I think after some of the later updates Phoenix became a CPU hog. I literally have to wait a few seconds so I can drag menu bars around and open tabs for categories of models. Or editing in the menu's there is a delay to erase and enter new numbers.

I agree with MSelig that creating a good model simulator is very complex endeavor, with no guarantee of success at the end. There are many factors that are hard or impossible to implement without significant simplification. The art of selecting the level of abstraction in the simulation is not obvious nor repeatable in different products. Developing realistic flight engine is long expensive and uncertain process that due to commercial pressure will always be beaten by simply pushing polished user interface and graphics (a process well understood and easy to do at scale). Visuals sell, internal qualities - not so much. Finally someone said all well implemented sims are pretty much the same - this is absolutely not the case, the major sims are quite different, with significant defects, and most have gone worse over time. I think because of new developers who don't have the skill set of the original ones. Not that they are bad, but not as good as the people who started the sims. These days good developers have more profitable and important things to do than write rc sim...

Stefan
http://rcflightsim.com

Originally Posted by skirtz

I agree with MSelig that creating a good model simulator is very complex endeavor, with no guarantee of success at the end. There are many factors that are hard or impossible to implement without significant simplification. The art of selecting the level of abstraction in the simulation is not obvious nor repeatable in different products. Developing realistic flight engine is long expensive and uncertain process that due to commercial pressure will always be beaten by simply pushing polished user interface and graphics (a process well understood and easy to do at scale). Visuals sell, internal qualities - not so much. Finally someone said all well implemented sims are pretty much the same - this is absolutely not the case, the major sims are quite different, with significant defects, and most have gone worse over time. I think because of new developers who don't have the skill set of the original ones. Not that they are bad, but not as good as the people who started the sims. These days good developers have more profitable and important things to do than write rc sim...

Stefan
http://rcflightsim.com

I would have to conclude that anyone who thinks the flight dynamics and aero are solved problems and more or less the same amongst all sims, provided they are "well implemented," must not actually know a lot about flight dynamics and aero. It's not simply a matter of crunching numbers. Lots of simplifications and engineering judgment are needed to arrive at something that will produce an accurate simulation. If you could really solve the problem with some number crunching on a PC in real time, the US government wouldn't spend millions (maybe billions?) of dollars on flight testing and wind tunnel programs every year. Also, I would have to start wondering why I'm wasting days of supercomputing time at work to simulate an aircraft in flight for a second or two.

In the old days, they had to design an airplane and test models in a wind tunnel to be sure it would fly, before they built the real thing. And even then, they would discover in test flights that the model didn't reflect reality and test pilots would die.

Now, the vast majority of the aerodynamic analysis is in fact done by computers. Computational Fluid Dynamics is what makes this possible. It's slow, it can't be run in real-time even on very fast computers. But when they actually build an airplane, they have tons of accurate modeling data and already know with an extremely high degree of confidence that it will fly exactly like it did in the computer. It's why the aerospace industry can design very fast or very large planes and they don't crash on their first test flight.

This math is not the same as the math you need to run a flight sim on a PC in real time. So your argument is specious. Someone who spends their days at work running simulations on a supercomputer would know this. I must conclude you don't really spend any time at work running simulations on a supercomputer and that last comment was sarcasm. If I'm wrong, I apologize.

You completely missed the point. The reason why less CPU power is needed to run a sim than CFD is because engineers have made simplifications and approximations to reduce the problem to something tractable. That is not an easy process, nor is it a given that it will be done correctly for a given desired flight envelope. It takes a lot of specialized knowledge and expensive-to-obtain data, and the problem only gets harder for RC sims where post-stall flight is very important. This has been my point the entire time.

Regarding CFD, having worked in the field my entire career, I think you give it too much credit. While it's true that real-world testing is increasingly being replaced by simulation, it's not currently close to being able to completely replace it. Moreover, the confidence in the solution is greatly reduced if there is no data available to directly validate the simulations. Modern aircraft programs always use CFD, wind tunnel testing, and flight testing together to fully characterize the aerodynamics, flight performance, structural loads, etc. In any case, I think it is unlikely that many of the RC flight sim teams make broad use of CFD in developing their flight models. The scope of runs, and associated cost, required to develop the flight models for a whole lineup of sim aircraft is out of reach still. If you want to read some good papers on how one sim handles it, take a look at the links here:

Originally Posted by MSelig

Ahhh ... I remember my days of working on FlightGear. That was fun. We (students and myself) put together many aircraft with the "UIUC" FDM model for FlightGear. Some of our FGFS work was published in 2004 -- Wright Flyer (AIAA Paper 2004-0211).

