The new and improved ASAS is way better than before.

[Dodgey]

Doesn't 0.90's SAS show the various options for what heading to hold? Why's this rocket not have that?

It depends on the level of probe core or pilot in board.

[Greenfire32]

Can't see anything wrong in the rocket, either.

Me neither. Mostly because you did the completely stellar job of recording at night.

Post a snarky OP, get a snarky reply *shrug*

Seriously though:

If I were to guess, I'd say it's the wing/fairing causing some weird areodynamics on your rocket. I bet all the peanuts that if you remove the fairing, it'll fly as intended (provided you're on stock aero).

Also, could be an issue of not having enough control authority. I know some of my rockets require more than one ASAS to function correctly so it's kind of backwards to assume that one ASAS unit can handle everything.

Edited February 25, 2015 by Greenfire32

[White Owl]

While I'm sympathetic to your cause, I have to point out a pretty big flaw in the argument. You say the old ASAS was better, and then show us the current ASAS failing hard. If you really want to illustrate your point, let's see the identical rocket launched with the old ASAS.

I personally haven't had any issues with the current system keeping my rockets flying straight, but it's next to worthless for spaceplanes at high AOA.

[Pecan]

[Replying to cpast's "Doesn't 0.90's SAS show the various options for what heading to hold? Why's this rocket not have that?"]

It depends on the level of probe core or pilot in board.

Which is to repeat what Tex_NL pointed-out in post #5 on page 1, which explains everything.

[SpacedCowboy]

Two words: Top Heavy (However,,,Excellent explosion) +1 rep

[magnemoe]

It depends on the level of probe core or pilot in board.

Yes, it lacked the icons on the left side of the navball, either this is not an 0.9 video or the SAS is not active, I have not been able to even do burns in orbit with it off.

Managed to launch an 0.25 spaceplane in 0.9 with only mechjeb as probe core, experienced pilot so did not notice anything wrong until later, more fun I let mechjeb do the deorbit burn so I did not notice anything was wrong before I aborted mechjeb landing autopilot. It don't work for spaceplanes at all so I set it 2 degree west of KSC and let it do deorit and the first adjustment, check one: can I circulate and do an launch with an pilot and more fuel? No chance, then just return to spaceport and loose the plane as this is totally hopeless.

This is an design who is use a lot in 0.25, no serious accidents after debugging.

[Flixxbeatz]

Old ASAS be like: "Lost a wing? No problem!"

Two words: Top Heavy (However,,,Excellent explosion) +1 rep

Thanks, I learned alot from

.

[cantab]

It's also flicking the control surfaces like crazy, much like the new superSAS options (eg hold prograde) do. At best that makes an unwanted infiniglider. At worst, ie in FAR or I expect in newstock, it needlessly increases drag. And either way, it looks ridiculous.

A fair point has been made that the current SAS is unwilling to use full control authority, but using full authority in response to the tiniest deviation isn't a lot better.

[Superfluous J]

A fair point has been made that the current SAS is unwilling to use full control authority, but using full authority in response to the tiniest deviation isn't a lot better.

This may be crazy but the system could use the angle to destination and current rotational speed and direction to determine approximately how much authority is needed.

[cantab]

This may be crazy but the system could use the angle to destination and current rotational speed and direction to determine approximately how much authority is needed.

I believe it tries to, but the response isn't "tuned" either manually or automatically to take into account how much torque is available and what the vessel's rotational inertia is. (And throw in the complication of aerodynamics). Hence regular SAS jitters with high control authority and is sluggish and lazy with low control authority.

A cleverer system ought to be possible, but I'd put it some way down the development priority list.

[Superfluous J]

A cleverer system ought to be possible, but I'd put it some way down the development priority list.

Me too. This isn't something I expect in the game until we get closer to 1.0

[Renegrade]

It's also flicking the control surfaces like crazy, much like the new superSAS options (eg hold prograde) do. At best that makes an unwanted infiniglider. At worst, ie in FAR or I expect in newstock, it needlessly increases drag. And either way, it looks ridiculous.

