Fran's Writings on Design and Engineering
Page 9

Return to the Design Writings Main Menu Page

Subscribe to my YouTube Channel!
List of articles on this page:
(April 2014)
The Vintage Computer Festival East, 2014
Reverse Engineering the Maillardet Automaton
Recreating The Original Maillardet Writing Instrument
Making instructional videos, writing articles, and maintaining my blog are all quite expensive.  Please help me continue to make more and better content by becoming a contributor!  Please support Fran's Blog - 
For Science!
Follow Fran on Twitter:


Subscribe to Fran on YouTube:

VCF East 2014
(April 2014)

I attended the Vintage Computer Festival on April 6, 2014 in Wall New Jersey, and there was so much cool stuff on exhibit that I have to break up the days events into three separate videos.  There was a lot of Commodore stuff on display, so part one is all Commodore - just for the Commodore fans! Special appearance by Bil Herd and many other Commodore nuts and their cool machines:

 This is part two of my three part series from the Vintage Computer Festival on April 6, 2014 in Wall New Jersey. Here I take a look at one of the Apollo Guidance Computers (AGC) with flown components from Apollo 14, and I take a tour through the great exhibits with running 8-bit and 16-bit systems and some demonstrations on how these 70's and early 80's systems worked. You can read about all of the VCF 2014 exhibitors and their computers on the website here:


Dave Haynie talks at VCF East on April 6, 2014 about developing the various Amiga systems, up into the last days of Commodore in April 1994. Introduction by Bil Herd. This was a fascinating hour of must-hear stories for any serious Commodore fan. Dave even wears his Commodore Death-Bed Vigil shirt! Wow - 20 years ago.... Where did the time go?



Reverse Engineering the Maillardet Automaton 
(February 2014)

I spent many weeks thoroughly examining video of the Maillardet automaton in operation that Andy Baron had taken, as well as his photos of the automaton in various stages of disassembly, and I became well educated in its operation and structure.  I came to the conclusion that the eyes of the automaton were originally fully animated, and while in operation the eyes used to follow the pen on the paper, complete with variable ocular convergence, which is a subtle trait not seen in any other known automata.  These features would have given the head of this machine a truly life-like realism that has long since been lost.   In its heyday the Maillardet automaton would have been singularly deep in the 'uncanny valley' as a result of this full articulation in the eyes.  I have reconstructed a reasonable arrangement for the missing linkages and mechanisms that once created convergence and alignment, based on what currently exists in the machine, and taking into consideration more recent additions and alterations, such as lead weights, cemented eyes with brass bar, and other reworked parts in the head that represent attempts to restore or stop various kinds of movement in the eyes at various times since the loss of the original components. 

I can see that originally there existed three small cables which transferred actions from the cam followers through points on control levers in the shoulder and a pulley in the body, which when combined with the motion of the head (tilt and rotation) created additive and subtractive movements which allowed the automaton to continually 'look at' and follow the drawing hand with its eyes.  The three cables came up through a wide opening in the back of the neck to a missing plate that once swiveled on a guide pin (still present), where springs and cams reduced the throw of these three cables to levers, transferring that motion to the eyes (X and Y axis) and eye lids (Z axis). 

Photo taken by Andy Baron of disassembled neck assembly and new neck ring he made to replace the damaged original, and my drawing showing where the original cable guide plate was located inside.

Based on what currently exisis in the head I deduced that the eyes originally existed in a Master/Slave arrangement, with the movement of the left eye controlled by the linkages and the right eye following.  A cam at the rear of the Master left eye (removed) pushed against the spring that is still in place there, which kept the eye looking to the right unless acted upon by a lever on the bottom of the left eye mount, since removed (this part has been reworked), which was acted on by an arm controlled by the cable on the left of the neck to the main X axis arm to the left of center in the shoulder.

Photos taken by Andy Baron of the remaining eye mechanism, removed from the head.

Reworked parts and mounts for lateral controls.

 The Slave right eye follows the lateral action of the Master left eye by the interaction of a spring-mounted transfer yoke that performed the combined task of keeping the eyes in register with one another and creating realistic convergence.  This variable convergence (near and distant seeing) happens as the semi helical "tails" (which can be seen hanging off the eye mounts) move through two slots in the transfer yoke relative to the vertical eye position (Y axis control). 

