Question about slow rotating craft (gyro? reaction wheel? what is gimbal?)

[Taki117]

It has to do with how fuel flow actually works. It doesn't feed all engines at the same time, but rather one at a time starting with the first engine placed. This means that the last engine is pushing fractionally less mass than the first, which causes that side to accelerate fractionally more. This in turn is amplified over the entire ascent, resulting in a slow roll. This is best countered with AV-R8 winglets or the Delta Delux Winglet for larger rockets.

[Streetwind]

Ooohh, I see. That actually makes a lot of sense.

The SAS is probably compensating the minor differences during straight flight, but when I am steering I am monopolizing the control resources (reaction wheels and gimbals) and so the spin can propagate uncountered.

[Taki117]

Ooohh, I see. That actually makes a lot of sense.

The SAS is probably compensating the minor differences during straight flight, but when I am steering I am monopolizing the control resources (reaction wheels and gimbals) and so the spin can propagate uncountered.

Exactly. The best way to avoid the roll is to either add more reaction wheels to stages you are planning on dropping, add winglets to the bottom of your craft to help with stabilization, or try to counter it manually.

[SSSPutnik]

Did you try locking thrust vectoring?

You can also assign this to action groups.

The fuel flow thing is interesting, but I would have thought SAS would counter this?

Edited February 4, 2014 by SSSPutnik

[Taki117]

Did you try locking thrust vectoring?

You can also assign this to action groups.

The fuel flow thing is interesting, but I would have thought SAS would counter this?

SAS does if you have enough wheels, otherwise it's just annoying.

[DrMonte]

It has to do with how fuel flow actually works. It doesn't feed all engines at the same time, but rather one at a time starting with the first engine placed. This means that the last engine is pushing fractionally less mass than the first, which causes that side to accelerate fractionally more. This in turn is amplified over the entire ascent, resulting in a slow roll. This is best countered with AV-R8 winglets or the Delta Delux Winglet for larger rockets.

I bet it's just all about gimbals in engines. Had lots of designs where turning gimbals off helped to stabilize a rocked. Even without gimbals four engines are not so stable as single Mainsail. Not to mention that T30 does not have gimbals at all and Mainsail does. Which would lead to vice-versa issue - lack of control in lower part of the rocket.

Fuel flow is a bit different topic. If you have a statement that X-couplers has issues with fuel flow you should provide some evidence.

[Taki117]

I bet it's just all about gimbals in engines. Had lots of designs where turning gimbals off helped to stabilize a rocked. Even without gimbals four engines are not so stable as single Mainsail. Not to mention that T30 does not have gimbals at all and Mainsail does. Which would lead to vice-versa issue - lack of control in lower part of the rocket.

Fuel flow is a bit different topic. If you have a statement that X-couplers has issues with fuel flow you should provide some evidence.

The gimbal helps with steering, not necessarily stability. 1 mainsail is always going to be more stable than a cluster of engines because of a variety of factors, fuel flow is only one speculation (Based on aircraft flameout). It could also have to do with the way the engines are gimbaling (If placed with symmetry they have a tendency to rotate round thier axis) or it is even possible that the rotation is coming for somewhere else and while the engines are trying to compensate by gimbaling the amount of gimbla needed is greater than the gimbal range of the engine and without other control surfaces (because reaction wheels are prohibitively heavy for many designs) it's just not enough.

[Claw]

Ahhh, okay. So this was bugging me so I ran a few tests. I don't have the conclusive answer, but I do think there is something more goofy going on than just fuel flow causing fractional weight differences.

-It is NOT caused by gimbaling engines (thrust vectoring, if you prefer).

-I am highly skeptical that it is uneven fuel flow or fractional weighting due to fuel flow.

As a test rocket, I built an RC-L01 with a large advanced SAS on top and an orange tank underneath. I attached four PB-NUKs with symmetry. I attached a quad adapter to the bottom of the orange tank and mounted engines from there.

