Getting your Trinity Audio player ready...
|
From almost the moment they were invented, autopilots have carried a reputation for being untrustworthy or even dangerous. New pilots learned to fear these mysterious boxes and most flight instructors simply ignored them. A New York Times article about a “robot-piloted plane” making a test flight across the Atlantic in 1947 is emblematic of the suspicious, pilot vs. computer mindset that was instilled in pilots for over five decades.
But while no technology can replace a well-trained pilot in the left seat of a general aviation airplane (at least not yet), the right technology can definitely make that well-trained pilot safer. That has proven to be true for terrain awareness and warning systems, which have dramatically reduced controlled flight into terrain accidents. Likewise for affordable datalink weather systems, which appear to have put a sizable dent in weather-related accidents ten years after going mainstream. Heck, even the nose wheel has cut down on crashes, in spite of many pilots’ best attempts to ground loop a Skyhawk.
It’s time to add the much-maligned autopilot to this list of life-saving technologies—and retire some old myths in the process. The latest generation of digital autopilots are nothing like the “mechanical brain” of 1947 or even the analog systems of the 1970s and 80s. They are the first models to truly live up to the grand title they are often given: “automatic flight control system.”
Not like the old ones
Many pilots don’t fully appreciate just how much autopilots have changed in the last 20 years. As in so many other areas, Garmin kicked off the revolution, launching the GFC 700 in the mid-2000s as an add-on to their G1000 glass cockpit. It quickly won plaudits for its smooth handling and long feature list, but it was limited to new airplanes. In 2017 the rest of aviation got access to this technology with the introduction of the GFC 600 and GFC 500, retrofit versions of the GFC 700 that have become wildly popular. Avidyne also offers the very capable DFC90 autopilot as an easy upgrade, and has carved out a particularly strong position in the Cirrus world.
While these include many of the same buttons you’d find on an old Century III or Cessna “Navomatic” autopilot (HDG, ALT, etc.), just about everything else is new. Consider:
- They are attitude-driven, so they’re rock solid compared to rate-based autopilots like the S-TEC 30 and 40 that depended on a turn coordinator.
- Instead of using a vacuum-driven attitude indicator, they typically run off an extremely reliable digital AHRS.
- All modern autopilots are two-axis (heading/track and altitude hold), and many have a third axis too (yaw damper)—an option once reserved for turbine airplanes.
- Built-in over/underspeed protection means the autopilot won’t fly the airplane into a stall if you leave it set on VS mode during a climb (older models can and do).
- They include some version of a Level button, which automatically returns the airplane to straight and level flight from almost any attitude.
- In most cases even the servos are upgraded, with brushless DC motors and no slip clutches, so it’s more than just software that’s new.
- Garmin models have Electronic Stability and Protection (ESP), which provides tactile feedback to the pilot even when the autopilot is off.
- Some include an automatic descent feature for pressurized airplanes, which will sense a pressurization failure and descend out of the flight levels without pilot input.
- Smart glide is an increasingly common feature, which provides detailed guidance in an engine-out scenario, including real-time glide range calculations.
Maybe most important of all, these new models are less expensive than older, less capable autopilots. While perhaps not “cheap,” the incredibly capable GFC 500 costs around $8,000 (for the hardware). That makes it affordable to install in 45-year old 172s, not just brand new turboprops. We see it all the time in our avionics shop at Sporty’s: a Garmin G5 flight instrument and GFC 500 can offer an older airplane a new lease on life without breaking the bank.
Three main threats
So much for the features—do they matter? A look at the GA accident record suggests they do. Far from being just sales sheet fluff, such powerful autopilot capabilities directly address three of the most common accident causes: loss of control in flight, VFR flight into IMC, and powerplant failure.
Loss of control is an unhelpfully vague phrase (as we have discussed), but many of these accidents are the result of the pilot losing situational awareness or becoming task saturated. Both of these can be prevented during cruise flight if a good autopilot has the controls while the pilot briefs an approach or talks on the radio—especially if the autopilot knows not to stall or overspeed. Closer to the ground, where many loss of control accidents happen, an ESP system is an invaluable aid that will push back if you overbank in the pattern or pitch up too much on takeoff. Critically, you don’t have to remember to turn it on.