Back to your original question: I have talked/written about some of the methods embodied in the FS One. You will find the papers listed in the link below for years 2010 and 2014 -- AIAA pubs.

https://m-selig.ae.illinois.edu/publications.html

The 1903 Wright Flyer paper is in the same list. I did the Wright Flyer model because I wanted to "feel" what it was like to fly -- as best I could in a sim. (I had some very good students working w/ me on the project!)

I am pretty sure that CFD is not used in any rc sim. Acording to MSelig article

Component Aerodynamic Models
In the model, the aircraft is divided into basic components, such as
the wing, horizontal tail, vertical tail, propeller, etc., and for each, a
separate model is developed to determine the contribution that each
component makes to the total forces and moments on the aircraft
at each point in time. For each component, the local relative flow
is determined, taking into account aircraft speed and rotations
(assuming rigid-body kinematics), together with wind, turbulence,
and any aerodynamic interference effects, such as propeller wash,
wing induced flow, shielding effects (e.g., tail blanketing in spin),
ground effect, and more. For most of the models, the component state
(relative flow, surface deflections, and other data as it might apply) is
then used to determine any related aerodynamic coefficient data. The
final component dynamic pressure, aerodynamic coefficient data,
and respective reference area are then used to determine the forces
and moments.

I am sure all rc sims including mine (ClearView) are using this approach.
The main differentiation is how this approach is implemented (very non trivial).
I can tell you that the aero computations take very small amount of the total processing power.
The physics engine alone uses few times more CPU than the aero math.
For modern computers or mobile devices the processing power is never an issue for rc sim (excluding the graphics).

Stefan
http://rcflightsim.com

<snip, ok>

sorry this message is empty.

Originally Posted by skirtz

I am pretty sure that CFD is not used in any rc sim.

To be clear, I didn't mean to imply that CFD could be used directly in the sim, I meant that it could be used to derive the aerodynamics data that goes into a sim flight dynamics model. (But it would not be practical / cost-effective for that to be the entire approach.)

Great input, mostly going over my head but good stuff coming out which is what this is all about.

Had some 'Man Time' last night (she was watching a film) so I downloaded and installed the AccuRC demo, I already had the ClearView and Heli-X trial as well as Phoenix 6 and RF7 so 5 sims to compare.

I worked out a little 'program' to test my skills and the sims which was fast backwards, funnels, hurricanes, inverted slow banked 8's and then just pulling full up and sawing the collective followed by the same collective work but using full aileron (roll).

Unfortunately it took a bit of fiddling and several test flights to get Idle Up active on ClearView and the AccuRC demo is only 5 minutes long so I couldn't repeat the teat as often as I would have liked and this is purely how they felt to me. I was also flying the ClearView, AccuRC and Heli-X model 'out of the box' whereas I have 'tweaked the RF and Phoenix models so that may have an influence.

I found that ClearView was the hardest to fly as in the models drop out of the sky when you mess up and so in that respect it won (I am looking for realism). RF came next and the others were quite similar however Heli-X was slightly 'robotic' giving it a bit of an arcade game feel.

However, really not enough between them to call a clear favourite and for basic muscle memory 'how should I move the stick when going backwards upside-down' training they all do a good job but as already posted here, you still have to learn 'for real'. I decided that I won't be buying a new Sim at the moment because of the 3 new ones that I tested none of them were 'that' much better. Given more time I may have found a clear winner but for what I want at the moment I will stick to RF for memory muscle training, Phoenix for nice 'flying around' and Battlefield 4 for arcade stuff.

Please continue the discussion about 'achieving realism', interesting....

Paul

@PaulB
Hint regarding the arcade feeling with Heli-X: I found that by default the camera is locked on the helicopter, resulting in an unnaturally direct reaction to inputs. You can adjust that by setting the camera parameters to 'soft cam' (think thats how its called) enabled, either with standard values or even increasing somewhat. If you want to add further realism you may even chose to enable wind .