MSFS flight model

[mrueedi 702]

14 hours ago, MrBitstFlyer said:

20 hours ago, Bobsk8 said:

In my opinion, if someone flies a default C172 in any flight simulator and says it is realistic, they have probably never flown in real life in a C172. 

Fixed it.

And yet we saw the videos with the real world 172 pilots, who were happy with the flight model of the default 172 in MSFS (one even back in August 2020). Or did you not watch the video?

 

5 hours ago, MrBitstFlyer said:

I've had XP12 for two days. In the first hour with a default aircraft I had none of the flight model problems being talked about here - no twitchiness in the rudder and fine pitch and roll control.  The settings for controller sensitivity curves are so easy to use too.  As somebody else said, if the flight model moved forward enough, MSFS would be hard to beat.  

Were you happy with the almost "arcady" lack of any delay between control input and response (not my words), which was demonstrated and complained about in this video?

 

5 hours ago, blingthinger said:

- The summed up force and moment values on all those force elements everyone is raving about, get squashed down such that their sums equal the FSX tables. I get the backwards compatibility aspect, but this is no badge of honor. And like robert young says, they got rid of a lot of stuff that is useful to make this work when table models are still involved.

- The algorithm for said element-squashing and scaling process starts by calculating the force and moment coefficients at 20 pre-defined control surface positions and making a big interpolation volume. This happens once at aircraft loadup. During the flight if the user has the controls at a position that is between those points (which is MOST OF THE TIME), the forces and moments generated are averages of the nearest points in the volume. This essentially means that the "personality" traits of an airframe (as defined by the FSX table model) get smeared out. 

I think these 2 are the big reasons why the dynamic responses of even the newest planes are still a bit off. I struggle to see even the likes of PMDG fully overcoming them fully even with their own flight model tweaks. Also note this largely influences the dynamics. Getting steady-state stuff like a cruise climb and associated fuel consumption are going to be just fine from the FSX tables.

Lots of buzzwords... What do you mean with the term "dynamic response"? Science does only bring together the terms "aerodynamic" and "dynamic response" in relation to train crossing and wind turbines. Please, be precise with terms and meanings.

Imho, you misunderstood why this normalization is required. This mechanism is required to get the dynamic behaviour of the new model while keeping the flight performance numbers (speed, range) of the classical aerodynamics theory (for which FSX stands for). It is clear, that calculating forces and moments separately on 640 surfaces, the resulting sum could be off to some degree, which affects the static perfromance numbers as ROC, speed, range, fuel burn. To solve that problem, the chosen approach seems reasonable to me. This 20-config lookup table is only used to normalize the forces and moments so the flight performance is approximating the old one.

It is not about PMDG, not about anything what you could mean with "dynamic response", you actually gain the personality of an airframe, because that comes from the 640 surfaces of its geometry. 

 

5 hours ago, blingthinger said:

No fuselage effect in the lift equation? Are the fuselage force elements only generating sideways drag?

Wrong. To simulate the fuselage, there are 120 surfaces spread over an elongated rectangular box, which contribute forces and moments. 

5 hours ago, blingthinger said:

- No drag equation.

Wrong. I do "Ctrl+F" with "drag" on the page and get 27 hits. Drag is considered thoroughly, you are right however, that the page is not specifying how exactly.

5 hours ago, blingthinger said:

- No roll moment equation.

Wrong. I have a large screen, but the formula under pitch moment (torque) is not fitting on my screen. Below that you find the Yaw moment formula.

5 hours ago, blingthinger said:

- At each element, they're generating force and moment coefficients independently and simultaneously

Wrong. Read under "This gives us the following general order for each time step of the simulation": first the forces are computed, next "Compute their associated moments"...

 

Edited September 9, 2022 by mrueedi

[mrueedi 702]

1 hour ago, jcomm said:

regarding the total lack of Coanda effect modelling in MFS

Imho the embarrassing part is on the side of the one who asks. For classic aircraft, the Coanda effect plays a role at the wings, not the fuselage... The response from the Asobo guy was correct. 

