Why Does Anybody Use Engine Gimbal?

[Geschosskopf]

I have been playing KSP since 0.20.  I've been to all the planets and back more times than I can count.  And I have never once had a good experience with engine gimbal, so I always disable it.  I do this for perfectly logical and valid reasons, and with only 2 exceptions I cannot find any logical or valid counterarguments in favor of using gimbal.  And even those exceptions are minor special cases that can easily be handled without using gimbal.  Thus, I am very curious as to why so many folks seem to think gimbal is a good thing, and why so many engines have it when it does so many bad things.

Here is why I find engine gimbal to be not only useless but counterproductive in the vast majority of cases:

First off, let's be clear on what gimbal is.  It is the rotation of the engine nozzle off the centerline of the ship.  As a result, gimbal creates off-axis thrust, which necessarily induces a rotational moment in the ship while the ship is already under thrust from the same engine.  This causes the ship's trajectory to curve.  Because of this, the game uses gimbal as a primary control mechanism accessible via user WASD and SAS, the same as it treats torque, aerodynamic control surfaces, and RCS thrusters.

The question then arises:  When is it desirable to curve your trajectory while you're under thrust?  There are only 2 situations that answer.

• Doing a gravity turn during launch
• Doing aerobatics or ACM in a highly maneuverable airplane

In ALL other cases, you want to your ship to be going in a straight line while under thrust.  This includes all orbital maneuvers, the ascent of SSTO spaceplanes, and airplanes trying to cover long distances.  Gimbal is counterproductive in these situations because SAS uses it like any other control system to handle the very minor course adjustments necessary to keep the ship pointed in the desired direction.  This results in wobble due to the frequent application of off-axis thrust, which will amplify and be amplified by any structural wobble in the ship.  While it's true that in 1.1.x you can now tune the amount of gimbal to reduce such wobbles, why even bother when you really don't want gimbal to begin with?  It's far easier to just disable it and have enough other control authority (torque, RCS, control surfaces) for SAS to do its job.  Besides, every time the engine gimbals, you lose fuel efficiency due to cosine loss, which is another reason not to use it.

So now let's look at the 2 special cases:   gravity turns and stunt-flying.  In both cases, gimbal can be substituted for by other control systems.  And in the case of gravity turns in particular, the excessive control authority of gimbal (at least without spending a lot of time tweaking it) is largely counterproductive again.  So that leaves gimbal desirable only for stuntplanes.

Now, if the game was smart enough to use gimbal as a TRIM mechanism instead of a primary control mechanism, my opinion would be very different.  Then gimbal would be great for asymmetrical ships because it could keep the thrust always going through the CoM.  But gimbal doesn't work that way so I absolutely hate it.

So tell me, do you like gimbal and why?

Edited May 16, 2016 by Geschosskopf

[Gaarst]

I like when I press a button and when that makes my rocket turn. I don't like using too much SAS (OP) and I usually don't want to bother with RCS or Verniers for the sake of part count (except when docking). So I use gimballed engines.

Try make a proper gravity turn with an Ares I design in KSP: no gimbal = rocket goes straight up.

And there are not that much losses due to gimbal; eg: a 10° gimbal still spits 98% of its thrust in the right direction.

[Snark]

Gimbal helps maintain stability during atmospheric ascent.  If you have a rocket whose aerodynamic stability is marginal, some engine gimbal can make all the difference.

And this is not just a newbie-who-can't-design-a-stable-rocket problem.  Even if I design my rocket to be overall stable, it may be that I have an upper stage that's marginal:  i.e. it activates at a speed & altitude such that aero stability is still a concern, but it doesn't have any fins on it (because those would hurt the overall aero stability of the craft during the early part of ascent).  Gimbal's a painless way to add stability while under thrust, while not hurting stability before staging.

It also can be useful for atmospheric reentry, again if there's a craft whose aero stability is marginal.  If I'm trying to maintain a retrograde attitude to let the engine take the brunt of the heating, and the craft's CoM placement makes that a marginal proposition, then having the engine gimbaled with a bit of retro-thrust can give SAS a lot more oomph to work with.

