MSFS flight model

[blingthinger 649]

7 minutes ago, rka said:

Seems quite obvious IMHO. Remember Mr. Meyers flight model explanations in those videos? If I were an XP advocate, I would be rather careful when trying to point out who has better public FM documentation or who has to catch up in that regard. 

Oh yes. I wish we could see XP's guts too. Kudos to Asobo for being far more brave in that regard.

[abrams_tank 7,426]

4 hours ago, blingthinger said:

That was funny. Literally laughing here.

This just might mean Asobo has done something unique that could be recognized by the aerodynamic community (you know, where Boeing and Airbus engineers hang out). It would be a stamp of proof that they are pushing boundaries. That's all. But I suppose you're right in the end. Though I don't comprehend what you're worried about. Seb and Co. could take the time to write a paper and you wouldn't notice the lost time. Nor would their market share suffer a bit. Though obviously the folks in these halls would get upset.

.

How can Asobo push the boundary of aerodynamic simulation on a desktop PC?  The boundary of aerodynamic simulation is done on much more expensive and faster computers, likely at places like NASA.  Asobo did a very good job, given the constraints of a desktop PC that the average consumer can afford, but why would they go publishing their algorithm in an aeronautics journal when it's likely inferior to whatever aerodynamics algorithm that NASA is running but NASA is probably using computers that are that much more expensive and faster than your average desktop PC and aren't facing the same constraints as Asobo?  

And yes, we would notice the time being wasted if Seb went to write a paper for an aeronautics journal.  He spends much of his full time job on MSFS already, and there are still complaints that stuff isn't being fixed fast enough and there are too many backlogged tasks.

 

Edited September 11, 2022 by abrams_tank

[NZ255 309]

Sometimes having to solve with constraints (like real time for games) means breakthroughs are made. I’m not saying this is the case here, I have no idea. 

[Glenn Fitzpatrick 3,798]

There is a lot of good CFD stuff about.

You can do quite good fluid flow simulation, including turbulent flow, laminar flow and boundary layers in Solid Works.  Just might take you a few days to generate one set of numbers.

There are others like Siemens FLOEFD.

The issue is not finding good CFD software, the issue is finding a way to do it in real time without making too many compromises.

[blingthinger 649]

4 hours ago, abrams_tank said:

How can Asobo push the boundary of aerodynamic simulation on a desktop PC?

It's not all about computational power but I can see how you'd think that given the discussions on CFD. Advancing real-time flight simulation capabilities could very easily happen on a PC. For example, what if there was some need to have it run on a low power embedded system? Think Arm CPUs. An application may not care about the visual graphics either.  

No you wouldn't notice. In this case Seb wouldn't be the only author but even if he did pull a jerk move like that, he's still going to be using an entire team who would contribute bits and pieces to the content. Journal articles aren't books. They're generally on the order of 10-pages long. Sometimes more, sometimes less. Not to be confused with a graduate thesis. Maybe they'll do it in a few years after more dust has settled. 

As for "pushing boundaries", I'm just quoting FS2020 fans. Good to know where you stand on the matter.

And as for a small team not making progress fast enough to feed the ravenous wants and imaginations of overly-entitled customers... mmm, no. I won't comment further on that one...

Edited September 11, 2022 by blingthinger

[abrams_tank 7,426]

6 minutes ago, blingthinger said:

As for "pushing boundaries", I'm just quoting FS2020 fans. Good to know where you stand on the matter.

For the average desktop PC.  Yes, MSFS is definitely pushing boundaries for the average desktop PC.

Not for the expensive computers they use at NASA.

Edited September 11, 2022 by abrams_tank

[blingthinger 649]

Just now, abrams_tank said:

For the average desktop PC.  Yes, MSFS is definitely pushing boundaries for the average desktop PC.

Whew! Nice recovery there.

