Get extra lift from AOPA. Start your free membership trial today! Click here

Autopilot Basics

Installed in few trainers, these are common in more complex aircraft

By Marc E. Cook

Wiley Post dubbed his 75-pound Sperry autopilot "Mechanical Mike." Though crude and poor-flying by today's standards, the contraption surely helped to make his second round-the-world flight a success. Imagine yourself solo at the controls of a Lockheed Vega for a 13,000-mile, 11-stop journey; that was Post's mission in 1933. And were it not for the assistance of his pneumatic autopilot, it's certain Post's fatigue would have been monumental.

Call the flight-control box what you like—George, Mike, Otto—it's become an indispensable part of long-distance air travel. (We'll let the fly-it-by-hand purists sneer and spew forth vaguely disparaging names. Hand-flying is well and good, but most pilots responding honestly are likely to say that a set of helping hands on long flights or in high-workload situations is worth the expense.)

Compared to Post's compatriot aboard the Winnie Mae, today's autopilots fly better, last longer, and cost relatively little. Electronic flyers are so prevalent that few high-performance singles go without them; in part, aggressive marketing from Century and S-Tec has resulted in a great quantity of modern retrofit boxes, even in elderly aircraft. Bendix/King focuses on remaining the dominant original-equipment supplier. The capabilities and relative low cost of these modern autopilots border on the fantastic.

At the heart of matters, autopilots attempt to mimic the actions of the pilot. In doing so, electromechanical systems use a variety of electrical and mechanical means to operate the flight controls. Pneumatic autopilots were also popular in the 1950s and 1960s, but they have largely been replaced in the field with newer, more reliable electromechanical models. The number of surfaces worked by the autopilot determines its so-called number of axes. A single-axis autopilot manages just one set of controls, usually the ailerons; this provides wing leveling through roll control. Adding autopilot control of the elevator and pitch-trim system makes the box a two-axis system. Finally, when a yaw damper is installed, the setup can be called three axis. It's common, by the way, to find advertisements that state a three-axis autopilot is installed aboard an airplane when, in fact, there's just a roll and pitch autobox on board. This misconception might well come from the fact that while the autopilot directly manipulates only two axes, for instance, it affects all three.

Each make of autopilot works slightly differently, but there are similarities. All need some way to determine the airplane's attitude. The simpler autopilots many of Cessna's original equipment boxes as well as the S-Tec line use a special turn coordinator that sends electric messages to the flight computer. Small brushes inside the gyro signal the yaw/roll attitude of the airplane. These are known as rate-based systems, since the turn coordinator is a rate-of-turn device. In fact, modern turn coordinators, which sense both rolling and yawing moments, were developed to better drive early autopilots; the previous turn and bank instruments were not sensitive enough for the task.

Let's say that you've activated your single-axis autopilot to keep your wings level. Even in the smoothest air, eventually a wing will dip. The sensors in the turn coordinator remind the flight computer that the airplane is turning. The computer, in turn, sends a signal to the roll servo a small electric motor fitted with a slip clutch that, through a bridle cable, grips one of the aileron cables. As the roll servo gently applies aileron against the turn, the flight computer monitors the progress, eventually removing the command when the turn coordinator signals that the wings are once again level. This loop works continuously, many times a second; sometimes the control inputs are too small to notice at the yoke. It's important to understand also that autopilots merely react to conditions; there is no way the devices can anticipate the movements of the airplane.

In addition to the computer's command of the servo, there is an additional circuit that informs the brain of the servo's actions. This is necessary to limit control-surface hunting and over-controlling. Imagine flying along in gusty conditions and the right wing drops. The computer commands a roll-left from the servo, but the airplane, thanks to inertia, keeps rolling right. Without a feedback loop, the computer might just command more and more left-turn aileron until the control hits the stops. At that point the airplane is likely to snap back to level flight with enough vigor to send Junior's Mighty Morphin figures flying through the cabin. Modern autopilot systems incorporate this servo-position feedback loop to limit control deflection under certain conditions.