Those are my own main use cases.  I almost always tweak it down quite a bit even for atmospheric ascent, and usually lock it completely for operations in vacuum.

It's also useful for a lot of people in muscling the orientation to what they want it during a gravity turn.  True, a well-designed gravity turn shouldn't require that.  On the other hand, nailing the gravity turn "just so" takes a lot of practice to develop the skill, and in the early stages of a newbie's KSP experience, I can imagine that this would be very frustrating.  Some gimbal can smooth the way.  (I'm just speculating here-- can't speak from first-hand experience, since I've never been in the position of "newbie trying to deal with aero stability."  When I was a newbie, aero wasn't a thing yet, so I already had a lot of rocket-handling practice under my belt when aero arrived and I had to learn to handle it.)

 

[bewing]

27 minutes ago, Geschosskopf said:

Now, if the game was smart enough to use gimbal as a TRIM mechanism instead of a primary control mechanism, my opinion would be very different.  Then gimbal would be great for asymmetrical ships because it could keep the thrust always going through the CoM.  But gimbal doesn't work that way so I absolutely hate it.

So tell me, do you like gimbal and why?

I don't think I completely agree with this statement? I don't see a real difference between primary control and trim.

In general, I want my engines to push straight, yes. Especially since engines with gimbal weigh more. Which is one reason why I prefer nukes and ion engines.

However -- the top 3 reasons I use gimbal/thrust vectoring:

1) In the early game, many of my upper stage landers are short on reaction wheel torque. So when I am landing one of these ships on the Mun, for example -- if I maintain a small amount of thrust with a gimballed engine (a terrier usually) then I can actually steer the thing. If it were not for the gimballed engine, then it would take 30 seconds to accurately rotate the thing 90 degrees, and it would be a funny looking crater and not a ship.

2) Making upper atmosphere turns in jets. When I'm flying along at mach 2.5, at 20km altitude in my panther-based jet, and I fly through my waypoint and complete one item of my contract checklist -- then I have to fly through the next waypoint. Which means I have to turn sometimes. I don't know how often you try to turn a supersonic jet in the upper atmosphere, but it's really hard. The plane wants to go straight, and you have very little control authority without the thrust vectoring.

3) I find myself often trying to launch rockets that are slightly unstable at certain airspeeds during ascent. I can't imagine that I'm the only one -- especially with all the threads around with people trying to get help with their rockets flipping. I can either try to add some fast-acting canard-style positive control at the front, or I can try to use some passive or active aerodynamic surfaces at the rear ... or I can launch with a gimballed liquid engine for a booster.

[Geschosskopf]

3 minutes ago, bewing said:

I don't think I completely agree with this statement? I don't see a real difference between primary control and trim.

Trim is entirely different from primary control, even though it might use the same physical part (i.e., a control surface).  Primary control is a transient thing.  You turn to the desired heading and stop rotation there.  Trim is a constant thing for maintaining stability without having to make frequent adjustments using the primary controls.  Trim provides a constant, internal, off-axis force to counteract a constant, external off-axis force.

For example, you have an airplane that for whatever reason wants to yaw slightly to the left while in flight (say due to the drag of a ladder only on that side of the cockpit).  So you can put in a bit of right rudder trim to compensate for this.  The trim offsets the rudder slightly so that its tendency to yaw the plane to the right cancels out the plane's natural tendency to yaw to the left.  Then the plane flies straight without needing periodic kicks back onto the desired course.

So now consider a rocket with asymmetrical thrust, like a space shuttle.  The engines are mounted off-axis so have to be aimed off-axis to thrust through the CoM.  You can set it up that way in the VAB but only for 1 fuel state, because the CoM moves as fuel burns.  It would be great if the engines automatically gimbaled to keep pointing at the CoM as it moves.  That would be automatic trimming, which I"m pretty sure is how they worked in real life.  But gimbals in KSP don't work that way, so the shuttle develops off-axis thrust as fuel burns, and the player has to keep kicking it back on course.   These kicks will involved gimbal if you haven't disabled it, but the gimbal is a transient, primary control force here, not a trim thing.

