Come back old ASAS - all is forgiven!

[boxman]

This is so odd... I am now testing it with a rocket i can fly fine manually without any SAS at all... But after reaching 10k as soon as i enable the SAS it just flips completely over to retrograde.. If i turn of SAS i can even regain control and get it back to prograde and keep it flying stable but again when i enable the SAS it flips over to retrograde within seconds.

It really makes no sense how it is just some rockets and how it seems like it does this on purpose.

Meanwhile other rockets work as intended even if they are less flyable manually.

[Sean Mirrsen]

ASAS in KSP has always done what real-life engineers with more freedom than we have in the game do manually - set engine gimbal angles, control surface trim, etc, automatically. ASAS technically doesn't exist in real life - there are flight computers and autopilots, and ASAS in KSP is somewhere inbetween.

And while the current SAS system is a massive improvement over the old ASAS in many important aspects, for many people it fails in its most basic function. Which is, do its level best to keep whatever you have built in whatever direction you've last pointed it. While some people seem to remain unaffected, the problem is absolutely unbearable to those that are, as it becomes simply useless, for all its advanced features. Yes, it is more efficient, yes it can remain always on, yes it is built-in. But without it actually doing its job, those improvements are no more useful than aerodynamic bodywork, ABS, GPS, and climate-control in a car with no engine. That is what the affected people see. Nobody is harping on the new SAS out of sheer spite because of its being changed, after all.

[Tw1]

If you spotted something wrong by doing this test, please list all your mods if any, and check the KSP/Parts folder to see if there's anything in there (on a stock install it should be empty).

Hmm. I had the jet engine inside the parts folder for some reason, both times I downloaded. They aren't meant to be there?

Somehow, It doesn't like me putting things in it, but I've still got to check things before I make a bug report.

[astropapi1]

Most of us aren't "haters", we're just trying to tell the devs it doesn't work in our games, it simply doesn't mantain your heading.

What's wrong with giving constructive criticism?

[Tiron]

Nope. I had zero problem launching my giant, 200ton to LKO lifter last night that had a total part count around 800. I am guessing this is an isolated bug; the new SAS works great for me.

It's too widespread, and from what I've seen of it, far too basic to be isolated. I think some people are just not bothered by it as much as others. Basically, it's not acting as an attitude hold at all, it's acting as a motion damper, and one that's very slow to kick in at that. It makes it almost impossible to hold a precise heading in space, because there's no drag helping it out. On a spaceplane it works spectacularly, because it's basically working as a drastically improved version of the avionics nosecone. But it's doing that in space as well, where you really need it to act like Mechjeb's Killrot.

[Guest]

Oh i see thank you for that, i just hope that people realise that it will be fixed, because tbh the old system was a load of arse and really don't want it back.

We have two devs telling us that our concerns are valid, why would we not realize that it will be fixed (in some manner or another)?

[robot256]

I just wasted two hours reading the entire thread, and though I've not actually tried 0.21 yet, I'd like to make some observations:

As a controls engineer, I would suggest that it seems the new SAS is not properly adjusting its gain based on the inertia of the craft and the available torque. Maybe this is a bug, maybe not, but it seems to be the crux of many issues people are reporting. If Squad is really concerned about making SAS useful without ripping apart large stations, they should make the gain and damping values user-settable, because even automatic re-tuning won't work in all cases (I adjust those values frequently in Mechjeb). Then we should do some tutorials on how to use the properly--a bit of control theory never hurt anyone.

As someone who actually builds spacecraft for a living, it seems to me that a lot of this heartache stems from the fine line Squad is attempting to walk between making a physically accurate simulation and a gamified version of reality. Real rockets are hard--really hard--to design, build, and control, and have severe limitations on what shapes and sizes are feasible with today's technology. But this is a game, so we want to make cool stuff that would never work in real life--that's why it's so fun. Unrealistic aerodynamics, magic torque, and over-aggressive controllers make this easier. I think you'll find that every time Squad takes a step toward a more realistic simulation, some of the "fun" designs will become more difficult to pull off, and whether that makes it more or less enjoyable will vary from person to person.

