8 hours ago, bendead said:

I am just wondering, on G1000 equipped aircrafts with GFC700 autopilot, do they use specific AP "code, logic, servo value" for each plane or it's universal?

I’m not sure. I used to work for a Garmin dealership in the early 2000s, and have installed and configured many GNS430 and GNS530 systems, but moved to corporate aviation in 2005 before the Gxxxx became widely available. I would assume that each installation is customized for the particular aircraft it is installed in, since they can be used in types ranging from the C172 to turboprops. 

5 hours ago, marsman2020 said:

Of course one could could a complete AP replacement using their own code for the AP, their own cockpit button/glass cockpit code updates to connect to that code, and directly driving the writeable variables for the control surface positions. But beyond that we are stuck as far as any additional updates to make the Asobo AP work better.

It appears that is the approach that the FBW team will eventually be taking for the A320 - replacing the default autopilot completely. A basic autopilot could be coded with simvars for current pitch and roll angle, turn rate, heading (and heading bug position) and altitude as read-only inputs, and aileron, elevator and pitch trim as outputs. A more complex autopilot would require more inputs, but even there, aileron, elevator and trim are the only required outputs.

I still think those 1000 points are a "myth"... in some way...

Probably they're actually pre-canned for specific aircraft types, don't know if hard coded and if client or server side.

Those raw data calculations were probably run in-house for a given set of typical aerodynamic parameters for given type of airframe, and the models added to the hangar are then classified in terms of their characteristics and overall performance obtained that way.

The same probably applies to other systems, the a320 and it's fbw protections and so on having a good chance of following this approach, so, imo, whatever it's done even by talented devs can easily be made useless each time a sim update takes place.

This are just my thoughts, not based on any official info !

 

Edited December 6, 20205 yr by jcomm

17 minutes ago, jcomm said:

I still think those 1000 points are a "myth"... in some way...

Probably they're actually pre-canned for specific aircraft types, don't know if hard coded and if client or server side.

Those raw data calculations were probably run in-house for a given set of typical aerodynamic parameters for given type of airframe, and the models added to the hangar are then classified in terms of their characteristics and overall performance obtained that way.

The same probably applies to other systems, the a320 and it's fbw protections and so on having a good chance of following this approach, so, imo, whatever it's done even by talented devs can easily be made useless each time a sim update takes place.

This are just my thoughts, not based on any official info !

 

This is wildly, wildly wrong.

The 1000 points model is real, although as I've mentioned numerous times in various threads, it isn't really about the dev themselves defining 1000 points manually; the dev defines the size and shape of the flight surfaces and the flight model engine subdivides those surfaces into hundreds of individual sections, each of which has the air simulation run on them separately.

All of these calculations are done in real time in the sim.

-Matt

Well, I was not commenting on how they're defined but rather on the thesis that, at situation / aircraft loading time these are calculated to set the "border" conditions for the given model.

Edited December 6, 20205 yr by jcomm

7 hours ago, robert young said:

True, you cannot stop the internal a/p doing this, but you can over-impose another wings rolling command as that is exactly what we are working on thanks to the lateral thinking and expertise of a very clever chap who is helping me. Early days as yet but it is working in principle. Refining it is another matter, but it looks promising.

This is one take that I didn't mention specifically, because it doesn't really fix the AP bank system.

Yes, for a 10 degree turn you could command a 40 degree turn instead and just cancel the extra when closer to your intended rollout heading to keep the bank angle up. It "works" for a very coarse definition of works, but how it works is going to be a bit on the brittle side, in terms of future compatibility. It also has some edge conditions to consider, like not going over the 180 degree mark with the commanded turn which will swap turn directions. You also then have to overlay custom ModelBehaviorDef and/or JS/WASM logic, which will make it a solution that can't be applied to every plane by one change, they'll all need individual changes.

Nonetheless, if folks feel it's worth the effort, that is one possible hack.

I know it sounds like I'm dissuading folks, and in a certain sense I am, but not out of stubbornness. I just hate to see solutions which may have future issues, so I just want to make everyone aware of the plusses and minuses. These are all good thoughts, though!

-Matt

Edited December 6, 20205 yr by MattNischan

4 minutes ago, jcomm said:

Well, I was not commenting on how they're defined but rather on the thesis that, at situation / aircraft loading time these are calculated to set the "border" conditions for the given model.

I don't believe this is true either, to my knowledge. The border conditions (like stall behavior and such), are emergent properties of different calc segments hitting critical angles of attack at different times, due to their relative positions in the airstream.

-Matt

1 hour ago, MattNischan said:

This is wildly, wildly wrong.

The 1000 points model is real, although as I've mentioned numerous times in various threads, it isn't really about the dev themselves defining 1000 points manually; the dev defines the size and shape of the flight surfaces and the flight model engine subdivides those surfaces into hundreds of individual sections, each of which has the air simulation run on them separately.

All of these calculations are done in real time in the sim.