Likewise, VFR-into-IMC is a persistent problem that will not be solved exclusively with technology, but it can make a difference. The quick adoption of subscription-free ADS-B weather proves that and has helped with prevention, but sometimes pilots still mess up—after all, forecasts aren’t perfect and there are large gaps between METARs. If prevention fails and a pilot does stumble into clouds without an instrument rating, an autopilot is a more reliable way to escape than the old fashioned 180-degree turn, which requires cool focus on the part of a panicked pilot. The level button is a great start, a sort of panic button that can prevent a rapid loss of control. From there, the pilot can turn around or “climb and confess.”
The third accident cause, engine failure, might not seem related to autopilots but it is for anyone flying with Smart Glide capability. Time is your enemy with a failed engine, because time equals glide range (and thus, options). With a quick push of the Smart Glide button, pilots can answer a lot of questions in seconds: what airports are within glide range, how much altitude will I have when I reach that airport, and what is the weather there? Then, a connected autopilot can fly that glide automatically, maintaining best glide speed to maximize range. Yes, you’ll have to actually land the airplane (unless you own a TBM or a Cirrus Jet), but by then hopefully the shock will have worn off a bit and the autopilot will have you in position on final approach.
No, they aren’t magic
Before anyone accuses me of being an out of control technophile, let me be clear that I view autopilots—even the top-of-the-line GFC 600—as an aid to the pilot in command, and nothing more. Garmin’s recent service bulletin, which required owners to pull the AP circuit breaker until a critical software update was completed, is a vivid reminder that no technology is perfect.
But let’s keep some perspective: there have been at most a handful of runaway trim incidents in these autopilots and no accidents (that I’m aware of), but every year over 100 people die from accidents that would be easily prevented by an autopilot. As pilots, we seem deeply uncomfortable with the former, but resigned to the latter. To be fair, pilots aren’t unique in this regard. American drivers are often skeptical of safety technology like collision warning systems or lane-keeping cruise control, even though the evidence shows they work (one study found that automatic braking cuts accidents by half). And yet once again the same bizarre mindset takes hold: if a robot kills one person, it’s a national outrage; if 40,000 people die every year on roads because of human mistakes, it’s just something we have to live with.
When it comes to autopilots, many pilots seem to choose the worst of both worlds: spending the money on capable avionics, but not understanding how to use them. The accident record bears this out—examples of out-of-control autopilots flying an airplane into the ground are vanishingly rare, but not so for confused or overconfident pilots.
Consider the Citation pilot who, in spite of flying a twin-engine jet single pilot and having very little time in type, preferred to hand fly the airplane. From the NTSB report: “During the training, he struggled significantly in high workload environments and had difficulty operating the airplane’s avionics suite, which had recently been installed. He revealed to a fellow pilot that he preferred to ‘hand fly’ the airplane rather than use the autopilot.” This proved to be a fatal mistake.
Or there’s the non-instrument rated Cirrus pilot who took off at night and quickly encountered marginal weather. He told the controller he was returning to his departure airport—probably because he had entered a cloud—but never thought to engage the S-TEC 55X autopilot on board, and crashed. This seems like a case where the Level button would have made a big difference.
You have to know how to use it
As usual, the binary choice of “better pilot skills or better technology” is wildly simplistic. The only correct answer is both! There are plenty of examples, from the airlines to the Cirrus community, where new technology and updated training have combined to significantly increase safety. The new technology part has been decided—pilots are voting with their wallets and modern autopilots are being installed at a rapid pace. It’s the second step, updated training, that requires immediate attention. We simply must stop ignoring the autopilot during training, whether it’s initial, transition, or recurrent.
It starts with primary training, and sadly the word “autopilot” appears exactly once in the entire Airman Certification Standards for the Private Pilot. Even this mention is buried in an appendix: “the applicant is expected to demonstrate automation management skills by utilizing installed, available, or airborne equipment such as autopilot, avionics and systems displays, and/or a flight management system (FMS).” My anecdotal evidence suggests that, since this is vaguely worded and not a required task, most instructors and examiners skip this part.
The flight training industry is working at maximum capacity right now, and every hour counts. But this unwillingness to even consider the autopilot sets the tone early on; it suggests that an autopilot is a “nice to have” feature that can be ignored until you move into higher performance airplanes. That’s not just wrong, it’s backwards: a new Private pilot launching on one of his first cross countries is exactly the mission for an autopilot.
The definition of good autopilot training will vary from airplane to airplane, but it starts with having a plan. I approach a new autopilot the same way I would approach a new co-pilot, for that is exactly what a DFC 90 or GFC 500 is: a potential helper, but one that must prove itself. That means a desire to learn as much about my new colleague as possible and a general attitude of skepticism for the first few hours.