The question also was about CFD, which has only be added to simulate prop wash and deep stall. The 640 surfaces, which dont block crosswind, are still there.

Also the video from Austin was hilarious. Measuring how the speed decreases while the plane is not moving itself though the air must result in a curve which is useless other than to clalculate propwash while the plane is parked.

[jcomm 5,939]

21 minutes ago, mrueedi said:

Imho the embarrassing part is on the side of the one who asks. For classic aircraft, the Coanda effect plays a role at the wings, not the fuselage... The response from the Asobo guy was correct.

http://www.thermofluids.co.uk/effect.php

Applies to any exposed / curved / convex  surface ofnthe aircraft, including wings OFC but also fuselage, engine nacelles, etc...

Regarding Austin's vídeo de was just validating the fórmula de uses to calculare propwash, which hás been um xp at least since xp10....

 

[Bobsk8 12,754]

6 hours ago, Sky_Pilot071 said:

I haven't seen these issues either in MSFS.

sp

Me either, much of the twitchiness complained about is poor control setup and lack of the user's ability  on how to properly fly an aircraft, 

[TrafficPilot 121]

12 hours ago, robert young said:

I managed to iron out a lot of these things in my Turbo Bonanza mod, but was not entirely successful at fixing ground behaviour because of the baked in extreme crosswind reaction.

I'm still flying the Turbo Bonanza in MSFS. In my opinion It's one of the only aircraft which doesn't exhibit "arcade game" handling. Yes there are issues with crosswinds as you pointed out but so far my rudder pedals haven't fallen apart.

Unfortunately it appears that, even after two years "realism" is not a priority for Asobo so I'll just stick with the modded Bonanza until they either improve the core flight dynamics and add weather or I get fed up and just go "real" flying more often..

Re FSX. I still have the steam version installed. What a difference it is flying in FSX compared to MSFS. So fluid, so smooth, so natural and with weather "effects" where they should be (with ASE) - not baked in like in MSFS. I think Asobo have a long way to go to improve on that.

Edited September 9, 2022 by TrafficPilot

[Alpine Scenery 1,609]

14 minutes ago, TrafficPilot said:

I'm still flying the Turbo Bonanza in MSFS. In my opinion It's one of the only aircraft which doesn't exhibit "arcade game" handling. Yes there are issues with crosswinds as you pointed out but so far my rudder pedals haven't fallen apart.

You didn't try the 310r yet?

I was skeptical and thought people were just over praising, but it's really good. To others, people should relax, I am all for improving the sim, not sure why people get offended from technical critique of the sim. If they provide sample equations or things you can call, that is a flight model, even if you can also write your own equations. Some may not even use any of the equations at all, but that doesn't mean no modeling exists. There is a flight model because there are pre-built equations you can use or not use, up to you.

That was my point though, they made it complex to model things, like you have to fix and do everything yourself, which is why only a small number of developers are able to master it.

 

[lwt1971 7,420]

7 minutes ago, Alpine Scenery said:

You didn't try the 310r yet? ... I was skeptical and thought people were just over praising, but it's really good.

The Milviz C310 really is the best modeled light/GA aircraft for MSFS right now in terms of flight dynamics for sure, and also probably systems and wear&tear modeling, etc (maybe the Sting S4 or the FSW C414 second?). Of course A2A will have something to say about that when their Commanche releases, and all the better for us 🙂 (really hoping they'll drop their long promised overview or deep dive video(s) on the Commanche soon, the release will likely not be till 4Q at the earliest I'm guessing).

Edited September 9, 2022 by lwt1971

[Bobsk8 12,754]

6 minutes ago, lwt1971 said:

The Milviz C310 really is the best modeled light/GA aircraft for MSFS right now in terms of flight dynamics for sure, and also probably systems and wear&tear modeling, etc (maybe the Sting S4 or the FSW C414 second?). Of course A2A will have something to say about that when their Commanche releases, and all the better for us 🙂 (really hoping they'll drop their long promised overview or deep dive video(s) on the Commanche soon, the release will likely not be till 4Q at the earliest I'm guessing).

