Orbital Accuracy at acceleration

[HiPotOk1978]

I guess this question is best aimed at Mr Martins, but anyone that knows the answer, please comment

some of you may know that I simulated the ReFuel companies operations. One of these important on-going missions is for the LEO station called EOP
http://www.orbiter-forum.com/gallery/showgallery.php?cat=559

I have wanted to simulate the station running so that the random events that would affect the orbit of the station would appear over time including nodal regression

Does anyone know how accurate this is at 10x vs 1x?

 

[simonpro]

It depends entirely on the speed of your computer. The faster your computer the more accurate it will be, so check the framerate. Also check the vessel information to see if 'orbital stability' (or whatever it's called) becomes active. That reduces accuracy.
10x shouldn't have any problems though, I can run it fine up to 1000x on my rather slow laptop.

 

[TSPenguin]

Let me quote myself here.

Orbiter works in time steps. The slower your processor the more time each time step has to cover.
When you use time acceleration the same applies. If you press the any thruster key for a tenth of a second then orbiter executes this over (let's say) 50 time steps. At 10x it does this for 5 time steps. At 100x the key press lasts for orbiter 0.5 time steps, but orbiter will only see that during that time step your key was pressed and executes the command for one time step and the burn lasts double that long. At 1000x this means that you burn your engines for 2 seconds instead of 0.1 second.

The same applies for a floating/rotating station.

 

[simonpro]

That's not really answering the question. He's not talking about user inputs, he's talking about raw accuracy of the physics subsystem in calculating the accelerations on an orbiting object at high time accelerations.

 

[tblaxland]

^ What they said. Also, dynamics.pdf in Orbiter\Doc\Technotes is worth a read.

I have wanted to simulate the station running so that the random events that would affect the orbit of the station would appear over time including nodal regression

Nodal regression is not random

 

[TSPenguin]

That's not really answering the question. He's not talking about user inputs, he's talking about raw accuracy of the physics subsystem in calculating the accelerations on an orbiting object at high time accelerations.

Indeed it is not, but it explains the time step system quite well and I am a lazy typer today.

 

[tblaxland]

He's not talking about user inputs, he's talking about raw accuracy of the physics subsystem in calculating the accelerations on an orbiting object at high time accelerations.

More specifically, the question is about the accuracy of the numerical integration over large time steps. I think that was the point TSP was trying to make, ie, time step size is roughly proportional to time acceleration factor.

BTW, large time steps in themselves are not the worst problem - worse are isolated long time steps because they are harder to correct for. So my advice would be to avoid large textures that may cause significant disk IO or make sure you have sufficient RAM to avoid it.

 

[DarkWanderer]

If you're annoyed by inaccuracies and have a decent CPU, you can increase accuracy by using higher-order methods.

For example, my parameters:
Angular state:
RK8 only, target substep 1.0 deg., max steps 180
Linear state:
RK6 up to 1sec, RK8 otherwise

Runs fine even on Celeron 3.00.

 

[Zatnikitelman]

What someone (read: me as soon as I figure it out) needs to do is begin comparing the various propagation methods in different combinations over different time steps.
http://orbiter-forum.com/showthread.php?t=6630
This week into next, my C# class will learn the basics of programming networking stuff and I figure even if I don't know everything, I should still be able to write something that can grab data off Orb:Connect and shove it into a csv file or something for calculation in a spreadsheet.