Missing convergence yoke and spring.

The transfer/convergence yoke kept the eye tails at a constant distance apart in a fixed horizontal plane relative to a mounting point at the front of the head (single screw mount).  As the eye assembly tilts by combined action of head tilt against another spring mounted lever that has since been modified (forward on neck, to the right) and controlled by the cable on the right side of the neck to the Y axis eccentric pulley that is just to the right of center in the body.  The semi helical shape of the eye mount tails caused ocular convergence (near looking) as they slid through the two slots in the transfer yoke, causing a mirrored rotation of 10°-20° difference in the alignment of the eye mounts when the eyes and head are both down at the lowest point and the ends of the eye tails are in the slots, and also causing alignment (distant looking) of the eyes when the head and eyes are both up and level with the higher parts of the eye tails are resting in the slots of the transfer yoke.  The transfer/convergence yoke would also yaw by flexing the spring which mounted it by a single screw to the front of the head, and this yaw action constantly transferred the lateral motion between the eyes in proper register. 

Another Andy Baron photo that shows the pathway of the missing vertical eye control cable through a missing guide wheel in the neck, from the front of the eccentric pulley which also acts as a variable Y-axis counter weight to balance the arm.

Lastly, a cable coming from the pin at the end of the Z axis arm in the center of the back controlled the eye lids through another actuator in the center of the head, again through that missing plate, controlling the lids through additive and subtractive movements of the head tilt combined with the lifting and lowering of the hand to the paper. 

Andy's photo of the partially disassembled shoulder showing Z and Y axis controls, which I have labeled here.


Recreating The Original Maillardet Writing Instrument
(February 2014)

The 1801 drawing in possession by Westminster Library shows us many things about the ideal rendering capability of the automaton, since this drawing was rendered under the guides of the Maillardet shop and the machine's settings and function at that time of the rendering are assumed to have been ideal, or as best the machine could have ever functioned. There is of course the incredibly fine line width and detail in this rendering of the Chinese Temple, which we assume is in Maillardet's ink and rendered at the original size with the original drawing instrument. 

Maillardet Automaton rendering, dated 1801, collection of the Westminster Library.

Two curious effects in the 1801 drawing are the numerous times that lines either skip or stop prematurely, and other lines that overrun stopping points at "T" junctions, which I will address in the second part of this letter.   The lines skipping is odd, as I had stated that this rendering would be under ideal conditions, and I will assume that the writing instrument would have been designed specifically to be reliable enough to draw a constant line, so the best explanation for the line skipping is that the tip of the pen is not consistently contacting the paper, due to the fact that the best paper or velum of that time was uneven and never completely flat.  The fact that a natural depression in the paper would cause a skipped line indicates that in the ideal settings for the automaton that the actual tip of the physically mounted pen was not ever intended to be in contact with the paper. Instead, I speculate that an extremely light stylus riding inside the pen transferred a specially formulated high flow ink to the paper (see attached drawing) through the means of capillary action down a fine tube at the writing end of the pen.  In my drawing the tolerance of sizing between the tube and stylus is exaggerated, and in practice this would have to be a rather snug fit with very minimal play between the parts. 

My simple drawing of the floating stylus pen, with exaggerated gap - 
the real thing would have fairly snug parts.

The stylus could have been lightly weighted by a ball at the top to also serve as the valve to stop ink flow then the stylus is raised high enough off of the paper.  The ideal distance from the end of the tube in the pen to the surface of the paper is set as the absolute minimum distance when the Z axis follower is the lowest points of the Z axis cam, when the automaton is drawing a line.  Analysis of the 1801 rendering shows variations in ink flow and line width, and when comparing that early rendering to the automaton's actions while rendering today it is clear that minute variations in the Z axis cams are present to control ink flow and line thickness by varying the specific distance between the tube and paper, thereby having fine control over the opening of the valve, and regulating the flow of thinned ink over the stylus of the pen.

