3.75 m parts + SAS = Death Wobble

[rkman]

Those are extremes. No wonder that even MJ does not work well with that.

I think it is not unreasonable to expect no excessive oscillation or wobble with average rockets build from stock parts - with use of the so-advertised stability augmentation system.

Not just under very favorable circumstances but generally for normal applications MJ demonstrates that a well behaved PID can make the difference between mission failure and a smooth ascent to orbit.

[Jouni]

Those are extremes. No wonder that even MJ does not work well with that.

Extremes are good for demonstrating why PIDs don't really work with non-rigid rockets. If you practice modular construction with more than 2 or 3 modules per ship, or use the Claw and/or asteroids for anything, you'll experience similar behavior in far less extreme cases.

[DancesWithSquirrels]

Jouni - The PID parameters are demonstrably different in the code for 'orientation hold' versus all other modalities, and the other vectors (normal, prograde, etc.) are too highly tuned and oscillate. This is reality - if all orientations used the 'orientation hold' default SAS values they wouldn't thrash so badly.

Are those values ideal for all ships? No, of course not. Different ships with different gimbal authorities and different flex between the engines and the probe/pod under control will have different optimal PID settings. But the basic problem of overly reactive SAS for most orientations is still there.

Any system has an ideal set of PID parameters - the default orientation hold isn't too bad for most Kerbal constructions, but the PID parameters for all other orientations is clearly and numerically too reactive.

Edited June 14, 2015 by DancesWithSquirrels

[rkman]

Extremes are good for demonstrating why PIDs don't really work with non-rigid rockets. If you practice modular construction with more than 2 or 3 modules per ship, or use the Claw and/or asteroids for anything, you'll experience similar behavior in far less extreme cases.

I'm not clear on whether you think PID controllers in KSP should be near useless, or that you think some fixing is in order?

[Jouni]

I'm not clear on whether you think PID controllers in KSP should be near useless, or that you think some fixing is in order?

Three things. PID controllers are good for rigid ships. SmartA.S.S. is only marginally better than stock SAS. Wobble stabilization would require more complex control systems, such as a separate PID controller for each part with control authority.

[rkman]

SmartA.S.S. is only marginally better than stock SAS. Wobble stabilization would require more complex control systems

My experience is different:

I fly all my rockets with MJ ascent guidance, and most of those can wobble but don't thanks to tweaking of Tf max.

600ton, 3 stages

34ton payload in 2 sections

(it's for 3.7 scale ksp universe)

It wobbles a lot with stock SAS and with MJ when it is incorrectly configured:

but it flies steady with MJ correctly configured (high Tf max).

[Jouni]

It wobbles a lot with stock SAS and with MJ when it is incorrectly configured:

http://i.imgur.com/LXC1tEt.jpg

but it flies steady with MJ correctly configured (high Tf max).

If you have ship that works with MechJeb but not with stock SAS, you can usually make it work with stock SAS after some minor tweaks. That's what I meant by a marginal difference.

[Warzouz]

Well oscillation (which is wobbling for me)...

I must admint I'm fed up with it. Maybe they are due to poorly designed ships. But then why are we able to fly those rockets manually and not with the stock autopilot ?

The problem with SAS is not that it overcompensates, but because it's too simplisitc. Heading is 90 : I'm in 80 : Full yaw until we cross 90. THEN it thinks about slowing down by reverting Yaw, then again until it stops.

That a very unefficient negative retroaction loop which can leed to positive retroaction (divergence/explosion...). That's basic automatism. In that matter SmartASS is far superior. It seems it looks at the target heading and angular speed it gets. As we do while piloting ships. We don't blindly change direction, we slowly coast our target.

Even without oscillation problems, on a stiff rocket, the SAS is a pain when you want to change direction. It always overshoot and you have to wait to come back. I usually deal with than by changing from "Node" to "SAS" several times. It feels I'm doing the SAS job...

Now, even MechJeb can't fly anything anywhere.

On wobble/oscillations : The problem is that they aren't that easy to fix. Adding struts and struts can't solve everything. Further more, they never disapear from the ship, increase the part count and look... ugly. Fairing should at least include autostruting of their payload.

I must admit I installed KJR when I saw a payload oscillating outside the fairing. I found that totally stupid and unrealistic. It nearly made me stop playing KSP, until someone kindly pointed me to KJR.

[vixr]

...they kidnapped a cab driver in Mexico, got him drunk, then hooked him up to a Kermancephalogram machine and cloned his brain waves while driving...

someone did this to my wife...

[rkman]

If you have ship that works with MechJeb but not with stock SAS, you can usually make it work with stock SAS after some minor tweaks. That's what I meant by a marginal difference.

You can call it minor tweaks but it is tedious, not much can be tweaked without ending up with a rocket that is un-steerable. MJ does a much better job than stock SAS.

Adjustment of PID parameters (in case of MJ via Tf max) can very much make the difference between death wobble and essentially no wobble. It is not a marginal difference.

Previous versions of the game also demonstrate that wobble is not caused simply by 'PID on a non-rigid structure'. Maybe you did not notice because you don't use SAS?

Now, even MechJeb can't fly anything anywhere.

That's not exactly true. Try adjusting Tf max (Attitude Adjustment), besides doing all the other things right: no to much gimbal, not to much reaction wheel, not to fast, gentle steering, not to many fins, not to much drag at the front.

Edited June 15, 2015 by rkman

[Jouni]

Even without oscillation problems, on a stiff rocket, the SAS is a pain when you want to change direction. It always overshoot and you have to wait to come back. I usually deal with than by changing from "Node" to "SAS" several times. It feels I'm doing the SAS job...

It's a stabilization system, not an autopilot. It's your job to turn the ship to the correct direction before activating SAS in any mode.

You can call it minor tweaks but it is tedious, not much can be tweaked without ending up with a rocket that is un-steerable. MJ does a much better job than stock SAS.

Balancing RCS thrusters is tedious. Setting up fuel lines for asparagus staging is tedious. Balancing center of lift vs. center of mass is tedious. Balancing control authority for stock SAS is usually a quick and simple task, because a few basic design patterns work most of the time.

Previous versions of the game also demonstrate that wobble is not caused simply by 'PID on a non-rigid structure'. Maybe you did not notice because you don't use SAS?

I didn't claim that a PID causes wobble. I claimed that it often makes the wobble worse by issuing commands to turn to the wrong direction.