 

3 minutes ago, bewing said:

1) In the early game, many of my upper stage landers are short on reaction wheel torque. So when I am landing one of these ships on the Mun, for example -- if I maintain a small amount of thrust with a gimballed engine (a terrier usually) then I can actually steer the thing. If it were not for the gimballed engine, then it would take 30 seconds to accurately rotate the thing 90 degrees, and it would be a funny looking crater and not a ship.

I find that 2 or 3 of the 0.625 reactions wheels are sufficient for this job.  Plus, stacking them atop the Mk1 pod provides a convenient surface to attach the small fins and radial parachutes needed to reenter and land safely with Material Bays attached below the pod.

 

3 minutes ago, bewing said:

2) Making upper atmosphere turns in jets. When I'm flying along at mach 2.5, at 20km altitude in my panther-based jet, and I fly through my waypoint and complete one item of my contract checklist -- then I have to fly through the next waypoint. Which means I have to turn sometimes. I don't know how often you try to turn a supersonic jet in the upper atmosphere, but it's really hard. The plane wants to go straight, and you have very little control authority without the thrust vectoring.

OK, I'll give you that one, although I consider that in the aerobatics / ACM category.  In any case, it's still turning an airplane instead of going in a straight line.

 

3 minutes ago, bewing said:

3) I find myself often trying to launch rockets that are slightly unstable at certain airspeeds during ascent. I can't imagine that I'm the only one -- especially with all the threads around with people trying to get help with their rockets flipping. I can either try to add some fast-acting canard-style positive control at the front, or I can try to use some passive or active aerodynamic surfaces at the rear ... or I can launch with a gimballed liquid engine for a booster.

Putting wings on the nose usually makes launch instability worse.  Rockets in launch configuration should be built like arrows:  weight at the front, tailfeathers at the back.  You want the CoM as far forward as possible and the CoL as far aft as possible.  This produces the most stability on launch.  Putting wings on the nose moves the CoL forward, reducing stability.  So instead, your non-torque control mechanisms all the way at the back.  This gives them the most leverage because the CoM is as far forward as possible, plus putting fins back there moves the CoL further aft as well.

[bewing]

9 minutes ago, Geschosskopf said:

Putting wings on the nose usually makes launch instability worse.

I totally disagree. You put wings on the nose for the same reason you put canards on the front of a plane. Positive control is more efficient. And anyway, if you put the control surfaces at the back end, then you lose them with the first stage.

Edited May 16, 2016 by bewing

[John JACK]

Landing. And matching speeds at rendezvous. Without excessive use of RCS fuel or OP/underP reaction wheels.

[Geschosskopf]

2 minutes ago, bewing said:

I totally disagree. You put wings on the nose for the same reason you put canards on the front of a plane. Positive control is more efficient.

Um, no   These applications have nothing in common at all 

Any control surface will generate a given amount of off-axis thrust based on its physical properties and the airspeed and air density at the time.  This force is then multiplied by the lever arm distance between its point of application and the ship's CoM.  The greater this distance, the more control authority a given control surface will have.  Because properly designed rockets have the CoM as far forward as possible, the lever arm is shortest at the nose and longest at the rear end of the rocket.  Therefore, control surfaces mounted as far back as possible will always have more control authority and be more effective than the same surfaces mounted on the nose.  Plus, putting fins at the rear is desirable for overall rocket stability because it moves the CoL aft, whereas putting fins on the nose decreases stability by moving the CoL forward.

The situation with airplanes is totally different.  Jet aircraft typically have their CoM more or less amidships, so control surfaces on either end of the plane have about the same amount of leverage.  Also, for stability, you want the CoL just slightly aft of the CoM instead of as far away from it as possible, but with delta wing designs the CoL is often too far aft.  Therefore, putting canards up front can move the CoL up close to the CoM where it should be.  Plus, if you don't have a huge wingspan, you might have run out of trailing edge space on the main wing to mount more control surfaces, so the canard can help there as well.

