PRELIMINARY REPORT .. .

DEVELOPMENT OF A POOR MAN'S VFR AUTOPILOT By Don Hewes (EAA 32101) Aero Research Engineer 12 Meadow Dr. Newport News, VA 23606

-L HIS IS A Preliminary Report on what I call a Poor-Man's VFR Autopilot that I have developed and installed in my BD-4 airplane. In more technical terms it is a limited authority single-axis wing leveler with a heading-hold feature that permits continuous hands-off flight in smooth and light turbulent conditions. "Single axis" refers to the roll axis, and "limited authority" refers to the fact that the maximum control power of the system is limited, in this case, to about 10% of that for the ailerons. The autopilot authority is purposely limited as a safety feature so as to minimize the problems caused by a possible hard over failure of the autopilot system. The system is intended to be used as a pilot aid only during normal cross country flights under VFR conditions, and not under abnormal or emergency conditions which could require full control authority about all three axes. It represents the simplest and lowest cost autopilot system, and relies heavily on the inherent stability of the airplane in the pitch and yaw axes to ensure stable flight. Total cost for the system was about $60. This report covers a general description of the system including the circuit details, installation in the air-

plane and its operation. However, it is not intended as a construction article because the system is still under

development. The heart of the system is the wing leveler which is the brainchild of H. Douglas Garner, a NASA Research Instrumentation Engineer, whom I have known for many years. In Reference 1, which was published in SPORT AVIATION about four years ago, Doug presented a humorous account of his early fluidic autopilot for general aviation type airplanes. If you were fortunate enough to sit in on his forum at the 1977 Convention at Oshkosh, or the 1978 Sun 'N Fun Fly-In at Lakeland, you heard him describe subsequent efforts devoted to applying this technology to the homebuilder's area of interest. Shortly after Oshkosh '77, Doug gave

me a set of the notes that he handed out at the Forum and I found that his system would operate an aileron trim system that I had planned to install in my BD-4,

and would also operate the same system as a wing leveler. (Note: a wing leveler will hold the wings level but cannot maintain a desired heading.) Working from Doug's notes, I was easily able to

assemble a breadboard system using parts obtained primarily from Radio Shack at a very modest cost. This was my first experience with these solid state chips but had no problem whatsoever. I found, however, that the circuits did need some modifications to work properly in the airplane. Doug had indicated at the time that he had not had time to thoroughly check out the circuits, so this finding was not unexpected. After installing my first operational unit with the few 24 MAY 1978

modifications, I completed a few flight tests that showed the wing leveler was working correctly. It was evident, however, that the roll-control tab was not powerful enough. (I had merely guessed at the size and mounting dimensions.) More important was the demonstration that the wing-leveler by itself was not what was needed. Doug had a magnetometer system he had been working with to provide the heading hold feature, but he did not have it developed for the homebuilt system yet. Therefore, I decided to add the heading hold feature to my wing leveler using the directional gyro that I had already had in the airplane. After spending a couple months learning how to design and build circuits for photo transistors, I was able to mount an acceptable heading sensor on the face of the directional gyro and couple its output with the ratesensor of the wing-leveler. I then spent a couple hours flight-testing the system to make proper adjustments and determine its operating characteristics. It was this unit that flew the plane, almost unaided by me, down to Florida and back for the Sun 'N Fun Fly-In this year, and did so without a single failure. At the time of writing this report there are several refinements that I am planning to incorporate and test. Consequently I am not going to present complete details of the system. Inasmuch as the interest in the system was so high at Lakeland, I decided that I ought to present some form of a preliminary report. There may be some experimenters who will be encouraged to proceed from here and apply some of their ideas to this very interesting and important field. There are, of course, many ways to solve a particular problem and who knows at the beginning which way is the best. A schematic diagram of the system is given in Figure 1 to show most of the elements of the system. The detailed circuit diagram is presented in Figure 2 but I won't discuss it in this report. Starting at the left side of Figure 1 we see the clock representing the fluidic rate sensor which measures a combination of the rolling and yawing motion of the airplane. This sensor takes the place of the normal mechanical rate gyro which is fairly heavy, expensive and sometimes unreliable. The fluidic sensor, as indicated in the sketch of Figure 3, is merely a very small jet of air passing between two very small thermistors which change electrical characteristics uniformly as the air jet is deflected to impinge one or the other thermistor. Deflection of the air jet is caused by the rolling and yawing motions of the airplane to which the sensor is attached. (This action is the same effect you observe when you are holding a garden hose and rotate the nozzle back and forth slowly; you observe how the water stream appears to curve in the direction opposite to the direction of rotation.) You get response to both rolling and yawing motion by mounting the sensor so

that its sensitive axis is tilted up relative to the longitudinal axis of the airplane. This combined response is desired for optimum effectiveness of this system. The construction of this sensor is extremely simple and can be accomplished merely with a saw and drill.

However, I'll not go into this at this time. A forthcoming article by Doug will cover these details and many others as well. A small trim control is provided so that the sensor can be adjusted to zero output when there is no turning rate. In the refined version this control may not have to be readjusted once it is set.