[jcomm 5,939]

Let's wait for the upcoming ( in 10 yrs ? ) quantum-based desktops  🙂

Meanwhile both MFS / ASOBO and LR / Austin are doing their best to bring to nowadays desktops / laptops / notebooks / game consoles and even smartphones the best that can be expected when using non table-based-only flight models.

I use yet another flight simulator - IL2 Great Battles - that IMO can go even further ahead in terms of accuracy and feel of flight, because it's players are mostly WW2 air combact "experts" and some know from their memmory the various performance figures of their preferred fighters, and indeed 1C / 777 has nailed it in pretty much every aircraft model available in their air combat sim, while using a BeT / CFD approach as well, as far as we can infer from available information about their flight dynamics model.

Edited September 11, 2022 by jcomm

[lwt1971 7,420]

@mrueedi thanks for some great insights and explanations on this thread! Are you an aircraft dev by any chance? I remember seeing that damping graph, was/is that in any current MSFS sdk docs or elswhere? In any case, hope you continue to contribute on these topics to bring more clarity to all the MSFS community. As I hope Matt Nischan, Seb et al continue to do as well.

As for MS/Asobo revealing more about how their aerodynamics/physics engine works under the hood, I'm sure their already well documented SDK will continue to get beefed up:
fde physics: https://docs.flightsimulator.com/flighting/html/Samples_And_Tutorials/Primers/Flight_Model_Physics.htm
fde: https://docs.flightsimulator.com/flighting/html/Developer_Mode/Aircraft_Editor/Tabs/The_Flight_Model_Tab.htm
flight model config overall: https://docs.flightsimulator.com/flighting/html/Content_Configuration/SimObjects/Aircraft_SimO/flight_model_cfg.htm
flight model config aerodynamics parameters: https://docs.flightsimulator.com/flighting/html/Content_Configuration/SimObjects/Aircraft_SimO/flight_model/aerodynamics.htm
defining a flight model: https://docs.flightsimulator.com/flighting/html/Samples_And_Tutorials/Tutorials/Defining_A_Flight_Model.htm
CFD: https://docs.flightsimulator.com/flighting/html/Developer_Mode/Aircraft_Editor/Debug/Debug_Aircraft_CFD.htm
Stall & other debug/visualization tools: https://docs.flightsimulator.com/flighting/html/Developer_Mode/Aircraft_Editor/Debug/Aircraft_Debug_Menu.htm

I'm hoping their Feature Discovery Series 15 and 16 planned for Oct 20th and 27th will dive deeper into the new aerodynamics improvements due to adding helicopter and glider support as well as the CFD driven atmospheric airflow enhancements within 20km cocoon around aircraft. And finally the ground handling/physics fixes and improvements can't come fast enough so hopefully they'll give more examples of how to use the new parameters introduced in SU10 and then further work hopefully in SU11 and beyond.
 

Edited September 11, 2022 by lwt1971

[mrueedi 702]

 

20 hours ago, blingthinger said:

By the way, has Asobo published any papers on this surface model + CFD in any aerospace journals or conferences of the AIAA or ASME? 

Not that I am aware. I fear, the SDK page is all we have.

3 hours ago, lwt1971 said:

Are you an aircraft dev by any chance?

Only hobby wise, I programmed an ArduPilot based drone groundstation which offers all autopilot modes a real aircraft offers (including VNAV and LNAV), and this software I am currently also transforming for use in manned aviation as a low cost EFIS for ultralight or experimental aircraft.

 

[blingthinger 649]

9 hours ago, lwt1971 said:

CFD: https://docs.flightsimulator.com/flighting/html/Developer_Mode/Aircraft_Editor/Debug/Debug_Aircraft_CFD.htm

@mrueedi

This is the page showing the default CFD spatial resolution. I couldn't remember the URL before. They're using upwinding which is last-gen but sufficient for their current low-fidelity usage. The CFD solution runs very quickly, but they do warn of increasing from the default 20x20x20. It gets expensive quickly! 

Also, there's a pic in there of a 3D visual model. The air is flowing through the fuselage in various places. This means they are not using the visual model in the CFD. The only other geometry definition I'm aware of in the sim are the flat hershey-bar wings and fuselage used for the force coefficient elements.

