Orbit Wobbles

[sal_vager]

I'd say it's because of the jump when coming off rails, though 5x is the minimum we can use on our active vessel in KSP, I have an orbit here that varies by a few hundred meters while "active", stops wobbling completely at 5x warp then leaps by 0 to 20 kilometers when I come off warp.

1x warp doesn't look to be any better (tested by switching to a distant vessel and back, comparing periapsis before and after), the jumps are much larger than the wobble on an active orbit, that'd happen every time you touched WASD (or your joystick sent a signal because of too small a deadzone). Edited November 18, 2015 by sal_vager

[Stupid Waldo]

As long as it wobbles back and forth, it isn't so bad, as it'll stay to a relative area. With some of my orbits, it only appears that they're travelling in a single direction. With my Minmus and my Mun intercept orbits, I'm only seeing apoapsis increasing, not decreasing. Also I think it's rather odd that they can calculate an accurate orbit at one point, but don't stick with that orbit. Wouldn't it be better to scrap to constant recalculation, since the original orbit path took into account basically every factor?

[sal_vager]

Almost every factor, when you're on rails your ship is just a part of the rail, the position of the rail, ie: its direction, magnitude (distance from the center) and facing, are relative to the parent planet, moon etc, the ships position is a product of elapsed time on the rail, there's no physics taking place, there's no interaction.

When you're off rails your ships direction, magnitude and facing are relative to the center, the rail orbit is gone, and the orbit you're on has to be determined from your speed, direction etc, and the lines drawn accordingly.

Two very different ways to handle orbits, but you have to have the second to handle the player controlling the vessel, and you need the second (you need to run the physics engine to know where you are going) to make the first.




Not going to post for a bit, sleep has to occur sooner or later...

[Red Iron Crown]

[quote name='Stupid Waldo']As long as it wobbles back and forth, it isn't so bad, as it'll stay to a relative area. With some of my orbits, it only appears that they're travelling in a single direction. With my Minmus and my Mun intercept orbits, I'm only seeing apoapsis increasing, not decreasing. Also I think it's rather odd that they can calculate an accurate orbit at one point, but don't stick with that orbit. Wouldn't it be better to scrap to constant recalculation, since the original orbit path took into account basically every factor?[/QUOTE]
When off timewarp, the vessel isn't following a precomputed path like it is when using on-rails timewarp. Instead, gravity is being applied to all the separate connected masses that make up the vessel (and other forces too, like thrust or drag) and the position and velocity is calculated for the next frame. Then the game takes that position and velocity to compute and display the new orbital path. Every frame, so that changes due to burns show up immediately.

I've found the best workaround for the imprecision issue is to spend most time in orbit at 5x warp, only coming down to 1x time to turn and burn.

[Claw]

[quote name='Stupid Waldo']I might have exaggerated a little, but as you can see, over 30 seconds the apoapsis of the orbit increases by about 20km.[/QUOTE]

Occasionally I see this. While there is some fluctuation in the orbit, and I see a lot of flashing intercepts (which is often the floating point and calculation errors), this particular one is usually a bit different (when the AP goes racing off). Check your debug log (ALT+F2) and look for lines that say something like "dT is NaN". When I see the orbits racing off this fast, that message usually shows up in the debug log.

I'm not sure what causes it. The few times that I've gone digging, it was another craft in the file that had bad ORBIT parameters (one of them was NaN). I had to terminate the flight or manually edit the Orbit parameters to fix it.