I read a review of the Chelton AP-3C recently, and it has me wondering
some
things.

It's attitude-based. The article I read seems to work on the presumption
that this is better than rate-based, but it does later say that one
shouldn't be dependent upon a vacuum-based AI for one's autopilot.

The one which is better depends on which kind is made by the person writing
the article. There are good and bad autopilots of both types. Vacuum or
electric AI for attitude based autopilots doesn't really matter. What
matters is that you have enough equipment to cross check so that an
instrument failure, especially one the autopilot depends on, can be detected
by the pilot.

In a spinning mass gyro, failures from bearing wear or other mechanical
related problems are as likely occur in air driven as electrically driven
instruments. But to compare the sources of power for those, an electrical
bus with multiple sources of energy (alternator, battery) is statistically
more reliable than a vacuum pump.


So what do people with attitude-based APs typically do? Electric AI?
Which
electric AIs don't tumble?

I'm also wondering about the ability of the AP-3C's ability to use GPS
steering in both horizontal (GPSS) and vertical (GPSV) planes. If on a
GPS
flight plan, what would occur if the AI failed? If the AP is working to
keep to the flight plan (in both planes), wouldn't the airplane stay
straight and level even if the AP thinks that the plane is in a descending
turn (or whatever it thinks from the failed AI)?


Attitude based autopilots depend on electrical signals from an attitude gyro
which may be the instrument in your panel or in EFIS equipped aircraft in a
box mounted somewhere in the plane. In addition to the pitch and roll
signals, there are validity signals that tell the autopilot whether its ok
to use the pitch and roll signals. If the gyro isn't spinning fast enough
or doesn't have power at all then the validity signal state will cause the
autopilot will stop using that instrument for guidance. Depending on the
model, it may disconnect, sound an alert, or both. It's remotely possible
for the attitude sensor to fail somehow and the validity signals still
indicate its working ok. In this case the autopilot will try to keep flying
the plane to the bad pitch and roll with the obvious result. The pilot will
have to notice this and disconnect the autopilot if necessary.

Normally an autopilot will not keep operating in nav mode if its basic
orientation sensors such as the attitude gyro or turn coordinator are broke.
GPS H and V steering represent a special case of nav mode because then the
FMS or GPS is literally telling the autopilot how much to bank the wings or
pitch the nose, which is different from telling the autopilot how much the
localizer or glideslope needle is deflected. Answers to your questions
about the behavior of GPS H and V steering during attitude sensor failure
are autopilot model dependent, and you should ask Chelton about the AP-3C.
I think you'll find that in most or all cases, the autopilot will cease to
function if the attitude sensor has failed, including roll steering modes.
Let us know what Chelton says.

This information is very general because autopilot interfaces and behaviors
and sensors vary widely. The FAA regulatory guidance for autopilot
certification is very lightweight and is basically oriented at ensuring the
pilot can detect and recover from an autopilot problem (or underlying
autopilot sensor problem) before it can kill him.

If any autopilot experts out there think I've said something erroneous, feel
free to correct me so we can all learn.