And as a proud Mechjeb user, I am amused by the number of people who say (or imply) "Mechjeb is cheating" and then complain when their automatic heading controller breaks. If it's so important that your rocket automatically hold its heading during a 10-minute burn, why not let it automatically calculate and execute the timing of the burn, too? The simple truth is that flying rockets without autopilots is as unrealistic (and fun) as anything else in the game. But don't start a flame war on my account :-)

[Giggleplex777]

I think we should keep the new SAS (with some bug fixes, of course) and get the old ASAS back with a different model. That way, everyone's happy.

[Tiron]

I think we should keep the new SAS (with some bug fixes, of course) and get the old ASAS back with a different model. That way, everyone's happy.

Well see, there's something of a misunderstanding going on. People are shouting for the old SAS back because it could actually *hold a heading*, not because it works better. Anyone who thinks it works better is daft.

What I'm sure almost all of us actually want is the new SAS working like it did in C7's videos, where it basically works as a better-in-almost-every-single-possible-way version of Mechjeb's Killrot Function (which is itself a vastly improved version of the old ASAS). Right now it seems to be working as a better-in-almost-every-possible-way version of the old Avionics Nosecone.

[RepoRogue]

Thanks devs! While many of the people in this thread are annoyed or frustrated, most of us are just frustrated with the bugs and not with you. Again, thanks for getting on this so quickly.

[Tw1]

As a controls engineer, I would suggest that it seems the new SAS is not properly adjusting its gain based on the inertia of the craft and the available torque. Maybe this is a bug, maybe not, but it seems to be the crux of many issues people are reporting. If Squad is really concerned about making SAS useful without ripping apart large stations, they should make the gain and damping values user-settable, because even automatic re-tuning won't work in all cases (I adjust those values frequently in Mechjeb). Then we should do some tutorials on how to use the properly--a bit of control theory never hurt anyone.

Could they be reduced to a slider or two? If they're is a simulator/test mode some day, perhaps they could let you play about there to get things just right.

[FleetAdmiralJ]

Did some more testing on quickly-built rockets and design definitely seems to be a factor. SAS doesn't make use of a craft's full control authority when it needs to (whereas the old ASAS did it all the time, which caused its own problems with wobble and such). This leads to a sluggish response and cases where the SAS simply doesn't fight hard enough to compensate for drift caused by slight instabilities in design, even if they're so minor that old ASAS handled them perfectly - or, indeed, manual flight without an excessive amount of input. On smaller, lighter craft - both airplanes and rockets - this isn't too much of an issue. But big rockets beware...

I too noticed that winglets on rockets seem pretty useless unless you are manually correcting. However, one thing that seemed to matter is that if you just had a rocket without any launch struts (or whatever they're called), if your rocket leaned and you turned on the SAS, it tended to "lock in" the way the rocket was leaning. So making sure your rocket is straight as an arrow when SAS is enabled seems to matter too. Or at least it did for me.

[Eric S]

Are you just using reaction wheels or are you using the advanced stabilizer? JUST reaction wheels won't automatically steer your ship.

There's a flight computer in the command capsule. Only one flight computer is needed per craft, so that's not the issue. If the craft had a probe part instead of a capsule part, then you'd be right.

[the-d-o-c]

OK, tested it again and got the slight downward drift around 140m/s. SAS keeps her rock solid. This jives pretty well with what I've seen so far, in that SAS works like a charm on fairly light aircraft but refuses to use the control authority of larger rockets to actually keep them stable. High TWRs and/or accelerations seem to make these problems escalate very quickly.

managed to stick 4 engines in asparagus mode onto the "space station core" (added rcs tank & thrusters) & launch it into orbit, had 1 fin on each rocket, 1 winglet on each rocket near the top & it flew stable

i have adapted the design for a refueller, this wont fly stable, it is bigger & has more TWR using the same rocket set up but in rockomax form, same amount of control surfaces in same place, straight up isnt too bad, little wander here and there that can be corrected, but as soon as i start the Gravity Turn and ditch the first stage it all goes pear shaped

[AfailingHORSE]

I think we should keep the new SAS (with some bug fixes, of course) and get the old ASAS back with a different model. That way, everyone's happy.