I tried pretty much every engine that had a combined TWR 1.0 or greater including: LVT-30, LVT-45, Poodle, Skipper, Mainsail, TurboJet, RAPIER (rocket mode), Aerospike. For the larger engines, I turned clipping on (so yes, 4x Mainsails). I won't run through every iteration, but here's what I noticed. This was all SAS on and off. Most movement was apparent with SAS off, but I could get rolling with some engines while executing the turn (due to saturated controls).

-Some engines caused the rocket to roll, some caused it to yaw, others were rock steady.

-Gimbals locked or free did NOT matter. LVT-30 and LVT-45 did a combined roll/yaw maneuver that was very similar to each other despite one being vectored and one not.

-Poodle was rock steady, Mainsail caused a roll much like the LVTs.

-Order of attaching the LVTs didn't seem to matter. Even when added one at a time counter clockwise, or clockwise by hand.

etc. etc...

However, the big surprise to me was the fact that when I ROTATED the LVT-30s, it completely changed the rockets reaction. If you notice the little stem sticking out of the side of the engine, where that pointed made a difference, although it wasn't always consistent.

-Pointing the stem inward (symmetry on) completely nullified any roll. So to do this you have to rotate the engine 45 degrees inward (and some combination of 90 degs based on where you mount).

-Pointing the stem outward (symmetry on) got rid of most of the roll, but not all.

-Pointing all the stems in one direction by hand (symmetry off) made the rocket yaw, though not always in a consistent direction away from the stems.

-Pointing all the stems the default direction (symmetry on, outward toward one side) wanted to roll. It always rolled in the same direction, regardless of rotating the engine any number of 90 deg intervals.

Strangely...

-Rotating the LVT-45 stem inward at a 45 deg angle inward (symmetry on) made no difference in roll.

-Rotating the LVT-45 stem outward at a 45 degree angle (symmetry on) nullified the roll (still wanted to yaw very slightly).

Anyhow, there were a bunch of other little things I noticed too, but the results of rotating engines caught me off guard. Makes me wonder if the thrust vectors (the vector itself, NOT the gimbals thrust vectoring) are getting some kind of weird rounding errors or something.??? (Sorry, Ihave no experience with KSP's code.)

Also, the rocket I tried this on was pretty small, so I don't know how it scales to something bigger (which I assume people are referring to in this thread).

EDIT TLDR: You can possibly fix your rolling rocket by turning the LVT-30 engine stem inward, or the LVT-45 engine stem outward.

[Guest]

Part clipping can cause phantom forces to occur, some of which affect the handling of the rocket quite dramatically. When constructing your rocket or rotating engines around, make sure there are not 2 parts trying to occupy the same space. If you've already tried everything else, such as struts, sas, gimbals on or off, control surfaces, etc, and are still having problems, look closely for part clipping. I personally like to see pictures of the problem craft for this reason. I can usually tell if there is a clipping problem from a picture.

[Tank Buddy]

Reaction wheels: They create torque that spins your spacecraft and helps it maintain heading. Manned capsules and probes come with their own torque, but sometimes it's won't be sufficient, often with the probes. If you really need to move your ship and you add more than 2 reaction wheels on a given stage, consider using RCS.

RCS- primarily for docking, it can also be used to rotate large ships when reaction wheels cannot compensate. It consumes fuel and requires careful placement.

Gimbal- Some engines have "gimballing" which means the nozzle rotates, allowing for thrust vectoring and greater control, but only when the engine is firing.

[SSSPutnik]

-It is NOT caused by gimbaling engines (thrust vectoring, if you prefer).

-I am highly skeptical that it is uneven fuel flow or fractional weighting due to fuel flow.

Forgive me, but I'm sceptical of this. I'm damn near 100% certain I've had asparagus staged rockets where turning off vectoring fixed rotation or greatly reduced it.

(Talking during vertical boost phase here). (I kept a single central engine with vectoring ON and all surrounding engines with vector off).

[Claw]

Forgive me, but I'm sceptical of this. I'm damn near 100% certain I've had asparagus staged rockets where turning off vectoring fixed rotation or greatly reduced it.

(Talking during vertical boost phase here). (I kept a single central engine with vectoring ON and all surrounding engines with vector off).

