Stock SAS hates FAR... why?

[Elmetian]

Since 0.25 arrived I've been struggling to get FAR and the SAS to play nicely. It seems that there are a few things going on:

• The stock SAS overshoots, causing an annoying pitch oscillation
  
• When stability is achieved with a PID controller, the SAS still 'slips' from a given heading rather than applying the required deflection
  
• The SAS doesn't use more than about 50% deflection of the control surfaces most of the time
  
• The SAS doesn't seem to make much use of reaction wheels even when outside of the atmosphere
  

So far I haven't received a satisfactory explanation for why this is happening. I've experienced this in four separate installs now, and tested with only FAR installed. I've tried regex's PID tune and the Pilot Assistant mod. Both have helped, but they don't completely get rid of the problem. I've tried making my Procedural Wings control surfaces larger, moving them farther from the CoM, and reducing the deflection tweakable but the SAS still refuses to use all of the deflection available to it. This is especially annoying as during ascent and reentry SSTOs often require large pitch inputs to hold them at the desired angle. I can hold 100% positive pitch manually with my 360 controller, so why can't the SAS?

I also have my suspicions that with FAR installed the stock SAS isn't using reaction wheels properly. When I'm ascending through the upper atmosphere (40-69km) where the control surfaces have little effect, the nose will slip from where I left it even with the SAS engaged. It only takes a small input from the reaction wheel to correct it, so why isn't the SAS doing this? I've noticed that in space the SAS sometimes doesn't fight against torque induced by slightly unbalanced RCS placement as well. What's going on?

I've been playing KSP since 0.14 and honestly I'm bored with the stock game. Without FAR, KSP seems almost ridiculously forgiving and I don't want to uninstall it or even downgrade to NEAR. Having to fiddle endlessly with PID tuners and manually flying things for the entire duration of flights is really frustrating me. It doesn't feel fun anymore, or even challenging, it's just another chore. I haven't sent anything past LKO in a couple of months now. I guess you could say this is me desperately hoping there's a community fix in the works for this problem.

[Wanderfound]

While not a complete solution...what settings are you using on the PID tuner? Cutting kp to 1/3rd of default sorts the wobble, but I also chop clamp to 1/2 default; this seems to improve the "hold" ability.

A bit of manual input is still required if you're trying to hold a large AoA (i.e. > 15°), but it'll happily hold a vertical climb if the plane has the TWR to maintain it.

[Elmetian]

Very similar settings. I only needed to reduce Kp by 1/5 to get rid of the wobble. Also increasing Kd and Ki helped with the settling time, and I reduced the clamp to 1/3 of the original value.

These settings keep my SSTO on course up until about 20km altitude, but after that the SAS slips and I have to keep my hand on the controls to hold the nose at 5° to 10° on the navball (AoA is smaller still). It's not as though it's particularly nose-heavy, or doesn't have enough control authority to maintain this attitude. I can hold the nose up without much effort with just the control surfaces. I've even slapped on a large reaction wheel for the high altitude flying, but the SAS hardly uses it, which is ridiculous.

Edited December 14, 2014 by Narcosis

[panzer1b]

While the aerodynamic models are neat, the reason i ditched FAR really quick was the overkill amount of work it takes to get a simple craft to fly properly....

Wish they release a version of FAR that has no GUI, no nothing extra, no features but pure aerodynamics (profile based drag, more plausible atmosphere, ect).

Guess its time to start coding another spin off from FAR (similar to NEAR, but with no features whatsoever but the atmospheric/drag changes)

[Elmetian]

the reason i ditched FAR really quick was the overkill amount of work it takes to get a simple craft to fly properly....

Don't get me wrong, I like FAR. It adds a level of realism that makes building aircraft much more interesting. Making them fly nicely isn't too difficult as long as you follow the same principles RL aircraft have to, but then more complicated aircraft have a lot of software to keep them flying straight, and without the SAS FAR becomes a real pain the arse to use.

I've noticed that in orbit the SAS often seems to only be doing half a job. If I apply full negative pitch and then take my finger off of the W key, the SAS immediately tries to stop this rotation by inputting full positive pitch (just as it used to). However, once the rotation has been stopped, the SAS sometimes removes this input and forgets where the 'target' direction was. Because it doesn't bother applying any other inputs, the nose slides straight past the target direction and the craft is left rotating as though there was no SAS at all.

[bandi94]

I started using FAR 2 days ago , i had the same problem : with SAS activated a nasty wobble gets place.

