here's an update to the b52 - it's just some relatively minor changes to the
fdm and panels.
I've been trying to improve the take-off and landing characteristics, or at
least get them to fit the pictures and film I've seen.
The best way to take off is by using the autopilot - at sea level just extend
the flaps, punch in the altitude hold on the autopilot (and heading if you're
lazy), and apply full power. It'll also take of with the default the default
elevator trim of -0.07 but it 'staircases' a little bit - the auto pilot
smooths that out.
I've included some notes and observations about taking-off and landing in the
readme.
While I wouldn't claim any sort of accuracy for it, I hope it manages to get
some of the characteristics right.
much what i've seen p-51d's do. Be careful of spins.
Added vstab incidence (real ones have it) thus improving takeoff behavior.
Reduced turbo-mul to what it probably should be.
Returned wing camber to Andy's estimate.
Increased flap drag. And tried making adjustments to get the thing to not
glide so impossibly far.
Corrected empty + pilot aircraft mass to 7190.
Upped "cruise speed" to maximum operational speed of 380 knots.
Reduced prop pitch at cruise to 0.8 (3000 rpm is only for takeoff).
Changed camber to 0.01 (found reference that said 1%)
Decreased tail surface effectiveness slightly just to get better numbers.
Changed turbo-mult to 2.5 and wastegate-mp to 30
Changes:
Fit the solver to the known stall speed instead of an abstract
approach configuration. That is, specify stall AoA and speed as
"approach" values. At approach weight (20% fuel) I can now just
barely hold the aircraft in the air without losing it at 87 kias.
Note that the IAS guage seems to have a slight calibration error, it
was reading something that looked more like 95 knots; I got the real
value out of the property browser.
Change the sign of the incidence value for the wing. It specifies a
rotation about the axis pointing out the left wing, so positive values
are "nose down".
Increase the "effectiveness" of the tail surfaces quite a bit.
Smaller surfaces do actually need higher values. YASim scales surface
force coefficients with their areas, which is correct within a single
wing. But between wing-like devides, generated forces kinda/sorta
scale linearly with their spans. This is particularly important for
tail draggers, since YASim's lack of prop wash modelling needs to be
offset by extra elevator authority.
Add a camber value to the wing. Most wing airfoils are asymmetric,
and produce non-zero lift at zero AoA. I picked 0.1 (10% of stall
lift at zero alpha) as a reasonable guess. If someone has airfoil
data for the Mustang we could look this up exactly. This was the
biggest change, which allows the cruise AoA to be much lower than
approach or stall.
Reduced the compression value for the tail wheel to 20cm. These
things are very stiff; they "compress" only as much as the tires do.
Even 0.2 is too much motion, but the numerics tend to go wacky when
you give them very high spring coefficients. This helped the ground
handling a little bit.
Removed the extra damping from the main gear. My impression of tail
draggers is that they tend to have "squishy" main gear. Again, this
(subjectively) seemed to improve ground handling to me. I also tried
reducing the spring constants to 0.5, but that ended up being too
squishy -- you could see the ship (stopped on the ground) bank to the
left by 2-3° when you pushed the throttle forward.
Ground handling is still pretty difficult; I get the best results by
holding the tail down until 90 knots or so and then very gently
lowering the stick. The aircraft bobs once or twice and then lifts
off. I don't think this is proper procedure, though.
Andy
Changes:
Fit the solver to the known stall speed instead of an abstract
approach configuration. That is, specify stall AoA and speed as
"approach" values. At approach weight (20% fuel) I can now just
barely hold the aircraft in the air without losing it at 87 kias.
Note that the IAS guage seems to have a slight calibration error, it
was reading something that looked more like 95 knots; I got the real
value out of the property browser.
Change the sign of the incidence value for the wing. It specifies a
rotation about the axis pointing out the left wing, so positive values
are "nose down".
Increase the "effectiveness" of the tail surfaces quite a bit.
Smaller surfaces do actually need higher values. YASim scales surface
force coefficients with their areas, which is correct within a single
wing. But between wing-like devides, generated forces kinda/sorta
scale linearly with their spans. This is particularly important for
tail draggers, since YASim's lack of prop wash modelling needs to be
offset by extra elevator authority.
Add a camber value to the wing. Most wing airfoils are asymmetric,
and produce non-zero lift at zero AoA. I picked 0.1 (10% of stall
lift at zero alpha) as a reasonable guess. If someone has airfoil
data for the Mustang we could look this up exactly. This was the
biggest change, which allows the cruise AoA to be much lower than
approach or stall.
Reduced the compression value for the tail wheel to 20cm. These
things are very stiff; they "compress" only as much as the tires do.
