Physics and Aerodynamic on Directional Stability - Part 2 - Getting to the Root of the Problem

Anyone already tried this? Im sceptical as it mentions nothing about what they changed and on what data or experience. But maybe one of you guys has tried it and can review it?

First of all, let me say that I appreciate, and can imagine also everyone in this Topic, that someone working with Asobo could share these informations!

My perception is that the Modern Flight Model has the tendency to overestimate Stability Across the range of Aircrafts. From the Ultralight, to CRJ and Airbuses, it always will by standard calculate by Default too much Stability, specially on the Yaw and Pitch Axis. Its interesting you mention the Scalars are not normalized, because thats exactly oposite of what is stated multiple times at the SDK. It always make the Point, for multiple atributes, that a value of 1 should be the most realistic solution found by the Simulator for the entered aircraft attributes. If thats not the case with the Control Surface Scalars, maybe I think it would be better for it to be more clearly stated.

Another issue that comes to my mind, is that the Simulator has taken this approach of the Virtual WindTunnel, specially because they wanted a system that would make developing easier and faster, by having a system that would use the real dimensions and data for the aircraft, and come with the solution on it’s own, resulting in a more realistic output that would be possible than by estimates and guessing from the Developer to approximate the desired result. That would be specially true for airplanes that has no Aerodynamic Data available online, which are actually the case for most of them. Not even the Airfoil for many of the modern GAs are available since that is all classified information.

If the Scalar are not normalized, and the Virtual WindTunnel is unable to achieve satisfactory results, note i am saying satisfactory, because achieving a spot on Aerodynamic Model for an airplane based on a computer is something that only professional grade CFD Simulations on SuperComputers can do, so that will never be a realistic expectation from a Desktop Simulator, but Satisfactory I mean around the ballpark of what an airplane that would be allowed to be certified by being having resonable and predictable flight handling, that should be a priority IMO for the Team for the next Sim Updates.

Its a simulator still in its infancy, an I dont think we should expect the same level of fidelity other Sims have achieved after over 10 years of continuos improvements and polishing. Having said that, when there is a trend of Overstimation of Stability by the Modern Flight Model identified on most, if not all, aircrafts available since launch, unless the developer has worked on the issue himself to overcome the initial undesirable result, it’s for me easy to wish that Asobo team worked themselves on why this is happening in the Core of the Calculations, and bringing it closer to what an airplane, even the most exotic ones, should present as derivatives, translated as terms of flying qualities, that should be within range of the most extremes oposite values that would be within bounds of what doesnt translate into a very undesirable flying experience. The same happens for real aircraft certification steps, and its not what the exactt nmbers that matters, but if its within acceptable parameters.

Hope these kinds of chat between users and Asobo team, increases, as it can be very beneficial for both sides to make this experience better!

2 Likes

Perhaps you could point me to the areas in the SDK where it makes this point about the scalars. I don’t disagree that this is the most intuitive expectation, but I’m not seeing where this is specifically pointed out and we definitely want to make sure that this is cleared up if the SDK docs are saying that.

I’m not sure I totally follow why normalized scalars are a requirement for accuracy (and again, I’m not saying having them normalized is a bad thing). They’re just variables into the flight model. If the internal code has a coefficient of 1.5 and the scalar is 1, that’s not any mathematically different than internal code with a coefficient of 3 and a scalar of .5. The output force calculations will be the same.

The team (well, mostly Seb) is working on additional parameters that will enable some scalars themselves to be more normalized. However, a number of planes have been built on the dynamics as they exist today; you can’t simply just change internal coefficients without breaking the world, so to speak. The knowledge that the individual planes have too much stability is known and is being worked on for each individual plane. And I can see what we can do about adding recommended values to the documentation. But indeed accurate stability derivatives can be achieved with the right scalar values, today.

-Matt | Working Title

4 Likes

I think what is missed in this topic… I think the decision was made to make this more of a ‘game’ than a ‘simulator’.

Maybe the devs will move more towards the simulator version as time goes by.

I think if there’s one thing this type of discussion really does not need is another simplistic ‘is it a game or simulator’ angle. Please don’t.

5 Likes

Totally disagree with your statement.

5 Likes

Hope this is a step in the right direction:

SDK UPDATE

Documentation:
    We made some changes on the Flight Model section, created a Flights & Missions section, added a page about ice on aircraft models, and updated flightmodel.cfg with new parameters. More to come!
3 Likes

It’s very difficult to accept that, though, when the geometry section is so ‘loose’ in its definition of the airframe. Never mind the lack of ability to actually define an airfoil section, there are some distinct problems with basic shapes:


Area divided by span should give a chord of roughly 4.36ft. Yet the sim’s interpretation is
Tailfin

And then there’s the fuselage - I do not see how the developer’s geometry can be ‘not very correct’ when the sim will not allow you to define anything more than length and an average fuselage diameter - totally inappropriate for the aircraft above (and for many other aircraft which are not airliners!)


As you can see, the ‘best fit’ I could achieve leave the hump of the centre fuselage unaccounted for and the nose as defined is much larger and unstreamlined. I could probably accommodate one or the other by altering the figures, but not both.

If ‘correct geometry’ is to be the way forwards, with the aerodynamic coefficients and other scalars just used to nudge things in the right direction, then we need much more ability to create that accurate geometry.

4 Likes

Makes me wonder how one would go about recreating a lifting body design…

2 Likes

Seb and I were just discussing this, actually. X-Plane suffers from the same issue, as you can’t always make the geometry parameters as accurate as you would like. Conversely, even in planemaker, when you had very accurate geometry, it sometimes didn’t give you the result you’re looking for either, and so you resort to scalars and other tweaks. It’s difficult to account for different shapes when the shape could be basically anything, so you have to start somewhere., programmatically. The normalization step is a perfect fit for your scenario presently, if the average fuselage you are inputting is not getting you where you want.