Or just make another key command to toggle the old SAS, like say... Ctrl+8+6

[MachXXV]

I just wasted two hours reading the entire thread, and though I've not actually tried 0.21 yet, I'd like to make some observations:

As a controls engineer, I would suggest that it seems the new SAS is not properly adjusting its gain based on the inertia of the craft and the available torque. Maybe this is a bug, maybe not, but it seems to be the crux of many issues people are reporting. If Squad is really concerned about making SAS useful without ripping apart large stations, they should make the gain and damping values user-settable, because even automatic re-tuning won't work in all cases (I adjust those values frequently in Mechjeb). Then we should do some tutorials on how to use the properly--a bit of control theory never hurt anyone. [...]

You're spot on - when you do get a chance to try 0.21, you'll immediately recognize that it's just a PID tuning issue by watching the ASAS' behavior. I've been poking around with the possibility of adding a plugin to tune the controllers parameters, but first I need some refresher theory on PID controllers, it's been a long time since school, and it's not something I ever had to use professionally.

[Tiron]

There's a flight computer in the command capsule. Only one flight computer is needed per craft, so that's not the issue. If the craft had a probe part instead of a capsule part, then you'd be right.

Well Originally, the Command Pod's SAS system was set up so it could only damp motion and not actually hold a heading, rather like what we're seeing, because it was supposed to be a demo version the Kerbals got trying to sell them the full version.

Problem is, they're now ALL doing that, in a weird way that I think is caused by it actually being a full attitude hold, except the 'target attitude' keeps resetting.

You're spot on - when you do get a chance to try 0.21, you'll immediately recognize that it's just a PID tuning issue by watching the ASAS' behavior. I've been poking around with the possibility of adding a plugin to tune the controllers parameters, but first I need some refresher theory on PID controllers, it's been a long time since school, and it's not something I ever had to use professionally.

I've heard secondhand(at least) that the devs said it's no longer PID based at all. That said, the way it's acting on mine is very similar to how the old PD-only ones were working. Only slower, and with a brief control reversal at the end of it settling in (as if it were starting to go back the other way to reach its target point...which is right there, so it only does it for a fraction of a second)

Edited July 25, 2013 by Tiron

[Quantum Jim]

I think that part of the problem is not the SAS or ASAS, but the engine gimbal. Locking the gimbal seems to help when I have trouble controlling a rocket.

[samstarman5]

I just got done with my first flight, and while having to add some manual corrections in did add to the challenge, I got my three-stage rocket into orbit without a hiccup, even achieving orbit easier than I have before. I find OP's reaction to be a bit unfair since he was testing it on a craft outside of the usual SAS parameters and didn't even follow up with a rocket before coming on the thread to flame about it. However, there have been many posters here who did test SAS on ballistic vehicles and shown that SAS could use a little bit more. As it is now, though, I don't see it as a game-breaker and just an added challenge to the game. Time warp still acts as a super-sas to the craft, so that can perform as an added stabilizer given enough time between maneuvers.

We did get rather spoiled to SAS holding our craft's position for us. It was high time we got our usual kick in the pants for being so complacent in a game that has been known to be rather punishing. Whatever comes to any further tweaks to the system, for now I don't see any reason to not continue our respective space programs. We may just have to forget a lot that we learned before and begin anew with how we treat SAS.

Excelsior!

