Dual architecture for FDE

[Dominique_K 10,149]

Not a specialist at all but  I understand we are passing from an architecture were all the forces are concentrated on 1 point to one where 1000 points will define how the aircraft behaves. Choosing where to spread the points and how much of a feedback each will generate does not seem so simple.  Thinking of three examples : a delta-winged needle-shaped Mirage IV, the burly AN2 biplane and a plane  with voluminous floats for instance.

  

Edited November 24, 2019 by domkle

[Julkensen 47]

All will really depend on how these 1000 points are working.

If the parameters like lift and stall threshold are calculated automatically by FS20 depending on the geometry of the wing this points  are bound to it should not be much harder for developers to implement this. But if it's not done automatically by the sim, then devs need to know the characteristics of a wing in very fine detail. Without data coming directly from the aircraft manufacturers, I don't know how they will be able to achieve that.

[RALF9636 1,956]

58 minutes ago, Anders Bermann said:

In a perfect physical and environmental (and high fidelity) simulation, wouldn't physical parameters such as drag and lift, for any given model, it's physical properties and dimensions taken into account, be automatically calculated?

I'm sorry for sounding simplistic, but in a world/simulation, where physical properties for air, temperature (and the interaction of these as well as density changes and forces, derived from the interaction of these physical properties) as well as effects like friction, are correctly simulated with relation to each other, objects passing through such a simulation, would automatically be subject to forces from the interaction of these physical properties. This would induce effects like drag, friction etc on the object - taking it's physical and dimensional properties into consideration... which would results in (at least the ground work) for the flight dynamics of an object.

Again, sorry for being simplistic.

These were my thoughts exactly.

But then again - being a complete layman - it might have been too simplistic.

 

50 minutes ago, FDEdev said:

Even if it would be possible, a precise simulation for the whole aircraft would require way too much computing power.

Apparently simulation is still by far not precise enough, otherwise aircraft companies wouldn't need to perform any flight tests and they would never need to modify their aircraft during testing.

When Boeing designed the new 747-400 wing, they couldn't even correctly calculate/simulate the bending moment for a single wing with all their computing power. 

Sounds plausible.

1000 points still can only describe a fraction of what is really going on aerodynamically around an aircraft, taking into consideration all kinds of different conditions like air pressure and temperature, pitch, bank and yaw, airspeed and g-load etc.

 

 

[FDEdev 955]

12 minutes ago, Julkensen said:

If the parameters like lift and stall threshold are calculated automatically by FS20 depending on the geometry of the wing this points  are bound to it should not be much harder for developers to implement this. But if it's not done automatically by the sim, then devs need to know the characteristics of a wing in very fine detail. Without data coming directly from the aircraft manufacturers, I don't know how they will be able to achieve that.

How should this happen automatically? You need to specify the airfoil(s), the incidence angle, washout, dihedral, vortex generators etc.

Even if the sim calculates the wing correctly, at least these data are necessary for a precise calculation. Some of these data are not easy to find.

If you don't have them, one will have to work with 'reverse engineering' and lots of trail and error as usual. 

[Dominique_K 10,149]

21 minutes ago, Julkensen said:

All will really depend on how these 1000 points are working.

If the parameters like lift and stall threshold are calculated automatically by FS20 depending on the geometry of the wing this points  are bound to it should not be much harder for developers to implement this. But if it's not done automatically by the sim, then devs need to know the characteristics of a wing in very fine detail. Without data coming directly from the aircraft manufacturers, I don't know how they will be able to achieve that.

There are thousands of wing 3D geometries and a close partnership with the manufacturer would be indeed necessary. When it still exists. If not a laser scanning of an exhibit in the Smithsonian or at Le Bourget air museum may help 😉

If FS20 automatically computed everything, for every wing, the consequences for the hobby would be mind boggling. 

 

Edited November 24, 2019 by domkle

[FDEdev 955]

8 minutes ago, domkle said:

There are thousands of wing 3D geometries and a close partnership with the manufacturer would be indeed necessary. When it still exist. If not a laser scanning of an exhibit in the Smithsonian or at Le Bourget air museum may help 😉

 

Not really. E.g. with an accurate 3 view plan, and knowledge of the above mentioned required values, a rather close simulation should be possible.

That said, the DR400 shown in the videos does have a significant model error and if the aerodynamics are actually matching the visuals, it will not fly like the real one. 

