Rails at 1x time warp

[Vector]

Suggestion: when outside the atmosphere and all thrusters are off, put the center of mass on rails even at 1x time warp and even when rotating via reaction wheels

Observations:

1. Orbits at 1x time-warp "vibrate" even when the ship is completely passive.

2. Orbits appear to be extrapolated from instantaneous control module velocity instead of center of mass. Rotating a ship introduces errors in the extrapolated orbit.

3. Orbits change when changing time-warp between 1x and higher time-warp.

[Person012345]

This is a bad solution imo. The fact that orbits vibrate when they shouldn't means there's a problem in the simulation that should be fixed I suppose (they didn't used to behave this way, orbits used to be stable) rather than a ham fisted forced "solution" that merely covers up the problem. Of course I don't know exactly what is causing it so I don't know how easy or hard that might be.

[Vector]

I don't see this as covering up a problem, I see this as a more accurate model.

If I have a linear function y = mx + b, and I increment each delta-x step using y = y + m * dx, then it's mathematically equivalent to the function but numerically the result is different. The errors accumulate when using a large number of small steps. In contrast, evaluating the function at different x accurately follows the function and roundoff does not accumulate. This is the difference between mathematics and numerical analysis.

I can't say for sure but it really feels like this is what is happening, where physics mode is introducing a large number of small errors. This is inherently less accurate than a closed-form solution for the trajectory of the center of mass.

When accelerating due to thrust or atmosphere, then sure, use incremental delta-v and compute each delta-position since you have to calculate each step anyway. But when your orbit is supposed to be a patched-conic, use the closed-form patched conic.

[Idobox]

I think what happens is that your ship wobbles, and it cause small movements of the center of mass. If you are around kerbin, calculating an intercept with a far away planet, the small speed variation can cause large errors.

So, with no thrust, the CoM shouldn't move normally, but you still get rounding errors. The solution you propose is not shocking, but shouldn't be used when RCS are active, as they can change the orbit.

[Vector]

I think what happens is that your ship wobbles, and it cause small movements of the center of mass. If you are around kerbin, calculating an intercept with a far away planet, the small speed variation can cause large errors.

So, with no thrust, the CoM shouldn't move normally, but you still get rounding errors. The solution you propose is not shocking, but shouldn't be used when RCS are active, as they can change the orbit.

I agree, except that it's not the center of mass that's moving, it's the fact that the center of control is extrapolated to compute the orbit, not the center of mass. This is one of the things I think is wrong about the model. If I put an Okto-2 on a long arm, rotating (using reaction wheels only) perturbs the orbit much more than when the control is near the center of mass, and in a predictable fashion. This tells me that it is not rounding error, it's the center of control vs. center of mass. There may be rounding error in addition. I agree that RCS can create unwanted but real translation forces, and that these should be allowed to modify the orbit. I am suggesting rails only for rotation from reaction wheels.