top of page

A Somewhat More Focused View



In the mid-1990s, I developed a balloon autopilot for Steve Fossett to use on his attempts to make the first solo circumnavigation of the earth.  I had been co-director for equipment preparation and launch of Fossett on his Pacific-crossing flight in early 1995.  On this flight, Fossett had used a rudimentary analog autopilot I had originally designed and tested in the late 1970s.  This analog device was unreliable and tended to drift off altitude.  After Fossett’s successful Pacific flight, I proposed that he contract with me to design and build for him a reliable balloon autopilot based on a digital computer that would keep his balloon at any desired altitude.  Fossett agreed. He used the resulting autopilot on all his round-the-world flight attempts, including his successful nonstop circumnavigation of the earth by balloon in 2002.  Fossett named the device I developed for him the “Comstock Autopilot.”


Design Specification


The design specification for this device included several requirements.  Important among these were that the autopilot would


  • fly a balloon more closely to a desired altitude than any human pilot could,


  • be able to fly any balloon controlled by a burner, that is, balloons where the lift is created by either hot air or a temperature-controlled lighter-than-air gas.  This would allow flight-testing to be done in convenient and inexpensive to fly hot air balloons,


  • require no input regarding what balloon it was flying.  That is, it would learn what it needed to know about the balloon it was flying by learning the flight characteristics of the balloon literally on the fly,


  • be rugged and reliable,


  • use an acceptable amount of electric power.


The resulting autopilot met these specifications.


The Main Components


The main components of the autopilot are a Z-World BL1100 single-board control computer, an accompanying Z-World relay board, a Ball Model 400 electric variometer, and an ancillary circuit board that includes the atmospheric pressure sensor and circuitry to support indicator LEDs and piezo alarm.   The Z-World company no longer exists, and much better single board computers are now available – several decades later.


The Central Algorithm


This autopilot operates on a 6-second or 15-second cycle, selected by the user.  Once each cycle, the autopilot determines how much heating is needed at that time.  A single line of code in the program calculates the amount of heat needed, from none to an amount equal to the full length of the cycle.  This calculation is simply a weighted average of acceleration, altitude error, and vertical velocity, with the sign reversed.  The approximate relative coefficients of these are, respectively, four (acceleration), two (altitude error), and one (vertical velocity).  If the result is negative, no heating occurs.


The Details


The devil is partly in the details.  Most of the computer program is code for obtaining the best input data for altitude, vertical velocity, and acceleration.  Altitude is calculated from barometric pressure as sensed by a pressure sensor intended for automotive engine controls.  To eliminate the effect of noise, this is read 200 times and averaged before it is converted to altitude.


Vertical velocity comes from a sensitive electric variometer, or rate of climb, again read 200 times and averaged to eliminate noise.  The particular variometer used in the autopilot is one that outputs a particularly clean, accurate signal.


Acceleration is calculated from change in the measured rate of vertical velocity over small time increments.


Although the atmospheric pressure sensor is compensated for temperature change, the variometer used is not well compensated.  In effect, the zero point of this device drifts with temperature.  Instead of measuring this drift and tediously calibrating each finished autopilot to the particular temperature drift of the variometer, the autopilot integrates vertical velocity over time and occasionally compares the calculated net change in altitude with the actual change in altitude from the atmospheric pressure sensor during the same period.   Any difference is assumed to result from variometer zero drift, a correction is calculated, and this is added to subsequent vertical velocity readings from the variometer.  In effect, the variometer is thus “zeroed” on the fly.


Performance of the Comstock Autopilot


Pre-delivery testing of each unit included a flight test on a hot air balloon, with altitude recorded on a barograph.  Typically, this testing would include disabling the burner unit until the balloon had fallen into a fast descent, followed by watching the autopilot fly the balloon back to the desired altitude.  Also typical, was the sudden release of 60 pounds of ballast, also followed by watching the autopilot fly the balloon back down to the desired altitude from the ascent that would result.


In service on large temperature-controlled helium balloons, as were being used in very long distance round-the-world flight attempts, the altitude error was less than 30 feet for low elevations and less than 300 feet in the thin air at 30,000 feet altitude -- usually much less in both cases.


Feasibility of a Balloon Autopilot


The success of this autopilot in flying both hot air balloons and large temperature-controlled helium balloons on many flights demonstrates that a balloon autopilot is feasible.  The Comstock Autopilot was developed over a period of several months by a person not previously trained in the design of control devices.

bottom of page