Even 0.2 is too much motion, but the numerics tend to go wacky when
you give them very high spring coefficients. This helped the ground
handling a little bit.
Removed the extra damping from the main gear. My impression of tail
draggers is that they tend to have "squishy" main gear. Again, this
(subjectively) seemed to improve ground handling to me. I also tried
reducing the spring constants to 0.5, but that ended up being too
squishy -- you could see the ship (stopped on the ground) bank to the
left by 2-3° when you pushed the throttle forward.
Ground handling is still pretty difficult; I get the best results by
holding the tail down until 90 knots or so and then very gently
lowering the stick. The aircraft bobs once or twice and then lifts
off. I don't think this is proper procedure, though.
Andy
I've done a new Sea Hawk. Well, the model and the fdm aren't new, just
slightly modified, but I've incorporated some more animation stuff and tried
a different way of texturing it.
The animation additions are things I worked out for the YF23 (still can't get
it to fly again) - mostly u/c related stuff.
This actual aircraft (WV908) is in the Royal Navy Historical Flight and so is
kept unusally clean;) In the photographs I looked at, hardly any fuseage
panel lines were visible so I've not tried to include any. Instead of the
normal fuselage side views and wing plan textures I've just used fairly
high-res detail 'patches' for the markings, applied to 'carrier' objects.
This is much more like the way I'd do things if I was working on a picture in
my 3d software - there I'd apply multiple textures to an object, as required,
using scope and priority settings to get what I want. With .ac format models
you can only apply a single texture to an object - hence the need for the
carrier objects. These are sections of the underlying 'real' object that
have been moved out/away from from the original object surface so that they
cover them.
The pros are 1) the resolution of the details is much higher - have a close
look at the Ace of Diamonds art on the nose and on the bit that sticks out of
the front of the tail-fin (I don't know the proper name for this), and 2)
less texture space is needed.
The most obvious con is that there's more potential for z-buffer problems. I
notice that whilst on the ground (in chase view), there appears to be
z-buffer fighting for the texture patches but once the a/c has taken off and
has got to about 200ft or so, it clears up. I've no idea why this should
happen as I'm not changing any of the view settings.
There are a few things I still need to figure out in this respect - I think
there's an issue with graduated textures - they seem to get reduced to 16bpp
and show a lot of banding that's absent on the texture and the model appears
inverted from tower views, at least in 16bpp, but ok from the chase view, and
stuff like that.
I said that the model and fdm aren't new but they have been modified a bit.
I've only reduced the approach aoa in the fdm but the model, whilst basically
the same, has had quite a few changes to accomodate the new u/c and texture
patches. It also got into FG by a slightly different route - I found that
AC3D would support polys with > 3 sides and after looking at and comparing
.ac and .obj file formats (I can export from RS3D in .obj format) I wrote a C
prog to convert .obj format files into .ac format files. It's a nasty bit of
hackery that even uses a temporary workfile (LOL), and it will only work with
the simplest of .obj files (e.g. no vertice normal or texture info) but it
allows me to export the model geometry in .obj format, convert it, and load
it straight into AC3D without turning everything into triangles or flipping
face normals.
The funny thing is that objects that consist of just a single rectangle (e.g.
an aileron or flap surface) are not recognised by FG and they have to be
split into two triangles, and once the model's loaded up into FG I can see
from some of the rendering artifacts that it's all still reduced to triangles
prior to display anyway. Still, it saves me from having to do a lot of face
flipping.
Really, the texturing method is experimental and I hope I get some feed-back
on it regarding display and performance issues. Ultimately, I'd like to try
to reduce the texture size down from 512x512 to 256x256 - nearly half of the
current texture 'space' is unused.
+ The wing stall angle was weird, as mentioned earlier. Values
greater than 45° don't have much physical meaning. I changed it to
25, which should be closer to a realistic value.
+ The lift="1.1" on the elevator flaps wasn't quite enough to get the
aircraft into the approach AoA. I changed it to 1.2, and it works.
It's still not as responsive as I'd have expected (then again, this
is a big aircraft...).
+ The hstab effectiveness was negative. This is wrong, as this is a
multiplier for all force generated by the surface. A surface with a
negative value will produce thrust instead of drag. :) Maybe this
was needed because the elevator was rigged backwards somewhere else?
I changed it
+ The wing and hstab had very small stall widths (very sharp stalls).
Low aspect wings should have gentle stalls, not sharp ones. I
changed it to 15 as a guess.
+ There's no rudder input mapped. This can be done with a "split"
input to the tails in exactly the same way that the aireron works on
the wing. It will simply add to the elevator input. V-tail
Bonanzas will require the same treatment.
elevator and trim give ridiculously high airspeeds.
