnified to any degree and output as a multipart printed piece that
could be stitched together and shown to executives back in Berlin as a giant wall hanging.
I took my wife with me to Berlin with the software and gave my technical German a
workout at lunchtimes for a couple of weeks, chit-chating with staff educated at the
Sorbonne and other prestige institutions in Europe as we sipped glasses of wine over sandwiches.
Beer was vended in the hallways there,
but I skipped it to keep a steady head on my shoulders. Polynomial splines were also used in two
dimensions to represent surfaces in computer graphics for automobile and 3D component design.
Using spline theory to generate two-dimensional characters had already
occurred to myself and other design engineers, but Dr. Peter Krumhauer had brilliantly
rediscovered some things that are hard to prove, and carried them further
forwards than ever with ingenious refinements and experiments on the
achievable density of the representations and compaction on his PC. I
proposed implementing the difference equations generating the splines
with a 68000 and actually
building a suitable typesetter graphics engine, but our principal results were magnified graphical
images that proved the utility of the method. The difference engine
approach to generating polynomials solely by additions of differences
led to fast algorithms for character generation from very compact data
sets. Flawless characters the size of billboards could be produced.
Perhaps the most entertaining days I had at that time were the ones
in which I investigated the way polynomial spline characters can be transformed
to produce various distortions, such as barrel distortion, pincushion
distortion, and other special effects.
The video memory planes and character scaling hardware in the pipeline
character processor were redone by me for The Fox series of typesetters
as standard format Multibus cards.
This way, our special-purpose card(s) could be mixed with standard-issue Intel CPU, memory, and I/O cards
to rapidly generate prototype system to be displayed a trade shows. Production systems might feature
custom CPU, memory, and I/O cards for cost reduction.
I worked on my piece of the typesetter graphics engine
while the other engineering staff worked on other components of the entire product line, which
featured more than one typesetter and a front-end system with multiple terminals.
I was in the same room during the
first year with Charlie Hayek, another MSEE working on an exotic servo scheme
that avoided a tachometer in the rate servo loop, but had poor holding torque. Charlie had
discovered this weak spot in an academic proposal written up in an applications textbook.
Without a tachometer in an inner rate loop, the sans-tacholmeter servo did not have a springy hold on the target
position, but could be broken loose into a spin. Charlie drilled it into
me to always put a tachometer in the inner servo rate loop to get solid holding torque,
and never mind theoretical savings from not using it. To that I owe my eventual success in
producing a better servo controller than Charlie could produce with his manager insisting on
a cost-saving scheme that turned out to have an unacceptable weak spot that could not be
repaired without adding a tachometer, although the up-front analysis and cost justification
looked at first quite reasonable. Thus Charlie fixed the book design by demonstrating the analytical solution
to the holding torque with small deflection problem, winning for inclusion of a tachometer in the system.
The third man in the room was an engineer who finished his project by jumping up and down on his card,
somewhat distressing our faith in the inevitable progress of cooperating international characters of
(Charles River Data Systems: UNOS, UNIX Version 7, Berkeley UNIX with the
CSHELL, also the Bourne Shell, digital video design, servo system design
and simulation, analog design, microprocessor system design, and
H.Berthold AG, once the largest manufacturer of phototypesetting equipment in Europe, folded in Berlin in 1993. I believe that pressure from inexpensive ink jet printers
and laser printers producting results comparable to those obtained with high-priced videosetters and old-fashioned phototypesetters
was a factor. Another was that inexpensive Windows PCs, with the introduction of the electronic mouse, made
it possible to do things with fast general-purpose PCs that had previously been limited to machines with special keyboards and tablets such as page makeup workstations like the massive and elaborate Berthold Magic System. Remnants of AlphaType and Berthold survive today as font shops. For Berthold in the movies, see The Adventures of Baron von Munchausen (1988). In 1878 H. Berthold defined picas and points with 1m = 2,660 typographic points.
During my first stint at AlphaType/Berthold my wife Jo and I lived in Palm Harbor north of Tampa,
then moved to King Arthur's Court in Dunedin while I worked at ABA Electromechanical Systems, which became
General Defense. We were still living there a year later when I returned to AlphaType/Berthold and applied servo methods I had practiced at General Defense to our line of phototypesetters. When I went
to Berlin to install character translation software for the Berthold Magic System, Jo came along. Like the Baron von Munchausen, I seemed to enter and leave the company riding on a cannon ball, toting off perhaps as much treasure as the strongest man could carry. Dr. Peter Krumhaur, who hired me to work with him on fast spline character generation techniques for a new generation of High-Speed CRS typesetters, sported a full beard that might remind you of the Sultan in the movie.