What effect does SAS placement have, if any?

[troyfawkes]

So, this is probably a really dumb question from someone with as many hours in this game as i have, but:

What effect does SAS placement have on control? E.g. if I put a wheel well outside of my CoM, what changes? What if I flip it vertically? What if it's off my CoM in two or three dimensions?

Edited April 11, 2016 by troyfawkes

[Foxster]

As far as I know it makes no difference. Wherever you place them, and with whatever orientation, they cause the craft to rotate about the CoM.  

[Kerbart]

While placement of the SAS wheels itself has no effect on the amount of torque it generates, it will cause twisting and bending of the parts that "transmit" the torque. For that reason it makes sense to place them close to the COM, just to prevent bending, vibrations and other Bad Things.

[HebaruSan]

Orientation (flipping) of the part definitely doesn't matter. Position seems to be a matter of some dispute.

As far as I can tell, a reaction wheel's torque force is applied from the position of its part, which affects both what it does and how effective it is. For example, if you have a long, bendy rocket and you put reaction wheels just at one end, it'll bend when you use them. (And if your command pod is also at that end, the directional SAS modes will set up a destructive oscillation, because SAS uses the command pod's current orientation to decide which direction to push, so the further off center it is, the longer it'll push in the opposite direction, and so on until the craft explodes.) But if both are at the center, then the outer segments will swing around, but less catastrophically. Similarly, if you have a big heavy craft, a reaction wheel at the edge will behave as if it's trying to strong-arm the entire thing around from that distance, whereas at the center it will have more leverage.

I'll try to remember to run some tests on this. It shouldn't be too hard to loft some representative test craft, but I'm not sure how to best measure angular acceleration. Might have to settle for counting rotations with a stopwatch.

[troyfawkes]

Hmm, maybe worth it to just create a modded part with extreme torque and throw is on the end of a very long series of long struts. Could probably do it clamped on the launchpad. Also, sounds like FUN

[Superfluous J]

Conservation of momentum tells us that it doesn't really matter where you put them. They will apply the same torque wherever they are on the ship, even though they technically break other parts of CoM*.

However, if a torque-inducing control device is too far from your command pod - and too far from other torque-inducing control devices, then your ship will tend to develop a wobble. If you're (un)lucky, this wobble could be at just the right frequency to shake your ship apart. The further apart they are, the more likely and more violent the shaking.

*Reaction wheels don't exactly break the Conservation of Momentum law, but they'd eventually be spinning faster than the speed of light the way they allow you to spin around an infinite number of times.

[Plusck]

1 hour ago, Kerbart said:

While placement of the SAS wheels itself has no effect on the amount of torque it generates, it will cause twisting and bending of the parts that "transmit" the torque. For that reason it makes sense to place them close to the COM, just to prevent bending, vibrations and other Bad Things.

I'm of the opinion that there is a lot of overestimating the power of torque wheels and their ability to cause problems, when in fact it's a combination of many factors (where the torque wheels are actually minor contributors) that causes this sort of problem.

So I agree absolutely to the first paragraph here:

1 hour ago, HebaruSan said:

Orientation (flipping) of the part definitely doesn't matter. Position seems to be a matter of some dispute.

As far as I can tell, a reaction wheel's torque force is applied from the position of its part, which affects both what it does and how effective it is. For example, if you have a long, bendy rocket and you put reaction wheels just at one end, it'll bend when you use them. (And if your command pod is also at that end, the directional SAS modes will set up a destructive oscillation, because SAS uses the command pod's current orientation to decide which direction to push, so the further off center it is, the longer it'll push in the opposite direction, and so on until the craft explodes.) But if both are at the center, then the outer segments will swing around, but less catastrophically. Similarly, if you have a big heavy craft, a reaction wheel at the edge will behave as if it's trying to strong-arm the entire thing around from that distance, whereas at the center it will have more leverage.

I'll try to remember to run some tests on this. It shouldn't be too hard to loft some representative test craft, but I'm not sure how to best measure angular acceleration. Might have to settle for counting rotations with a stopwatch.