Actually after you get the basics it's not hard to lift things.

- use a launch TWR around 1.4-1.6

- when you reach 50-100m/s perform a 5-10 degree turn

- wait for your prograde vector to catch up

- follow the prograde vector up to 25-30km

- you can do whatever you want.

I had to send 2 probes to Mun , after a bit of re-design , the rocket launched all alone , after i did the 5 degree turn the rocket by himself ( no SAS or anything ) followed the prograde vector. So it was something like "look no hands"

Second thing i lifted was an MKS module for my Mun base , same thing after doing a 5 degree turn the rocket almost followed the prograde vector hoverwer this time i was needed to do some adjustments ( the rocket tended to deviate a bit from the prograde vector but with some adjustments it was easy to hold it manually).

Edited December 14, 2014 by bandi94

[Hodo]

While the aerodynamic models are neat, the reason i ditched FAR really quick was the overkill amount of work it takes to get a simple craft to fly properly....

Wish they release a version of FAR that has no GUI, no nothing extra, no features but pure aerodynamics (profile based drag, more plausible atmosphere, ect).

Guess its time to start coding another spin off from FAR (similar to NEAR, but with no features whatsoever but the atmospheric/drag changes)

So you want FAR but not FAR?

Sorry FAR isn't that complex, if it will fly in real life it will fly in FAR. I have built some amazingly stupid designs that work in FAR, some better than others. The trick is knowing how things actually work and taking your time with them. If you slap together something willy-nilly like and don't pay attention to the CoM and CoL then yes you will have issues. But some of the best aircraft on Earth fly well because someone took time to do the right thing and figure out how it works, why it works and how to improve on it.

I flew before the PID tuner and the pilot assistant, and I still can fly without it, I just turn off SAS and fly manually using trim, you know like a real plane.

- - - Updated - - -

I started using FAR 2 days ago , i had the same problem : with SAS activated a nasty wobble gets place.

Actually after you get the basics it's not hard to lift things.

- use a launch TWR around 1.4-1.6

- when you reach 50-100m/s perform a 5-10 degree turn

- wait for your prograde vector to catch up

- follow the prograde vector up to 25-30km

- you can do whatever you want.

I had to send 2 probes to Mun , after a bit of re-design , the rocket launched all alone , after i did the 5 degree turn the rocket by himself ( no SAS or anything ) followed the prograde vector. So it was something like "look no hands"

Second thing i lifted was an MKS module for my Mun base , same thing after doing a 5 degree turn the rocket almost followed the prograde vector hoverwer this time i was needed to do some adjustments ( the rocket tended to deviate a bit from the prograde vector but with some adjustments it was easy to hold it manually).

You don't want to launch at that high of 1.4-1.6, you may want a little bit lower.

[Wanderfound]

Very similar settings. I only needed to reduce Kp by 1/5 to get rid of the wobble. Also increasing Kd and Ki helped with the settling time, and I reduced the clamp to 1/3 of the original value.

These settings keep my SSTO on course up until about 20km altitude, but after that the SAS slips and I have to keep my hand on the controls to hold the nose at 5° to 10° on the navball (AoA is smaller still). It's not as though it's particularly nose-heavy, or doesn't have enough control authority to maintain this attitude. I can hold the nose up without much effort with just the control surfaces. I've even slapped on a large reaction wheel for the high altitude flying, but the SAS hardly uses it, which is ridiculous.

I'm not entirely sure what's going on, then; I just don't get that problem unless I'm trying to climb to extreme altitude with insufficient speed. How fast are you usually moving when you hit 20,000m? I normally flatten off at ~15,000m and start cranking through the supersonic, with the aim of hitting hypersonic (>Mach 5) as I pass 30,000m.

Maybe try a test flight? A basic Kerbodyne ship (Wedgetail XXV is probably the smoothest-handling option) combined with the same PID settings I use:

That might let us narrow down whether the problem is piloting, design or something else.

Edited December 14, 2014 by Wanderfound

[Elmetian]

How fast are you usually moving when you hit 20,000m?

Normally about 800 to 900m/s by that point. Even with a 20 ton payload there's enough lift and control authority to keep my nose up. The SAS just refuses to for some reason.

Maybe try a test flight? That might let us narrow down whether the problem is piloting, design or something else.

That sounds like a good idea. It's been so long since I built anything with stock parts, so this may well be a problem specifically with the way the SAS utilises Procedural Wings control surfaces when FAR is installed.