Still worried about the flight model ... how planes move

[jcomm 5,939]

Probably those 1000 max points available for lift / drag /thrust calculations are also being severely restricted at the moment for the default aircraft since using many could possibly slow down the simulation ?

If of those potentially available 1000 points they're actually:

.) only using a few;

.) simplifying whatever is being used as lift / drag tables for each surface;

then the end result can well be what we have.

Maybe more detailled models, where the number of points used grows considerably, and other parameters are fine tuned, can show a completely different behaviour ?

But I would really like to understand who within development, and with which "credentials" is behind the flight dynamics, systems modelling, etc...

With XP ate least we know who is, and we can find interesting videos of the 1 man behind that component of XP where he gives some insight into what's being done, or what plans are there for the future.

The same probably applies to weather modeling too.

ASOBO says they use Meteoblue as the source for aloft weather, in the form of forecast model data, and METAR for observation. Question is - how is the merge being implemented ?  How do they "phase-out" from aloft weather into local weather and make the merge, so dependant in factors like sector altitude for the METAR sources, which can differ significantly between, say, the USA and Europe.

Fine tuning of the Global Weather Model and it's merge with GRIB data, SIGMET, etc.. will most probably get addressed with updates of the sim.

That's my Bet, and I am Betting on MFS 🙂

 

Edited August 7, 2020 by jcomm

[2reds2whites 827]

7 hours ago, ThrottleUp said:

I just saw this video the other day. Just press play because I've shared it at the timestamp I want to talk about.

If that didnt work then skip to 21:00 and watch from there till 21:26

See how the A320 NEO does that rapid very toy-like roll. It really looks & "feels" like an RC model. Seems to be lacking the "heavy feel" of a tubeliner.

 

There's no problem with the roll and responsiveness of that 320. He's just being too aggressive with his inputs.

People would be very surprised by the agility and dynamism of passenger jet aircraft. They only move slowly because we fly them that way. 

Edited August 7, 2020 by 2reds2whites

[MattNischan 3,778]

2 minutes ago, jcomm said:

Probably those 1000 max points available for lift / drag /thrust calculations are also being severely restricted at the moment for the default aircraft since using many could possibly slow down the simulation ?

If of those potentially available 1000 points they're actually:

.) only using a few;

.) simplifying whatever is being used as lift / drag tables for each surface;

then the end result can well be what we have.

Maybe more detailled models, where the number of points used grows considerably, and other parameters are fine tuned, can show a completely different behaviour ?

 

That's just not how any of this is going to work.

Picture a 172, high wing, fuselage mounted tail. You define the wing surface. You only say, the wing is at Y height, it's this big, it has these wing characteristics (chord, twist, etc). Then, the simulation takes that surface and breaks it into a grid. So, the 172 wing is about 11m by 2m, so if you break that into 100 segments you get 11cm x 2cm segments. The engine then looks at the airflow coming across each segment as asks, how much lift is this segment generating given the airflow hitting it? How much drag?

The airflow, which is not static, will generate a slightly different answer for each segment for the airflow's contribution, and the part of the wing will also generate a little different answer (is it the leading edge, the back of the wing, etc). That's how you get the stall behavior you see in the sim, where you actually see aileron instability and such, as each individual wing segment will stall independently of the others, since their lift, drag, aoa, and such is all calculated for each segment. Then the combination of those forces calculated is fed back into the physics system to move the aircraft.

This is actually not hugely dissimilar to X-Plane's blade element theory, except that my understanding is that blade element theory only effectively considers a sampling of 2d slices of whole flight surfaces as opposed to more of a grid sampling approach. They each have their pros and cons, I would have to imagine.

Again, the 1000 points things is not a thing flight modelers define themselves. They don't place by hand 1000 points of individual flight surfaces each with their own properties, and anyone in the flight modeling community who expected that is very curious to me. It would be nearly impossible for an individual to come to a flight model solution that functioned by defining that many surfaces manually. The idea is, you define the whole flight surface, the sim breaks it up into individual contributions for you given the air sim in real time. This isn't a computing challenge: it's trivially easy to run 1000s of lift and drag calcs well down into the microseconds range of speed of execution on today's hardware (and by today's I mean since the Ghz era of computing).

Real fluid dynamics sims that break that sampling into hundreds of thousands or millions in a 3D grid? That's where you get into non-real time analysis.

