Flight model, upcoming CFD/physics enhancements, etc

[robert young 2,897]

10 minutes ago, scotchegg said:

I know next to nothing about the physics, but from a language pointe of view is it not common to distinguish between the simulator’s overall ‘aerodynamics model’, and an individual plane’s ‘flight model’? 
 

I try to follow the conversations as much as possible but sometimes it seems like people are using both terms interchangeably, and I’m wondering if I understand them correctly…?

Indeed. I suppose to be accurate, one should mention potential or scope of an aerodynamics system within one sim. And yes a flight model nearly always refers to the characteristics of one aircraft within that system.

[robert young 2,897]

1 hour ago, lwt1971 said:

Any particular thoughts on the CFD stuff they showed? Seems pretty interesting (and not a trivial improvement) given the way surrounding air flow is modeled ................

Yes it is interesting but I have to wonder why these and other aerodynamic elements have to be so convoluted, or maybe that is the nature of public relations in order to impress people. I tend to look completely dispassionately at these things. My first question is always: IS THIS ELEGANTLY SIMPLE or is it needlessly complex, and can it be applied without difficulties? If the answer is yes (elegantly simple) then I'm all for it. Elegant simplicity is at the heart of nearly all good flight modelling. The more convoluted the parameters the more room there is for unnecessary white noise and cluttered, clashing elements. If all you wish to do is to change the angle of a prop and make it feather you definitely do not need hundreds of computations based on every inch of the prop. I doubt even $50 million military simulators get anywhere near this. All you need is a simple system to feather the prop and, if you are flying at the time, to emulate almost zero airstream over the prop. So it stops. Elegant simplicity.

Where helicopter blades are concerned there is even more potential for over-complicated systems. All you need are as FEW lift points as possible, sufficient to simulate retreating and advancing blades and their quickly changing angles of attack. More points than necessary doubles or triples the potential for getting it wrong. I cannot emphasise enough that too many parameters can ruin a flight model. Indeed I would say my best ones have fewer than 10-15 basic elements that are juggled with excruciating care, rather than hacking around with endless numbers of parameters.

Edited February 23, 2022 by robert young

[abrams_tank 7,473]

24 minutes ago, robert young said:

Yes it is interesting but I have to wonder why these and other aerodynamic elements have to be so convoluted, or maybe that is the nature of public relations in order to impress people. I tend to look completely dispassionately at these things. My first question is always: IS THIS ELEGANTLY SIMPLE or is it needlessly complex, and can it be applied without difficulties? If the answer is yes (elegantly simple) then I'm all for it. Elegant simplicity is at the heart of nearly all good flight modelling. The more convoluted the parameters the more room there is for unnecessary white noise and cluttered, clashing elements. If all you wish to do is to change the angle of a prop and make it feather you definitely do not need hundreds of computations based on every inch of the prop. I doubt even $50 million military simulators get anywhere near this. All you need is a simple system to feather the prop and, if you are flying at the time, to emulate almost zero airstream over the prop. So it stops. Elegant simplicity.

I disagree with you here.  The holy grail of a flight simulator is that the  physics engine is so advanced, that the 3rd party developer merely defines the 3D model of the plane, how much thrust the plane generates by propeller(s) or jet engine(s), and the physics engine takes care of the rest.  There is zero tweaking by the 3rd party developer on the plane’s flight model if this holy grail of a flight simulator is attained.

In essence, such a flight simulator would have perfectly (or almost perfectly) emulated the physics of the real world. Perhaps we don’t have the computing power, or the code complexity to produce such a flight simulator at this time.  But this is the holy grail of what a flight simulator should do, if one day we do have that computing power and the code complexity to achieve it.

Edited February 23, 2022 by abrams_tank

[jcomm 6,005]

Unfortunately, even as advanced as today's desktop hardware is, such an holy grail is not attainable, so... we are left with:

.) approximations to the geometry of the aircraft - quite rough in MFS where we can't at it's present state define such "simple" aspects of aircraft geometry as multiple wings / tails, or even assymetry in their sizes or some other shape characteristcs and their impact on the wing performance;

.) approximations in the calculation of airflow interactions with different surfaces, either by lack of definition or lack of effective computation of the variations in flow and how that flow interacts with shadowed surfaces...