After flying the C 310, I wouldn't go back to the Comanche, if they gave it to me for free. 

[Bobsk8 12,754]

57 minutes ago, TrafficPilot said:

I'm still flying the Turbo Bonanza in MSFS. In my opinion It's one of the only aircraft which doesn't exhibit "arcade game" handling. Yes there are issues with crosswinds as you pointed out but so far my rudder pedals haven't fallen apart.

Unfortunately it appears that, even after two years "realism" is not a priority for Asobo so I'll just stick with the modded Bonanza until they either improve the core flight dynamics and add weather or I get fed up and just go "real" flying more often..

Re FSX. I still have the steam version installed. What a difference it is flying in FSX compared to MSFS. So fluid, so smooth, so natural and with weather "effects" where they should be (with ASE) - not baked in like in MSFS. I think Asobo have a long way to go to improve on that.

I agree, flying in FSX is like riding in a train simulator, not like flying at all, compared to MSFS. 

[blingthinger 649]

5 hours ago, jcomm said:

allow for fine tunning of their MFS's CFD approach

CFDDDD. I sometimes call it "Colorful Fluid Dynamics"! I once heard a NASA engineer call it "Constant Frustration and Disappointment" 😉 It's a necessary tool, don't get me wrong. We wouldn't have the aerospace world in front of us now, without it. But it's full of some fairly significant approximations. And chasing numerical convergence will make your hair fall out from "What The Heck!" moments when it doesn't! 

I have mixed feelings on this one. Asobo's the first to use it in this way in a product for sale, maybe at all. +1 for creativity and ++1 for reaching for the holy grail. But I think this is solidly in the category of "wayy ahead of its time". It shouldn't be breaking things. I have two theories on why they are having problems.

First is that the geometry they are simulating doesn't represent any actual airframes. The air momentum is interacting with a fuselage and wings that are rectangle flat plates (per SDK): hershey bar wings and empennage. The intent of CFD is to simulate the actual geometry. They're fighting a scenario of running a CFD model that in no way represents reality. Good luck!

Second is that their computational mesh is way too coarse for this. They'll get some swirling air effects, etc, but it feels more like a bandaid covering up the shortcomings of the FSX+force elements. And they admit it's breaking things. I do not envy Asobo's devs right now. So, yes it's a novel concept, but no, it's not a physically consistent approach at the moment, and may not be an option in a pure sense in our lifetimes because of the computational cost. 

Edited September 9, 2022 by blingthinger

[Alpine Scenery 1,609]

Well, it is a hard thing to code, no doubt. I wouldn't want to do it. That said, there are ways of splitting the difference, as always. You don't need to achieve perfection, and you can pre-generate thousands to millions of LUT tables in advance for common calculations that slow things down. 

It was a decent effort, but they need a standardized system to make it easier for all developers to have a decent flight model. Too many developers are making planes that fly like a hot air balloon more than a plane.

 

Edited September 9, 2022 by Alpine Scenery

[blingthinger 649]

5 hours ago, mrueedi said:

What do you mean with the term "dynamic response"?

How the system responds to a step function input. If I flick the control stick, how does the airframe respond? How fast does it move? What motions are induced? Are there motions about more than one Cartesian axis? A steady cruise climb is a static behavior. FSX can get that just fine. 

You are correct that the force elements are there to generate dynamic responses. And they do just that. In this case it's like putting springs on a rocking chair. My point is that they are not realistic. Not only are they acting on a geometry that doesn't represent reality, but they are being squashed and scaled around to match the FSX curves. The SDK specifically states they are 2D grids of elements. Look at the text above the airplane schematic on that URL. Fuselage is described as 40x3. I'm unable to turn that description into a 3D volume. Their descriptions are of 2D flat plates. Hershey bars for wing, empennage, and fuselage. If that is the case, the fuselage elements contribute nothing more than side drag forces.