In this basic mode, the autopilot cannot determine aircraft heading. A so-called heading bug mounted inside the heading indicator (or directional gyro) or horizontal situation indicator (HSI) is used to command the computer to maintain a given heading. Typically installed with a heading-hold system is some means of channeling navigation information to the same circuits that execute the heading-hold function. In the days of VOR-only navigation, few pilots invoked this nav-tracking function; the wiggly signals from the VOR often resulted in the nose of the airplane sniffing back and forth like a malamute in search of chow. With the advent of loran and GPS, however, the nav-tracking function has become eminently more useful. In this mode, the autopilot uses information from a course deviation indicator or HSI to determine a reference magnetic heading. Then the computer senses the deflection of the left/right needle and commands turns both to intercept and to maintain a course that keeps the needle centered.

Considerable programming talent goes into ensuring that the autopilot follows navigation signals smoothly. Most systems employ some form of predictive computation and alter the control scheme and pitch/bank limits, depending upon rate of closure on the desired course or the distance from the approach aid in use.

If you take the single-axis, roll-only autopilot and add control of the elevator, you'll have a two-axis system that can maintain a given attitude or altitude. Again, depending upon manufacturer, various instruments may be used to supply attitude information. For instance, Century, Cessna, and Bendix/King all use a special attitude gyro with pickoffs to provide both roll and pitch information. In this type of system, the turn coordinator is not used. S-Tec, on the other hand, employs a small accelerometer and a sensitive barometric sensor to provide pitch reference.

When you engage the attitude-hold feature of a two-axis autopilot, the computer will attempt to maintain the pitch attitude at the time of engagement. Attitude-indicator-driven autopilots do this by operating the elevator servo to hold the attitude signaled by the indicator. Real simple: Engage at five degrees nose-up and the autopilot will work vigilantly to hold that attitude. Using the attitude-hold feature in these autopilots allows the airplane to climb or descend at that reference pitch attitude; the pilot must manage power to adjust descent rate and airspeed. All of the current autopilots ensure that the roll portion is activated and functioning before any pitch authority is issued. In most cases, it's not possible to operate a two-axis autopilot in roll-only mode. The exceptions are the earlier Century models and the S-Tec System 50; this is a handy feature, in that if there's something wrong with the pitch side, you don't lose the use of the entire system.

In altitude-hold mode, the autopilot simply tries to maintain the set pressure altitude present at time of engagement. Most systems don't have any way of sensing barometric pressure changes, and so you may have to reset the altitude manually when moving through rapidly changing pressures. Because the autopilot will try to maintain altitude in the mode, it's good form to disengage the function in moderate or severe turbulence. You're better off allowing the altitude to drift to save the airplane from excessive speed and pitch variations. In the bumps, use the attitude mode only; your airplane will thank you, even if ATC doesn't. (Pilots flying IFR remember to ask for a block altitude if the ride is vigorous enough to spawn big altitude deviations.)

Some autopilot installations employ a subsystem that allows for climbs or descents at a preset rate and level-off at a predetermined altitude. So-called preselect setups can be a bit confusing at first, but the ability to crank in the final altitude and let the autopilot do the work can be a real asset. Of course, the pilot must be aware of airspeed limitations in rate-of-climb or rate-of-descent modes.

Still other autopilot setups use what's called a flight director (FD). A set of movable triangles mounted on the face of the attitude indicator serves two purposes. First, in the FD mode, the command bars direct the pilot to make the appropriate control movements for the desired mode, be it heading-hold or glideslope-intercept. Second, when the autopilot is engaged and actually flying the airplane, the flight director acts as a window into George's mind. Supposedly, the servos react to the motions of the FD bars; if, for example, the indicators show that the airplane should be rolling right but it isn't, chances are you've got a servo problem. If, by comparison, the airplane is following the FD's cues but it's not what you want it to do, then there's probably something amiss in the autopilot's central brain.

