Where to put Reaction Wheels?

[Orc]

Hi all,

 

I am playing the demo version of KSP, and having an awesome blast.

 

What follows is a  small question from a complete noob.

 

Where should you put the Small Inline Reaction Wheel in your stack?

 

With the number of parts in a rocket being at a premium (upgrading the VAB without access to unlimited contracts to pay for the upgrade is impractical), where is the best place to put the reaction wheels?

 

I struggle with keeping my orbit-capable rockets stable and aimed in the right direction. One in three launches of the identical rocket ends in a narrow escape for the pilot and half of the remaining flights don't make orbit, or make such a marginal orbit that it does not matter ( periapsis below 30Km).

 

Extra stability, particularly in the 10 km to 30 km is needed. As I understand it the Small Inline Reaction wheels should help but even when I include three in a stack (one near the capsule, one attached to the Stack Coupler between the first and second stage, and near the bottom of the first stage) things still go wrong.

 

What am I doing wrong? Or to Kerbal Space Rockets just have an appauling fail rate?

 

many thanks and regards,

Orc 

[LN400]

I will suggest putting the reaction wheels as close to center of mass as possible. More than one and you will have the wheels fight each other. You might even want to disable the reaction wheel in the control pod while you use the main reaction wheel.

A few thoughts on general design:

Long and narrow rockets (not too long or you end up with a very wobbly contraption), not short and wide. Top heavy with fins at the bottom for stability in the atmosphere combined with "the right amount" of thrust (like an arrow if you like).

High drag at the front is a very bad idea. Streamlined rockets are good. Grand pianos are bad.

Gimbaled engines makes life a whole lot easier. The T-45 is a good choice unless you absolutely need the extra power from the T-30.

Mind you, if you have gimbaled engines, moving fins and reaction wheels, you could very quickly find yourself with a rocket that tries to wobble itself to shreds. With a well designed rocket, you shouldn't need more than static fins and gimbals (OR a reaction wheel), alternatively non-gimbaled engines combined with either a reaction wheel OR controllable fins.

Limit the thrust and speed until you reach about 36km altitude. Going too fast in the denser atmosphere will make controlling the rocket very difficult or just plain impossible.

Edited December 12, 2015 by LN400
typo

[Reactordrone]

The demo version is very lacking in control compared to the full version. With no steerable fins and no gimballing engines you'll have to balance the ship well to have effective control with the reaction wheels. Too many fins at the back and you'll have no ability to change direction, not enough and it will want to flip at mach 1.

Something like this should just about fly itself to orbit.

 

 

[Plusck]

I find that reaction wheels help, and they help immensely in space. They are quite effective at countering an off-balance centre of mass when landing on the Mun, for example.

But in the atmosphere they only contribute so much - if you drift too far off prograde (i.e. you're trying to power through a cross-wind) there is a point they just don't have enough force and you will flip.

So yes - as others have said, it is first a matter of making your craft aerodynamiclly stable, then a matter of getting your gravity turn right.

For the gravity turn - you basically need to follow the exact flight path in reverse that you'd get from re-entry, except with a bit more vertical speed in the upper atmosphere to avoid all that heating. So a rather sudden and probably fatal (due to parachutes burning up) de-orbit in reverse, if you prefer. That means you'll always be facing exactly prograde, reducing all that supersonic drag that hits the side of your nose and flips you around.

[John FX]

Do you get RCS with the demo version?

If so you could fit some RCS to your first or second stage (whichever runs out around 40km) and some RCS ports at the front and back of your craft and have RCS help you maintain control when the air gets thin.

[Red Iron Crown]

Regarding reaction wheel placement: It doesn't matter where they are in terms of torque applied to the vessel, they work equally well in all locations. What does matter, though, is structural rigidity: If the rocket is a bit wobbly then a cluster of reaction wheels anywhere is a bad idea. Better to spread them out throughout the ship to distribute the torque applied.

Contrary to some belief, reaction wheels never "fight each other", they all simultaneously correctly obey the orders from the control point.