While I agree with the general principles, I think that the game gives a misleading impression of the problem and, specifically, I think that reaction wheel placement merely compounds a pre-existing problem. That pre-existing problem generally consists in having far too many weak joints between the controlling pod and the CoM of the vessel, and having reaction wheels just next to that controlling pod, and using an active SAS function and excessive gimbal, and trying to go sideways against the airflow.

Certainly, if you have a large space station held together with docking ports, reaction wheels at the edges can cause significant flexing of the different parts. However, that is a problem that can be easily fixed by just letting go of the controls for a short while.

The major problem generally occurs on atmospheric ascent. It looks like the reaction wheel is "bending" the ship but in fact the forces exerted by reaction wheels are far too weak to do that. Take a few reaction wheels, stick them on top of a heavily-clamped stack of three or four fuel tanks and you'll find that they hardly budge even if you try to develop a rocking motion.

So what must actually be happening is that in the atmosphere, and particularly with active SAS, the reaction wheel induces a minimal amount of flex which then multiplies the far larger forces exerted by the engines (giving slightly different vectors to the momentum of each part) and by drag. That is (IMO) the real culprit. And that flex leads the controlling pod to overcorrect, which leads to oscillation that can rip a ship apart. However, it's the engines and aero forces which provide the necessary force to do that - reaction wheels on their own simply can't.

What reaction wheels do do, however, is add joints, which can weaken craft as a whole. Putting a reaction wheeel in the middle of a tall stack (especially in the 2.5m scale) can actually make bending worse by doubling the amount of flexing in the middle of the stack. And that's why (IMO) moving the reaction wheel to the centre - thereby creating a double joint in the middle of a stack -  is (IMO) a fatal solution to a wobble problem.

[Kerbart]

21 minutes ago, Plusck said:

The major problem generally occurs on atmospheric ascent. It looks like the reaction wheel is "bending" the ship but in fact the forces exerted by reaction wheels are far too weak to do that. Take a few reaction wheels, stick them on top of a heavily-clamped stack of three or four fuel tanks and you'll find that they hardly budge even if you try to develop a rocking motion.

So what must actually be happening is that in the atmosphere, and particularly with active SAS, the reaction wheel induces a minimal amount of flex which then multiplies the far larger forces exerted by the engines (giving slightly different vectors to the momentum of each part) and by drag. That is (IMO) the real culprit. And that flex leads the controlling pod to overcorrect, which leads to oscillation that can rip a ship apart. However, it's the engines and aero forces which provide the necessary force to do that - reaction wheels on their own simply can't.

And yet... when I have a tallish-stack with a SAS unit in the payload (on top, not in the center), I find the whole bunch a lot easier to launch (less noodling) when I turn SAS off, then when I keep it on. If noodling is the result of a SAS wheel in the center, then why is it causing noodling when it's placed at the top of the stack, and do the problems disappear when I turn the SAS wheel off?

Edited April 12, 2016 by Kerbart

[Plusck]

19 minutes ago, Kerbart said:

And yet... when I have a tallish-stack with a SAS unit in the payload (on top, not in the center), I find the whole bunch a lot easier to launch (less noodling) when I turn SAS off, then when I keep it on. If noodling is the result of a SAS wheel in the center, then why is it causing noodling when it's placed at the top of the stack, and do the problems disappear when I turn the SAS wheel off?

When you say you turn SAS off, do you mean you disable the wheel or you deactivate SAS generally? I tend to avoid noodling by toggling engine gimbal on and off (except with the Vector, which needs to have gimbal reduced to about 10% to start with anyway).

Still, I'm not saying that an SAS wheel in the centre causes noodling, but rather that it makes it worse because noodling is due to excessive forces on joints. It's the flexing joints which make engines and lift/drag apply opposing forces to different parts of the ship, which are emphasised rather than damped by the SAS/autopilot function.

The actual force applied by torque wheels adds slightly to that problem - but it's further compounded by the wheels being in the command pod too. The command pod goes very slightly off its original heading and tries to correct. The reaction wheels by the pod start to correct this quickly, but the engine/control surfaces at the base of the rocket have already sent a ripple of movement up the stack, so when that reaches the pod it swings the other way (and in the worst case, this swing is emphasised by the straining reaction wheels which haven't yet managed to correct the heading from the command pod's PoV) and starts trying to correct the other way.