In this sense the Y axis compensation ratchet spring in the body of the automaton performs a role identical to the counterweight on the tone arm of a modern turntable, which regulates the tracking force of the stylus in the record groove.  The ratchet set spring in the automaton interacts with an eccentric pulley as a variable compensating counterweight on the drawing arm through the Y axis control, and it is critical to being able to set the precise counter balance to the weight of the arm, which varies according to the leverage throughout the full travel of the hand at all points, both near and extended.  This is so that the tip of the tube of the writing instrument remains within the range of necessary spacing that allows the stylus riding inside the tube to contact the paper when a line is intended to be drawn without the tip of the tube in the writing instrument ever making contact with the paper, as this would ruin the rendering, but also allowing the stylus to be lifted off of the paper for various brief fine detail starts and stops in some complex high detail parts of the drawings.  Andy pointed out this kind of detail to me early in the very tiny face of the figure he found in the 1801 Temple drawing.   I do suspect that some if not much of the distortion expressed in current renderings by the automaton are due to the drag of using a modern pen, the incorrectly set counterbalance, and the applied weight of the arm on the paper, in combination with the obvious wear in all points of the articulating mechanism from time and use.

Collection of the Westminster Library

In close up: on the left is a paper print of the original copper plate etching (mirrored here) used to create the cam programs by use of a 3-dimensional pantograph, which produced the 1801 rendering by the Automaton on the right.  The rendering is drawn in ink without any dipping, the detail is astounding, and the drawing is tiny - The base of the Temple that you see above is just under 1 inch across.

Weight Is Everything:
The Y axis compensating ratchet spring is critical to understanding all of these components in the refined rendering system, and evidence to this is in the fact that no such compensation was engineered into the automaton to deal with the inherent problem of X axis inertia that causes the horizontal lines to often overshoot the stopping point if the line is long.  Balancing the issues of inertia and a necessary minimum speed of the motor governors was a problem that likely became apparent by the time of the 1801 rendering, as the overshoot on X axis lines in that early drawing indicates that the speed of the machine was probably much faster than was originally intended when the cam programs had been created.  The lack of any mechanical compensation for the lateral inertia of the writing arm is a strong indication that the great care that went into engineering such a precise Y axis compensation means that this setting probably exists solely for the purpose of insuring the correct balance to keep tolerance of the gap between the tip of the pen's tube and the paper, in order for the low mass stylus to function properly inside of the writing instrument. 

Alignment Procedures:
The large hole cut in the in the body of the automaton right in front of the Y axis eccentric pulley is the final key to understanding the floating stylus writing system, and suggests a simple alignment procedure.  A reference position in a program would be marked at a point where a line is to be drawn (pen down on paper) and where the arm is at the full extended position, such as at the highest point of the arc over the Chinese Temple.  With the machine stopped, an operator standing behind the machine can look through the large hole in front of the eccentric pulley to view both the specific rotational position of the pulley and the spacing of the pen tube to the paper at the same time, to properly set the  balance of spring tension in the Y axis compensation with a key in the ratchet spring.

I had pondered a great deal over weeks as to the reasons why if Maillardet had come up with a high definition ink pen for the automaton, why had it not been reproduced and marketed for use all around Europe?  Such a discovery would have been revolutionary in the literate world.  But as I put the pieces together the answer became crystal clear this specific instrument that allowed the automaton to draw such an unprecedented fine ink line could not really be manipulated consistently in a human hand.  Only a precision machine like the automaton, with very narrow operating tolerances, could maintain the necessary level of control to hover the tip of the writing tube above the paper in such a way as to consistently manipulate the tiny stylus, weighing only a few grams at most, that rests within this pen. 

Andy Baron's blog about his work restoring the Maillardet Automaton in 2007 -

Andy Baron's documentation of the 2007 restoration is here -

Return to the Design Writings Main Menu Page

 Frantone Home  ||  About Frantone   ||   Contact  ||   Fransworld Daily Updates

©1994-2014 Frantone Electronics  All rights reserved.
All images and text are copyright Frantone Electronics.  No content of this website may be published or distributed without prior written permission from Frantone Electronics and any reproduction or manipulation of the content of this website for any purpose is strictly prohibited.