Pendulum SAS

[Falkenherz]

Not sure if in the correct forum topic:

I have the problem that, very often, SAS won´t keep my plane or rocket stable. This is especially true after strong course corrections, be it in atmosphere or in space. It seems that SAS just tries to cancel out a given course input by oversteering in the opposite direction. Like this, a very strong pendulum movements is induced.

The most annoying situation is when I lift my plane from 10 to 40 degrees ascend angle, already after liftoff @~200m/s. It starts to pendulum up and down, but never ever holds the 40 degrees which I steered it towards. I always have to counter-counter steer with multiple waves of manual input. At greater heights and speed, even then SAS cannot hold the new ascend angle and I have to press "S" continuously in order to keep the heading (thus, risking a stall).

A similar thing happens when you have relatively little torque on a vessel in space. Once you stop your steering input, SAS is supposed to stop the movement and hold position. What it actually seems to do is trying to get back to the position exactly where the steering impluse stopped. Like this, I can never achieve a stable change of course and always have to manually deal with the obligatory resulting, unnecessary, annoying pendulum movement.

This is terribly annoying and makes flight a chore to deal with. Does Squad knows about these annoyances and intend to do something about it? It is just about changing the way SAS "is thinking", I believe.

Edited May 29, 2015 by Falkenherz

[Starwhip]

SAS sucks. I'll say that right now. It overcompensates a third of the time, undercompensates half the time, and in the remaining fraction works just fine.

I usually use SAS as a dampener, and press "F" a lot to temporarily enable/disable it.

Hopefully it gets fixed soon.

[Ferdoni]

SAS worked 100% of the time fine in 0.90 right?.. Now you say it, I have the same problem most of the time now

[Falkenherz]

SAS worked 100% of the time fine in 0.90 right?.. Now you say it, I have the same problem most of the time now

In my 0.9 it did have the same problems, especially for vessels in space. It just seem excarberated now in atmospheric flight where you have additional drag and tear on your planes which SAS tries to over- or undercompensate.

[Ferdoni]

In my 0.9 it did have the same problems, especially for vessels in space. It just seem excarberated now in atmospheric flight where you have additional drag and tear on your planes which SAS tries to over- or undercompensate.

Weird because I didn't encountered the problem in 0.90.

Overcompensating is because the SAS issues a turn and don't slow it down when the chosen point is almost reached. This causes the ship to shoot over the point. This effect is also increased by the weight of the vessel making it almost inpossible to slow the turn down.

I don't know what the reason for undercompensating is but I think that it has something to do with that the available torque (or control surfaces) is to less for a reasonable turning rate.

[eddiew]

Reducing your control surfaces may help; SAS goes nuts if it has too much leverage to work with.

[Prepper-Jack]

Reducing your control surfaces may help; SAS goes nuts if it has too much leverage to work with.

This. SAS doesn't really engage in micro-movements as a whole, and frequently overshoots the target in a compensation. In lower atmosphere, you'll have to either tone down your control surfaces or remove gimbal and torque wheels to have a steady and controllable ascent. Aerodynamics causes issues at 150+ m/s there, and SAS will overcompensate.

For a number of recent (relatively smaller) rockets of mine, I've been using small delta wings at the bottom - and it's much more stable than using tail fins.

Also - reduce drag as much as possible to maintain stability. A long and thin rocket with little hanging on the outside is less likely to need SAS correction at all.

Edited May 29, 2015 by Prepper-Jack

[OddFunction]

The pendulum happens to me if I use both RCS and SAS, but if I disable RCS it generally holds course without overcompensating.

[Aethon]

You can also try right clicking the engine and limiting engine gimbal.

[DancesWithSquirrels]

If the SAS is overcompensating (oscillating), reduce the control authority. Disable redundant reaction wheels, reduce engine gimbal, use fewer active control surfaces, etc. That will give your ship time to respond.

There's an existing thread on this subject here.

[pincushionman]

SAS has problems generally, as others are pointing out, but since you mentioned planes specifically, I'll weigh in with this: SAS is not very suitable in particular to aircraft flying in atmosphere. You're better off using trim ("mod" + WASDQE) to keep your attitude stable. That and a joystick.

[thereaverofdarkness]

I haven't encountered any of this in any versions of stock KSP, however I did experience a strange bug the first time I used FAR in which it would devote torque on the vessel toward trying to spin out of control. People said I just didn't understand the aerodynamics, but I discovered that my vessels would work much better if I disabled torque and RCS completely and relied entirely on lifting surfaces and control surfaces to determine my orientation. When I tried FAR again on a different computer and different version of KSP (both times during alpha) it worked fine.

[adinfinitum]

One thing you mentioned is that the SAS always tries to return to where the last steering input stopped, I believe this part was an intentional design choice.

It was a while ago, so I might not be exactly right, but in earlier versions of the game the SAS only held the heading at which it was engaged, and to change to a new heading you had to disengage the SAS and reengage once you had reoriented yourself. It was then changed that SAS would be automatically disengaged and reengaged when there was steering input. I don't have any sources other than vague memories, so if I'm wrong someone please correct me.

The easiest way around this issue is to just tap the F key once you reach your desired heading, that resets the SAS so that it doesn't try to return to where steering input stopped.