Custom Autopilot Constants

X-Plane's default autopilot setup does a pretty decent job with "average" aircraft. However, if the control response of your plane is unusual, or just to get the autopilot really right, you'll have to set up custom autopilot constants.

Basically, the autopilot applies control inputs when the plane is not flying the way is should, given the current autopilot settings. For example, let's say you're set the autopilot for a 500 fpm climb. If the plane is currently flying level, the autopilot will apply up elevator to pitch the plane up and make it climb. The autopilot constants account for two critical aspects of the plane's behavior:

1. How much response you get for a certain amount of control input.
   
2. How long it takes for the plane to respond to control input.

If the settings are too weak, you have a "lazy" aircraft that doesn't fly to the autopilot settings. On the other hand, if the settings are too strong, the plane will overshoot its corrections and might even oscillate indefinitely.

There are two basic approaches: You can adjust the autopilot constrants in PlaneMaker, then save the plane, open it in X-Plane, and test fly it. Or you can adjust them in real time while you fly the plane in X-Plane. Guess which is more convenient. However, settings you change in X-Plane are not recorded in the aircraft file! You still need to go back to PlaneMaker when you're done to make your changes permanent.

To adjust them in X-Plane, from the menu bar select Special->Set Autopilot Constants.

To adjust them in PlaneMaker, open the plane in PlaneMaker and from the menu bar select Expert->Artificial Stability; then select the Autopilot tab. If the menu is blank, check the "Use custom autopilot constants" checkbox.

Either way you'll see the same menu with settings. (In X-Plane, you can drag the menu to the bottom of the screen to get it out of the way while you test-fly your current settings.) You'll see three major sections, for throttle control, roll, and pitch.

Speed error for full throttle: This is your sensitivity control. The autopilot will apply full throttle if the plane is this much slower than the set speed. More generally, decrease this value to increase the sensitivity, i.e., to make a larger throttle correction in response to an error in speed.

Throttle control per second: This value controls how fast the autopilot changes the throttle setting. Keeping this value small damps out the throttle jiggling in response to turbulence, but slows down the throttle response.

Speed prediction: This value tells the autopilot how much delay there is in the aircraft's response to a change in throttle. Increase this value if you find the aircraft overshooting speed corrections; i.e., it's too slow, the autopilot opens the throttle, the plane gets too fast, the autopilot backs off on the throttle, the plane gets too slow, etc. Decrease this value if you see the throttle open, and then back off before the plane is up to speed.

Roll error for full aileron: Just like with the throttle, this is your sensitivity control. The autopilot computes what your roll angle should be, based on your current heading and what it ought to be, and then applies the necessary aileron input to get the aircraft to the desired roll angle. As with the throttle, decrease this value to get a stronger aileron input.

Roll prediction: This value tells the autopilot how fast the plane rolls. As with the speed prediction setting, increase this value to keep roll angle changes from overshooting and decrease it if the autopilot's roll behavior feels "lazy".

Roll tune time: The autopilot won't get the roll behavior exactly right all the time, since control response depends on the plane's speed and other factors. After this time period has passed, the autopilot fine-tunes the roll angle by applying small aileron corrections.

Localizer CDI gain: This value and the next control the autopilot's response to a runway localizer. The CDI gain value tells the autopilot how fast to turn the plane in response to being off the localizer centerline. Localizers are the one LNAV mode for which the autopilot does not know the correct course heading, so it needs to "feel for the localizer" by watching changes in the CDI while it turns toward the deviation indicator. For heading select, VORs, and GPS routes, the autopilot knows the on-course heading. VORs and GPS legs are intercepted at a 30 degree angle.

Localizer CDI prediction: This value tells the autopilot how fast the plane can turn. It is used for all LNAV modes to turn the plane smoothly onto the on-course heading as it approaches the course centerline. As with the other prediction values, too small a value will cause the plane to turn too late and overshoot the course centerline; too large a value causes the plane to turn to the course heading before it is on course, and so never get quite on course.

Pitch error for full elevator: As with speed error and roll error, this value controls the sensitivity of the autopilot's elevator control. Decrease it to get a stronger elevator response for pitch control.

Pitch prediction: This value tells the autopilot how long the aircraft takes to change its pitch attitude. As with the other prediction values, increase it to tune out overshoot and oscillation; decrease it to make the aircraft hold pitch more closely.

Pitch tune time: This is the same idea as roll tune time. After this time period, the autopilot makes small elevator corrections to get the pitch angle exactly right.

Glideslope CDI gain: This is the same idea as the localizer CDI gain. Increase this value to get a stronger elevator response to the plane being off glideslope.

Glideslope CDI prediction: This value predicts how long the plane takes to change its descent rate. Increase it to eliminate overshooting and oscillating on the glideslope; decrease it to hold glideslope more closely.

Pitch degrees per knot: This value controls the rate of change of pitch for the speed with pitch and flight level change modes. In these modes, the autopilot pitches the aircraft upwards if it is flying too fast, and downwards if it is flying too slowly.

The settings are complicated and they interact. For example, if you increase control sensitivity (i.e., reduce the roll or pitch error values) you will probably have to compensate by slightly increasing the prediction time to account for the aircraft's more aggressive response.

The best way to get everything set is to start with the defaults. Then inspect the settings for "sanity" - i.e., set the prediction times according to how fast the aircraft responds to normal control inputs. Set the sensitivity values according to how you would control the aircraft - e.g., "I would apply full up elevator if the aircraft is pitched too far down by X degrees".

Then set up some test scenarios for each of the modes, test-fly the plane, and make adjustments until it responds correctly.

The pitch and roll control values need to be set up in stages. First set the sensitivity (error value) so that you get reasonable response - for example, pitch response to vertical speed changes in VS mode or roll response to heading changes in HDG mode. ("Reasonable" depends on the aircraft. A 747 rolls a lot slower than a Piper Cub.) If the aircraft overshoots or undershoots pitch and roll attitude, increase or decrease the prediction time, respectively. If you can't get rid of oscillation, reduce the sensitivity by increasing the error value.

Set the pitch and roll tune time to be 3x or more larger than the prediction time.

Now you can go on to the secondary settings (CDI gain and prediction). Here you can select the prediction value first by using known targets. For pitch, use ALT HOLD and set the prediction so the aircraft settles in neatly to the set altitude. Similarly, for roll, use HDG select and adjust the prediction time so that the aircraft rolls out properly to the selected heading. Finally you can adjust the CDI gain by intercepting  a localizer and glideslope. Localizer intercept is one of the hardest to get right. Set up a realistic worst case situation (e.g., for a typical airliner, a 30 degree localizer intercept angle, 10 miles out, at 250 kt) and make it work. You will probably have to go back and tweak settings you made previously. Saving and reloading a situation makes it convenient to rerun the same test with different settings.

If you've made your autopilot adjustments "on the fly" in X-Plane, they have not been permanently recorded! Even reloading the plane (or any other plane) will erase them. When you're done, write them down, open the plane in PlaneMaker and use the autopilot settings screen to make your changes permanent.