The heading sensor consists of a thin plexiglass disc

held to the face of the directional gyro by two small clips that permit the disc to rotate freely. The disc bears against the thin metal ring that is the retainer for the glass face cover so as to eliminate direct rubbing of the plastic against the glass. A small plastic block attached to the disc houses a small 12-volt light bulb and a photo transistor. The only source of light for the photo cell is

a pinhole in the housing toward the face of the compass. The only light to enter the pinhole is the light that is reflected by the compass card. The plastic block is positioned on the plexiglass disc so that the pinhole is on the same radius as the bottom of the direction letters (N, E, S, and W). The letters, being white on a dull black background, reflect more light into the photo transistor when they are aligned directly under the pinhole than when they are displaced to one side or the other. The photo transistor, therefore, is literally made to "see" the letters and to generate a signal that is proportional to the relative rotational displacement of the sensor and compass card. The letters E and S are used for sensing the position of the compass card because they are both formed by lines at the bottom that extend circumferentially about 10 degrees, and thereby provide better sensing signals. Actually the objective is to sense one of the edges of the letter and not the letters as a whole. The system can be made to home in on only one edge of each letter. Therefore, it is necessary to have the two edges space as far apart as possible. In operation, the airplane is placed on the desired

heading and the sensor disc is rotated so as to place the sensor directly over the "homing" edge of either the E or the S whichever is the most convenient to use. To ensure correct alignment of the sensor, the disc has a reference line scribed on it about 180 degrees from the sensor. Aligning this line so that it is directly over the

N or W guarantees that the sensor is properly aligned over the E or S. If the airplane departs from the desired heading, the rate sensor detects the turning rate and generates a

signal to oppose this turning. The heading sensor also

detects the displacement of the edge of the E or S which moves from directly beneath the sensor. It either gets

more reflected light or less depending on which direction the airplane turns and therefore is able to determine the direction of turn. Of course, if the heading changes

so rapidly due to a large gust that the other edge of the letter, or the next scaling mark, passes beneath the sensor, then the sensor will think that the airplane has turned in the opposite direction. However, this will

happen only in moderate to heavy turbulence and,

therefore, is not a problem for normal flying conditions. The two sensor signals are passed through amplifiers with their outputs indicated by very small micro-

ammeters of the type used on small portable tape recorders. The two signals are then combined through a potentiometer that is used to change the relative gain of the two signals. The rate-heading pot output is fed through a buffer amplifier (amplification factor of 1) to a limiter which keeps the signal within specific limits so that the servo will not be overdriven. A selector switch can be set to pass either the limiter output or the output of a manually adjusted potentiometer on to the pulse width modulator (PWM).

T - T R I M POT K - GAIN POT G - LINEAR AMPLIFIERS

B - MICROAMMETER PWM - PULSE WIDTH MODULATOR

FIGURE 1

The manually adjusted pot is used to control the servo system when the autopilot is turned off. In this way the system provides a convenient aileron trim system for use when the autopilot function is not being used. The output of the PWM is a 5-volt square-wave pulse train that is used to position the servo unit located out on the left wing tip. The servo is a standard unit normally used in small radio controlled model airplanes and weighs about 1% ounces, and is about the size of half a pack of regular cigarettes. The servo positions itself with an accuracy of about 1 percent of its normal travel and will exert a force of up to about 6 pounds while drawing under 1 amp of current. Power for the whole system is derived directly from the airplane's systems. The air jet of the rate sensor is produced by the vacuum system used to drive the gyro, and the 12-volt electrical system is fed directly into the two small internal voltage regulators that power the servo and the electronic components. The servo is used to drive a small separate control surface or tab that is mounted to the trailing edge of the wing tip outboard of the aileron as indicated in the sketches of Figure 4 and shown in the photo of Figure

5. There are several other possible methods of producing the necessary rolling moment to control the airplane. However, this particular method was selected because it appeared to be uniquely suited to my particular airplane. In the first place, a rather high level of friction in the aileron system ruled out the possibility of coupling the autopilot directly into the ailerons. Friction is a deadly enemy of most automatic control systems and must be avoided. Secondly, since I was working with a fairly low-powered servo, it was necessary to minimize the forces required to move the control surface. Thirdly, I wanted to limit the maximum control authority of the system so that it posed no problem if a hard over failure occurred. And fourth, I wanted to make the system easy to install, modify and remove because I did not want to put my airplane out of flying status for too long a period. Of course, I had to be sure not to alter the basic integrity of the airplane, otherwise I would have to go through the whole bit of recertification with the FAA. The tab is attached to the trailing edge of the wing tip by rivets with back-up plates beneath the fiber-glass skin. A partial rib was also riveted to the inside of the tip to provide a mounting surface for the small servo. The tab was made by wrapping a sheet of .020 aluminum around three '4 inch thick ribs cut from a sheet of reinforced Bakelite that was handy. The aluminum was riveted at the trailing edge and the ribs were held in place by small screws. The two end ribs were drilled with 3/32 inch holes at the quarter chord station. A short brass machine screw with the end ground down to 3/32 in diameter for a length of about 14 inch was attached to each support bracket to serve as pivot for the tab. The linkage between the servo was standard RC model hardware which can be purchased at most hobby shops. SPORT AVIATION 25