[wumpus]

39 minutes ago, bewing said:

I totally disagree. You put wings on the nose for the same reason you put canards on the front of a plane. Positive control is more efficient. And anyway, if you put the control surfaces at the back end, then you lose them with the first stage.

While "positive control" might technically be more efficient, in reality "positive feedback" is almost never stable.  The point of putting controls on the canards was a means to overcome some of KSP's earlier aerodynamic difficulties (the souposphere) and presumably was carried over to minimize the breakage of otherwise working spacecraft.

Don't try this in real life.  NASA/DoD might do this with a ton of computers, but stable negative feedback doesn't rely on bug-free software.

[Kerbart]

40 minutes ago, bewing said:

I totally disagree. You put wings on the nose for the same reason you put canards on the front of a plane. Positive control is more efficient. And anyway, if you put the control surfaces at the back end, then you lose them with the first stage.

More efficient is not always more desirable. There's something else positive about wings at the nose, and that is "positive feedback loop" which is generally a Bad Thing (generally; not everyone wants to stay in control of their vehicle when launching it).  Controls at the nose? Pitch up, and your control surfaces' angle of attack increases. So now they generate more lift, resulting in more torque, resulting in a steeper pitch, increased angle of attack, more lift, etc.

Fins behind the COM will have a stabilizing effect; fins (and surfaces) in front of the COM will result in an unstable vehicle. Nimble perhaps, but frightingly unstable.

[OhioBob]

2 hours ago, Geschosskopf said:

It's far easier to just disable it and have enough other control authority (torque, RCS, control surfaces) for SAS to do its job.

I use gimbaled engines during launch so that I don't have to have a lot of other control authority.

[Geschosskopf]

1 minute ago, OhioBob said:

I use gimbaled engines during launch so that I don't have to have a lot of other control authority.

Yeah, other than turning an airplane, that's the only other situation where I find it the least bit useful.  Do you use gimbal anywhere else?

[T-Bouw]

Well I have this probe-core without SAS whatsoever hanging in space right now that rather likes its gimballed engine.

Sure, it's trajectory is imprecise, but it's light and costs almost nothing.

[OhioBob]

31 minutes ago, Geschosskopf said:

Yeah, other than turning an airplane, that's the only other situation where I find it the least bit useful.  Do you use gimbal anywhere else?

Once I'm in space, then generally no.  On a couple rare occasions I put into orbit a very large and unwieldy interplanetary vessel that lacked adequate reaction wheel torque.  In that case the gimbaled engines helped to get and keep the vessel properly orientated for and during the ejection burn.  That and during launch are really the only times where I've found engine gimbaling to be really useful.  (I don't fly planes so I can comment about that.)

[A_name]

Despite your experiences, I find gimbal useful both during launch and for orbital maneuvers, essentially to have more control over my attitude. During launch, I agree that aerodynamic control surfaces could do the job, but sometimes you just need an extra bit of control, for instance in the upper atmosphere. For orbital maneuvers, it also helps when your craft doesn't have RCS or a bunch of reaction wheels.

[pincushionman]

Landers, especially large and unwieldy ones. If there's any time you absolutely must have bucketloads of control authority, and you need it now, a badly-thought-out suicide burn is it.

[FyunchClick]

4 hours ago, Geschosskopf said:

It would be great if the engines automatically gimbaled to keep pointing at the CoM as it moves.  That would be automatic trimming, which I"m pretty sure is how they worked in real life.  But gimbals in KSP don't work that way, so the shuttle develops off-axis thrust as fuel burns, and the player has to keep kicking it back on course.   These kicks will involved gimbal if you haven't disabled it, but the gimbal is a transient, primary control force here, not a trim thing.

