New SAS functionality and You! [0.23 Update]

[DMagic]

MechJeb has been updated. I would like to see what happens when people use MechJeb's Smart ASS to control their crafts instead of the new SAS.

If you were having unexplained flight behavior before it would be interesting to see if MechJeb can compensate where the new SAS can't. I'm not suggesting that you should have to use MJ from here on out, but it would help to know if these unexplained forces are enough to overpower MJ as well as the stock SAS.

[stupid_chris]

Okay, I can't keep my eyes open. Off to get rest people. Hopefully Chad will post an a blog or something to help out later.

[astropapi1]

Okay, I can't keep my eyes open. Off to get rest people. Hopefully Chad will post an a blog or something to help out later.

'Night, Chris. Thanks for trying to help us.

[4 IN 1]

I am going to support Chirs here, no matter how many **** I put on that test rocket, as long as everything is symmetrical, it will just fly right up.

[Sid]

I can't figure out this new ASAS at all.

Have the joins between parts been made more elastic ? I ask because I tested the Gemini-Titan and it was wobbling around on the pad like a piece of half cooked spaghetti, which isn't going to help. It never did that before.

I've got huge problems controlling any rocket now during the initial launch, they seem to have a mind of their own. Sometimes it's like one of the gimbals has gone hard over, other times they do as they please.

What they are like in orbit - when RCS and reaction wheels would actually come into play - I don't know. I haven't managed to get there yet and all my Kerbals have resigned

I reckon I'll go back to 0.20 until everything settles down. I was really only interested in the new Mun with the spiffy looking craters anyway.

[Vanamonde]

I'm back, and things are NOT going well this morning. That same little test ship

is unable to simply fly straight up. Those canards are capable of this degree of deflection under manual control,

but this is the extent to which the SAS calls upon them during flight.

The canards do twitch, and even deflect slightly, so clearly the SAS is telling them to do something. But what is it trying to tell them to do? And why isn't it telling them harder?

I would like to ask for some official word on whether this ship SHOULD be able to fly straight on ascent without active steering from the player. But as it stands, the ship is simply chasing its prograde marker, wherever the prograde marker happens to roam.

Potential factors: Fresh download from KSP store, running Windows, joystick unplugged.

[DMagic]

Have you filed a bug report? If others can fly the same craft perfectly straight and you can't then it seems that there is a bug somewhere and it should be reported.

And not that I think this should matter or cause such an issue, but you probably shouldn't use the inline advanced stabilizer. The command pod has SAS functionality, so there shouldn't be any reason to use another SAS part.

[Mulbin]

I would like to ask for some official word on whether this ship SHOULD be able to fly straight on ascent without active steering from the player.

Here's your official answer - no, the player shouldn't have to help SAS

The new SAS might be a lot of things, but it most definitely isn't a thing that requires constant input to maintain attitude. It's been pretty consistently doing that on all the test cases we did here, and there were a lot of them.

If the SAS isn't holding attitude properly for you, then it very well may be in a buggy state, because that is certainly not the intended behaviour.

[4 IN 1]

Again I am using the exact same design as yours and it will fly straight up without any input, I even switch out the LV45 and put on both an aerospike and LV30 with the same result.

[DMagic]

I just tested your design too. I throttled up, pushed T, then space and took my hands off the keyboard. It went perfectly straight up, no wobble, no veering off course, all the way until I ran out of fuel; it continued straight on until it neared it's peak of about 210000m, where it started very slightly veering to the west (maybe 1 or 2^o off course. The canards made very slight movements, but I had to zoom in just to see them.

If you haven't already done so, submit a bug report. There is no way that your craft should have the kind of problems you describe.

Edit: Here is another thing to try if you can get a command pod into space (it doesn't have to be in orbit, just above 70000m or so. Decouple the rocket so that you just have the command pod. First check for any no-input phantom rotation, both with and without SAS on. Then turn off SAS and stat spinning, you should continue spinning in the same direction. Next turn on SAS, you should top almost immediately. You shouldn't need any batteries or solar panels. The command pod has enough power for a few minutes of screwing around like this.

Edited July 25, 2013 by DMagic

[Tw1]

So, the reaction wheels have no SAS functionality of their own, but does the SAS make use of the torque they provide?

If not, that'd be handy.

[blenderman]

I built a rocket exactly like Vanamonde's.... it holds vertical all the way up not touching controls... and I can also turn to 90 and it holds position at angles... no problems with new ships. old craft files however... have varying degrees of success. I think old craft need redesign from scratch or at least editing and modification.

[DMagic]

Just a few more things about testing here.