5 hours ago, mrueedi said:

I do "Ctrl+F" with "drag" on the page and get 27 hits.

You misunderstand my intent here. I'm not suggesting there is no drag or lift or whatever. I merely want to SEE the equations. Why are they hiding them??

 

5 hours ago, mrueedi said:

This gives us the following general order for each time step of the simulation

That section is after they are back in FSX land. After the forces have been squashed and summed up to be the FSX tables. I'm talking about this section: "We then initialize the algorithm by attributing - to each surface element - local lift, drag and moments coefficients". Force and moment coefficients are generated simultaneously. Uncoupled from each other. This isn't necessary for a 3D model (flat plates or not). Mostly I don't get the moment solution. Why not forces alone?

 

5 hours ago, mrueedi said:

This 20-config lookup table is only used to normalize the forces and moments so the flight performance is approximating the old one.

The second half of that sentence says alot. Yes, I agree with you there. My point is that the 20 config lookup table is going to smear out the dynamic response of the airframe. Put a different way, the interpolation volume is likely too coarse. What's more, I bet most of these points are in the happy range of the operating envelope. Good for airliners and autopilots. That's where most people will be operating, too, so no harm there, but inherently this is unable to realistically represent the airframe. It will get you most of the way there, but the final 10-15% of the dynamic response is lost.

Edited September 9, 2022 by blingthinger

[mrueedi 702]

1 hour ago, TrafficPilot said:

exhibit "arcade game" handling

Define "arcade game" handling. This term is useless when discussing the merits of a flight model.

 

1 hour ago, TrafficPilot said:

Unfortunately it appears that, even after two years "realism" is not a priority for Asobo

It could appear as well, that this is just your foregone conclusion. You need to substantiate your criticism. Dont you think that prop physics in SU8, CFD in SU 9 is a testimony of their efforts?

 

23 minutes ago, blingthinger said:

The air momentum is interacting with a fuselage and wings that are rectangle flat plates (per SDK): hershey bar wings and empennage.

That is used for the 640 surfaces simulation, not for the CFD. CFD only covers prop wash and deep stall. CFD is an approximation and addresses to some degree the intrinsic inability of FSX to handle deep and spinning stalls. 

[blingthinger 649]

1 minute ago, mrueedi said:

That is used for the 640 surfaces simulation, not for the CFD. CFD only covers prop wash and deep stall. CFD is an approximation and addresses to some degree the intrinsic inability of FSX to handle deep and spinning stalls. 

I'm confused here, then. CFD requires some sort of geometry definition. It's not CFD otherwise. Why wouldn't they also use the same geometry being used for the force elements? If they aren't, what are they using? Even if one thinks they are using the beautiful 3D models you see on the screen, it would then get reduced down to a 20x20x20 volume grid (those are their default numbers). Sooo essentially that would be back down to a coarse, blocky mess that doesn't at all represent reality and would actually be far worse than using the force elements layout because you'd have surface grid cells pointed in all sorts of directions after subsampling the 3D display model.

[Alpine Scenery 1,609]

Too much theory, in my experience, theory is theory and real-world implementations often vary when trying to simulate stuff. I don't think more theory is what they need, just more testing - rework - testing - rework. Unless they got that savant aerodynamics coder that can apply it all correctly in one shot, then the problem is you got too many things interacting and only one thing slightly off throws the whole thing out of whack. It's one thing to have the math right on paper, but it's quite another to get it into a simulation. 

I don't think the flight model feels arcadish at all, it feels the opposite to me, it feels overly calculated and slightly miscalculated, like it is trying to do too many things at once without proper multivariate cross-weighting. Arcadish flight model would be like Star Wars Squadrons or Elite Dangerous or something, where things just move on a mostly linear set. MSFS definitely feels less linear than every other sim I've ever flown, but that is both good and bad at times.

 

 

 

Edited September 9, 2022 by Alpine Scenery