In the videos we can see basic aspects of even propwash effects apparently not taken into consideration. A couple examples:

.) The lift / drag vectors over the wings aren't affected by a stopped engine with it's associated prop feathered, so you get the exact same lift / drag from the area of the wing behind that prop irrespective of it being or not in rotation and / or the associated engine(s) working or not...

.) The sideslip effect over even a free-rotating prop like the one in the 350i not being modelled - the prop stays put if you put the aircrfat into a sideslip, yet it rotates axially when you roll the aircraft... as if in one case the relative airflow matters, but not in the other...

MFS has still a long way ahead to get at least to the level of another flight simulator that has long been using these features and blade element theory for the base of it's flight modelling and, even with all of those bells & whistles sometimes still needs a lot of fine tunning because the myopia of these approaches can play a negative role in the outcome of the flight model.

A fully parametric flight model such as the one used in FSX / P3D, or even way better in Flight Gear with JSBSim, provided the associated flight test data is available, can actually display closer to real performance from the flight model, inside the normal flight envelope. Outside of it some simulators can do a more acceptable job such as DCS World or IL-2 Great Battles, or even in some aspects Aerofly FS. I used ELITE IFT for ages for IFR trainning & proficiency and with it's table-based approach it was "On Spot" regarding the performance and even the feel of flying provided for it's hangar of GA singles and twins, although post-stall was a "word not allowed"... but it wasn't supposed to be it's purpose anyway.

So, what Rob says is, as usual, something I have to agree with, no matter how much I would like to believe MFS and XP12 will be closer to that "Holly Grail" of flight simulation abrams refers to...

Edited February 23, 2022 by jcomm

[robert young 2,897]

8 hours ago, abrams_tank said:

I disagree with you here.  The holy grail of a flight simulator is that the  physics engine is so advanced, that the 3rd party developer merely defines the 3D model of the plane, how much thrust the plane generates by propeller(s) or jet engine(s), and the physics engine takes care of the rest.  There is zero tweaking by the 3rd party developer on the plane’s flight model if this holy grail of a flight simulator is attained.

In essence, such a flight simulator would have perfectly (or almost perfectly) emulated the physics of the real world. Perhaps we don’t have the computing power, or the code complexity to produce such a flight simulator at this time.  But this is the holy grail of what a flight simulator should do, if one day we do have that computing power and the code complexity to achieve it.

Ha Ha! Interesting theory and ambition. The notion that you can simply enter data into a flawless system and out comes a perfect flight model is something that has been discussed for decades. A near equivalent is in the realm of F1 racing where the similar ambition is to design a car with perfect downforce and perfect front and rear wings, all done by entering data into an algorithm. Unfortunately it doesn't work which is why F1 teams pay millions to aerodynamicists whose job it is to tinker with settings.

The data-in perfection-out theory has been tried ever since the Wright brothers and it has never worked, and even less so for simulators. They can be roughly right but it is the last 10% of detail that often creates the difference between a pig and a swan. Data in/out tends to produce extremely poor pitch stability and an excessive amount of bounce in pitch, hence the video below of the lady professional Dash 8 pilot who kept blaming herself for over-control - whereas it was not her fault but was the grossly excessive pitch sensitivity and bouncing motion of the extremely poor standard flight models of most of the standard aircraft in MSFS.

For this reason I cannot stand flying the standard aircraft in this sim and privately tweak every single flight model to get rid of the ridiculous and crude, unharmonised controls.

 

[turbomax 3,307]

2 hours ago, robert young said:

The data-in perfection-out theory has been tried ever since

if it did work perfectly, why would Airbus, Boeing, NASA etc. still make extensive use of expensive wind tunnel tests? their flight model calculations seem pretty accurate though, they claim the actual flight tests of the real aircraft perform within tiny margins of their pre-calculated values. But that is easy, they don't have to support anything from a hang glider up to an A380. I suspect  they must secretly be using x-plane and some bleed elements?