This is my 600 part CPU performance test rocket (which is a story for another thread). The design is simple and very symmetrical, it just has a ton of parts, but it flies perfectly straight. For control points it has the command pod, two inline reaction wheels, the lander can (it's unmanned though, so I don't think it does anything), twenty winglets at this stage, three or four rings of RCS (which are never really needed), and only one, central gimbaling engine in all stages.

It can turn at relatively low altitudes, too. It is sluggish, and doesn't hold course perfectly, but I think that's to be expected. Only minor manual inputs are needed to keep it steady. During the most unstable section of flight, shown on the right, there is a bit of wobble and some rotation. You can see that the RCS does fire to keep it fairly steady, but manual input is necessary to keep it from rotating and to hold it on course. This really isn't surprising though, that short section of the flight is not designed to handle large changes in direction and that central stack on the bottom left flops around a lot. This craft was created in 0.20, so it could also have some instability related to that, but nothing major.

And here is another thing that I mentioned earlier. Mechjeb was updated so I tested out the Smart ASS function on a tall, somewhat wobbly rocket. With the stock SAS function there wasn't really any wobble at all and this stayed straight up. Mechjeb is definitely more aggressive in keeping the rocket on course. The winglets move quite a bit more and the rocket starts to wobble. And it is definitely better at holding the rocket on a fixed course once you start turning. Even when I decouple the lower stage Mechjeb is better than SAS at holding course.

Edited July 25, 2013 by DMagic

[tobjv]

(Steam version) I had this bug too. I deleted KSP from my computer before installing it with the delete local content and even basics ships were a problem. I uninstalled the game and went to

C:\Program Files (x86)\Steam\steamapps\common

And deleted the entire folder. Then I redownloaded it. I have yet to have the problem again.

[draeath]

(Steam version) I had this bug too. I deleted KSP from my computer before installing it with the delete local content and even basics ships were a problem. I uninstalled the game and went to

C:\Program Files (x86)\Steam\steamapps\common

And deleted the entire folder. Then I redownloaded it. I have yet to have the problem again.

When you remove games in steam, only the data that steam placed is removed - files created by the game remain.

[ChefCookChop]

Everything is working fine for me now. Just had to remove all the mods, i could add most of them back though. Good update.

[Anizer]

This is what I noticed after the patch;

-While exiting the atmosphere my ship no longer shakes

-turning on SAS in orbit stops the ship from rotating more quickly

-there are still moments where you can see SAS not doing it's job and letting the ship move around

-follows maneuver target better

Also i'm just wondering if turning on SAS on top of your target location and orbit direction (on the spinny ball) locks on.

[KSK]

Just my 2c worth after a quick test last night. This was a clean install, Mac OSX version from the Squad website. Keyboard and mouse controls.

No pictures I'm afraid but my test vehicle was: one person capsule (SAS and reaction wheels included), service module for RCS and battery, short narrow tank plus LV-T30 for the upper stage, largest grey Rockomax tank plus four LV-T30s plus four winglets for the lower stage.

Reasoning: Capsule provides SAS control and a bit of torque. Winglets for control authority in atmosphere, torque + RCS if required for control authority out of atmosphere.

Granted this is hardly the largest or most challenging test vehicle for the new SAS but it all seemed to work fine. No problems getting to orbit and it was nice to be able to toggle the RCS on during ascent without it spewing monoprop all over the place. RCS system was triggered occasionally but only in short bursts. On orbit, the capsule would hold its attitude without any problem, even during a rather lengthy circularisation burn to correct for the less than stellar orbital insertion. What the SAS wouldn't do is follow the direction of flight (prograde vector), so manual correction was required for a longer burn, such as when de-orbiting with RCS only.

This is nothing new though as far as I know - or at least it isn't with my designs. I rarely used ASAS on orbit though due to the monoprop wastage, so perhaps I've always been missing something. It makes sense to me though - if the capsule attitude is being held fixed, then it just can't follow a constantly changing direction of flight.

What I did notice was that the new SAS was much more precise for small corrections, so powered landings are going to be a lot nicer now. Not to mention that I can finally steer a capsule in atmosphere.

Overall - a cautious thumbs up. I like it so far but as I said, this wasn't a particularly exacting test. Time to launch something bigger!

Edited July 26, 2013 by KSK

[dpraptor]

Thought I would share my test of a flight that i have been doing. From what I read and what I have seen, It seems it might be the rockets that are the problem. I used Zenith's SuperNova rocket, which launched just fine with no problems. But I added 1 more asparagus set (3 stacks & rockets) to be able to get a large fuel load up.

Here is the rocket and the test flight - I did this a number of times and had very little problems with the rocket staying on course.