-Matt

But Matt, they don't make a blind bit of difference in practice. If they did I would fall over in celebration. What matters is the practical use of things. Yes, I realise time is needed. But frankly, I see nothing whatever in practice that is an improvement. Apart from the improved ground friction every single aspect of the flight models is not just easier to tune using FSX legacy values and design (and NOT just because there is something new to get used to), but the scope is better in the legacy system. So I do not care if there are a trillion lift points. What matters is whether the behaviour of aircraft is either more flexibly tunable or believable.

This is nothing whatever to do with prejudice, lack of embracing of the new, and nothing whatever to do with ANY bias whatsoever. I simply look at the practical possibilities and given the choice I would go back to the legacy code. It is more refined, less convoluted, more flexible, more tweakable, less confusing, and has more practical parameters that are editable. I will give an example. The prop wash in FSX is tunable in three or possibly four dimensions. So is for example the relationship between yaw, slip, roll and stability. You cannot do this in MSFS so well. If it was possible I would be the first to celebrate.

I'm not interested in PR. I'm interested in what works. There is no point in having even 10 extra lift points if the aircraft oscillates uncontrollably, or cannot hold a simple ILS heading, or cannot track a heading without yawing all over the place. ELEGANT SIMPLICITY is best! Give me FOUR lift points and they can be tweaked better than 4 trillion lift points if the parameters are direct, functioning and straightforward. LESS is MORE!

1 hour ago, robert young said:

I'm not interested in PR. I'm interested in what works. There is no point in having even 10 extra lift points if the aircraft oscillates uncontrollably, or cannot hold a simple ILS heading, or cannot track a heading without yawing all over the place. ELEGANT SIMPLICITY is best! Give me FOUR lift points and they can be tweaked better than 4 trillion lift points if the parameters are direct, functioning and straightforward. LESS is MORE!

The issues you're mentioning are bugs with the default autopilot system, not the flight model. The new flight model introduced in MSFS (which makes use of blade element theory like X-Plane) is far superior than the one in FSX/P3D, unless flying on rails is your personal preference. Anything is possible if you're willing to put the time and effort into it, and I expect the flight model to improve even further over time.

 

And speaking to the default autopilot system, aside from tuning the PID values in ai.cfg, there's not much else to be done to tweak it - that's up to Asobo. Since we want full control over autopilot, we're rolling out our own from scratch for the A320neo.

Edited December 6, 20205 yr by IcemanFBW

2 minutes ago, IcemanFBW said:

The issues you're mentioning are bugs with the default autopilot system, not the flight model. The new flight model introduced in MSFS (which makes use of blade element theory like X-Plane) is far superior than the one in FSX/P3D, unless flying on rails is your personal preference. Anything is possible if you're willing to put the time and effort into it, and I expect the flight model to improve even further over time.

 

And speaking to the default autopilot system, aside from tuning the PID values in ai.cfg, there's not much else to be done to tweak it - that's up to Asobo. Since we want full control over autopilot, we're rolling out our own from scratch for the A320neo.

I'd rather have any of  @robert young "rail" flight models than Asobo's 1000 super-duper stuff they have provided.

1 hour ago, evaamo said:

I'd rather have any of  @robert young "rail" flight models than Asobo's 1000 super-duper stuff they have provided.

Perhaps, but the maximum possible accuracy of the legacy model is way, way lower than the maximum possible accuracy of the new flight model.

With identical ability and time, one can make a much more dynamic and realistic flight model with something that integrates a ton of individual surface calcs than with something that does just table lookups and scripts the boundary conditions, no matter how well you do it.

With all due respect to the greats of the previous sim generations, I'm not speaking from some PR perspective. There's just no way the legacy model can model certain things. We asked for a model with as much potential as something like XP, and when we got it, the first thing we want to do is go back to the old table-sim. It's very odd to me.

The sim community has a strange dual personality: always and ever unsatisfied but then when faced with change almost invariably resists it. Sometimes change requires going backwards a touch before going a long way forward. It is my belief that that's where we are right now.

-Matt

3 hours ago, IcemanFBW said:

The issues you're mentioning are bugs with the default autopilot system, not the flight model. The new flight model introduced in MSFS (which makes use of blade element theory like X-Plane) is far superior than the one in FSX/P3D, unless flying on rails is your personal preference. Anything is possible if you're willing to put the time and effort into it, and I expect the flight model to improve even further over time.

 

And speaking to the default autopilot system, aside from tuning the PID values in ai.cfg, there's not much else to be done to tweak it - that's up to Asobo. Since we want full control over autopilot, we're rolling out our own from scratch for the A320neo.