Increase the elevator effectiveness slightly.
Increase the vertical stabilizer effectiveness, and give it a 1.5
degree incidence angle, to compensate for asymmetric propeller
effects.
Increase the rudder effectiveness to help with cross-wind takeoffs and
landings.
Here're some updates for some of the instruments, the TSR2 and the B52.
The instruments are just parameterised/non-specific versions of the engine
and fuel gauges. The old gauges will still be used by the a10s and the sea
hawks until I get them done.
There're a couple of tweaks to the tsr2 yasim config and amendments to the
model file to change the angle at which the airbrakes operate, correct the
direction that the nose gear retracts/extends and to rotate the main-gear
carts during retraction/extention. There're also new panels, using the new
instruments.
The b52 yasim config has some big changes due to finding some more info i.e.
wing incidence of 6 degrees, fuel capacities and aileron changes (removed
from G & H versions!). It's now based on an 'F' model and I've put some
comments in the yasim config about it. The b52-readme.txt includes a
suggested method on getting airborne;). I've included an amended model -
B52-F.3ds, which is the default in the new model file, and a couple of panels.
I'm quite pleased with the b52 - the take-offs resemble the photographs I've
seen and I think I've got reasonable values where I've had to guess at stuff,
but the results from the yasim solver look as through the model is
approaching the limits of acceptability for yasim. By this I mean that the
model is probably a bit dodgy, not that the yasim solver is limited.
speed/cruise speed setting was better and I've reverted to that - they were
too fast at altitude:) They are closer using Andy's figure.
They don't quite have the required range though:(
There's now another a10 variant - a ferry load configuration - no weapons but
3 x 600 us gal external tanks. This seems to be the heaviest confguration
for an a10. I also found a performance graph for the a10 in clean
configuration and it showed max speeds of 275 knots at 45000 ft and 398 knots
at 1200 ft. As things seem to get a bit squiffy above 35000 ft, in my
experience, I used the 1200 ft figure. This is pretty close the the figure
that Andy Ross found (380 knots at 0 ft) and I've also tried to incorporate
his mods into the new configs, with a better balanced wing/hstab settings.
After getting a good clean config, I added the extra masses/fuel
tanks/fueselages but didn't change any of the rest of the config, and it
still seems to work properly. The external tanks were done as fuel tank
entries, with corresponding fuselage entries, whereas I'm not sure how to
handle the 'weight' entries in yasim yet so the weapon load is still done
with an increased total mass and ballast entries.
There'll be some stuff in the archives that haven't changed i.e. the cl & wl
a10 and seahawk models, but until I get using cvs for my own bits and pieces
it was easier to do the whole directory. I hope this isn't a problem.
Finally, I've set up some panels, based on the c310 vfr and mini panels. The
vfr panels have a lot of additional instruments on them but impose quite a
high frame-rate hit on my system. The mini panels also have some additional
instruments but seem to work without noticable penalty here. I've set these
panels as the defaults for these aircraft but I'm not sure it's a good idea
to have the vfr panels by default due to the performance hit. Using them as
default will should give some feedback though;)
To go with the new panels are a number of rough and ready instruments I
hacked, mostly out of one of the rpm gauges, but I've adapted the throttle
quadrent (jet-throttle-quadrant) to funstion for throttle, flaps and reheat.
All the gauge faces were based on an existing ffgfs instrument texture, so
there's no problem with copyright there. All the other bits that I've done
may be distributed under the same conditions as fgfs.
The only real changes to the seahawk stuff were to add the panels, but I
found that Sea Hawks have three fuel tanks, with a 'saddle tank' over the
engine feeding fuel into the second/rear tank, so the yasim config's changed.
Here are some more updates, and bits and pieces.
The tsr2 is mostly a re-write, with more reasonable figures for the wing and
hstab. I've also tried to reduce the poly count on the model whilst adding
the airbrakes and canopies.
Finally, I've set up some panels, based on the c310 vfr and mini panels. The
vfr panels have a lot of additional instruments on them but impose quite a
high frame-rate hit on my system. The mini panels also have some additional
instruments but seem to work without noticable penalty here. I've set these
panels as the defaults for these aircraft but I'm not sure it's a good idea
to have the vfr panels by default due to the performance hit. Using them as
default will should give some feedback though;)
To go with the new panels are a number of rough and ready instruments I
hacked, mostly out of one of the rpm gauges, but I've adapted the throttle
quadrent (jet-throttle-quadrant) to funstion for throttle, flaps and reheat.
All the gauge faces were based on an existing ffgfs instrument texture, so
there's no problem with copyright there. All the other bits that I've done
may be distributed under the same conditions as fgfs.