[LN400]

2 hours ago, Red Iron Crown said:

Contrary to some belief, reaction wheels never "fight each other", they all simultaneously correctly obey the orders from the control point.

This surprises me a great deal. I have had wild wobbles that would appear if I had multiple reaction wheels and they would go away if I only used 1 reaction wheel, The same wild wobble would appear if I used steerable fins at the rear and reaction wheels on top, or gimbaled engines with reaction wheels on top. I always assumed it would be the same reason behind the wobble as the wobble would go away if I chose one control and disabled the other.

Now I wonder what could cause these wobbles then, seeing how they appear and how I have remedied it in the past. 

[Red Iron Crown]

3 minutes ago, LN400 said:

This surprises me a great deal. I have had wild wobbles that would appear if I had multiple reaction wheels and they would go away if I only used 1 reaction wheel, The same wild wobble would appear if I used steerable fins at the rear and reaction wheels on top, or gimbaled engines with reaction wheels on top. I always assumed it would be the same reason behind the wobble as the wobble would go away if I chose one control and disabled the other.

Now I wonder what could cause these wobbles then, seeing how they appear and how I have remedied it in the past. 

IME this happens when the control point is at an extreme end of a flexy vessel (a common arrangement), the control point starts wobbling back and forth making it quickly reverse its commands to the wheels. More stiffness is the solution (if feasible), otherwise reducing total torque or controlling from a point near CoM is helpful.

[LN400]

10 minutes ago, Red Iron Crown said:

IME this happens when the control point is at an extreme end of a flexy vessel (a common arrangement), the control point starts wobbling back and forth making it quickly reverse its commands to the wheels. More stiffness is the solution (if feasible), otherwise reducing total torque or controlling from a point near CoM is helpful.

This is great info. I'll need to test this (after the exam that is)!

[swjr-swis]

10 minutes ago, LN400 said:

This surprises me a great deal. I have had wild wobbles that would appear if I had multiple reaction wheels and they would go away if I only used 1 reaction wheel, The same wild wobble would appear if I used steerable fins at the rear and reaction wheels on top, or gimbaled engines with reaction wheels on top. I always assumed it would be the same reason behind the wobble as the wobble would go away if I chose one control and disabled the other.

Now I wonder what could cause these wobbles then, seeing how they appear and how I have remedied it in the past. 

@LN400, I've had similar observations with some rockets, even in 1.0.5 which I am still running completely stock. Very strong wobble at times that didn't get resolved until I removed wheels or opted for smaller ones. But I haven't been able to reproduce it consistently... same number of wheels on other craft, sometimes considerably smaller, would be rock solid stable, almost snapping to any angle I would steer it to and keeping it there. Very annoying when it doesn't work and the ship shakes itself senseless.

I am not sure they're 'fighting each other', it may simply be that too much torque at the ready makes it constantly 'overshoot and correct'. But that doesn't explain why with other craft it is stable. Still haven't figured this out.

[tg626]

One thing I found helped with simple rockets was to have a small ballast tank behind the capsule.  A fuel tank, with the flow shut off (right click, the click the green arrows beside fuel and oxidiser). This acts like the heavy front of a dart and prevents flipping.  The fuel is never used.

FYI, Mercury-Redstone had ballast in the same place, which is what gave me the idea.

[Snark]

32 minutes ago, LN400 said:

I have had wild wobbles that would appear if I had multiple reaction wheels and they would go away if I only used 1 reaction wheel, The same wild wobble would appear if I used steerable fins at the rear and reaction wheels on top, or gimbaled engines with reaction wheels on top. I always assumed it would be the same reason behind the wobble as the wobble would go away if I chose one control and disabled the other.

Now I wonder what could cause these wobbles then, seeing how they appear and how I have remedied it in the past. 