Interesting, you mean like having a toggle option of some sort to cut the gimbals out of regular craft control entirely and just have it compensate thrust torque. I usually spend a lot of time with RCS build aid and KER tweaking my designs to have as little thrust torque as possible for any given fuel state; I'd love to give such an option a try.

[cantab]

6 hours ago, Geschosskopf said:

 It's far easier to just disable it and have enough other control authority (torque, RCS, control surfaces) for SAS to do its job.

All those control options add mass. Most engines already come with gimbals, why not use them?

When you get into large ships, engine gimbal can offer seriously powerful control. For instance a single Rhino might offer 700 kNm of torque by gimballing, equivalent to a stack of twenty three of the 2.5 m reaction wheels. And that reaction wheel stack would weigh 4.6 tonnes, half of the Rhino's own weight.

Maybe there are cases you don't want gimbal, but to discard that kind of power out of hand seems premature.

[Geschosskopf]

1 minute ago, cantab said:

All those control options add mass. Most engines already come with gimbals, why not use them?

Because the do nothing but produce off-line thrust, which is exactly what you don't want for 99.9% of all the times you'll be running an engine, both from an efficiency and a wobble standpoint.  Gimbals are occasionally useful for jet engines, provided those engines aren't on an SSTO trying to reach orbit.  Gimbals are occasionally useful on 1st stage lifter engines to help with the gravity turn.  They are counterproductive in all other cases.

 

1 minute ago, cantab said:

Maybe there are cases you don't want gimbal, but to discard that kind of power out of hand seems premature.

Becasse, apart from lifting some godawful Whackjovian thing with no pretensions of aerodynamic design, there's absolutely zero need and even less desire for such large amounts of off-axis thrust.  Even for rockets weighing a couple thousand tons on the pad, you've got plenty of control with Vernors and tailfins.

 

[bewing]

6 hours ago, wumpus said:

While "positive control" might technically be more efficient, in reality "positive feedback" is almost never stable.  The point of putting controls on the canards was a means to overcome some of KSP's earlier aerodynamic difficulties (the souposphere) and presumably was carried over to minimize the breakage of otherwise working spacecraft.

Don't try this in real life.  NASA/DoD might do this with a ton of computers, but stable negative feedback doesn't rely on bug-free software.

You guys can just go argue about it with Burt Rutan, you know. Yes, a canard/positive control design requires a good reliable redundant fly-by-wire system (powered by an intel 286 chip at least!), and they exist all over the place IRL.

 

7 hours ago, Geschosskopf said:

Because properly designed rockets have the CoM as far forward as possible, the lever arm is shortest at the nose and longest at the rear end of the rocket.

Wrong! On a 3 stage rocket, the CoM starts out someplace in stage 1. Once you stage, the CoM is somewhere in stage 2. By the time you stage that one away, you are above the atmosphere. So the best lever arm is always to have the control at the nose.

[cantab]

2 hours ago, Geschosskopf said:

Because the do nothing but produce off-line thrust, which is exactly what you don't want for 99.9% of all the times you'll be running an engine, both from an efficiency and a wobble standpoint.  Gimbals are occasionally useful for jet engines, provided those engines aren't on an SSTO trying to reach orbit.  Gimbals are occasionally useful on 1st stage lifter engines to help with the gravity turn.  They are counterproductive in all other cases.

 

Becasse, apart from lifting some godawful Whackjovian thing with no pretensions of aerodynamic design, there's absolutely zero need and even less desire for such large amounts of off-axis thrust.  Even for rockets weighing a couple thousand tons on the pad, you've got plenty of control with Vernors and tailfins.

 

But what's your Vernor doing? Using fuel to produce off-line thrust! What's your tailfin doing? Using aerodynamics to produce off-line drag!

In fact Vernor will almost always use a lot more fuel to produce a given torque than an engine gimbal. Suppose I want 10 kN of sideways force and 1000 kN of upwards force, then using a gimballed engine I need sqrt(10²+1000²) = 1000.05 kN of total thrust. But using a fixed engine and a Vernor I need 10+1000 = 1010 kN of total thrust. The Vernor is using two hundred times as much fuel on turning as the gimballing engine is.