 

[MattNischan 3,867]

3 hours ago, robert young said:

For this reason I cannot stand flying the standard aircraft in this sim and privately tweak every single flight model to get rid of the ridiculous and crude, unharmonised controls.

Robert, do you mind going into some detail about how you accomplish the reduction in pitch sensitivity and bounce? What cfg parameters are you changing and how are you changing them?

[lwt1971 7,457]

12 hours ago, abrams_tank said:

In essence, such a flight simulator would have perfectly (or almost perfectly) emulated the physics of the real world. Perhaps we don’t have the computing power, or the code complexity to produce such a flight simulator at this time.  But this is the holy grail of what a flight simulator should do, if one day we do have that computing power and the code complexity to achieve it.

That indeed would be the holy grail.. And if the air/ground/water physics were improved to the point of being able to plausibly handle various simulations (flight, trains, vehicles, watercraft) that would be the ultimate and satisfy all my simming needs 🙂 In the current world MS has created backed by cloud/AI/other tech, a solid foundation is definitely there to keep improving it to the point of being able to realize all this, especially if they improve the fidelity of on-the-ground depictions within the MSFS globe "clone"... I wouldn't be surprised if this is the path they're on and hopefully we'd be treated to train/driving/watercraft simulations in the same world in the coming years!

[jcomm 6,005]

Well,

being a very cautious guy at times regarding the "new" flight simulators, and having written what I wrote above, I still have faith in ASOBO regarding the development of MFS towards a more sound and 3pd-friendly platform.

I would LOVE to see the apparently "impossible" merge with Weather Software add-ons like my Beloved Active Sky, or even see tallented developers like Rob Young and guys from the other platform like the developer behind that AMAZING !!! CL 650 be able to offer their products for MFS.

They - ASOBO - at least, show proof of commitment to the cause. And they have that beautiful team leader by the nickname of "Sonant Alpaca" 🙂

If the end result will fulifill my expectations I really don't know, but so far it is doing it's thing acceptably !

One thing is for sure though: Better than any other sim before MFS is the perfect tool to train soaring tasks over non-familiar terrain, and that, together with a future "plugin" that could output NMEA to any of my PDA / PNA tools ( LK8000, XCSoar, Top-Hat...) will be the cherry on top of the cake...

If weather for soaring, as announced by ASOBO for a future soaring title in the line of the "Reno Air Races" turns out to show at least consistent effects, even if not fully covering the full range of soaring weather support for glider tasks, then I will surely keep using MFS with a lot of pleasure because I seriously doubt I will be able to get the same World Scenery from other platforms.

Better, more fine-tuned soaring flight dynamics would also be welcomed... but that Discus-2c is already very capable for a start.

In the meanwhile I am having Great Time simply flying some short / mid hops in the FBW a320 Neo, and looking forward for all the upcoming features, such as VNAV 🙂

MFS is SURELY NOT A SIM TO UNINSTALL 😉

Edited February 23, 2022 by jcomm

[robert young 2,897]

4 hours ago, MattNischan said:

Robert, do you mind going into some detail about how you accomplish the reduction in pitch sensitivity and bounce? What cfg parameters are you changing and how are you changing them?

Hi Matt,

Nice to hear from you (enjoying the CJ4 immensely). If you PM'd me I can take you through it.

[Krakin 5,846]

56 minutes ago, robert young said:

Hi Matt,

Nice to hear from you (enjoying the CJ4 immensely). If you PM'd me I can take you through it.

Oh nooo. Have the discussion in the open so we can all learn!

[tweekz 2,165]

20 hours ago, robert young said:

Sorry if this comes over as negative. I'm full of admiration for the 3d aircraft modelling in MSFS and there are some stunning landscapes, regions and sceneries. But the idea that Asobo is somehow revolutionary with their very impressive PR relating to aerodynamics  falls a bit flat for me. Most of the innovations they discuss were already implemented in many other sims (not just the big three) years and years ago.

Sure you can do a lot with lookup tables. Even way better if you know what you're doing vs someone who just throws in some basic geometric shapes and hopes for MSFS to do the job for him. MSFS still needs a lot of tuning via look-up-tables. But in general, I don't think it's a bad thing trying to make things more dynamic.