Another thing that happens is that once a joint starts to flex significantly, it seems to require much less effort to flex back the way. Perhaps reaction wheels alone are enough to send a pre-flexed joint back to opposite flex? I'm sure this could be tested with large stacks of short fuel tanks.

(edit - one final thing when you said "in the payload": fairing joints are terrible, so anything with a lever arm to flex that joint is going to cause problems. You didn't mention using a fairing but it's also a compounding factor. So basically, I suppose I'm trying to say that reaction wheels should always be placed right next to the weakest joint, preferably on the side facing the CoM, so that they don't add weakness unnecessarily and don't have leverage to add to flex which is going to happen anyway).

Edited April 12, 2016 by Plusck

[HebaruSan]

You guys have never choked up on a baseball bat? I think the physics name is Moment of Inertia, but it's been a looong time and I need a refresher...

[Plusck]

23 minutes ago, HebaruSan said:

You guys have never choked up on a baseball bat? I think the physics name is Moment of Inertia, but it's been a looong time and I need a refresher...

In space, nobody can hear you scream apply leverage.

It's counter-intuitive, but in space it makes no difference where you place your reaction wheels. The torque applied has exactly the same force no matter where the source is located compared to the CoM.

That means that the only consideration for placing torque wheels is what to do about bendy joints and confused "control from here" parts. Which is enough for plenty of disagreement ; )

[HebaruSan]

Update: My attempt to create comparable test crafts for this is being undermined by the absurd structural weakness of the Advanced Reaction Wheel Module, Large. After I put the end-grip version in orbit, the mid-grip variant needed additional parts to avoid ridiculous flex on ascent. Hopefully four cubic octagonals and four strut connectors won't be enough to distort the results.

Can the large reaction wheel part be made less ... silly?

[Red Iron Crown]

Torque applied to the spacecraft is the same regardless of placement.

Concentrating all the wheel is one place is bad, though, as it can cause bending. My rules of thumb are:

1. If there's only one wheel put it as close to the CoM as possible.
2. If more than one wheel, distribute them as evenly as possible throughout the craft.
3. Alternately, strut the craft enough that bending is a non-issue, then place wherever is convenient.

[Vegetal]

15 hours ago, Kerbart said:

And yet... when I have a tallish-stack with a SAS unit in the payload (on top, not in the center), I find the whole bunch a lot easier to launch (less noodling) when I turn SAS off, then when I keep it on. If noodling is the result of a SAS wheel in the center, then why is it causing noodling when it's placed at the top of the stack, and do the problems disappear when I turn the SAS wheel off?

The problem here has nothing to do with reaction wheels, but SAS PID configuration and engine gimbals.

In simpler terms: SAS is tuned very aggressively in KSP, and will gimbal your engines HARD during launch, which has much more torqueing capability than reaction wheels, causing the noodle effect. The distance between the command module and the actual engine is also a big factor, as it tends to create a constructive wave across the length of the rocket.

[Snark]

A note to @HebaruSan, who's having some skepticism about the "placement doesn't matter" (in terms of torque effectiveness);

As several other posters have said here:  it really doesn't matter.  Really.  Put your reaction wheels anywhere, they will have exactly the same effect in terms of rotating the craft.

That's what the laws of physics say.  It's also what the game actually does (I've experimentally verified it).

Understanding just why that is the case can be tricky, so I really don't blame you.    For a detailed explanation of exactly why that is, take a look here:

So, reaction wheel placement can have an effect in terms of making your ship wobble, so you should use appropriate strategy there if you have a big, bendy craft.  But other than wobble problems, no, it really doesn't matter.

[HebaruSan]

1 hour ago, Snark said:

A note to @HebaruSan, who's having some skepticism about the "placement doesn't matter" (in terms of torque effectiveness);

Yeah, at this point I accept that I'm probably* wrong, since this community is rarely unanimous if there's any real room for doubt. But as you say, it's tricky, and I need to see it for myself. I've thought of a better test craft design than my initial plan, which should lend itself to a decent imgur album.

*: You can ignore this word if you want to work with normal standards of evidence. I'm really just withholding judgment till I perform an experiment that I find satisfactory.