I'm not going to argue with you, if you mean the potential. But this "flying on rails" cliche really needs debunking. I'm surprised you have fallen for it. What does the phrase mean? Does it mean sim aircraft are somehow too stable? If so, in what circumstances? Do aircraft twitch around by themselves in still air? Why is there a comparison with a train? Actually, if you look at the victim of this flying on rails myth - FSX- the default FSX aircraft (except for the larger airliners) twitched and bucked in pitch all over the place, because for some reason people think aircraft should somehow be pitch unstable. The opposite is true and sorry to repeat, but almost all aircraft are REMARKABLY stable in pitch. So FSX default aircraft NEVER flew on rails anyway.

All contemporary aircraft that are not FBW or military types, are designed to be inherently stable. That doesn't mean it equals "flying on rails". It isn't a railway train. It is WEATHER and winds that destabilise aircraft, not still air. It is turbulence that disturbs otherwise remarkably stable aircraft. The aircraft left to its own devices in still air DOES, if you really want to keep promoting the myth, "fly on rails", in the sense that it is by default stable unless an outside force acts upon it. Aircraft designers spend hundreds of hours making machines that have decent vertical fins, ventral fins, balanced ailerons, decent C of G range, and most of all, and I say again, REMARKABLY stable in pitch unless something else is acting upon it.

The recent autopilot update is not a bug. It is a design decision to make the return to level wings far too soon to supposedly address other issues. Those issues might exist but making a heading capture take 3 times longer than it should is not the answer. That's all I have been saying all along.

 

Edited December 6, 20205 yr by robert young

I, for one, am glad that I do not have to understand what you guys are talking about, because I do not. Although I do find it fascinating and my daughter, who is an engineer and works for Boeing might make sense of it. I just get to enjoy the "fruits" of all your labor, for which I thank you for when this works. 

3 hours ago, MattNischan said:

Perhaps, but the maximum possible accuracy of the legacy model is way, way lower than the maximum possible accuracy of the new flight model.

With identical ability and time, one can make a much more dynamic and realistic flight model with something that integrates a ton of individual surface calcs than with something that does just table lookups and scripts the boundary conditions, no matter how well you do it.

With all due respect to the greats of the previous sim generations, I'm not speaking from some PR perspective. There's just no way the legacy model can model certain things. We asked for a model with as much potential as something like XP, and when we got it, the first thing we want to do is go back to the old table-sim. It's very odd to me.

The sim community has a strange dual personality: always and ever unsatisfied but then when faced with change almost invariably resists it. Sometimes change requires going backwards a touch before going a long way forward. It is my belief that that's where we are right now.

-Matt

I'm not buying that Matt! You are implying here quite clearly that a few old fashioned farts with a luddite view of the world and well past their sell-by-date somehow just don't "get" how 1000 lift points is the future. They are SO yesterday!

Well, forgive me but I'll link you to a video of me flying the Marchetti Sf260 which I co-designed back in 2004 for FS9 and updated in 2007. It hardly "flew on rails". OK the graphics look clunky compared to today, but for a "flying on rails", look-up tables, old fashioned, past-its-sell-by- date aircraft, it could spin more or less perfectly, side-slip up to 35 degrees and still have rock solid yaw stability, could pull 6 positive Gs but was still stable in pitch, had stall buffet, could do loops, hammerheads, stall turns, immelmans, cuban eights, barrel rolls, snap rolls and could even do INVERTED SPINS (do watch the video to the end if you could stand it!). In 2007, in FS9 and FSX, all with those old fashioned "look up tables" and only one theoretical wing. Surely not! And it had an autopilot with full IFR capability that did a near perfect ILS approach and totally accurate roll, VS, alt and Heading capture, as did many FSX aircraft with those out-of-date look up tables and old fashioned autopilot params.

Actually, even the suspension (look closely) was more flexible and malleable than most sims offer today. It took a lot of work, but the least of the challenges were the out-of-date lookup tables that is the fashion nowadays to write off as useless. How did I get it to do all that stuff? By using what was available and making the best of it. It wasn't perfect but it flew better than anything I've seen so far in MSFS, if you'll allow me to say so! The point I am making is that a decent flight model is not dependent on thousands of lift points. It is dependent upon using actually a small number of variables and tuning them to their best potential.

If you gave me four million trillion lift point calculations I doubt it would be any better, and possibly worse, because the more variables and lift surfaces you get, the more potential there is to get bogged down in clashing parameters, Oddly though, despite its age, the legacy flight model had a ton of features not found in other contemporary sims. For example to get a really good spin, you could do a completely bespoke combination of dihedral, roll, yaw and lift tables to get a near perfect wing drop. In MSFS, Asobo try to emulate this by triggering a left/right oscillation in roll. No multiple lift points there. I'm not knocking it. The potential is possibly there too, but the best flight models are done with unpretentious means and an approach that is elegantly simple. You do not need 1000 lift centres. You probably don't even need more than about six. It's PR overkill.

 

 

 

Edited December 6, 20205 yr by robert young

3 hours ago, evaamo said:

I'd rather have any of  @robert young "rail" flight models than Asobo's 1000 super-duper stuff they have provided.

So why did you buy MSFS?