The problem is not the reaction wheels "fighting each other."  The problem is that you have too much torque.  The PID tuning for stock SAS isn't great, and it gets the collywobbles if there's too much torque relative to the ship's moment of inertia.  You can observe this if you have a reaction wheel on a long, skinny ship and turn up the physics warp-- it starts wobbling around the ship's long axis, but not the other axes.  That's because the moment of inertia about the ship's axis is much smaller than about the lateral axes, so there's "too much" torque on just that one axis.

 

[Drew Kerman]

16 hours ago, Red Iron Crown said:

Regarding reaction wheel placement: It doesn't matter where they are in terms of torque applied to the vessel, they work equally well in all locations.

Interesting. I always placed mine towards the ends of vehicles as far away from CoM to get the most torque possible. You know, as in swinging a hammer from the base instead of the head. If the torque wheels generate a force shouldn't you want it at the long end of your craft to have the most effect? Or am I thinking about this wrong?

[Pecan]

On 12/13/2015, 10:47:04, Gaiiden said:

Interesting. I always placed mine towards the ends of vehicles as far away from CoM to get the most torque possible. You know, as in swinging a hammer from the base instead of the head. If the torque wheels generate a force shouldn't you want it at the long end of your craft to have the most effect? Or am I thinking about this wrong?

The force applied is torque - twisting, not pushing.  Anyone who wants to think this through can put a broom handle on the ground to simulate a long, thin rocket and check the difference.

If you want to rotate the stick clockwise, for instance, you can push, which works better at the ends, or twist which works better in the middle.  If you push in the middle instead it tends to push the whole thing sideways, instead of rotating it around the centre.  If you twist at the end instead you make that end the fulcrum and have to swing the whole length/mass of the stick.

The whole issue of reaction wheels has been debated a few times and, like RIC, my understanding is that it doesn't matter where you put them in KSP.  That said; floppy ends will naturally trail and then swing past the main body of the vessel with wheels in the centre, so causing endless wobbles.  Wheels at the ends would tend to bend the vessel, which may or may not be more controllable but will definitely be more vessel-dependent.

Please note that this is an exaggerated view of what causes the wobbles only - and at the moment the middle image is wrong anyway - as Snark pointed out there is no translation of the CoM.

Edited December 14, 2015 by Pecan

[Snark]

5 hours ago, Gaiiden said:

Interesting. I always placed mine towards the ends of vehicles as far away from CoM to get the most torque possible. You know, as in swinging a hammer from the base instead of the head. If the torque wheels generate a force shouldn't you want it at the long end of your craft to have the most effect? Or am I thinking about this wrong?

What Pecan said.

It doesn't matter where you put the reaction wheels, the result is the same.  Here's why that seems counterintuitive, i.e. why your brain wants to think that it matters:

Let's take your hammer example.  Very heavy head, long skinny light handle.  If you hold it at the handle end and try to waggle it back and forth rapidly by rotating your wrist, it's hard, right?  Because you're swinging that massive head back and forth.  But if you hold it by the head (so that you're waggling the handle), it's very easy and you can waggle it back and forth very rapidly.  So clearly it matters where your wrist is applying the torque, right?

Here's why the analogy doesn't hold:  you're not thinking about the fact that you're changing the center of rotation..  In scenario A, the hammer is rotating around the he handle end; in scenario B, the hammer is rotating around the head end.

"But of course the center of rotation changes," I hear you thinking.  "The center of rotation will be wherever I apply the torque."

Wrong.  That's what trips up our human brains.

When you're holding the hammer and waggling it, you're not just applying torque.  You're also applying an external force.  Your wrist is attached to you, and you mass a lot more than the hammer does.  For that matter, you're standing with your feet securely planted on the ground, which holds you stationary with respect to the Earth, so in effect the hammer is anchored to the Earth.  And the Earth masses a whole bunch more than the hammer.  So you're using your arm muscles (and the leverage of your own and the Earth's mass) to keep the pivot point (where you're holding the hammer) stationary.  That's an entirely separate effect from the torque you're applying.

It's only natural to think that way, because our "intuition" is based on our everyday experience, and virtually all of our everyday experience works that way.  But a free-floating rocket in space, not attached to anything, is different.