..or burning ALL the RCS ~

[r_rolo1]

I believe it tries to, but the response isn't "tuned" either manually or automatically to take into account how much torque is available and what the vessel's rotational inertia is. (And throw in the complication of aerodynamics). Hence regular SAS jitters with high control authority and is sluggish and lazy with low control authority.

A cleverer system ought to be possible, but I'd put it some way down the development priority list.

I guess that, adding to that, the devs from a certain time ago shelved all fixes that had something to do with atmo to after they got a new atmo model, because all the tune in they could do would go out of the window as soon as the new atmo model got out. This kind of SAS response issue , that is practically only noticeable in atmo ( because ships in space will normally not have changes in control authority, so they will either work or not ) was probably shelved along with stuff like showing the effect of atmo drag on a ship trajectory , atleast until the new atmo comes out :/

That said, the math behind this is not that hard to do if you forget for a moment that the atmo drag will change with orientation, and even then you can change the drag force input interactively ...

[Crzyrndm]

This may be crazy but the system could use the angle to destination and current rotational speed and direction to determine approximately how much authority is needed.

That is exactly what it is doing.

Each axis of Stock SAS (in stability mode) is run by a PID controller (I believe this is still the same as the original, but I wasn't around that far back to verify that ). The default configuration has integral (I) response set to zero, so all it is doing is saying:

a) How far am I from my target (P control)

How fast am I moving towards my target (D control)

Integral control is disabled by default because it's not useful in space where there is no force preventing the target being reached. A whole other story for planes ofcourse, this lack of Intregral action is what prevents them holding pitch where you want them to.

The "I give up and will just track prograde" is something Squad added on top of that to prevent some general insanity from what I've read here.

A cleverer system ought to be possible, but I'd put it some way down the development priority list.

It is possible, just somewhat annoying to handle automatically. You need an estimate of how much torque you have available (which depends on speed, atmospheric density, stalling, and location for control surfaces in a sane atmosphere), the current rotational inertia of the vessel, and an estimate of the level of natural aerodynamic damping at current conditions (which is probably related to the seperation of CoP and CoM).

From that you could recalculate Kp/Ki/Kd fairly well with a bit of experimentation. It likely would still fall over occasionally, but thats just going to happen with as many variables as you have in craft design.

NOTE:

I built my own SAS system for Pilot Assistant (because stock SAS just can't handle planes) and have tried to tackle this exact issue several times. So far it has proved itself beyond my levels of understanding so take what I have to say with a grain of salt

Edited February 26, 2015 by Crzyrndm

[Superfluous J]

That is exactly what it is doing.

Then I may be misstating what I'm asking for, or agreeing with something I am misunderstanding.

What I want is a system that does NOT, when you're in space and click "hold prograde," do the following:

1. Slowly accelerate rotation toward prograde (which it doesn't even do but that's a different topic IMO), not using all the available torque.
   
2. At the halfway point, keep accelerating toward prograde.
   
3. At prograde, keep accelerating toward it, causing the ship to swing wildly past it, usually several times before slowly getting what I wanted in the first place and could have done myself in a single motion.
   

And if it could NOT shimmy-shimmy-shake the ship apart when I ask it to hold prograde, that would be a bonus.

[keeper]

I usually use about 2 or 3 reaction wheels throughout my rockets.

Usually my upper stage will have a large one, as well as usually i have a smaller one on the payload, and if theres a capsule of sorts depending I may add another.

[Brotoro]

Then I may be misstating what I'm asking for, or agreeing with something I am misunderstanding.

What I want is a system that does NOT, when you're in space and click "hold prograde," do the following:

1. Slowly accelerate rotation toward prograde (which it doesn't even do but that's a different topic IMO), not using all the available torque.
   
2. At the halfway point, keep accelerating toward prograde.
   
3. At prograde, keep accelerating toward it, causing the ship to swing wildly past it, usually several times before slowly getting what I wanted in the first place and could have done myself in a single motion.
   

And if it could NOT shimmy-shimmy-shake the ship apart when I ask it to hold prograde, that would be a bonus.

I HOPE they are going to fix this. I also hope this was not intended as a 'feature' to represent the imperfect piloting abilities of the kerbals...because it's very annoying.