J

fl

'4 ^L-vw

£**

A i

H*

< I d

/

2ZK

A ^.>V^\ • 1 680A

ts >-

ISOK

AA/A^-

—•^WVW-WVWA-* r


TO SERVO A — RS-324 OP AMP (276-1711)

P — PHOTO X'lSTOR (276-130/140)

B — RS-1458 OP AMP (276-038)

SERVO-HEATHKIT-HIGH TORQUE GDA-1205-8 OR EQUAL

C — RS-556 DUAL TIMER (276-1728) D — SEE DIAGRAM UPPER CORNER

Tv T2 — THERMISTOR - FENWAL GB32LI OR SIMILAR (MIL COMM, 1616 COTNER AVE., L.A., CALIF. 90025)

E — RS-7805 VOLT REG. (276-1770) LED — LIGHT EMIT. DIODE (276-042) L — 12V LIGHT BULB (272-1141)

26 MAY 1978

FIGURE 2

The main uncertainty about the tab was just how much area was required. The span was determined by the length of the tip trailing edge which permitted a maximum tab length of 12 inches. I wanted to keep the tab as small as possible. The first try of 2'/4 inch chord

proved to be not quite enough, so I subsequently increased the chord to 3% inches. This appeared to be just about the minimum size required to produce acceptable performance in smooth and light turbulent air. Control authority was found to be about 1TO of the aileron control by measuring the amount of stick lateral displacement required to maintain wing level flight with the tab

fully deflected. I currently plan to install a second servotab of the same size on the other wing to see if two acting together will be adequate in the more turbulent air occasionally encountered especially in the spring and fall. The tab linkage was set up so that ± 45 degrees travel of the servo produced about ± 30 degrees of the tab, however the limiters were adjusted so that servo could be driven to a total of ± 90 degrees. Of course, this over travel produced some additional tab travel but the near dead-center geometry did serve to reduce the torque load on the servo to almost zero whenever large

FIGURE 4

input signals were applied. With this arrangement the

position of the servo was directly proportional to the input signal over the full range but the position of the tab was linear for only the middle half of the range as illustrated in the sketch of Figure 6. The rate and heading sensors are extremely sensitive devices and yet are quite easy to work with and adjust. The rate sensor was checked out on a rate table and was found to produce 45 degrees travel of the servo when a rate of % degree per second was applied to the sensor. The heading sensor appears to produce the same

travel for something less than a 1 degree change in heading; however, I have not actually measured the sensitivity. The actual values of these sensitivities are not too important as long as they are high enough.

The sensitivities are changed by the gain pots during

(Photo by Mark Hewes)

FIGURE 5a — View of roll control tab installed on Don Hewes 1 BD-4, N632DH.

flight testing to provide optimum performance. To do this, the gain pot which combines the two different

signals is adjusted to give maximum rate output and the rate gain pot is adjusted until the system forces the airplane to oscillate at the same time. Once this is done, the rate-heading gain pot is turned to increase the heading gain very slowly. To see the effect of this, the compass card is displaced a few degrees and the response of the airplane is observed. It should bank immediately in the direction of the heading change and then stabilize about the new heading. As the gain is increased beyond the desired setting, the airplane will tend to oscillate about the new heading and finally diverge. The setting is then backed off slightly to get the optimum setting. A photograph showing the installation of my prototype unit is shown in Figure 7. The total weight of the system including the control tab is under 2 pounds. As indicated previously, I have several refinements in mind for the system and plan to be testing these in the next few months. A final report on the system with construction notes will be submitted to SPORT AVIATION following these tests. THERMISTOR

THERMISTOR (SAME TEMP.) NO ROTATION FIGURE 3

+v

(HEATED)

(COOLED) -V WITH ROTATION

(Photo by Mark Hewes)

FIGURE 5b — Close up roll control tab installed on wing tip.

Reference 1 — The Saga of the Plastic Autopilot, SPORT AVIATION, March 1974. (Editor's Note - Sorry, this issue is no longer available.) SPORT AVIATION 27

-1001-1 I I > I i I I I I I ' I

3.b

3.8

4.0

PWM INPUT VOLT FIGURE 6

(Photo by Mark Hewes)

FIGURE 7a — Heading sensor, control unit and rate sensoi before installation.

(Photo by Mark Hewes)

FIGURE 7b — Heading sensor and control unit installed on instrument panel.

Ray Brown (EAA 31004), Rt. 1, Sky Harbor Airpark, Webster, MN 55088 completed this 125 h.p. Lycoming powered Daphne SD-1A September 1, 1977, after 6 years 8 months construction time.

Richard Bourque (EAA 95635), 8 Grimes Brook Road, Simsbury, CT 06070 restored this metalized Stinson 108-1 and hopes to have it at Oshkosh '78. Since Univalr still provides the parts, the 108 series may go on forever.

(Photo by Dick Stouffer) A pair of T-28s rumble by our cameraman — N3313G, left, owned by George Enhorning (EAA 90409) of Wolcott, CT and N28100 owned by Dick Dieter (EAA 56531) of South Bend, IN.