If you have a free-floating rocket, and you apply internal torque to it (e.g. via reaction wheels, not by expending reaction mass such as RCS thrusters), it will always rotate around its center of mass.  Always.  It is physically impossible for anything else to happen, because of the law of conservation of momentum.

If you put the reaction wheel at the center of the rocket, the rocket will rotate around the center of mass.  If you stick the reaction wheel way out at one end of the rocket, the rocket will still rotate around the center of mass.

To help illustrate this, go back to the hammer analogy:  Yes, if you hold the handle end still and apply torque to waggle the head back and forth, it's hard.  But suppose you do this:  when you waggle the hammer, you also move your wrist back and forth in a semicircle, so that you keep the head stationary while you waggle the handle end (where you're holding it) back and forth.  Suddenly it gets a lot easier!

If you want to test this principle in KSP, it's pretty easy to do.  Build something like this and put it in orbit:

Turn off SAS, and disable the two reaction wheels on the ends so that only the center one is working.  Using a stopwatch, apply exactly 10 seconds of torque with the reaction wheel, then let go.  Now time how fast it's spinning (e.g. how long it takes to make a rotation).

Now do the same thing again, but disable the center reaction wheel and enable one of the wheels on the ends.  Do the same test.

Result?  Exactly the same.  It still spins at the same speed as a result of 10 seconds of torque, regardless of where that torque was applied.

Edited December 13, 2015 by Snark

[Jimbodiah]

If launching is an issue, use Vernier engines near the top of the rocket to keep it from toppeling, and/or stabilizing wings at the bottom. Reaction wheels work better for maneuvering in space.

[Snark]

3 hours ago, Pecan said:

Both explanations are right, and wrong, because the real situation is more complicated than we've shown as you can imagine.  The most important truth is that forces (translation or rotation) can always be pictured as acting on the CoM, as Snark explained.  If a rotation is not applied at/around the CoM itself, however, it will also introduce translation as a side-effect. 

...

• SAS twist off-centre = rocket turns around CoM and runs away (although given the strength of reaction wheels/RCS it's more of an amble)

Side note:  Be very careful about your terminology-- "rotation" is not a thing that one can apply.  Torque, not rotation.  (Normally I wouldn't be so nit-picky and pedantic about nomenclature, but in this case, confusion over what exactly one is applying is a major part of the confusion.  People hear"rotation" and tend to think in terms of things pivoting around an axle, and axles apply both torque and translation force.)  I'll just assume you mean "torque" where you say "rotation".

Unless I'm misinterpreting what you mean here:  this is incorrect, specifically the statements that I've bolded in your quote for emphasis.

Applying a torque, with no external forces, to a free-floating object will never induce any translation whatsoever in the center of mass.  Not one millimeter.  Doesn't matter how big the torque is, doesn't matter where on the body the torque is applied.

If you have a rocket floating in space, and it's using only reaction wheels (no RCS thrusters) for torque:  the reaction wheels will not budge the rocket's CoM one millimeter no matter where they are or what they do.  They can rotate the rocket around its CoM, but they can't move the CoM.  At all.

This is easy to demonstrate, using the experimental setup I pictured in my earlier post.  Before running the experiment, park another ship right next to the experimental rig, with zero relative velocity.  You will note that when you spin the rig with its reaction wheels, it rotates around its CoM and does not move relative to the ship parked along side-- and that holds true regardless of whether you're using the reaction wheel in the center or at the end of the experimental rig.

(Apologies if I've misinterpreted your meaning.)

Edited December 13, 2015 by Snark

[Plusck]

3 hours ago, Snark said:

Applying a torque, with no external forces, to a free-floating object will never induce any translation whatsoever in the center of mass.  Not one millimeter.  Doesn't matter how big the torque is, doesn't matter where on the body the torque is applied.

