The two delightful emails transcribed below are a rare treat of story telling about the design and use of analog computers, as applied to flight simulators. Much of this design experience is directly relevant to the contemporary field of robotics.

Take note of the personal involvement in data gathering and attention to circumstances far outside mere equations.


-Joe Sousa, Webmaster.

From: "Ed Lyon"  

To: <> 

Subject: ac analog computers 

Date: Mon, 6 Aug 2007 07:22:22 -0400 


Enjoyed the historical Philbrick material.

Wondered if you ever found any of the military documentation of the huge analog aircraft flight simulators of the 1950-1970 era.  Many of these simulators used a-c-signals rather than d-c, and most used 60-Hz signals at that.  I designed a number of models, the most prolific being that for the F-86D single-seat interceptor aircraft.  Our company (ERCO) built 33 of these F-86D simulators for Air Training Command, and we placed them in training centers, like Perrin AFB, Texas, Davis-Monthan AFB, Yokota AB, Japan, Tyndall AFB, FL, among others.  They solved the equations of flight (third-order differential equations), using a-c operational amplifiers (I think the F-86D model had about 180 op amps) and a-c integrators and servos (60-Hz two-phase motors and drag-cup generators driving gangs of precision pots for multiplication and function-generation).  Some of the later models had cockpit motion, but the realism of even the non-motion models was so high that most pilot-trainees came out of the cockpits soaked in sweat and wobbly on their feet after a chase of enemy aircraft, especially through simulated stormy weather.  All the original aircraft instruments were used in the cockpits, some unmodified and operated through 400-Hz synchros geared to the servos for that particular function. 

The op-amps had four tubes, and had push-pull outputs providing an output signal normal (inverted WRT the input) or uninverted (for those functions that needed the oppositely signed quantity).  The standard output signal was +/- 50 V rms.

Servo/integrator amplifiers were big power amps that drove the Kollsmann drag-cup generator-driven servo-motors.  These had integral gear boxes that coupled the motors to the stacks of T-I-C precision potentiometers.  Some of these pots had upwards of ten or twelve taps, custom placed to match the designer’s need for functions of the servo variable.  When the required number of function pots exceeded 15 for an integrator, the drag was too high for the integration precision needed, so the number of pots was dropped to a few, and a “follower” servo that was a closed-loop position servo carried the load of needed pots.  The KC-135 aircraft simulator needed three Mach servo followers, for example.  That aircraft simulator used a new form of flight equations that did not solve for the aircraft-axis-oriented three velocities, but instead, solved for Mach number, which was more intimately connected with altitude, fuel flow, throttle position, and aircraft attitude than was the normal longitudinal airspeed, as was calculated in my F-86D simulator.

I am looking among my old literature for a handbook for the F-86D simulator, which, if I find it, I will reproduce for you.



Science and Technology Associates, Inc. - Arlington, VA 22203

From: "Ed Lyon"

To: "'Joe Sousa'" <> 

Subject: RE: ac analog computers 

Date: Tue, 7 Aug 2007 09:08:29 -0400 


Glad you like the idea of including flight simulators (analog) in the archive.  I find essentially zilch on the internet on them.  However, I found the attached picture of one that I personally designed and ran through its flight tests in acceptance trials for NATO.  It is an F-86K interceptor, built for the French, Italians, and Germans, who couldn't get the F-86D because of the secrecy of its fire control system (Hughes E-4) The F-86K used the same radar as the F-86D, but coupled its output to the display in a  simple lead-pursuit computation that steered the pilot to a tail-on shoot-down attitude, since the F-86K had guns, not rockets.  The F-86K simulator was in two trailers, one for the student's cockpit (the view shown in the attachment) and instructor's consoles, and the other trailer was for the analog computers.  Incidentally, the cockpits for all the F-86 series of aircraft flight simulators were taken from crashed aircraft, which our shop had to dismantle and rebuild as necessary.  The F-86K system had to be able to run on 60-Hz or 50-Hz power (and signal), since it operated in Europe, most of the time.

I recall that the joystick forces in the F-86K were strongly a function of airspeed and altitude, and the pilot had hydraulic assist in the elevators and ailerons, only.  We had to reproduce those joystick (and rudder, too) forces for the pilot, and they changed dramatically with engine flame-outs, of course (loss of hydraulic assist).  The joystick in the real aircraft had a bob-weight on the front of the pitch-control pivot, to tend to neutralize g-forces the pilot might inadvertently pull.  Thus a jerk back on the stick, causing a sharp nose-up tendency, would pull positive g's, and the bob-weight would then try to pull the stick forward to reduce the g-forces.

We had to simulate that effect as well.  These forces were all applied through a counter-rotating fluid-drive system that used a magnetic fluid. Magnetizing coils then caused more or less viscous coupling in one or the other direction of rotation, and an answer potentiometer then servoed the magnetizing currents.  Best thing about these clutches was the slight hydraulic whine they made, and it sounded just like the hydraulics in the real aircraft.

I noted a comment on the web by a veteran of some USAF fighter-interceptor squadron who said the F-86D simulators were "beautiful machines that made you immediately feel at home in the actual aircraft."  Most commented-upon training aspect was from those who had had flame-outs in flight, since the engine controls, sounds, and reactions we built into the simulation were very much like those in the GE engines, especially the delicate way the pilot had to treat the "emergency fuel control system," used after a flame-out, to prevent a compressor stall and a belch of flame forward, out the intake scoop.  The way I learned how the actual plane behaved under all these weird conditions was unusual.  The pilot's handbook was too terse, so I went to Andrews AFB, where there was an F-86D squadron, and introduced myself (a brash 24-year-old junior engineer), and asked if they would fire up a tethered F-86D for me to record all the engine characteristics.  They gave me a set of earphones to prevent me going deaf, and brought out a plane to play with.  They taxied it to a hard-stand with an exhaust deflector, and chained it down, started it up and let me watch for a while.  When the pilot learned that I had more cockpit hours than he had, he climbed out and let me run it through all its paces.  I went through two tanks of fuel, tape recording everything, and taking copious notes.  When I was finished, he let me taxi the plane back to the hangar.  Can you imagine such a thing today, with planes costing 50 to 100 million each?  The F-86D was a $250K aircraft.

But that experiment was the basis for the engine simulation in the F-86K.

Regards, and apologies for the war stories.


 Click cockpit image to enlarge.