re: [Flightgear-devel] Ground Effects

[<js...@ha...>]

> This is very impressive.  Within a wingspan or so of the ground, the
> plane seems to float almost like a hovercraft -- in fact, I had to
> switch to external view to confirm that the wheels were not on the
> ground (this is where squeal and rumble sounds would help a lot).  It
> takes a bit of work to climb or descend out of this band.

Should it be easier or harder to get off the ground with ground effect modeled? 
I don't know, myself, but it seems to me that it should be easier to at least 
get off the ground.

> I had to make some changes to JSBSim.cxx to make things compile, since
> Curt has moved the global control structure into globals, but it
> wasn't tricky.  I'll send the edited JSBSim.cxx file back to the
> JSBSim people and will send my FlightGear and base package patches
> integrating the JSBSim changes to Curt

Excellent. I'll be interested to see exactly what is different.

> 1. Proper static friction for the gear, so it won't slide around in
>    the wind (I think Jon is already working on it).

Yes, lots of thought getting put into this and perhaps soemthing due for 
inclusion this week.

> 2. A port of Dave Luff's engine model (someone promised that for this
>    week).
> 
> 3. P-effect on takeoff and climb.

Yes, I think I know how to model this with the prop. I have given lots of 
thought to this as well, and for including gyroscopic effects, too.

> [pro] more realistic piston engine model
> [con] ground effect not modelled
> [con] obvious jitters on the ground; plane drifts (sometimes sideways)

I noticed this, too. For me it actually seemed to only weathercock into the 
wind with no translation.

> [con] does not respond realistically to variable winds in flight (no
>       allowance for ground-referenced momentum)

Hmmm. I had thought LaRCSim was supposed to correctly model steady winds and 
dynamics.

> [con] not configurable at runtime

True.

> JSBSim
> ------
> 
> [con] less realistic piston engine model
> [pro] ground effect modelled
> [pro] no obvious jitters on the ground; plane does not drift

I sometimes *do* notice jitters on the ground, unfortunately. I am not sure if 
might be fixed by using doubles instead of floats.

> [pro] responds realistically to variable winds in flight (i.e. the
>       plane's momentum causes it to lag behind wind changes)

I am not happy with ground handling at takeoff with winds. I have a suspicion 
that this is due to too much power given to the nosewheel, however.

> [pro] configurable at runtime (i.e. you can substitute a different
>       engine and see what happens)

Very configurable.

> Now, consider FGFS running at a tradeshow, and a real-world pilot
> sitting down to try it out.  Which model will make the better
> impression?  I vote for making JSBSim the default model today, as soon
> as the newest patches go into the CVS -- the support for ground effect
> far, far outweighs the weak engine model.

It's up to Curt and the whole group I suppose, but I'll weigh in with some more 
thoughts, here. Tony and I are putting a lot of effort into JSBSim, right now, 
with Mr. Larson working on the engine model pretty hard, it sounds like. I have 
said for some time (to myself and out loud), let's wait just a little longer to 
make JSBSim the default FDM. There's always "just a bit more" we'd like to add: 
more robustness, config file sanity checking at startup, better engine models, 
more debug features, higher fidelity prop model, better gear model, better 
documentation, etc. I suppose there will not be a time for quite some time 
where there isn't something being improved or added into JSBSim. JSBSim does 
have deficiencies, as does LaRCSim, but it is also being actively worked on and 
improved and documented. The fact that it is actively being worked on, however, 
is a double-edged sword. It might be possible to introduce instabilities into 
FlightGear. I guess this could be avoided with care.

Perhaps it is getting very close to being my opinion, also, that anytime Curt 
and the FlightGear team wish to make JSBSim the default, we are ready, too 
(speaking for myself, at
least).