Although not wholly relevant, there have been a number of self-trained genius inventors who have claimed to have invented a torque-producing machine that does induce a translation in the centre of mass. And I think the whole idea of spinning flying saucers came from this idea, too.

[Drew Kerman]

@Snark @Pecan thx both for the clarifications. I'm pretty sure all mental faculties left me this morning and I did in fact equate reaction wheels with RCS thrusters. If you want to use RCS thrust to make a craft rotate, those are what you would want to place as far away from the point of rotation as possible right?

I'm going to go write "reaction wheels != RCS thrusters" on my whiteboard 100 times now

[Snark]

12 minutes ago, Plusck said:

Although not wholly relevant, there have been a number of self-trained genius inventors who have claimed to have invented a torque-producing machine that does induce a translation in the centre of mass. And I think the whole idea of spinning flying saucers came from this idea, too.

Yup.  Emphasis on the word "claimed." 

If you could actually do that, then you'd have Star Trek technology and all our problems are solved.

4 minutes ago, Gaiiden said:

@Snark @Pecan thx both for the clarifications. I'm pretty sure all mental faculties left me this morning and I did in fact equate reaction wheels with RCS thrusters. If you want to use RCS thrust to make a craft rotate, those are what you would want to place as far away from the point of rotation as possible right?

I'm going to go write "reaction wheels != RCS thrusters" on my whiteboard 100 times now

Yes.  If you want RCS thrusters as a rotation aid, then far from CoM is best, since that gives them the biggest lever arm to work with and they'll therefore generate more torque.

Tangentially related:  if you're using RCS thrusters for docking, be sure to turn on fine control mode, it's cooler than you realize.

[Plusck]

15 hours ago, Snark said:

Yup.  Emphasis on the word "claimed." 

If you could actually do that, then you'd have Star Trek technology and all our problems are solved.

I've been vaguely interested in this since the 1980s. This guy managed to get a number of people interested who probably should have known better. He was paranoid about people "stealing" his idea, so his invention was always a "black box", but I'm pretty sure he was merely exploiting angular momentum somehow, and any actual translation effect would therefore be either:

(a) false, and derived from a tiny force applied to whatever was holding the box up, or

(b) virtual, meaning that the process of spinning up his device did actually allow the box to generate an apparent force, but spinning it down would necessarily produce the exact opposite apparent force. Looking at it another way, this would mean that his box was using angular momentum to create a "virtual" CoM which was slightly offset from the "real" CoM, but only for as long as whatever wheels in there were actively being spun up. Once maximum rotation was achieved, of course, the system would be neutral, and friction in the wheels would necessarily bring the whole system back to where it started.

I strongly suspect that the real answer is (a). It would certainly be interesting if it were (b), and would excuse those University people for being duped, but its practical applications would be virtually zero.

 

edit: Apologies for going off-topic on this, however I do think that this invention (or non-invention, if you prefer) could help in intuitively grasping what reaction wheels are doing, and what they can or cannot do.

And another fascinating example of how to use reaction wheels to move around if you have the Earth to push against: The MIT's cubic robots

Edited December 14, 2015 by Plusck

[pincushionman]

If you have a rigid body, pure moment ("torque") can be applied anywhere and it makes no difference.

As others have said, however, we don't usually have rigid bodies. We usually have long, flexible ("bendy") structures where it might make sense to distribute moment along the length. This would reduce the stress between parts, but it shouldn't change the global response (rotational acceleration of the craft as a whole).

One thing where location does make a difference, however, is the "control from here" point. You want this close to the CoM if you can. Because SAS measures the rotational acceleration at that location, and at the ends of long bendy structures, it not only inherits the CoM's rotational acceleration, it also has it's own accelerations relative to it. These can either add to or subtract from the global rotational acceleration, depending on what exactly is happening.

In extreme cases…well, look at Pecan's last image:

On December 13, 2015 at 8:43:28 AM, Pecan said:

…

…

…The rotational acceleration at the end may well be in the opposite direction of the vessel as a whole. And then everything goes to hell and SAS shakes your ship apart.