Perhaps I should qualify my statement a little. This is in reference to launching a rocket using the large to small quad adapter. Since no pictures have been provided, I was going off of the various descriptions Streetwind posted. Most specifically, he said:

I'm using a quad adapter, so the alignment is perfectly even. I've also added Kerbal Joint Reinforcement and the rocket behaves no different than before.

So I used a quad adapter, with perfectly even engine alignment. I used LVT-30s (which have no thrust vectoring) and it rolled around the vertical axis. I then tried LVT-45s with gimbals free and gimbals locked. It still rolled around the vertical axis in both conditions. (Yes, gimbals free or locked for all four engines.) I then rotated the engines as described, and the rolling stopped. This happened with both the 30s and the 45s (gimbals locked and gimbals free).

I have also experienced rockets that rolled with control surfaces on, and when I locked those it stopped rolling. So there is certainly a lot of things interacting here, but I thought I would post my findings on the limited information Streetwind provided.

None of that is meant to say that you CAN'T experience a similar rotational problem on an asparagus staged rocket with gimballing.

[SSSPutnik]

Sorry, my bad

[MrAnonymous]

The more reaction wheels you put in your ship, the easier it is too control the ship but the more electric power it consumes

[DrMonte]

Quadcoupler part looks perfectly layed out:

node_stack_top = 0.0, 0.5, 0.0, 0.0, 1.0, 0.0, 1

node_stack_bottom01 = 0.625, -0.4875, 0.625, 0.0, 1.0, 0.0, 1

node_stack_bottom02 = 0.625, -0.4875, -0.625, 0.0, 1.0, 0.0, 1

node_stack_bottom03 = -0.625, -0.4875, 0.625, 0.0, 1.0, 0.0, 1

node_stack_bottom04 = -0.625, -0.4875, -0.625, 0.0, 1.0, 0.0, 1

but I would not decline possibility of rounding errors in further calculations. Minor rounding errors would lead to very minor fluctuations in trust vector. We know that second by second these minor forces sum up to visible trajectory changes if not compensated.

For really from my quite long KSP rocket building practice I have never bothered about this. Just add a SAS or few winglets and you are done. I'm not sure if it's even possible in real world to build a rocket without an additional controls to be stable enough. We can always say that it's something about uneven trust from engine tube and side winds of the Kraken.

[Streetwind]

Very interesting, Claw! This would imply that there is some sort of directional weirdness going on with the physics simulation of the engine part.

Case in point, the roll also happens with mod-added engines - KW Rocketry's Wildcat-V's are a great replacement for the LV-45T's. The Wildcat's model does not feature a one-sided exhaust, I believe. As such, I highly doubt that the game is trying to simulate minor thrust from exhaust gases or anything like that. Even the location of the exhaust should be irrelevant; for the LV-30T it was probably pure coincidence that the model was aligned in a way that made sense with the behavior. It's likely related to the part itself, internally and model-agnostic.

I will test and confirm/deny whether rotating the engines fixes my roll (both for the LV-45T and the Wildcats) when I get home today. If people are interested, I can also provide the craft file (stock variant) and/or screenshots. Just didn't think it was important since the phenomenon is easily replicated.

[Smidge204]

Quadcoupler part looks perfectly layed out:

Joints are flexible. Even if things are perfectly assembled, once you're under the influences of gravity, lift and thrust your whole craft warps and twists because the joints are not infinitely rigid. One theory is that any engine that it not directly in line with the CoM of your rocket might end up ever so slightly twisted/off-center, and even if the thrust and mass is balanced by symmetry this offset is not mirrored and can cause odd behavior.

Since most rockets have plenty of pitch/yaw control with gimbal'd engines, they usually lack in roll control... so your rocket starts to roll, which is also self-stabilizing to a point since any pitch/yaw induced by an offset engine gets averaged out over a full rotation.

That's the idea, anyway. You can add lots of struts to see if it helps (usually does), plus there's promise that 0.24 will improve joint mechanics.

=Smidge=

[Claw]

Yeah, I agree that the stem on the 30s & 45s really has noting to do with it. It was just a convienent way to reference rotations. The poodle has similar side graphic pipes but it is solid as a rock. The TurboJets are symmetric but also rolled.