[Crzyrndm]

Then I may be misstating what I'm asking for, or agreeing with something I am misunderstanding.

Note when I said stability mode.

The new vector holding stuff is on an entirely different system (and that system is FUBAR compared to the stability assistance mode). Not to mention that whatever control system they're using for it is locked up tight so I can't even modify it's behaviour without completely rewriting it.

Problem 1: Roll control => damping only (has no target) and much, much too aggresive about it

Problem 2: Pitch/yaw control => Aggresive P control (probably), no where near enough damping

tl;dr

Someone re-invented the wheel with corners and threw that in for the new modes (while the original system would have done the job just fine). It's probably doing the same thing, just (even more) terribly tuned

Edited February 26, 2015 by Crzyrndm

[r_rolo1]

TBH that behaviour is very much alike to similar issues that early mechjeb had. Maybe Squad is simply refusing to learn with r4m0n mistakes? :/

[TimePeriod]

If you want fairings, get a mod. All this stock nonsense is making your life a whole lot harder then it needs to really be.

[Woopert]

The problem I have with SAS is that it doesn't properly utilize what's available to it. I was testing a Space Shuttle with Mk55 engines, which now have 5 degrees of gimbal range, and the SAS system doesn't even use that! It just sits the engines in their original positions or close to it; meanwhile, using manual control is enough to keep the Space Shuttle flying straight. I just wish it was as simple as SAS holds whatever position you tell it, nothing more, and it will utilize engine gimbals to point them in whatever direction will keep the rocket stable. Better use of Vernor engines by SAS would be great to fix as well.

Edited February 26, 2015 by Woopert

[brdavis]

...the response isn't "tuned" either manually or automatically to take into account how much torque is available and what the vessel's rotational inertia is. (And throw in the complication of aerodynamics)...

Hmm. Do you think it would have been a good use of resources (time and manpower) to fully test and refine and carefully tune such a system… when you know in the next update you'll be restructuring the entire aero framework?

[Merinsan]

Hmm. Do you think it would have been a good use of resources (time and manpower) to fully test and refine and carefully tune such a system… when you know in the next update you'll be restructuring the entire aero framework?

If the next update was 0.91, I would say no, not a good use of resources. But this should definitely be done in version 1.0

[bakanando]

This is indeed a problem of SAS. I have had it happen with different ships that do have enough torque to control the ship, but for some reason it gets locked or broken in some instances and will rotate your rocket towards prograde even if you hold full torque against. It will eventually point down and make your rocket crash.

Posted about it some time ago, but at the time I thought it had something to do with the atmosphere. I recently saw a similar post to this one too.

[Crzyrndm]

Problem 2: Pitch/yaw control => Aggresive P control (probably), no where near enough damping

Ok, so I took some time to examine this a little closer and while I was mostly correct, I got this a little backwards.

If you observe the indicators/pips in the bottom left corner while one of the target modes is active, you'll notice that it is extremely jiggly. This is generally a symptom of excessive damping in a PID loop, and sure enough, I can replicate behaviour almost exactly by raising the Kd of the stability SAS mode (by a HUGE amount. Someone may have added an extra zero HUGE amount)

Video to show my testing

0:30-0:45 is stability Assist tweaked to behave terribly, then a target hold for comparison, then stability assist with it's default tuning from 1:10 (all done with infinite RCS so there's no change in mass or anything like that)

Just for comparison:

[TABLE=class: grid, width: 500]

[TR]

[TD][/TD]

[TD]Stability normal[/TD]

[TD]Stability target[/TD]

[/TR]

[TR]

[TD]Kp[/TD]

[TD]15k[/TD]

[TD]20k[/TD]

[/TR]

[TR]

[TD]Ki[/TD]

[TD]0[/TD]

[TD]0[/TD]

[/TR]

[TR]

[TD]Kd[/TD]

[TD]2k[/TD]

[TD]60k[/TD]

[/TR]

[/TABLE]

Seriously considering adding an override to make those modes use the stability settings now

Edited February 27, 2015 by Crzyrndm