AC 120-123, confusingly titled Flightpath Management, is a good place to start for a new autopilot owner. While it’s targeted at airline pilots, it has a number of helpful sections for GA pilots as well, including the key differences between manual flight and autoflight modes. The FAA’s Advanced Avionics Handbook is also worth reading, although it’s now fairly dated. The key lesson to learn from these two publications is how to create an automation strategy, because flying with an autopilot is much more than just pushing a few buttons. It’s really an approach to workload management and aircraft monitoring, one you have to figure out before you ever set foot in an airplane.
From there, studying should move to model-specific resources, because each autopilot has its own unique features and procedures. With digital pilot guides and detailed videos available online for free, there’s no excuse not to do your homework. Fortunately, you haven’t burned any avgas yet.
Only after the book work is complete should the learning move into the cockpit, but once again this should be structured and deliberate. Truly mastering an autopilot requires both time with a knowledgeable flight instructor (type clubs are invaluable here) and a few cross countries in good weather to experiment with different modes. While an exhaustive list of autopilot topics to learn could fill a book, here are a few critical skills:
- Embrace the essential autopilot habit: anytime you push a button on the flight director or autopilot, confirm the proper mode is active. This can be done on the mode controller (the autopilot box in the panel) or on the “leaderboard” across the top of the primary flight display. Either way, make sure that if you want heading and altitude hold, it is actually engaged.
- Also remember that autopilots are a “garbage in, garbage out” cycle, so you must confirm not only the mode, but also the bugs and active flight plan. If you followed the habit above, the autopilot may indeed have altitude hold engaged, but what altitude? I always connect a data source (altitude bug, flight plan active leg) to an autopilot mode (ALT, NAV).
- Understand all the ways to disengage the autopilot. Typical options include the mode controller, a red button on the yoke, and the AP circuit breaker (which should have a red collar on it for easy identification). Practice each of these options and know how to do them without looking.
- Find the minimum altitude for autopilot engagement. Many pilots don’t even know this exists, but it’s there in the manual. Note that it very well might be different for departure and approach.
- Practice flying with an out-of-trim condition. One of the most dangerous failure modes on a modern autopilot is an electric trim problem. If the trim starts to run away, you might find yourself flying a very heavy airplane. Go up with an experienced flight instructor to see what that feels like.
- Practice handling unexpected autopilot disconnects. The accident record shows a number of crashes that resulted from a pilot suddenly being handed an airplane after the autopilot malfunctioned or kicked off in turbulence. Stay current on taking the yoke and maintaining basic aircraft attitude.
- Always have a big picture view of where you are in the automation pyramid: raw data (everything off), flight director on but no autopilot, autopilot on but controlled by heading, autopilot on with NAV/APCH, autopilot on with auto throttle, etc. This is not an all or nothing choice, and sometimes hand flying with a flight director makes sense, so be ready to step down if things get confusing.
A lot of that might sound obvious, but that’s the point. An autopilot is just another part of the airplane, like the fuel system or the GPS. It requires good maintenance, good training, and a watchful eye. With that, a modern autopilot can make flying both safer and more comfortable, a rare combination these days. If you’re looking for the next big thing in aviation safety (at least from a technology perspective), it’s not angle of attack indicators or heads-up displays. It’s already here, in the form of the GFC 500 or DFC90—and it’s time we learned how to use them.
- Christmas reading list: 24 books for pilots to read in 2024 - December 6, 2024
- Weather flying means learning to read clouds - November 11, 2024
- What’s wrong with the teardrop pattern entry - September 25, 2024
I agree with everything you said above about safety and capability, but I think your discussion about cost isn’t quite on point. By the time you add up the costs of the autopilot, a G5/GI275 to drive it, and installation, a $20K bill isn’t out of the question… and that’s for a simple single. It gets dramatically more expensive as your installation becomes more complex. I’m not blaming Garmin alone; it’s not like an STEC 3100 is any less expensive.
The question I’d really like to see you address is this: in 2016, the FAA said that NORSEE was going to solve all of our problems, and it specifically mentioned autopilots (https://www.faa.gov/newsroom/faa-policy-helps-modernize-ga-airplanes-and-helicopters). Where’s my inexpensive NORSEE-certified autopilot?