So I don't know if that will fix your issues, but I thought it was pretty strange. You might try rotating them in different ways if the direction I suggested doesn't work.

[Vanamonde]

Given how hard it is to completely eliminate roll, my approach is not to try to stamp it out, but simply to control it once it arises. A few control surfaces will do it in atmo, and in space, roll is nothing more than a minor nuisance if you're flying by the navball, which you should be anyway. But anyway, once you shed some launch stages, your existing control means become proportionally better able to stabilize the remaining mass of the ship, and the problem goes away.

[Streetwind]

Well, here is the craft file of the stock variant. It is not very good by most standards, but considering that I built it on my first playthrough ever, in career mode, without reading a single design guide (a rule I tend to set for myself in games that rely on player skill and knowledge) I was still fairly proud of it. Though not the first one to achieve a Mun landing (that goes to variants VI for a one-way trip and VII for a return trip), this rocket was the heavy muscle that earned about two thirds of the total amount of science needed for the stock tech tree through many consecutive (and often parallel) missions to both moons.

After the addition of mods it has since been improved in many areas, most notably by adding more thrust so it isn't underpowered anymore (back then I didn't know about ideal speeds inside the atmosphere).

I did a few test flights just now. Rotating the engines by 45° definitely did something. I wasn't able to fully eliminate roll, maybe because I have other sources contributing, but it was noticably reduced. And in one case, even switched directions! Quite funky, but good to know for the future

[Claw]

I think your rocket looks great.

It's a bit long for the amount of control that's on it, so what it looks like when it's in the gravity turn it has a bit of yaw wander to it. (I don't think you need to change anything about it to "fix" that.) It seemed like when the SAS was trying to counter the yaw and you do your gravity turn, the controls get saturated and it can't stop the resulting roll. It also looks like that yaw is still somehow associated with the weird engine characteristics.

I also rotated the engines and noticed a big difference. Seemed like the roll was nearly gone, and it matched what I saw on my tests. It didn't want to roll at all unless I was actively turning at full rate. (Not sure if that's what you also saw.) Now that I think about it, I've seen similar rolls with long rockets controlled by MechJeb during the initial part of the gravity turn, but never really thought about what it was caused by.

So I would say if you're happy with how it controls, I wouldn't bother adding anything else. It didn't seem like much roll to counter after I rotated the engines and it was pretty easy to control.

[Guest]

I agree that this is a nicely designed rocket.

I added the big sas module just above the quad coupler and it eliminated what little roll there was. I've flown rockets that rolled much worse than this one that were still functional. I think Smidge204 is spot on with his assessment. If you zoom in to the quad coupled engines while in flight you can see the engines gimballing in different directions. It does seem logical that this could introduce some roll.

It's interesting that rotating the engines helps the problem. I'm going to test that out later for myself.

Edited February 7, 2014 by Otis
grammar

[Claw]

I think Smidge204 is spot on with his assessment. If you zoom in to the quad coupled engines while in flight you can see the engines gimballing in different directions. It does seem logical that this could introduce some roll.

Except that you can lock the gimbals and still cause the rocket to roll by pitching/yawing with just the reaction wheels. In fact, it's worse.

The engines are going to gimbal around because the SAS is trying to maintain a set attitude. This could also be causing some roll, but I think in the case of the quad coupler it's something else. I think it might actually be some minor programming bug or artifact since it only affects some engines, is pretty sporadic, and is affected by rotating engines.

[AlexisBV]

Read somewhere here that in theory, SAS should be near center of mass, but in game, you can whack em anywhere, no difference.

Oh it has. SAS can even smack your ship to pieces if you put it wrong.

For a rigid body, it makes no difference where you apply torque, the resulting angular acceleration will be the same and it will still rotate around the CM. Naturally, if the body isn't rigid, and if you apply torque to a structurally weak spot, unplanned separation might occur...

Also, keep in mind that while reaction wheels/SAS apply the same amount of torque in all 3 degrees of freedom AFAIK, the moment of inertia of your vehicle isn't the same for all 3 axis. For a long pencil-like body, the moment of inertia for it's long axis is much smaller than the others. That's why it's easy to roll with SAS, but not so much to pitch/yaw.