Once again, the SAS is to weak

[Temeter]

To clarify, I'm talking aoubt the System activated with T. Not the modules itself, they are plenty powerful.

Seems like 0.24 brought an old issue back, afair from around 0.22. The first implementation of the new ASAS (post-UPDOWNLEFTRIGHTLEFTDOWNBOOM-era) wasn't aggressive enough. Engine vectoring could balance out ships to some degree, but generally imbalanced weight distribution would shift the heading, while manual control was more than able to spin your ship around.

If you don't know the effect: You can easily reproduce it with a craft with small fuel tank, a gimballocked engine/aerospike, and the stabilizer sas modul. Then put one of the bigger radial rcs-tanks on one side of the ship and look how it behaves in space. When flying straight, the ship has trouble keeping a stable course and swaying around. Then manually move the rocket for a short time, and the SAS often completely loose it's ability to stabilize the rocket after releasing controls.

In the beginning, I had some smaller ships with weaker gimbal/SAS (but manually more than powerful enough), where i had to constantly readjust orientation during a burn. Similarly space planes feel notably less stable, and often can't deal e.g. with slight instability or short physical judder. It's especially bad when using time-acceleration during flight.

Did Squad comment on trying to change SAS? Because atm it's unnecessary unpractical, especially on long plane-flights, for time-acceleration and less symmetric ships. I'd really like to see a revert to a working state, or - even better - the option to manually set the agressiveness of SAS. Different ships work optimal under different SAS-settings, but to strong reactions are usually better than weak ones.

--------------------------------------------------------------------------------------------

To explain it better, pictures:

Javascript is disabled. View full album

There are basically two symptoms on the example craft:

1. SAS alone can't ceep a craft straight despite having more than enough force.

2. Short command inputs cause the SAS to be even less reliable for the rest of the burn/flight (or until 'resetting' SAS).

Edited July 20, 2014 by Temeter

[sal_vager]

SAS isn't supposed to lock you solidly on a trajectory, the old ASAS behaviour was a placeholder.

Besides, you can always add more

Moved to dev suggestions.

[mike9606]

Well the SAS used to be so aggresive it would tear space stations apart, and who would make a ship asymmetrical like that on purpose?

[Temeter]

You're misunderstanding me, this is all about the improved versions. The first (progressive) SAS-rework wasn't aggressive enough, critizised, and changed in the next patch. Everything was fine. With the last update, the system got weaker, to the point where it doesn't reliable holds a ships course, especially combines with short manual inputs.

That's the task of SAS. And it does not do what it's supposed to do. Not sure what adding more SAS would change. The issue is that my ship is turning right while accelerating, and the counter-movement via SAS is only half the available power and therefore to weak to stop spinning.

Edited July 20, 2014 by Temeter

[Roastduck]

You're misunderstanding me, this is all about the improved versions. The (progressive) SAS-rework wasn't aggressive enough, critizised, and changed in the next patch. Everything was fine. With the last update, the system got weaker, to the point where it doesn't reliable holds a ships course, especially combines with short manual inputs.

That's the task of SAS. And it does not do what it's supposed to do. Not sure what adding more SAS would change. The issue is that migh ship is turning right while accelerating, and the counter-movement via SAS is only half the available power and therefore to weak to stop spinning.

Adding a second SAS unit would double the amount of torgque available (i think), but they weakened the first SAS part available in order to balance it with the ASAS part you unlock later

[Temeter]

Adding a second SAS unit would double the amount of torgque available (i think), but they weakened the first SAS part available in order to balance it with the ASAS part you unlock later

Dammit, explaining stuff in english was never my strength. Added pictures to OP.

The SAS does not use enough of the available force. I can perfectly stabilize the ship via keyboard input, but that's the task of the SAS. Especially spaceplanes often tend to change course under time-acceleration.

[Master Tao]

Dammit, explaining stuff in english was never my strength. Added pictures to OP.

The SAS does not use enough of the available force. I can perfectly stabilize the ship via keyboard input, but that's the task of the SAS. Especially spaceplanes often tend to change course under time-acceleration.

Thanks for clarifying. I can confirm this, though I'm not sure whether it's the intended behavior.

As Temeter said, SAS does not use all available torque to maintain heading. It's possible to build crafts that you can manually hold steady using torque alone, but that veer off heading without control input with the SAS on.

[Temeter]

I want to add that i finally decided to make this thread, when the issue occured with an 18 ton fuel tank+skipper engine+SAS module. The culprit of asymmetry? A single, 0.2 ton hardpoint.

I'm not sure whether it's the intended behavior.

As i said, the first version of the current progressive SAS had the same issue. People wondered if that was intended. The dev-comment was as following: "It doesn't work correctly if it can't keep the course."

I mean really, why would you use SAS when you sometimes can't even fly straight through space?

Edited July 20, 2014 by Temeter

[Sky_walker]

I build a rockets twice the size of yours with significantly more asymmetry - and didn't have any problems.

You're playing with mods?

If not then I don't know what you are doing wrong, sorry. For me it works fine with every spaceship I've build so far. And I even hauled an asteroid on Kerbin orbit (took a bit too few RCS for that, so turning took a while, but still: I made it).

[Temeter]

It's an old and known problem retuning. It has nothing to do with mods (only using KER atm), and I've indeed brought up 5 times the weight without noteworthy issues.

As i said, it depends on the craft, sometimes the system is strong enough, sometimes it's to weak. The former case doesn't matter, the latter one has to be fixed.

Edited July 20, 2014 by Temeter

[Random Tank]

SAS should be nerfed to oblivion, because it's too powerful actually.