[new display name 2]

22 hours ago, Murmur said:

I don't think most other people in this thread get the point, rambling about how real aircraft are always different from virtual ones, etc.

Noted.  I thought my comments were valid and germane.

I'll avoid "rambling" in any future posts and stick to the usual "cudda, wudda, shudda" and speculation.  Sorry for the intrusion in such an esoteric thread.

[odourboy 1,289]

2 hours ago, MattNischan said:

That's just not how any of this is going to work.

Picture a 172, high wing, fuselage mounted tail. You define the wing surface. You only say, the wing is at Y height, it's this big, it has these wing characteristics (chord, twist, etc). Then, the simulation takes that surface and breaks it into a grid. So, the 172 wing is about 11m by 2m, so if you break that into 100 segments you get 11cm x 2cm segments. The engine then looks at the airflow coming across each segment as asks, how much lift is this segment generating given the airflow hitting it? How much drag?

The airflow, which is not static, will generate a slightly different answer for each segment for the airflow's contribution, and the part of the wing will also generate a little different answer (is it the leading edge, the back of the wing, etc). That's how you get the stall behavior you see in the sim, where you actually see aileron instability and such, as each individual wing segment will stall independently of the others, since their lift, drag, aoa, and such is all calculated for each segment. Then the combination of those forces calculated is fed back into the physics system to move the aircraft.

This is actually not hugely dissimilar to X-Plane's blade element theory, except that my understanding is that blade element theory only effectively considers a sampling of 2d slices of whole flight surfaces as opposed to more of a grid sampling approach. They each have their pros and cons, I would have to imagine.

Again, the 1000 points things is not a thing flight modelers define themselves. They don't place by hand 1000 points of individual flight surfaces each with their own properties, and anyone in the flight modeling community who expected that is very curious to me. It would be nearly impossible for an individual to come to a flight model solution that functioned by defining that many surfaces manually. The idea is, you define the whole flight surface, the sim breaks it up into individual contributions for you given the air sim in real time. This isn't a computing challenge: it's trivially easy to run 1000s of lift and drag calcs well down into the microseconds range of speed of execution on today's hardware (and by today's I mean since the Ghz era of computing).

Real fluid dynamics sims that break that sampling into hundreds of thousands or millions in a 3D grid? That's where you get into non-real time analysis.

^ this.

[jcomm 5,939]

@MattNischan,

 

interesting view, and I would like to think that's the way it's being done, but nothing I know about - and I do not even have access to the aircraft SDK - could give me that feedback.

It's for sure similar to X-Plane's approach, but if 1000 points were always used, no matter what aircraft was being modelled, that would mean a lot in terms of finite element analysis depending on if you were modelling an A380 or a Land Africa...

There's also the chance of it being simplified in terms of some of the aerodynamic parameters. A mix of approaches, like we had in Flight Unlimited 1 ( 2, and latter 3 diverged a bit from it ) and even in Fly! can also be the case in MFS, I really don't know.

 

P.S.: OTOH, given what's mentioned in the following excerpt from Episode 3 ( Aerodynamics ) video, you might actually be right in your view...

Also, while the legend reads "1000 fully simulated surfaces", the programmer says "thousands"...

 

Edited August 7, 2020 by jcomm

[Kilo60 660]

4 hours ago, fogboundturtle said:

or maybe they don't want default aircraft to be accessible for everyone and let the 3rd party dev take full advantage of the flight dynamics.

If the Devs themselves can't code it in yet what makes you think 3rd Party developers are going to know how to use it?

 

I really don't think it is that simple to code in or it would have been done by launch...

[HiFlyer 9,360]

This will probably just show my naivete since I really know nothing about it, but if I was a modern designer doing a aircraft that had 1,000 points as suggested by the videos by asobo, and I was working on a project that had artificial intelligence and computer power almost literally coming out of my ears, I would be assigning, if it all possible most of the flight model creation to an automated process, and then  tweaking that manually as necessary, to taste......

[Bert Pieke 7,585]

When have we ever had default airplanes in a sim that were not improved upon by various 3rd parties over time.. 😉

[bonchie 2,305]

2 hours ago, Kilo60 said:

If the Devs themselves can't code it in yet what makes you think 3rd Party developers are going to know how to use it?

 

I really don't think it is that simple to code in or it would have been done by launch...