Here is the ship ready for takeoff

1st stage - staying on track

2nd stage - still on straight heading

Gravity Turn - still no problems

With a rocket this complicated - I cant blame the SAS system for being the problem. I used an ASAS part (the Inline Torque Wheels), maybe that helped. I did not turn off the gimbals on the rockets, although I did on earlier tests and didn't see much of a difference. I think the SAS controls them, so maybe that is no longer an issue.

I also noticed that the rocket started to rotate, which happened a lot before the update. This time it kept correcting itself and returned to 90 degrees, even from both directions as it rotated to the left and right, and it continued to correct from rotating.

Also my PC is an 17 6 core, I have 16GB ram and a 2GB video card, so it is also prudent to see if the PC may be the problem.

Also I had the original update from steam and never reinstalled the game, so I dont see that as too much of being the problem either.

Edited July 26, 2013 by dpraptor

[shuttle099]

Good Guide!

[tobjv]

When you remove games in steam, only the data that steam placed is removed - files created by the game remain.

Yeah, I know. I am saying that some how seemingly corrupted the entire game when it shouldn't be possible.

[Corbald]

After reading threads like these, both after the primary release and after the hotfix, I decided to do a bit of poking around. I wasn't having any issues with the new SAS on my own desktop Win7 machine, but I decided to see if any of my other machines would. Of the five, very different machines I own, I was able to reproduce the SAS bug on the second one (my Wife's laptop). Using the Store Download, copied directly from my own install folder, with Save and Ships intact, NO joystick has ever been installed, first install of KSP ever on that machine, my test rocket, which flew just fine under SAS on my own machine, spun out of control very quickly, exhibiting seemingly random drifting, inability to maintain 'lock,' and appeared to have random inputs being fired, though I didn't touch the controls.

Very Strange, indeed! However, this (at least) confirms that there is a bug present for some users, which appears to be computer dependant... I'm trying to figure out what might be different between these two test cases that would cause this issue...

My Olde-But-Bolde desktop is significantly weaker of a system than my Wife's laptop, so I doubt it's the physics slider, as that would suggest that slower machines would spend more time in 'inaccurate' frames. My own spends greater than half of every minute in 'yellow' physics mode. :/

My own machine is a Dual Core with 8gb ram + the windows patch to use all 8gb in the Win7 32 bit environment and an ancient nVidia GPU.

Wife has a Quad Core, 6gb, 64bit Win8 install, with AMD Radeon...

Anyone match some of these specs? Is it 32 vs 64 bit? Win7 vs 8? Ram? Radeon vs nVidia?

[regex]

On my install, completely clean directory-wipe from Steam, I have an issue with the SAS holding a heading while under thrust if my craft is even slightly out of balance. Given that HarvestR has stated SAS should holding attitude I believe it is not working properly at this time and I have a bug report in detailing my issues.

If the current behavior I am experiencing is, in fact, the intended behavior of the new SAS then I would have to declare it nearly useless for any actions while the craft is under thrust if the craft is in the least bit unbalanced (a single radial parachute on a 20-ton vehicle would introduce a heading error over a long, slow burn). Not only does it not hold attitude but it actively resists my attempts to correct heading. The only use for it regarding rockets is to stop rotation, but only while the craft is not under thrust. For aeroplanes I find it works fantastically.

[Tiberion]

Well I have used the new SAS quite a bit both during testing and afterwards while updating mods.

I am.... almost completely sure that the issue is with COM/COT alignment. I can build ridiculously massive rockets in Novapunch that weigh more than 500 tons on the pad, and other than a little wobble due to physics it flies fine and SAS can manage it.

I can build a shuttle in my shuttle mod (which is about as imbalanced a launch vehicle as you can make) and SAS doesn't really work at all. It just seems to not respond.

That made me suspect the imbalance issue, but I am more convinced after re-enabling the Center of Mass Compensation plugin that endlesswaves made for me a while back. it basically is attached to the gimbal module of a single engine and adjusts its vector to align the overall Center of Thrust with the ships Center of Mass, in short counter-acting the off-center thrust and making it fly straight.

With the Compensator enabled the SAS was able to slightly hold the shuttle on course, though it was still.

I am really confused about the people saying its not an attitude hold system, that's completely the point.

it should work as thus:

When activated, it saves the heading you were on, and minus any player input, used control authority (gimbals, winglets, rcs) to hold it. If you add input, it lets you "slip" on the axis of input until you release, and then once again takes a heading snapshot to hold. At that point, it will counter-steer to remove any remaining angular velocity and return you to the snapshot heading.

it should not just hold the heading your own when it finally stops any drift, that would be pretty much unpredictable, and thus useless.

Edited July 27, 2013 by Tiberion

[stupid_chris]

That's exactly the issue I found too Tiberion. It seems like when the CoL and CoM aren'T aligned, the SAS just "assumes" everything is actually fine and it's going the right direction.