Why does my plane's vector drift to the right when I turn on SAS?

[Tricky14]

So this will happen I turn on SAS after flying a super light, long distance glider plane completely straight and horizontal.

 

 

The wings are slightly curved dihedral, which provides some balancing. They're also mounted above the center of mass to take advatage of additional balancing advantage from the keel effect. SAS doesn't care, it starts gently turning the rudder elevon more and more to the right. Also I chose elevons for tail wings to save mass. It shouldn't make it particularly manoeuvrable, but this is a glider so manoeuvrability should not be an issue. Either way, elevons are not a great choice for lift purposes, but I think the wing placement relative to center of mass should compensate for the lack of that. Either way, I don't think it should affect my vector in SAS.

So what is going on here? What could I do to fix it?

 

Craft file: http://s000.tinyupload.com/index.php?file_id=94594297542320929992

Note: The landing gear are a bit poorly placed, but with reaction control it should be easy enough to prevent it from tipping during takeoff.

 

Edited May 4, 2016 by Tricky14

[T-Bouw]

Two things come to my mind:

Do you have a joystick/gamepad connected? If the deadzone is not large enough, then this could be it.

Is the plane build perfectly mirrored? No exeptions.

[Tricky14]

6 minutes ago, T-Bouw said:

Two things come to my mind:

Do you have a joystick/gamepad connected? If the deadzone is not large enough, then this could be it.

Is the plane build perfectly mirrored? No exeptions.

 

No joystick is connected. Also this only seems to happen when I enable SAS *after* having already stabled out the plane. I do use Mouse Aim Flight, but SAS overrides its input, plus I immediatly disable MAF after turning on SAS.

The plane is perfectly mirrored. All parts are connected and positioned with angle snapped. The wings are curved, but are still mirrored part by part.

Edited May 4, 2016 by Tricky14

[T-Bouw]

Perhaps disable MAF before turning on SAS.

I just thought of another thing:
Make sure to turn off Roll and Pitch in the properties window of the rudder.
And for good measure, do the same thing to the other surfaces (but the other way around of course).
In other words, make them dedicated control surfaces.

[Boris-Barboris]

1 hour ago, Tricky14 said:

So what is going on here? What could I do to fix it?

Yawing left doesn't make your plane go left, because there is no wing to push you left and, if I understand picture correctly, lots of flex to force that tiny rudder to push you right too much. Strut that thin tail to cockpit, and if it doesn't help, stop using single control surface as rudder, try deluxe or similar, or place some small wing segment as a fin right on top of CoM.

[Tricky14]

1 minute ago, Boris-Barboris said:

Yawing left doesn't make your plane go left, because there is no wing to push you left and, if I understand picture correctly, lots of flex to force that tiny rudder to push you right too much. Strut that thin tail to cockpit, and if it doesn't help, stop using single control surface as rudder, try deluxe or similar, or place some small wing segment as a fin right on top of CoM.

I agree that the rudder doesn't do much. But I am not trying to yaw, I'm trying to fly straight. I suspect it's the reaction control doing something wrong and SAS is overcompensating.

They are awesome tips for control though, but I'm not there yet. Early carreer mode, struts and deluxe control surfaces aren't available. I already made a functional glider with the engines mounted under the wings, but I wanted to see if I could make it energy effcient enough to fly reliably around the world. I have the first tail wings, but I wanted to save on mass so I thought I'd try an elevon instead.

I'll try putting a wing on the CoM, but I'm very close to the part limit. Still 30, sadly.

[Stoney3K]

Are the torque wheels in your cockpit off? SAS may be fighting the torque wheels with the control surfaces.

[pincushionman]

There are longstanding bugs regarding symmetry and the relative stiffness of parts and joints. Which causes planes to slowly roll in the air and is partially responsible for the craft drifting to one side during the takeoff and landing rolls.

I suspect your plane shows a very slight roll due to this, which is not apparent during manual flight because it is so small and your design self-stabilizes. This can be corrected (some) by increasing the stiffness (using KJR or strutting).

Also, SAS is not very savvy when it comes to planes, and is not smart enough to know that yaw and roll are coupled. Since roll causes turning, the appropriate response to an apparent "yaw-like" deviation is to apply aileron, not rudder. Rudder can even make it worse. Sort of like how pitch deviations might be corrected by throttle adjustment rather than elevator.

TL:DR: Planes are asymmetrically floppy, and this causes roll, which causes turning. Also, planes are complicated. Also, SAS is not a very good autopilot when it comes to planes.

Edited May 4, 2016 by pincushionman

[Tricky14]

2 hours ago, T-Bouw said:

Perhaps disable MAF before turning on SAS.

I just thought of another thing:
Make sure to turn off Roll and Pitch in the properties window of the rudder.
And for good measure, do the same thing to the other surfaces (but the other way around of course).
In other words, make them dedicated control surfaces.

Excellent points, all of them. I'll give that a shot. Thanks.

 

1 hour ago, Stoney3K said:

Are the torque wheels in your cockpit off? SAS may be fighting the torque wheels with the control surfaces.

It's just the stock cockpit. Can you even adjust the torque of controls of that?

26 minutes ago, pincushionman said:

There are longstanding bugs regarding symmetry and the relative stiffness of parts and joints. Which causes planes to slowly roll in the air and is partially responsible for the craft drifting to one side during the takeoff and landing rolls.

I suspect your plane shows a very slight roll due to this, which is not apparent during manual flight because it is so small and your design self-stabilizes. This can be corrected (some) by increasing the stiffness (using KJR or strutting).

Also, SAS is not very savvy when it comes to planes, and is not smart enough to know that yaw and roll are coupled. Since roll causes turning, the appropriate response to an apparent "yaw-like" deviation is to apply aileron, not rudder. Rudder can even make it worse. Sort of like how pitch deviations might be corrected by throttle adjustment rather than elevator.

TL:DR: Planes are asymmetrically floppy, and this causes roll, which causes turning. Also, planes are complicated. Also, SAS is not a very good autopilot when it comes to planes.

I see. I noticed some of that already, but it's good to have it confirmed. The roll/yaw was a new one for me though. First plane I build to do this (I tried maybe a dozen times before)

[Stoney3K]

Just now, Tricky14 said:

Excellent points, all of them. I'll give that a shot. Thanks.

 

It's just the stock cockpit. Can you even adjust the torque of controls of that?

I see. I noticed some of that already, but it's good to have it confirmed. The roll/yaw was a new one for me though. First plane I build to do this (I tried maybe a dozen times before)

Right-click the cockpit and hit "Toggle Torque". There should be a reaction wheels indicator in the context menu that shows the status.

The Mk1 cockpit *does* have reaction wheels IIRC.

[Tricky14]

29 minutes ago, Stoney3K said:

Right-click the cockpit and hit "Toggle Torque". There should be a reaction wheels indicator in the context menu that shows the status.

The Mk1 cockpit *does* have reaction wheels IIRC.

Ooh. Lol, I think I misunderstood Stoney3K. I thought by 'off' he meant 'slightly skewed'. No, the reaction wheels are default enabled.

[Zhetaan]

This may be superfluous, but don't forget to hit Mod-X to be sure you've killed any trim.

[Tricky14]

I figured out where the drift came from. The wings meet at the middle of the plane, but also slightly overlap to take the most advantage of the vehicle width limit of the runway. The rudder or yaw wing is mounted on top of that, not the fuselage below it. So while it looks like it's nice in the center, it's only actually attached to one of the wings. Once I fixed the attachment, the drift was gone.

Thanks for the help everyone! Could not have reached this conclusion without your suggestions.

Edited May 4, 2016 by Tricky14