The F-16 is the only one of four Air Force aircraft in which I was qualified that had an autopilot. It had two modes of operation 1) Altitude Hold and 2) Attitude Hold. I would use the Altitude Hold when flying from Point A to Point B while in the PCA and cruising at .9 Mach as it gave my wingmen (had 4 of them once on a flight from MacDill AFB to NAS Miramar and back) a steady platform on which to ‘fly’ as we cruised at FL 450 and above.
I only used Attitude Hold one time when I was leading a four-ship taking off from MacDill. The weather was pretty low with bottoms at 500 feet and tops above 6000 feet. As we were heading to the range to drop bombs, we had to take 20 second spacing on takeoff (one doesn’t do formation takeoffs with bombs on the airplane as a bomb could possibly come loose on the takeoff roll — a lesson learned MANY years ago, but one that nobody wanted to learn again!). Once airborne, the wingmen would fly a radar trail formation — locking up the aircraft in front of them and maintaining one mile in trail behind them until on top and a visual rejoin could be made. After getting airborne and at the end of the runway, I began a right turn to our departure heading. As I wasn’t yet in the weather, I turned my head to look back at my wingmen as some were rolling or they were airborne. That is when I punched into the weather, so I started my rollout on the departure heading as I also turned my head back to my instruments. I then had the worst case of vertigo I have ever experienced — it literally felt like I had done a backflip! I struggled to read the instruments as my eyeballs were rapidly moving back and forth in their sockets. However, I knew where the autopilot switch was and quickly selected Attitude Hold as I knew I was on heading and at the right pitch angle to maintain the climb. It seemed like an eternity until I popped out on top and that was when my eyeballs and my head instantly ‘erected’ to the horizon. Thank goodness for Autopilot!
Pity that you only comment on the Garmin models and not the STEC 2100 and 3100. Any views on those models?
An excellent article, John. Hopefully it will be widely read and inspire pilots to embrace the modern AP/FD products now available and to become knowledgeable and proficient in their use. These units, as you point out, are components alongside other essential equipment that can greatly enhance safe operation and risk mitigation when properly utilized thereby aiding in the transition from “pilot” to “flight Manager.”
Having an autopilot in ifr is as important as being ifr current. Always, always verify the mode of the autopilot selection. A button push is not verification.
Excellent article. When I started my flight training not so long ago pundits in online forums stated with derision that those who would use autopilots or airframe parachutes shouldn’t be flying an airplane. Having spent 25 years writing software before going into medicine I thought autopilots and parachutes were the holy grail. And I was right.
Thanks for good article.
Some of the autopilots that can be used on S-LSA or experimental such as on the Garmin 305 or 307 combined with the G3X can provide a superior autopilot with considerable cost savings from those excellent autopilots on certified aircraft.
John – Great piece (as always!) One Q: Possible you intended “VS” rather than “IAS” here: ?“ Built-in over/underspeed protection means the autopilot won’t fly the airplane into a stall if you leave it set on IAS mode during a climb (older models can and do).”
You’re absolutely correct – thanks for pointing that out. I’ve fixed it.
Great article. However your original comment in the article about VS was correct. With the older APs if you left in in VS it would keep pitching up to try to maintain the set VS while the airspeed bled off. IAS does not have that issue.
Good article and I am a big fan of APs and their correct use. I have a KFC200 and use an Aspen PFD with the EA100 adapter to drive the AP. I am two weeks past my EA100 crapping out on me in cruise in IMC. Luckily, I regularly hand fly my airplane so I simply continued the trip and hand flew the airplane to my destination landing just above minimums for the RNAV approach that I had to fly.
No big deal and a non event. If you fly long enough this happens. I have had a few AP failures over the years that I have flown and it has never caused an issue.
At this stage in my piloting career I would never start a trip in IFR without a working AP but I always assume that I will have to finish the trip flying the old fashioned way.
I do, however, think that the new glass panel PFDs make hand flying much easier than the old steam gauges, though it did take me awhile to get to that point too.
John, you’ve captured the capability approach to automation perfectly. You won’t be surprised if I bring up Blair and Helms’ differentiation between the capability approach, in which “machines are amplifiers of human will, better enabling them to make something of their world…by exercising dominion through technology, people gain greater command over their environment”, and the cybernetic approach, which “encloses people within closed control loops that regulate systemic variables within set parameters”.