[Tiron]

I just got done with my first flight, and while having to add some manual corrections in did add to the challenge, I got my three-stage rocket into orbit without a hiccup, even achieving orbit easier than I have before. I find OP's reaction to be a bit unfair since he was testing it on a craft outside of the usual SAS parameters and didn't even follow up with a rocket before coming on the thread to flame about it. However, there have been many posters here who did test SAS on ballistic vehicles and shown that SAS could use a little bit more. As it is now, though, I don't see it as a game-breaker and just an added challenge to the game. Time warp still acts as a super-sas to the craft, so that can perform as an added stabilizer given enough time between maneuvers.

We did get rather spoiled to SAS holding our craft's position for us. It was high time we got our usual kick in the pants for being so complacent in a game that has been known to be rather punishing. Whatever comes to any further tweaks to the system, for now I don't see any reason to not continue our respective space programs. We may just have to forget a lot that we learned before and begin anew with how we treat SAS.

Excelsior!

Well C7 has said that they're looking into the cause of the problems a few pages back, so I'm pretty sure the added-difficulty isn't intentional. and the current behavior isn't what was seen in previews either, they were showing an attitude hold closer to Mechjeb's Killrot than the old ASAS, only with vastly improved use of thrust and much more stable.

[Quantum Jim]

What happens if you just add more ASAS (or more SAS) modules to a craft? Just wondering if it would help? (Moar boosters, moar struts, moar sas!)

[Feldynn]

Thanks C7 for the post a few pages back (and Harv for the one many pages ago)!

I did a bit more testing with some rockets just now using the stock Kerbal X and a basic Mun-rocket of my own design, generally the new SAS seems to be working as intended but isn't using nearly as much of the available control force as it could (and should IMO). I did a few simple rotation type tests in a 100k orbit around Kerbin and it does slow rotation and wotnot just like in C7's test video a few weeks back with his space station (though I was using a rocket), but it's just a bit too sedate as it's not using enough of the torque / RCS thrust.

When done manually I can rotate, turn and cancel out the movement pretty quickly but when left to SAS alone it seems almost to use only as little force as necessary to try to stabilize the craft, in a way I might liken it to an old /soft joystick that doesn't return to center as quickly as a brand new stiffer / tighter one does. Perhaps all that's needed (at least for rockets) is a simple increase in the initial amount of available thrust / torque / control force that SAS uses to stabilize the craft?

[Mr Sinn Sinny]

I just download the update from steam - played for maybe an hour, and registered on here just to say this.

I now can't get any ships into orbit. The SAS just doesn't work. Even with fins, rcs, control wheels, everything/anything or nothing. The ship begins to roll completely uncontrollably and I'm unable to reset. Problem is, I did always reply on SAS, but simply because its SO hard to fly manually. Yes I'm sure people will say I need to practice, get better, whatever. But the game has become completely unplayable for me. Considering that I put hundreds of hours into this game already, I now find myself really angry. I'm sure what they have done is better - but I can't figure out how to fly the damn ships. Maybe I'm missing something, but as far as I'm concerned, the games broken for me and I may never play it again.

Such a a shame, because I was REALLY beginning to get into it - I can't even get a basic ship into orbit, let alone dock!

Really, REALLY annoyed right now.

My two cents/rant.

[reiyuki]

Did some more testing, for any Devs/users that care to hear.

- Tested a simple VAB rocket with RC-L01 (large remote control chassis). Rocket drifts from launch and spins to ground with SAS.

- Put an Mk1 cockpit in front, same thing.

- Set 'Control From Here' on Mk1 cockpit before launch, works perfectly.

---The problem seems to occur in non-kerbal controlled systems---

With the 'Remote' control modules, it appears to be dampening movement as normal, but not applying any extra force to 'maintain heading'.

Perhaps the AI version is failing to set desired heading?

[Duxwing]

In my experience, holding a course with both the old and new ASAS systems has required patient finesse because the old system was too rigid to accept tiny course corrections, and the new system is too feeble to maintain them.

-Duxwing