At the default 20³ resolution, it wouldn't make sense to use the visual model. The default air volume extends 25% beyond the maximum and minimum extents of the airframe in each direction. For the length of the fuselage for example, that would leave 13-14 cells for the airframe out of the 20 along that axis. The CFD equations will treat the surfaces exactly as you could imagine them: the resulting fuselage model would be more blocky than the F117! Obviously this would be very bad for aerodynamics if you're trying to simulate the PC12 in that picture. Better off just using the flat plates and applying all the fudge factors and corrections (they even mention turning some of the fuselage elements OFF if needed because it's a massive flat plate, not a cylinder-like shape) to make it not entirely break the core flight model.

It's messy to work with and they are very honest about that fact in that URL. It will require a lot of unrealistic tweaking of elevator/rudder behaviors as well as the deep stall, for a very long time into the future. Maybe until we get those quantum computers @jcomm has ordered!

Edited September 11, 2022 by blingthinger

[Alpine Scenery 1,609]

On 9/11/2022 at 5:30 AM, Glenn Fitzpatrick said:

You can do quite good fluid flow simulation

Yes, the issue is they don't even have that much % of the CPU available for equation processing without slowing the whole SIM down, as the CPU is already the bottleneck for most people on higher-end GPU's. That said, I think there is a fundamental disconnect between finding the balance of ease-of-use for the average aircraft dev and "achievability" for the master modeler. Both types of aircraft developers need an easier system if we want to see more planes that feel a bit more realistic. As is, a lot of stuff is fudge tweaked to no end from my understanding, or people just build their own equations entirely.

 

[Christopher Low 5,359]

Is it even possible to create a realistic flight model on a PC using real world physics equations without the framerate grinding to a halt? How many parts does a 3D model of a plane need to be broken down into before the end result gets anywhere near the real thing? I mean, a thousand surfaces sounds impressive, but that's still a rather blocky representation of a real world smoothly curved airframe.

[jcomm 5,939]

7 minutes ago, Christopher Low said:

Is it even possible to create a realistic flight model on a PC using real world physics equations without the framerate grinding to a halt? How many parts does a 3D model of a plane need to be broken down into before the end result gets anywhere near the real thing? I mean, a thousand surfaces sounds impressive, but that's still a rather blocky representation of a real world smoothly curved airframe.

Very True, but I think ASOBO's approach is producing some very interesting results, even at such a level of discretion.

I still prefer the overall feel of flight and flight characteristics reproduced by X-Plane, and even better IL-2 Great Battles, but I see MFS's FDM in the right way and progressing really fast, and I do believe in the near future, each of these simulators in it's own will achieve an optimal ( given the constraints ) simulation of flight, well IL-2 probably not getting updated that much other than for fine tuning of specific aircraft performance figures.

I am really looking fwd for the additional levels of detail coming with MFS's Rotary Flight Dynamics.

It's a Time of Amazing Flight Simulation endeavour!!!

Edited September 13, 2022 by jcomm

[Scottoest 1,561]

22 minutes ago, Christopher Low said:

Is it even possible to create a realistic flight model on a PC using real world physics equations without the framerate grinding to a halt? How many parts does a 3D model of a plane need to be broken down into before the end result gets anywhere near the real thing? I mean, a thousand surfaces sounds impressive, but that's still a rather blocky representation of a real world smoothly curved airframe.

True physical "realism" could essentially be broken down into near-infinite complexity computationally, so no.  And beyond that, are you calculating the drag from every bit of grime on the fuselage or wings? No? #gamenotsim 😄

The trick with flight sims is just to have the computational fidelity to provide a dynamic and "realistic enough" experience, and then do some fudging around the margins to get the rest of the way there in terms of specific "feel" for the individual airframes etc.

More surface-points being processed in that approximation of reality is certainly always better, but ultimately a good flight model is as much a work of art as it is a work of math.