Edited November 24, 2019 by FDEdev

[Julkensen 47]

We'll see how this  works when the SDK comes out.

This would surprise me a bit that you need real  world data for this method to work. I don't think microsoft would expect 3rd party developers to rely this heavily on manufacturers data. The wing parameters and geometry is one of the most guarded secret for airbus and boeing because it directly relates to performance, fuel consumption, and thus, range and efficiency of the aircraft. The economic implications of this data is much to important for them to just give away like that.

[jcomm 5,936]

Too soon to get a sound idea about how it's going to be done in MFS, I believe...

It looks a lot like what X-Plane has been doing for ages, but I believe taken to a more detailled level in some aspects, as for instance the force vectors I could observe over the various surfaces, like when we enable the "Visual Flight Model" mode in XP. It looks like their distribution is more detailled, namely over the wings, and they talk about interaction of the stream with the various components and with the effects they cause over downstream ones...

DCS uses, so they say, an approach where a very expensive CFD software is used to deduce the aerdynamics data when there aren't sufficient RW documents accessible, or those that exist are somehow limited in scope or even dubious ( based on tests performed on damaged and repaired aircraft, etc... ).

The latest developemnts in the X-Plane front, that we can learn about from Austin's latest videos, focus primarilly on various types of soft tuning of the base FDM, and the tests I've run, at least with the default aircraft and one of the commercial helicopter add-ons did show the differences, towards what I might consider a more plausible "feel of flight". But fine tunning a specific aircraft using the approach XP uses, and the one MFS is apparently also using can be rather tricky. For instance, I remember the Van's RV-6 in MS FLIGHT correctly pitched down when flaps were lowered. The default ( experimental ) Van's familly in X-Plane 11 still pitch up on flap deployment, so, either a "scripted" approach is followed, like I believe is what happened in DCS after the initial release of their Spitfire Mark IX that also pitched up on flap deployment, or in the recent IL-2 Battle of Bodenplatte update that fine tuned the effect on their Spitfire too, or other sources of "magic" must be operated on the set of surfaces used for the computation of the "force resultant".

Cooperating with the manufacturers, more than with rw pilots who aren't simmers and sometimes answer using a completely distinct mindset to simmer questions, can make a difference in trying to approach the final feel to what it really feels like in reality.

Even DCS fails miserably sometimes. Take their Yak-52, and the various posts contributed at their Forums by a RL Yak-52 owner / pilot that pointed aout some really ridiculous innacuracies... And... just imagine what a WW2 pilot might comment on the accuracy of their ww2 modules ( and yes, I do know they took Eric to test the Fw-190 once ... and 1C / 777 a Russian Hero ww2 pilot too to test their LagG-3 ).

Basically, I can't expect MFS's flight dynamics to be any more miraculous than XP's, DCS's or whatever... It's all a combination of having a good framework to develop over, and then tallent and a lot of patience and available resources to do the fine tuning.

Same applies to aircraft systems and their modelling, which in Episode 3 have been identified as one of the major updates. They talked about the "turn & slip" indicator, which I actually only found correctly modeled in ELITE IFT ( now XTS ) among all simulators and add-ons I have tried in my simmer life, as well as in rather old IFT-Pro that I used around 1993-95. I hope we can finally perform a plausible rate 2 ( standard rate ) or rate 1 ( in the case of most gliders ) or rate 4 for most airliners and some choppers, using the turn & slip as a reference.

I would also really like to see the "leaning bug" fixed, and finally be able to use the mixture on reciprocating engines as it is supposed to be used IRL... 

And a more detailled model of a turboprop, with a free-running turbine, that none of the civil simulators ( less ELITE IFT and XTS ) do correctly in many details including the relationship between Prop RPM and Fuel Flow on free-runing PT-6 like turbines ( XP11 did fine tune it, but it's still out of sync with reality, and it puzzles me since Austin owns and aircraft propelled by one of those... ).

Let's wait & see 🙂

Edited November 24, 2019 by jcomm

[Dominique_K 10,149]

As far I know, just speaking of airliners, Airbus and Boeing use similar design software like CATIA and reverse engineering is certainly also facilitated by data collected through their airlines customers which run both aircraft lines.

A big question to which we have no answer at that time, is what kind of partnership MS signed with aircraft manufacturers. Just for the use of their names or something going much deeper ?