You could have said the same thing for FSX. Yet, 3rd parties did far more than ACES ever did.

Believe it or not, I'd believe someone from A2A has far more ability to tweak the flight model than anyone working for Asobo.

Edited August 7, 2020 by bonchie

[Murmur 3,941]

3 hours ago, ex-N9142P said:

Noted.  I thought my comments were valid and germane.

I'll avoid "rambling" in any future posts and stick to the usual "cudda, wudda, shudda" and speculation.  Sorry for the intrusion in such an esoteric thread.

Sorry for the tone. Actually the fact that there are differences between real and virtual aircraft, due to differences in flight controls, etc. is a well valid and real point. It's just that the observations made by Mr. Young are not related to that, and many people in this thread are missing it.

[MattNischan 3,778]

3 hours ago, jcomm said:

P.S.: OTOH, given what's mentioned in the following excerpt from Episode 3 ( Aerodynamics ) video, you might actually be right in your view...

Given that they made the time to bake the flight surface force visualization into the sim for developers means that it certainly is not a simplified flight model.

I'm extrapolating based on what Asobo has said, the video clips like you've posted above, and from other things I can't necessarily talk about now. Does this mean the default planes have 100% nailed it? Not necessarily, but the allegations of it just being a mildly tweaked FSX flight model are totally nutso. But don't take the exact 1000 number too specifically; they're referring to the fact that the flight surfaces are multisampled across the flight surface at small intervals (don't forget there's a tiny bit of language barrier here, too). The exact amount is going to be dependent on the size of the flight surface and the amount of them total.

Again, I do agree that the planes look to be lacking some inertia in the videos and possibly a few other small bits and bobs. Nothing that concerns me, though, as being incriminating of the flight model itself being incapable of accurate aircraft. I think the future looks great.

[robert young 2,897]

18 minutes ago, Murmur said:

Sorry for the tone. Actually the fact that there are differences between real and virtual aircraft, due to differences in flight controls, etc. is a well valid and real point. It's just that the observations made by Mr. Young are not related to that, and many people in this thread are missing it.

Not sure quite what you are getting at, but I made it very clear that there are 3 seperate strands to my comment: 1) Control Input. 2) Reaction of the aircraft to the control input 3) Reaction of the aircraft to weather or airstream/gusts. Perhaps you got that but I'm confirming that I think both control input and core flight model are equally important.

[Chock 17,769]

7 hours ago, 2reds2whites said:

People would be very surprised by the agility and dynamism of passenger jet aircraft. They only move slowly because we fly them that way. 

This is absolutely true. I've been to lots of airshows and flown quite a few planes, but of all of those experiences, the best thing I've ever seen, was a brand-new B757 still yet to be delivered to the airline, at an airshow over Woodford in the UK.

I watched it lift off runway 24 at EGCC (before it changed to 23R) and then turn for EGCD Woodford, then beat up the airfield in a flight of such aerobatic exuberance, it was like a Spitfire coming back from a mission after having just downed ten 109s. Obviously the 757 isn't stressed for a lot of G and even if it were that'd have taken a lot of cycles off its airframe lifetime, so they were being careful not to put too much G loading on the thing, but the 757 is rather famously overpowered, so it was genuinely pretty aerobatic, doing chandeles, power zoom climbs and such.

An amazing sight. Wish I'd filmed it. My wife at the time said she watched me watching it do that display and she said I couldn't stop smiling whilst seeing it. Wouldn't have missed that for the world, seriously.

Edited August 7, 2020 by Chock

[Murmur 3,941]

16 minutes ago, robert young said:

Not sure quite what you are getting at, but I made it very clear that there are 3 seperate strands to my comment: 1) Control Input. 2) Reaction of the aircraft to the control input 3) Reaction of the aircraft to weather or airstream/gusts. Perhaps you got that but I'm confirming that I think both control input and core flight model are equally important.

I understand your main observations are the excessive twitchiness, especially noticeable on pitch axis (wobbling, underdamped nose), excessive elevator authority even at low speed, excessive reaction to turbulence. I agree with your observations.

I think many people in this thread have the idea that changing control sensitivities is all it takes to resolve the issues, whereas my idea is that those issues are intrinsic to the current flight models of default aircraft (and I reckon you think the same?).

However I'm not pessimistic, rather I'm eager to know more on this new flight model, how can it be tweaked and edited by 3rd parties, etc.