The trick is to recognize the drift from capability to cybernetics, which is what we see when we consider the human as moving from the role of the pilot to a role as a systems manager. If we are using automation to gain greater command over the environment, the role and responsibility of the pilot has not changed. Only in the cybernetic approach do we consider piloting skill and ability as less necessary.
And nothing more tragically reinforces your point regarding handling unexpected autopilot disconnects than the disastrous Challenger accident near Hartford a couple of weeks ago.
Thanks (as always) for your thoughtful comments, Steve. I agree that the fundamental debate seems to be “is this a pilot question” or “is this a technology question?” I lean towards the first answer.
And yes, the Challenger accident is a tragic (and nearly unbelievable) example of losing control not just of the airplane but also the automation systems. I suspect that one will be taught in ground schools for many years to come.
This all sounds like a great topic for a book…
I once encountered an unusual situation in a brand new (120hours) Mooney. Cruising as 8000ft VFR on autopilot when the plane reared up into an impossible climb, much adrenaline and a hasty push of the red button and a return to manual flight and slowing heart rate. A quick panel check showed the AH almost at the bottom of the scale and a vacuum failure. On landing an immediate look at the engine and plumbing finally found the cause. The heat had caused the clear plastic hose from the vacuum pump to collapse and hence no vacuum. It all happened so quickly as the previous scan had shown no problems and there was no drifting up or down by the AP
not sure if there was a mod produced by Mooney, but that plane sure got one.
Hi John,
Excellent article! As autopilots become more ubiquitous in small GA airplanes, we as instructors have to become more familiar with their operational and failure modes, and devise a plan of action for teaching pilots the best practices for using them effectively and safely. The conundrum I face is that out of our club’s seven airplanes, five have autopilots, and those five planes are a mish-mash of Garmin, S-Tec, and King units. Capable boxes all, but each have their quirks, and unique “Button-ology” that has to be ingrained into pilots.
I’ve observed that one one the most difficult aspects of autopilot training is to get the pilot being trained to just let go of the yoke. More than once, I’ve walked the pilot through the autopilot setup, mode selection and activation, only to have their hand stay firmly clamped on the yoke and actually resist the autopilot inputs. There’s a big psychological barrier to be overcome for some people. One fellow in particular was so reluctant to relinquish control that I turned off the autopilot, and assumed control of the plane. I told him to fold his arms across his chest while I flew. Then I had him close his eyes and keep them closed. I hand flew for a minute or so, then engaged the autopilot. We flew along like that for a couple minutes and I asked him if he could tell who was flying the plane. He couldn’t, so I had him open his eyes, but he wasn’t allowed to touch the yoke. After a seconds of near hyperventilation, he calmed down and I allowed him to steer the plane around using the heading bug. He was still a little on edge for a few minutes but eventually settled in and began to trust that “Jorge” wasn’t going to kill him. From that point on, training went smoothly.
I have my pilots watch this video on automation dependancy: https://youtu.be/WITLR_qSPXk
Thanks for a great article!
Cheers, Drew
Great article. My approach to AP use is to balance safety and maintenance of my hand flying skills. I have a personal rule that I always hand fly climb out until 1000 ft AGL and decent after the FAF. I personally never do coupled approaches because this is where I want my skills to remain sharp in IMC. While I know that this goes against Cirrus and great instructors like Gary Reeves, this way I am always keeping up with my hand flying skills which is just as important as automation control. One last rule, the moment the AP does something I don’t expect I disconnect it. This way I don’t bust an ATC clearance or have an unrecoverable trim situation (trim to the stops that cant be hand corrected). Aside from this you can count on me having my AP on so I can pay attention to everything else that is important.
Excellent article. You mention the safety benefits of the new generation of autopilots (in particular the envelope protection of the Garmin systems). Do you know if there are any data to back up the assumption that these systems measurably contribute to reducing risks? For someone contemplating the cost of ditching a perfectly good previous generation autopilot, it would be nice if there was something more than conventional wisdom to justify the expense.
Great question Tom – I am a big fan of using data whenever possible. Unfortunately it’s hard in this case, for two reasons. First, these autopilots are pretty new so there just aren’t that many out there and many of the ones that are flying were installed in the last year or two. We’ll probably have to wait for more data.
The other issue is that we’re mostly tracking things that didn’t happen: a loss of control that was caught before it became a problem, a go around that didn’t get too slow, etc. That’s hard to find data on, because there aren’t any NTSB reports to review. I will say that – so far! – I haven’t seen any incidents in those reports about newer autopilots. But we will definitely have to keep watching.