The attitude indicator is very important for any sort of instrument flying, so it must be accurate at all times regardless of the plane's movements.

I presume that the indicator would be connnected to a gyroscope. However, gyroscopes drift, and a drifting attitude indicator would be pretty useless. Of course, you might have a reset button, but safely using it would require a visual reference to a horizon, which obviously you don't have during IFR flight.

Another idea might be to periodically reset the indicator based on gravity --- many model aircraft use this something like this approach to drive an attitude calculation; however, during coordinated turns, the gravity vector is indistinguishable from the vector during level flight. If a plane continually turns in a coordinated way, such an attitude indicator would be confused in exactly the same way that a human would be, resulting in bad things like the autopilot doing a graveyard spiral. A gravity sensor + accelerometer wouldn't be able to detect the difference between a gradual turn into a steeper and steeper coordinated spiral and simply a gradual gyroscope drift while remaining in level flight.

How are attitude indicators kept from drifting?

  • 1
    $\begingroup$ gyro drifts are corrected by external sensors, such as GPS. Could you rephrase your question to focus on one of the following: 1) how are (mechanical) gyros of planes reliable after long flights? long years of use? 2) how can mems gyros / IMUs be kept reliable? $\endgroup$ – Gürkan Çetin Jun 28 '15 at 21:14
  • 2
    $\begingroup$ How can GPS correct the gyro of the attitude indicator? I'm not talking about INSs here, and I think this is very obvious. $\endgroup$ – ithisa Jun 28 '15 at 21:14
  • $\begingroup$ Do you want to know about gyroscopic AI, or EFIS AI? Then do you want to know about preventing errors, or correcting (resetting) errors in flight? $\endgroup$ – mins Jun 28 '15 at 22:59
  • $\begingroup$ If I am not mistaken, an EFIS still uses a gyroscope, albeit connected to a computer instead of directly to a ball. My question is more a theoretical one: how can an AI be accurate if no sort of inertial measurement seems to be able to distinguish between a long coordinated turn and level flight? And in IFR flight, there are no outside cues that can be used to reset errors in flight (imagine 8 hour international IFR flight at night) $\endgroup$ – ithisa Jun 28 '15 at 23:41
  • $\begingroup$ The answers are all in this question: If I fly a balanced turn long enough, will the AI show I'm level? $\endgroup$ – NathanG Jun 30 '15 at 0:13

A gyroscopic AI has an erection mechanism, which continuously corrects the AI to be upright based on the local level, or downward acceleration vector. The correction rate is generally 3-5 degrees per minute.

The way that the AI corrects itself is a system of pendulous vanes. When the gyro is not upright relative to the local level, centrifugal force pushes open vanes on the gyro's case. Air escapes through the uncovered holes, applying a force to the case, and correcting for precession. Gyroscope emphasizing pendulous vanes

(FAA AC 65-15A)

The second half is somewhat counter-intuitive, but flying in a coordinated turn will not continuously increase the error in the AI. For the first half of the turn, the AI will add error; in the second half, it removes the error. After a 360 degree turn, the precession error will be removed completely. For more detail, see another answer: If I fly a balanced turn long enough, will the AI show I'm level?

  • $\begingroup$ I'm not confused over the correction of precession error. However, let's say somebody stays in a coordinated turn for an hour. Then the erection mechanism (which I do know about) would be confused, right? It would "erect" to a slanted position because in coordinated turns, the acceleration vector always points down? i.e. the gyro is always "upright relative to the local level". $\endgroup$ – ithisa Jun 30 '15 at 19:21
  • 3
    $\begingroup$ That's the counter-intuitive part: the erection mechanism does get "confused" by the different down-vector, but averages out across the entire circle. Think of it this way: in a 360 degree circle, the "down" vector sweeps out an entire arc - so on average, "down acceleration" is in fact straight down. (Not a perfect metaphor, but should help your mental picture.) $\endgroup$ – NathanG Jun 30 '15 at 19:24

I've had this same question and even asked my flight instructor. He said the attitude indicator does not drift because it is on a 2-axis gimbal and rotates freely and will not precess or become incorrect unless it tumbles (reaches the physical limits of its movement), such as happens when doing aerobatics. A directional gyro, however, will drift due to the fact that it is only on a 1-axis gimbal and is deflected against when the aircraft accelerates, turns, or maneuvers. I'm still not sure how it is not subject to precession as it is a mechanical system and thus subject to friction and forces, but that's what the CFI said. He looked puzzled by my question (the same one you are asking) and answered as if the attitude indicator is never wrong.

  • 1
    $\begingroup$ Welcome to the Aviation Stack Exchange! Unfortunately, your CFI is confused - the AI definitely does need correction for precession. The description you're giving is a decent comparison between a DG and a wet compass, but even a perfectly frictionless gyro needs to be corrected for apparent precession. $\endgroup$ – NathanG Jun 30 '15 at 0:15
  • $\begingroup$ Do you have a technical description of how this works/how to deal with this in flight? How long does it take in normal flight for such errors to manifest themselves? $\endgroup$ – Pugz Jun 30 '15 at 0:29
  • 1
    $\begingroup$ In straight & level flight, the errors don't manifest themselves because the AI is constantly correcting itself. See my answer (and the answer linked to it). $\endgroup$ – NathanG Jun 30 '15 at 0:31

The simplest answer is that the error introduced by gyroscopic precession is not considered significant during normal aircraft operation (maintaining relatively normal flight attitudes, i.e. within 30* of level flight). There is a mechanism designed to automatically re-orient the gyroscope during flight, called a "caging mechanism", which normally works very well but during sustained maneuvers it can actually exacerbate precession error. The caging mechanism can usually be activated manually in VMC (or at least when the horizon is visible) with a "push to cage" button to ensure the AI matches the real horizon.

Theoretically, the error induced by precession over time could add up to be significant. Violent maneuvers increase the effects of friction and precession in the gimbal system and can cause it to lock or tumble. These situations are unlikely to happen undetected by the pilot, for several reasons:

  • Pilots flying by instruments don't use just one of them; there is some overlap in what the instruments should be telling you. For instance, at an indicated "cruising" airspeed where AOA is low, if the AH indicates level flight but the VVI or altimeter say you're sinking, or the TC indicates bank or any of your course indicators (DG, CDI/OBI etc) indicate a change in heading not mirrored by the slip indicator, that casts doubt on the accuracy of the horizon.

  • During IFR flight, the pilot is required to coordinate with ATC. In most cases, ATC can detect on radar that a plane is deviating from course, and will advise the pilot of this and coordinate with him to resolve the problem, including diagnosing any instrument failures that may have led to the deviation in the first place.

  • Pilots, no matter how experienced with instruments, don't like flying in IMC. It's just no fun; it requires a high level of concentration and a basic departure from the "trust your senses" mentality that has been the survival mantra for the human race for millenia. The potential for pilot error is high. Pilots will therefore seek to minimize their time in IMC by working with ATC to climb above or navigate around regions of IMC.

  • Modern instruments incorporate solid-state accelerometers and magnetometers as well as gyroscopes as the basis for the inertial navigation system, especially for the AHRS subsystem of glass cockpits. Gyroscopes may still be present in the system but they can be much more easily detected to be out of calibration and corrected by the computer. The horizon indicator included in modern "primary flight displays", therefore, is pretty trustworthy.


Your Answer

By clicking "Post Your Answer", you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.