Wobble and Thrust Vectoring?

[Geschosskopf]

I've never had much if any problems with wobbly rockets, even when they're held together with multiple, unstrutted docking ports. However, I keep reading how many people have serious wobble issues. Now, some of the designs I see are asking for wobbles, what with obvious imbalances in weights or forces, or with severe stress concentrations at wasp-like waists, but most rockets I see that are claimed as wobbly don't appear to have these problems. In fact, they look much like mine. So for a long time I was puzzled as to why they wobbled and my generally similar rockets didn't. Was it related to computer horsepower, different mods, different flying styles, or what? Finally, however, I noticed that the main difference I could see between these wobbly rockets and mine was that mine always have tail feathers and these did not.

This got me thinking. All rockets require something that generates control forces so they can do the gravity turn during launch. Reaction wheels (unless you have an insane number of them) don't provide enough torque to turn big rockets in a timely manner. Very few people seem to use RCS during launch, and as mentioned these rockets don't have tail feathers. So the only thing left is thrust vectoring.

2 variables affect the control force generated by thrust vectoring: the angle the nozzle points and the throttle setting of the engine. During launch, the throttle is generally wide open (or a hair less to prevent overheats) so that's really not a variable here. That leaves the angle. I really don't know how the game moves the nozzle but observations suggest that instead of using fine increments between 0 and full deflection, the nozzle goes from 1 extreme to the other without stopping in between. If true, this means that during launch, thrust vectoring is slamming the rocket with off-center thrust as hard as it can each time it activates. And because of this, thrust vectoring is not a precision control system.

This lack of fine control becomes critical if you have either SAS or MechJeb running during launch. Both of these systems try to keep the rocket on a precise heading using whatever control systems are available to them. This means they want to make very frequent, very small control adjustments to stay on course. But if all they have to work with is thrust vectoring, they're going to have big problems, especially during launch. At full throttle, thrust vectoring will almost certainly cause an overshoot, requiring an immediate correction in the opposite direction, which will also overshoot, repeat. Thus, the rocket is subjected to a constant stream of strong, off-center thrusts in different directions, which can't help but result in significant wobble. OTOH, if the rocket has tail feathers and/or RCS, even if it also has thrust vectoring, SAS and MJ seem smart enough to use them for their fine control needs instead of using thrust vectoring. Thus, no wobble.

Bottom line: The biggest source of wobble during launch that I've found is when you have no control systems except thrust vectoring AND are using either SAS or MJ. If you fly totally manually with SAS off, you're not giving a constant stream of small control inputs, just making major inputs at long intervals. Thus, you don't get wobble even if all you have is thrust vectoring. So, it seems to me that to avoid wobbly launches, you should either add some tail feathers and/or RCS, or don't use either SAS or MJ during ascent.

Anyway, that's my theory. Comments?

[lazarus1024]

Pretty much true. My general control strategy is to disable gimbal on most of my boosters except the center engine. This keeps overall wooble to a minimum, but still provides enough control authority. For smaller rockets with no side boosters, I generally disable gimbal as well, but I slap on some canards for aerodynamic control. 4 canards plus the built in reaction wheel is generally enough to keep most designs pretty stable up through around 20,000m and generally I've ditched the lower stage by then anyway. Upper stages with lower powered engines the wobble often isn't a big deal.

[shand]

Yeah, Locking gimbling to reduce control authority is a great way of reducing wobble. I've started using a mod that improves the gimbling a lto - but it isnt compatable with mechjeb (in that they both try and control the throttle and crazy things happen).

equally aero tail feathers are useless on the later stages where there is little atmosphere, so its a trade off.

[SaturnV]

But thrust vectoring is way more effective than RCS/winglet/reaction wheel for heavy rocket, if you can handle the wobble

[Skorpychan]

Generally, I DON'T disable gimbals on my 7-mainsail lifter. I need all the engines to make my gravity turn in time, and it's stable enough to not wobble so much.

The reaction wheels are picked to be able to move the transfer stage and lander with a decent authority. Adding enough to turn the lifter as well would screw with things, and make control inputs so strong I'd spin out and slam into my landing upside-down. Best to continue steering the lift stages on vectored thrust.

[allmappedout]

It's certainly nowhere near the issue it used to be since the new SAS system.

[arq]

The new SAS works wonders against wobble. However, I find that thrust vectoring is more than sufficient, even for 3000t+ rockets with 40+ mainsails, for controlling rockets. The only time where it may *need* aerodynamic surfaces is when using gimbal-less engines like the T30. Though disabling gimbals can help, on large launchers it's usually too much of a pain for me to mess with, and i really don't seem to suffer too much for it. A few (by which I do mean just a few) struts is really all it takes to keep most of my rockets stable.

The only thing that I find aerodynamic surfaces helpful for on rockets is making launch turns faster, since they can then use lift forces to keep you prograde. Also, if your rocket is (for some reason) significantly drag-heavy on the top, lift surfaces will pull that back to keep the rocket from tumbling. Also, if overcontrol is an issue, you can activate gradual control mode with CAPSLOCK and repeatedly tap the controls to turn yourself very gently.

[BahamutoD]

When I launch large and delicate rockets where even the new SAS causes deadly wobble, I just keep vectoring on but turn off SAS and use as little control as possible. You're supposed to use gravity to steer your ship in a g-turn anyways

Edit: Also, like the guy above me said, using aero surfaces (I just use non-controllable ones) will keep the rocket pointed pro-grade. Gravity will naturally pull the prograde marker toward the horizon if you do a small steer in the right direction.

Edited September 5, 2013 by BahamutoD

[AmpsterMan]

I've often heard that control surfaces above COM can cause imbalances. Does this mean that reaction wheel forces above COM may cause some instability?

[samstarman5]

A little fun fact that you guys might find handy to apply to your rockets: In the Saturn V 1st and 2nd stages, the center engine's gimbal was locked, leaving just the exterior four to help direct the ship. So you can rest assured that, in a way, you are just applying the same strategies as real rocket scientists that got us to the Moon.

[-Velocity-]

I've often heard that control surfaces above COM can cause imbalances. Does this mean that reaction wheel forces above COM may cause some instability?

No, they should not. Theoretically, I know from mechanics that it doesn't matter where a moment (a twisting force- a torque) is applied to a structure, the end result is the same. So it shouldn't matter.

Control surfaces do not apply moments, they apply a force, so it can matter very much where you put them, and if I remember correctly, there was also a bug where the control surfaces were canting in the wrong directions in some cases, can't remember what those were, but I thought that was fixed?

Edit: all that said, you wouldn't want to place your reaction wheels somewhere that was structurally weak, but that's just common sense.

Edited September 5, 2013 by |Velocity|