Two. Hundred. Times.

And that's if the Vernors were as efficient as the main engines, which they aren't, and have the same distance from the CoM, which might not be the case but the Vernors sure aren't going to be 200 times further from the CoM than the main engines unless you painstakingly contrive that.

I don't see any advantage to using Vernors when a gimbal-capable main engine is running. They're useful if you need control when the main engines are off though.

Edited May 17, 2016 by cantab

[FullMetalMachinist]

1 hour ago, bewing said:

You guys can just go argue about it with Burt Rutan, you know.

I would say that the situations where canards are helpful on a rocket are few and far between. Here's the thing about Burt Rutan's canard designs. They're almost all lightweight, subsonic, and low flying. That doesn't mean you should take his design philosophy and apply it to supersonic space craft. (also notice how the one space ship he did design is noticeably lacking canards) 

That would be like saying big heavy wheels work great for dump trucks, so they must be great for a sports bike. 

[Geschosskopf]

1 hour ago, bewing said:

You guys can just go argue about it with Burt Rutan, you know. Yes, a canard/positive control design requires a good reliable redundant fly-by-wire system (powered by an intel 286 chip at least!), and they exist all over the place IRL.

Which again has nothing at all to do with a rocket

 

1 hour ago, bewing said:

Wrong! On a 3 stage rocket, the CoM starts out someplace in stage 1. Once you stage, the CoM is somewhere in stage 2. By the time you stage that one away, you are above the atmosphere. So the best lever arm is always to have the control at the nose.

Hmmm, really depends on the design.  But have it your way for the sake of argument.  All the while you're staging, you still retain any wings on the nose of the rocket.  So if they weren't making the CoL be ahead of the CoM to begin with, they probably will after a stage or 2.  At which point your rocket becomes highly unstable because it then has the exact opposite configuration required for stability.

 

43 minutes ago, cantab said:

But what's your Vernor doing? Using fuel to produce off-line thrust! What's your tailfin doing? Using aerodynamics to produce off-line drag!

You cannot turn without using an off-axis force of some sort.  One of Newton's laws says that, IIRC That's not the point.  The point is, Vernors and control surfaces can both provide small, carefully controlled, and smoothly continuous off-axis forces, more like aerodynamic trim than how KSP does primary controls.  This is way more efficient than the wobble-inducing, jerky,  full-stop slamming of the overly powerful main engine, which can't help but over-control and end up fighting itself slamming back the opposite way.  Each such massive kick tosses the whole rocket off the prograde line, resulting in drag, and stresses and bends all its joints, for even more drag and more gimbal-slamming.

Rockets using gimbal on ascent look like a sperm cell pushing a wet noodle (if you don't have KJR) or a sausage (if you do).  But a sperm cell at the rear end either way, wildly, spasmodically flailing it's tail in all directions..  That constant gimbal flailing in all directions is a sure sign it's over-correcting.  If it was not, then you'd just see an occasional twitch in only 1 direction as needed to bend the trajectory over.  Nowadays, we can reduce gimbal movement so that maybe, after much trial-and-error, we can tune a gimbal to behave as it should for a specific rocket, with a specific payload, on a specific ascent trajectory.  But if any of those variables changes, you have to do it again.  So even if you use standard lifters instead of building a custom one for each shot, you're still in for a huge amount of work trying to make gimbals behave properly.  And even then, that's just for the 1st stage or 2.  After that, you have no further use for gimbal on any other engine.

 

43 minutes ago, cantab said:

In fact Vernor will almost always use a lot more fuel to produce a given torque than an engine gimbal. Suppose I want 10 kN of sideways force and 1000 kN of upwards force, then using a gimballed engine I need sqrt(10²+1000²) = 1000.05 kN of total thrust. But using a fixed engine and a Vernor I need 10+1000 = 1010 kN of total thrust. The Vernor is using two hundred times as much fuel on turning as the gimballing engine is.