The other stuff isn't really an update but something I did for a bit of a
laugh. The idea ocurred to me after animating the folding wings on the
seahawk and seeing that the aileron and flap axis still worked...
Here're a couple of updates. Well, one's an update - the tsr2 and this now
has reasonable wing and hstab values. The roll rate was very high and I
figured that as it was going to be a hack anyway, due to the all moving
tailplanes, I just made them very ineffective and linked them top the
tailplanes in the model file. I think I may have over factored them a bit
too much but it's interesting watch them in chase view. If the autopilot
roll-out-deg is dropped to about two it'll start oscillating, and from behind
it looks like it's paddling:)
and crash on startup. The defaults tend to assume that load is borne in
proportion to the gear's distance from the c.g., but taildraggers are more
complicated. Increase the spring strength.
Here's a B52-H that I've done and it can be distributed under the same
conditions etc. as fgfs. It seems to fly ok but I'm sure it could do with a
bit of adjustment - perhaps Andy R. will have a look and give it a few tweaks.
It wouldn't need too much work to convert it to an -A model (different
engines, taller fin etc.) and then it could be teamed up with the X15 :)
I still can't get textures on the models but it occurred to me that if
someone could apply blank texture maps to the models I could then fill in the
blanks, as it were.
I also noticed that there don't appear to be models included for the F-104,
F-15 & F-16 and I think I could probably do some and animate them for fgfs if
they'd be useful.
Re the A10 - Andy said he'd had a look at it and made a few changes to fix a
few things that I'd done incorrectly and asked if I wanted to keep tinkering
with it or should he roll his changes into the fgfs cvs? I said to him that
I'd be happy to go with his changes as they'll be more realistic/accurate than
the stuff I've done.
I've certainly no problem with anyone doing anything with any of the stuff
I've done, so long as it's an improvement :)
I've taken the liberty of attaching a .tar.gz file containing a .3ds model of
a BAC-TSR2, a yasim config file based on the correct figures (where I could
find them) and the -set.xml and model.xml files to fly it.
I'm primarily a 3d'er and originally did the TSR2 for a picture I'm working
on but when I got fgfs running (Debian Linux) I couldn't resist loading it in
and trying to get it to fly. The model was created in Realsoft3D and
exported as .3ds.
I've been able to tag the various sub-objects, to animate them but the export
process appears to 'flatten' any object hierarchy I set up and I'm guessing
this is necessary for sequential animations - I couldn't set up the correct
sequential rotations to properly bring the main u/c in. Also, in real life,
there are several other u/c doors that should open and close in sequence to
get the gear in and apparently the sequence was quite complex.
On the ground though, it is as shown (so I didn't need to model the extra
doors for my picture anyway;)
It could do with some airbrakes too, both for the model and for the fdm. As
with the extra u/c doors, I didn't need them for the picture and they haven't
been modelled.
As well as not being able to preserve object hierarchies when I export from
Realsoft3D's native object format to .3ds, I'm not able to preserve textures
or colour-mapping either, so the aircraft appears all white.
Hopefully, someone might like to add the extra doors and airbrakes, which
shouldn't be too difficult, and put some texturing on it - mostly white
anyway, for the prototypes, or a contemporary RAF scheme if someone wants to
pretend that it entered service.
The yasim fdm model, as said, cannot be regarded as accurate. However, while
it is based on the specs for the real aircraft, where I could find them, the
measurments are probably only accurate to about 1 metre. That's assuming I
was measuring the right things in the first place;) Other bits that I wasn't
sure about i.e. flaps, ailerons etc. have been hacked out of the a4 or the
747.
It could do with some 'refinment' by people who know what they're doing, but
it seems to fly about right, or rather, as I'd imagine:) (me want a
forward-looking ski-toe terrain avoidance radar:)
Anyway, I'm happy for the whole lot to be released under the same licence and
conditions as the rest of the fgfs stuff, either as a part of fgfs or by
anyone else who will also follow those same licence and condition terms.
I've also got a reasonable yasim b52 flying but no moveable bits on the model
yet, a vulcan with a similar simple 3d model using a grossly hacked c310
jbsim fdm (right numbers where I could find them but powered by a couple of
XLRs) and a Saunders-Roe SR45 Princess flying boat model and yasim fdm (can't
get it into the air with propellors but substituting equivilent jets (2.5x
factor) got it flying. I've started on a EE/BAC Lighting FMk6 model but I'll
probably be doing a Fairchild A-10 and an Antonov An-225 first.
I figure this is the best way I can contribute to the fgfs project, and l'd
like to be able to offer something.