[mrueedi 702]

 

22 hours ago, robert young said:

I'm afraid many make the mistake of assuming that just because certain parameters have been opened up or implemented by any core developer, that equals ALL flight models are going to be better.

Indeed, one thing is the flight model (the mathematical formulas, that are implemented to simulate the physical world) and the other is the configuration or parametrization of the flight model for a particular aircraft (all that is stored in cfg files). These two things are consistently not kept apart.

I would go so far, that most people, who report a problem with the flight model, actually have an issue with the parametrization of the flight model.

[abrams_tank 7,473]

13 hours ago, robert young said:

Ha Ha! Interesting theory and ambition. The notion that you can simply enter data into a flawless system and out comes a perfect flight model is something that has been discussed for decades. A near equivalent is in the realm of F1 racing where the similar ambition is to design a car with perfect downforce and perfect front and rear wings, all done by entering data into an algorithm. Unfortunately it doesn't work which is why F1 teams pay millions to aerodynamicists whose job it is to tinker with settings.

The data-in perfection-out theory has been tried ever since the Wright brothers and it has never worked, and even less so for simulators. They can be roughly right but it is the last 10% of detail that often creates the difference between a pig and a swan. Data in/out tends to produce extremely poor pitch stability and an excessive amount of bounce in pitch, hence the video below of the lady professional Dash 8 pilot who kept blaming herself for over-control - whereas it was not her fault but was the grossly excessive pitch sensitivity and bouncing motion of the extremely poor standard flight models of most of the standard aircraft in MSFS.

For this reason I cannot stand flying the standard aircraft in this sim and privately tweak every single flight model to get rid of the ridiculous and crude, unharmonised controls.

So I think that Asobo continuing with trying to advance their physics engine is the right path.  What you are asking for is more control over being able to tweak individual planes, which I don't disagree with and I already said so as much earlier.  Asobo continuing to advance their physics engine, while opening up the SDK to more tweaking on a plane's flight model, is the best of both worlds, IMO.

Like I said, I think the most productive way is to just post what types of tweaks you would like to see available in the forum for the Twitch Q&A, before the next Twitch Q&A.  Just let us know when you post the question for the next Twitch Q&A, and I would be happy to upvote your question to get the attention of Jorg/Seb/Martial at the next Q&A (I'm sure there are others here at Avsim who would upvote your question as well).  In addition, Matt seems to have followed this thread so perhaps he has also read your concerns and may take this information to Asobo.

[robert young 2,897]

1 hour ago, abrams_tank said:

So I think that Asobo continuing with trying to advance their physics engine is the right path.  What you are asking for is more control over being able to tweak individual planes, which I don't disagree with and I already said so as much earlier.  Asobo continuing to advance their physics engine, while opening up the SDK to more tweaking on a plane's flight model, is the best of both worlds, IMO.

Like I said, I think the most productive way is to just post what types of tweaks you would like to see available in the forum for the Twitch Q&A, before the next Twitch Q&A.  Just let us know when you post the question for the next Twitch Q&A, and I would be happy to upvote your question to get the attention of Jorg/Seb/Martial at the next Q&A (I'm sure there are others here at Avsim who would upvote your question as well).  In addition, Matt seems to have followed this thread so perhaps he has also read your concerns and may take this information to Asobo.

I'm all for Asobo developing more sophisticated approaches to f/ms. The issue so far for me is that the practical implementation is no better (and sometimes worse) than the FSX/legacy Asobo inherited. Also I fear that they seem to announce unnecessarily convoluted methodologies and I can't work out whether this is just PR to impress or genuine stuff. All I can tell you is that the main practical parameters by which f/ms can be currently tuned are no better in scope and no more sophisticated than what was in FS9 over 20 years ago, but that is no bad thing. It is a much maligned system which, if you know it well, can still deliver very sophisticated results. After all it gave us a lot of notable aircraft, from Mad Dog, to the SR71, PMDG 737s and even some quite impressive helicopters.