Correct geometry is to be the starting point, and generally, for most GA and commercial aircraft, is going to get you really quite close on its own. My point was only that if the normalization step is adding crazy amounts more lift and/or drag than the geometry implies, its possible something is off. It could also just be a funky plane shape. The designer has the best knowledge of which of these things are true.

Obviously the sim is not there yet to accommodate every single aircraft design possibility. But more arbitrary shapes and designs are things being talked about, right this very moment (more wings, more stabilizer surfaces, different profiles, more fuselage options, etc). This is all going to evolve significantly over the next decade.

-Matt | Working Title

4 Likes

One of the big omissions is presently the lack of a slats simulation IMO.
It’s the only current sim which doesn’t simulate them which makes an even remotely correct simulation of aircraft with slatted wings impossible.

I hope that implementing this feature doesn’t take a decade!

4 Likes

I can safely say that, at the moment, this particular aircraft is proving very problematic - geometry of the wings is set as accurately as possible, coefficients and damping levels are realistic and I think the Bleriot may roll faster! So something is going to have to move away from ‘accuracy’ to get the required response levels.

This is a point where, having ALL the scalars set to 1 as the default, normalised level would be particularly useful, as I am just going to have to work my way through them and see what results I can get rather than knowing a defined starting point on which they do not affect anything.

1 Like

I don’t think that it will be ever possible in any desktop sim to simply punch in RW numbers and expect the sim to automatically achieve an even barely acceptable realistic result.

A kind of ‘reverse engineering’ will be always required.

1 Like

There will always be a bit of reverse engineering required, totally agreed.

However, taking Jan Roskam’s figures for the T37 and putting them into P3D (just as a single example) and then giving the result to a couple of T37 instructors to throw around resulted in very little modification required. Fine-tuning, so to speak.

What I am seeing here is a situation where the geometry, supposedly the most accurate way with this sim, is not at all working. Then the fallback of the coefficients etc are not providing the required ‘nudge’, so it looks as if the entire flight model is going to have to revolve around random numbers to try and chase the performance and I just do not see that in any way as being a step forwards or a move towards realism.

Maybe, over the next decade, it will get better but people want aircraft now and developing flight models in this current situation is really not enjoyable.

5 Likes

Sounds like a pretty good reason to allow external flight models :slight_smile:

4 Likes

That’s the reason why I stopped designing aircraft and FDEs for flightsims after 40 years :frowning:
Looking at the first SDK version it was obvious that it will be a very long road until one could start developing FDEs in a comparable way to previous versions.

Although the planned improvements are highly welcome, it puts MSFS into the same bad situation as x-plane designers.
After each revision you have recheck and recalibrate all your aircraft.

6 Likes

Actually it looks like the new Flyinside Bell 47 heli is using an external flight model.

Yes, but:

  1. It looks like it. We don’t know what they mean by “custom flight model”. It could be an actual external flight model, it might be a highly tweaked config file for the default model.
  2. It doesn’t mean that Asobo are ok with it. It might mean that there are loopholes that haven’t been closed yet. And if that’s the case, what happens when they are closed and the addon stops working?

I can only go by what Asobo have said: They do not want third party developers to override the default flight model. Whether they misspoke, or they changed their minds, or this is not the case presently but it will be in the future, that’s all a bit too speculative.

1 Like

Yeah, I guess we’ll see as it progresses further along we’ll learn more about the model and also how Asobo reacts to it.

1 Like

Unfortunately, this leads me to believe MSFS2020 will face the same limitations as X-Plane do, as it’s just not possible to have accurate output of Flight Dynamics Data for an Airplane, by geometry and dimension information alone, in a Desktop Simulator. Not in the next years at least. Only a CFD Simulation is able to run such a complex solution, taking at least some hours, and for just a few specific Flight Condition. All other shortcut to do similar, with the available computational power a regular user might have at home, and under seconds, not minutes or hours of baking time. This is a fantastic solution for games / simulations such as Kerbal Space Program, which is perfect for the goal of that game, you have just enough realism to make it a fun Sandbox, where you can see what happens when you add another rocket, too much weight on the Nose, etc. But this just will never work for a Flight Simulator.

This Approach to the Flight Model leads to a simulator that represents Cessnas Flying like Bonanzas, which flies like Cirrus, and so on… Apart from different Powerplants, Drag Profiles, Stall Speeds, Cruise Performance, it ends up being around the same flying characteristics shared between all. They are just too close in dimension, geometry, for a cylinder shaped fuselage to differentiate between that. The small and key details that set them apart are just invisible to a Computer that doesn’t deal with hundreds of thousands of triangles.

Now, we can’t have those, Users won’t accept that, and Developers will have to step up their game to provide that the costumers expect. Like GrimPhoneix9349 said so well, they don’t want to wait for a decade. What I predict will happen, again what is already common between X-Plane Developers to being able to achieve satisfactory results from this Limited way of estimating Flight Aerodynamics, addons being released with addicional Wings, Tails, Wheels, Ailerons, etc, all hidden from the users. This is the way most find to being able to shape the output to where they need it to be. Not a good solution at all if you ask me.

What has being mentioned as reverse engineering, is one possible solution to this issue, if you have the real data expected, you fill the tables, and the sim will then iterate that, along with the Virtual WindTunnel, to achieve a Proper Solution , allowing the Dynamic Aerodynamic Simulation of the Modern FM to coexist with the Accurate and Reliable Dataset of Flight Derivatives.

5 Likes