To assist the pitch servo, most autopilot systems come with an electric means of trimming longitudinally. Some systems can be installed without electric pitch trim, and they use annunciators to warn the pilot that manual trimming is needed to relieve the strain on the main pitch servo. In other setups, the pitch-trim servo acts in concert with the pitch servo to keep the airplane in trim and to ease the work load of the pitch servo.

To certify an autopilot, the manufacturer must prove that a runaway condition will not exceed certain flight loads. Typically, it's 1 G either side of normal flight, or 0 to +2 G. The big concern with a trim runaway is that it could leave the airplane in a marginally controllable state, well on its way to departing the normal flight envelope, even after the autopilot's been disengaged. During certification, a runaway trim condition must be initiated during test flying and the pilot must wait three seconds after noticing the runaway before initiating a response. In those three seconds, the airplane must not exceed those G limits nor its normal operating airspeeds. If you've ever wondered why electric pitch trim in some autopilot-equipped airplanes runs so slowly, you now have your answer.

As important as understanding the hardware, a savvy autopilot manipulator needs to have a firm grasp of Otto's operating schemes and limitations. The best place to find this is in the airplane's flight manual, which will have an operating handbook for the installed autopilot. Read this document carefully. Know how to engage each feature of the autopilot and, more important, be sure to understand how to turn George off. Typically there's a big red autopilot disengage/trim interrupt button on the yoke, as well as a power switch on the autopilot control head. In some cases, operating the electric trim while the autopilot is engaged will also take it off line. Finally, you can always resort to pulling the autopilot circuit breaker and in an instant you should be able to find this one in a forest of breakers or shutting off the avionics master switch. If all else fails, you can just kill the aircraft or avionics master.

Why is it so important to know how to put George to sleep? A number of accidents have occurred because pilots didn't know how to disarm the autopilot. Moreover, a common thread running through these incidents is the pilots' attempting to override the autopilot without first disengaging it. One of the more common autopilot faux pas is trying to force the airplane to an altitude or attitude while the hold functions are engaged. Consider this scenario. The airplane has settled into cruise about 100 feet above the desired mark, in altitude-hold mode. You attempt to get it back to altitude by pushing on the yoke. The autopilot, sensing a dive from the selected altitude, cranks in more nose-up elevator. Soon the yoke gets quite heavy in hand, as the pitch servo and the pitch-trim servo both spool against the pilot's efforts. If the pilot suddenly comes to his senses and lets go of the yoke, the airplane will attempt to climb quite rapidly to the selected altitude. The rule is: Never fight the autopilot. If it's not doing what you want, disengage it and set the airplane up by hand.

What are the usual autopilot trouble spots? We asked Tom Rogers of Avionics West in Santa Maria, California, about the common problems. "Depends on the system," says Rogers. "We see the same kinds of problems out of the same models. For example, we often see an S-Tec system develop a slight wing rock. The solution is to run the servo on the bench at a higher current than it normally sees in service." Forcing the servo to absorb the greater current cleans the motor brushes. "Most of the Bendix/King [problems] come down to some kind of servo failure," Rogers continues, "and in the Century installations we see lots of problems with bad connections." Seems the female connectors used throughout the Century autopilots, particularly the III and IV, are prone to spreading, making electrical contact intermittent. Finally, Rogers says, "In the Cessna/ARC systems, it is usually that a component has been replaced but the system hasn't been aligned on the bench. Then the autopilot will fly poorly, particularly in pitch."

Carefully consider regular autopilot maintenance. A misaligned, sloppy box will fly the airplane badly and can lead to some interesting unusual attitudes in turbulence. Moreover, remember that a change of gyro instruments will probably require realignment of the computer's settings. An old, hamstrung attitude indicator, for example, will respond much more sluggishly than a new one, so the autopilot's reactions might become a tad sporty if you don't compensate with realignment. Instrument manufacturers also advise that frequent replacement of pneumatic-system filters is the way to prolong gyro life.

Even in the worst shape, though, today's autopilots are a remarkable step up from those at Wiley Post's disposal. And while many of us might rather be flying Lockheed Vegas, few would admit to much in the way of autopilot nostalgia.