BTW: SAS is for ...

*sigh*

The SAS is a bit weird; the fact it doesn't use all the torque availible is a bit annoying, but what I really find frustrating is the bounce back you get when maneuvering. This is where you turn your ship 90 degrees, and when it stops turning the SAS will then turn the ship back 10-20 degree as it over does the momentum cancellation... Sometimes it can be really bad where the SAS moves the ship the wrong way more than you moved the right way :/

Edited August 11, 2014 by Climberfx

[Kevin Kyle]

SAS wont even keep a ship in orbit and not under power from spinning now. Stacking on more doesnt fix a thing nor does AltX. Seems to be something isnt quite right with it to me. Not sure what to think about it. Why does a motionless ship in space start rotating no matter the effort you go through to stop it?

[Stargate525]

The problem I have with it is that the darned thing is too gooey. That stupid dampening mode is useless. If I have the SAS on, I want it to hold the heading I'm on. At the moment, you have a slight error, move to fix it, and then suddenly the SAS is useless because it's stuck in dampening mode until you're done firing (at which point you no longer NEED it).

[Kegereneku]

I cannot say I have problem with the current "new-SAS" and its dampening mode. Although I'm rarely using unbalanced rocket, plane maybe but they have all sort of other reason to not accelerate straight (in Atm or Space).

I don't seem to be able to reproduce the bug sorry.

[Ruedii]

The SAS isn't that bad.

However, it would be nice to have a "Trajectory lock" mode as well as a "Dampening" mode, and a way to switch between them on your active command module.

In fact it would be nice to have the following:

Trajectory Lock

Angle Lock

Dampen

As well as relative modes of:

Surface

Surface Trajectory

Orbit

Orbit Trajectory

Craft (A.K.A. Static)

This basically would make SAS be a basic "hold" autopilot system with several modes.

It might be a good idea to have Trajectory lock, and every relative mode (except craft/static) require at least one ASAS unit, while when only having SAS units and/or the in-command-module SAS you could just switch between Angle-Lock and Dampen mode.

Exceptions could be made for some of the more advanced items (like the large probe cores, the airplane cockpits and the cupola unit.)

This would give the ASAS more advantages over the SAS.

Edited August 12, 2014 by Ruedii
Elaborating some more

[Vanamonde]

SAS wont even keep a ship in orbit and not under power from spinning now.

If your ship is experiencing sourceless rotations, SAS isn't the problem. Have you checked the usual things? Joystick calibrated and zeroed? Trim cleared? Are the instruments indicated input calling for a rotation? SAS isn't designed to stop a rotation if the ship thinks, for whatever reason, it's being told to rotate.

[Wanderfound]

Not a problem here, either. The issue isn't the SAS, it's the designs. Spacecraft thrust needs to go through the centre of mass, in both KSP and reality. The shuttle used quite a bit of fancy vectoring tech to deal with its asymmetric design.

[Laie]

@Wanderfound:

The OP has enough torque and/or gimbals for his craft to fly straight. His beef is that SAS won't make full use of them.

[pizzaoverhead]

SAS uses a PID controller. To work correctly, it needs tuned to the specifics of what it's controlling. Because KSP players can make whatever craft they want, the single SAS PID has to be tuned to work with the majority of designs. (Older releases had two PID settings depending on what part you use, the Sensor Array Computing Nose Cone being the second configuration). This means that there will be craft unsuited for whatever values Squad pick.

Previously, it was tuned to lock its heading as much as possible. This had the effect of oversteering, missing the heading, going past it, and then oversteering in the opposite direction, causing wobble and often craft destruction. More recent releases made the handling softer, reducing this wobble and craft destruction for the majority of designs, at the cost of accuracy.

Three potential solutions:

1. Retune SAS again to increase accuracy but also increase wobble.
   
2. Add a variety of preset SAS settings selectable via either having a variety of different parts or tweakable menus on a single SAS part.
   
3. Allow users to customise the values for P, I and D for each craft via tweakables.
   

Edited August 12, 2014 by pizzaoverhead

[DaMichel]

Thanks for the explanation Pizzaoverhead.

2) sounds good to me. Perhaps it could be auto-tuned according to the available torque/mass ratio or whatever it needs to depend on. 3) is no good at all. The least thing i want to do is spend my time on is tuning three obscure coefficients.

I noticed the strange SAS issues, too - piloting planes (with FAR) but anway. During reenty SAS holds the attitude very well and counters the increasing aerodynamic forces up to full pitch input. But in atmospheric level flight it uses maybe 10% pitch input in an attempt to stay level and fail miserably at it.

[UberFuber]

Thanks for the explanation Pizzaoverhead.

2) sounds good to me. Perhaps it could be auto-tuned according to the available torque/mass ratio or whatever it needs to depend on. 3) is no good at all. The least thing i want to do is spend my time on is tuning three obscure coefficients.

I noticed the strange SAS issues, too - piloting planes (with FAR) but anway. During reenty SAS holds the attitude very well and counters the increasing aerodynamic forces up to full pitch input. But in atmospheric level flight it uses maybe 10% pitch input in an attempt to stay level and fail miserably at it.

Option 3) would be for more advanced rocket builder. For most others, one can go with default value.

I wonder if it's possible to have a button in tweakable that sets some sort of default setting.

For example:

1. Heading Lock - Primarily a PI controller with D for damping.

2. Kill Rotation - Primary a D controller (to zero out rotation) and nothing in PI.

3. Custom - Allow player to set individual PID values within reason.