Difficult to imagine that Boeing gave its name for a botched 747. Difficult to imagine that Asobo, as good as they are, could develop a 747 alone. 

[jcomm 5,936]

7 minutes ago, domkle said:

Difficult to imagine that Boeing gave its name for a botched 747. Difficult to imagine that Asobo, as good as they are, could develop a 747 alone. 

In that particular case they would better consult with Hardy Heinlin... There's simply no other like the Aerowinx PSX when it comes to modeling the 744-400 ( all variants ).

If just it could be brought into MFS 😍

[FDEdev 955]

30 minutes ago, Julkensen said:

This would surprise me a bit that you need real  world data for this method to work. I don't think microsoft would expect 3rd party developers to rely this heavily on manufacturers data. The wing parameters and geometry is one of the most guarded secret for airbus and boeing because it directly relates to performance, fuel consumption, and thus, range and efficiency of the aircraft.

That's why I mentioned reverse engineering and trial and error.  The beauty with the simulator is that I don't even need to know e.g. the precise drag figures of a certain airliner wing, since I can adjust the drag figures until it matches the RW performance data.

[Macpah 46]

41 minutes ago, FDEdev said:

That said, the DR400 shown in the videos does have a significant model error and if the aerodynamics are actually matching the visuals, it will not fly like the real one. 

Thank you for that!

Finally someone who can spot such errors, too. I think their DR 400 model is sadly pretty bad.

Except for the cockpit part which looks pretty spot on.

Thank you also for your FDE observations. Pretty much how I see it, based on my experience with X-Plane!

[Paraffin 2,197]

10 hours ago, DocBird said:

If MS does allow developers to use the "legacy mode" couldn't that be a backdoor for the early entry into service for helicopters? 

I know that the flight dynamics for helicopters in FSX were considered to be really bad. But my understanding is that eventually there have been some reasonable helicopters in FSX/P3D, like DodoSim, etc. ...

And there would be nothing like experiencing the new scenery in a helicopter in VR ...

As I understand it, the problem with helicopters in FSX code is that there simply isn't enough default flight modeling. Too much heavy lifting has to be done by the 3rd party developer, basically hacking the flight model. Only a few are able or willing to do that, so it's not a path to getting many developers involved and an extensive fleet of well-modeled helicopters. 

Compare that to X-Plane, where Laminar includes all the essential helicopter dynamics like vortex ring state, ground effect translation, retreating blade stall, mast bump, etc. in the base flight model. That makes it much easier for 3rd party developers to build great helicopter models. They can focus on the stuff that's actually unique to each model like 3 or 4 axis autopilot, GPS/FMS nav integration, special features like sling loads and so on. 

This is why I find it a bit worrying that MSFS apparently won't ship with a helicopter. I hope they're thinking ahead on this, and not just relying on 3rd party devs to bolt on an aftermarket flight model like in FSX. That will severely limit how many great helicopters are available in the sim.
 

10 hours ago, DocBird said:

Having said this:

It was possible to port over FSX/P3D planes with some work to XP 11. If MS opens a “legacy mode” why only for FSX-models and not for other sims (XP 11, DCS, etc.)?

Wouldn’t it be easier to port e.g. XP-11-flight models over since they are more accurate and closer to the new flight model than FSX/P3D flight models?

Well, it isn't just "some work" to port an FSX/P3D aircraft model to XP. It's a major effort. 

For example, Carenado develops their models first for FSX/P3D, then hands them over to Thranda for conversion to XP. The only things that can be carried over are the visual 3D models and textures, and the sound files. That's it. The flight model and systems modeling (including plugins), have to be developed from scratch. And then the model has to be tested and tweaked so it hits the numbers when flying in XP, the systems plugins aren't crashing, and so on. I wouldn't be looking for an easy import function here.

[vortex681 806]

On 11/24/2019 at 8:46 AM, RALF9636 said:

1000 points still can only describe a fraction of what is really going on aerodynamically around an aircraft

But it's still hundreds of times the number of points over what's currently available in the FSX-based sims!

Edited November 25, 2019 by vortex681

[n4gix 4,947]

2 hours ago, vortex681 said:

But it's still hundreds of times the number of points over what's currently available in the FSX-based sims!

More like a 1,000 times better. F9, FSX and P3d V.x has only 1 point...