Greater initial burn accuracy IMFD with D3D9 client?

[Zatnikitelman]

So I decided to pick up Orbiter again and actually fly it after years stuck in a development cycle. I thought that the first thing that might be interesting, would be to fly to the moon, using the DG until I got my bearings again. This time however, I decided to use the D3D9 client which gives massively higher framerates on my system (higher fps == higher accuracy of course). The first time I flew, I barely had to make any midcourse corrections. Thinking that was odd, I checked my parameters after the flight and sure enough, accidentally left nonspherical, radiation pressure, and gravity-gradient torque off. The next flight I flew, I made sure those were on...and only had to make a couple of real small corrections, and even then my lunar periapsis was only off by about 5km. So I downloaded D3D9 3R tonight, and flew a slightly different mission, but still, I only had to tap the thrusters once or twice throughout the entire mission before making my orbital insertion.

Now, I have done only two things differently between this, and other flights where I had made three or more MCCs at various stages. I set IMFD's map program to use the RK7 integrator, and I used D3D9. The only obvious differences aside from the lack of having to make MCCs, has been the massively increased framerates. Using the inline graphics, I got 50s-70s, but under D3D9, while IMFD was running, framerate remained in the mid-200s the entire trip. Have the higher framerates really had that much effect on the accuracy of executing IMFD's burn solutions? Or is there something possibly with settings in Orbiter itself that I'm just missing?

 

[Tommy]

IMFD automatically assumes non-spherical gravity sources, so that may make some slight difference if you don't have it enabled in the launchpad.

As to the effect of framerate, that likely depends on the burn guidance mode. If you are using "Off-Axis" it shouldn't matter much as the burn is calculated before the burn is made (and "locked" at t-3 seconds). If you are using "Realtime", the calculations are updated during the burn and framerate MAY have an effect (Jormo would know for sure if the calcs are made during the frame interrupt or by simstep.

In either case, the AP that handles ship orientation is dependent on framerate, so it will have a very small effect on accuracy of the direction. However, I suspect that the main difference is in the way you refined the set-up - and the use of the different solver.

 

[Cras]

I am for what ever interested very intrigued by this.

So to sum up, you think that using this alternate solver leads to better predictions on the burns?

Either way, I am off to the documentations to read up on this, another part of IMFD that my knowledge is not where I would like it to be

 

[Tommy]

Which solver works best will depend very much on the particulars of the transfer. To say that one works better than another in general would be incorrect.

In this particular case, I suspect the framerate is more significant since TLI's must be made in "realtime" mode where the direction of the burn vector is constantly being recalculated - so the ability of IMFD to keep the nose pointing in the right direction is dependent on it.

Also, I've noticed that there is a difference in accuracy that depends on the latitude of the TLI burn - with burns nearer the equator being more accurate - so the date of the trip matters. Also, the Sun will affect your trajectory, so accuracy will vary depending on the Moon-Earth alignment relative to the Sun (Target Intercept uses two-body solvers).

As well, the more carefully you refine the TEj and TIn the more accurate the transfer will be.

For what it's worth, I perform my TLI's with the Delta-V program (with Map program) and often don't need any MCC's at all to get a PeA within +/- 5km of target.

 

[Zatnikitelman]

Hrm, some interesting thoughts here, and that's exactly what I wanted when I started this topic! Actually, I've been using Off-Axis burn mode becuase I thought it was more appropriate for long duration burns. But I also thought that changing the integrator would only change the Map's estimation of the trajectory, which I could see being a factor when using offset mode and using the values in Map to tune the approach.
But it sounds like some empirical data is needed. I guess tomorrow I'm going to be trying out the same IMFD flight plan, all perturbations enabled, with the inline graphics, with the D3D7 external client, and with the D3D9 client. When flying, I'll have to make sure I engage and disengage time warp at the same times, but I'll record how many MCCs I make, and how strong they are.

 

[Tommy]

I'm not online at home, so I can't reference my Orbiter while I'm at my sister's. As far as the integrator goes - where are you selecting the different integrator? Is it on the main config page, or on Map's config. If it's the latter, then it only affects Map. If it's on the main config, I believe it affects all course programs.

To gain the best data, be sure you start from the same scenario - with the vessel already in it's parking orbit, etc. You could probably even have a scenario with IMFD all set up, just 4 minutes to the TLI, to make sure it's not a difference in set-up causing the discrepancy.