I don't agree with your math..  First, the total thrust of the gimbal engine is always along the hypotenuse and it can't ever produce more than its max thrust.  Thus, forwards thrust is reduced due to the sideways thrust.

Second, you're assuming the gimbal provides EXACTLY the desired amount of sideways force at EXACTLY the right time.  This never, ever happens.  The gimbal is all-or-nothing and the angle it can move is always (unless you take the time to tune it) considerably more than you really need at the time.  And the powerful lifter engine is running at full thrust.  The result is that you get WAY more sideways force than you want or need, causing the over-control oscillations and tail-wagging typical of gimbal-using rockets.  This reduces forward thrust to the frequent off-line thrusting,and increases drag considerably due to the rocket frequently being off the prograde heading in the oscillations and wobbles.  The result is needing a fair amount more fuel than not using gimbal.

Third, have you ever even used Vernors?  They really consume utterly negligible amounts of fuel even on rockets weighing a over a thousand tons.  I'm talking maybe 5m/s worth in total for an SSTO lifter that big from the pad to a 100km orbit.  And the vast bulk of that is spent orienting the monster for the circularization burn at Ap, not during the gravity turn.

So the thing is, even for really big rockets, you only need a tiny amount of force, carefully applied, to do a gravity turn.  Certainly it's way less than you get from the gimbal of a powerful 1st stage engine at full thrust.  And Vernors give you that, in a smooth, carefully controlled way so you avoid the wobbles and shakes of using gimbal, and thereby stay more on the prograde line so have less drag.  Either way, you need less fuel keeping the main engine pointing straight ahead than you do by letting it rattle around.

 

43 minutes ago, cantab said:

I don't see any advantage to using Vernors when a gimbal-capable main engine is running. They're useful if you need control when the main engines are off though.

Well, you should give them a try.  I myself was very leery of them at first.  I tend to design lifters with tight margins and was afraid they'd eat into my main engine fuel.  So it was the better part of a year before I actually tried them, but WOW!  They're great.

[Superfluous J]

Aha. So your argument isn't that gimbaling engines are inherently bad, but that Kerbal Space Program's IMPLEMENTAITON of them is bad.

I'm on line with that.

The two main problems with gimbaling engines is that they're far from your command pod and that they are all-or-nothing. Those two things together create the wobble you are complaining about. If you put a probe core way down at the bottom of your ship and control from there, it's not even remotely as bad.

[FullMetalMachinist]

5 minutes ago, Geschosskopf said:

Rockets using gimbal on ascent look like a sperm cell pushing a wet noodle (if you don't have KJR) or a sausage (if you do).  But a sperm cell at the rear end either way, wildly, spasmodically flailing it's tail in all directions..  That constant gimbal flailing in all directions is a sure sign it's over-correcting.  If it was not, then you'd just see an occasional twitch in only 1 direction as needed to bend the trajectory over.  Nowadays, we can reduce gimbal movement so that maybe, after much trial-and-error, we can tune a gimbal to behave as it should for a specific rocket, with a specific payload, on a specific ascent trajectory.  But if any of those variables changes, you have to do it again.  So even if you use standard lifters instead of building a custom one for each shot, you're still in for a huge amount of work trying to make gimbals behave properly.  And even then, that's just for the 1st stage or 2.

I don't think it's necessarily the fact that the engine gimbals that causes these issues. Don't get me wrong, I agree that there's a problem. I just don't think it's solely the gimbal that is the problem. It just looks that way because engine gimbal is so powerful. But if you take a craft that has loads of aero control surfaces (but is still just barely stable), the stock SAS will still wobble all over the place. 

I think the real root cause is that the PID controller for SAS is very badly tuned. Sure, engine gimbal makes it stand out more due to the large forces and the "instant" nature of its motion, but it's not the root problem. Stock SAS is. 

13 minutes ago, Geschosskopf said:

After that, you have no further use for gimbal on any other engine.

Agreed, once on orbit, engine gimbal is largely useless.