#250 crosswind conniptions

[Anonymous]

Originally created by: g...@av8n.com
Originally owned by: g...@av8n.com

Have you ever tried to take off with a crosswind?

I started the sim with the following options:
--runway=19L
--disable-real-weather-fetch
--metar="XXXX 012345Z 07315KT 50SM FEW015 SCT025 BKN035 OVC090 15/M01 A3092"

I'm using the flagship c172 at the default airport (KSFO). 
The wind is a crosswind from the left at 15 knots, with a
tiny tailwind component.

After the sim initializes, before starting the engine I
observe that the aircraft is in a nose-high attitude.
This is wildly unrealistic.

After starting the engine, by manipulating the throttle,
elevator, and brakes I can get the nose to go down to
the normal 3-point attitude.  I can also get it to go
back to the nose-high attitude.  All of this is wildly
unrealistic.

While holding the brakes I fully open the throttle, then
release the brakes.  The aircraft immediately pivots to
the right, i.e. downwind.  This is wildly unrealistic;
note that in the real world the aircraft will weathervane
*into* the wind.

I conjecture that the root cause may involve
*) the lift and drag coefficients for the rudder being
   completely wrong at angles of attack near 90 degrees
*) the lift and drag coefficients for the elevator being
   completely wrong at angles of attack near 180 degrees
*) the model for the effect of propwash being wrong also

The basic set of problems is 100% reproducible, although
details differ.

The c182rg is beset by similar problems -- to an even worse
degree.  Hold the brakes, open the throttle, release the
brakes, and it turns violently to downwind.

On the other hand, the SenecaII (another JSBsim aircraft)
and the pa25-250 (yasim) seem to behave much more realistically
under crosswind conditions.

See attached barf for details of version and configuration.

[Anonymous]

Originally posted by: anders.g... (code.google.com)@gmail.com

(No comment was entered for this change.)

Labels: Aircraft

[Anonymous]

Originally posted by: zakalawe@mac.com

(No comment was entered for this change.)

Labels: JSBSIM FDM

[Anonymous]

Originally posted by: stuar... (code.google.com)@gmail.com

The nose-high attitude due to tailwind has now been addressed. I'm not sure about the other issues.

John - could you re-test please?

-Stuart

[Anonymous]

Originally posted by: gijsrooy

(No comment was entered for this change.)

Status: Testing

[Anonymous]

Originally posted by: bre... (code.google.com)@gmail.com

(No comment was entered for this change.)

Summary: c172: crosswind conniptions

[Anonymous]

Originally posted by: cumuluni... (code.google.com)@gmail.com

The nose-high attitude due to tailwind _is_ fixed. The turning downwind is still there (which is a bug). Prop-wash is not modeled (but could be - this is a feature request).

Labels: Priority-Low
Status: Accepted

[Anonymous]

Originally posted by: bcoco... (code.google.com)@gmail.com

The interpretation of the turning downwind is incorrect and is therefore not a bug.

It is actually due to the propeller P-factor which generates a moment bigger than the "weathervane moment". As a result, the aircraft turns right at first but if you wait a bit you will see that the "weathervane moment" overwhelms the P-factor moment and the aircraft starts turning left as expected.

This can easily be confirmed by observing the 2 following properties as the aircraft starts

/fdm/jsbsim/moments/n-aero-lbsft
/fdm/jsbsim/moments/n-prop-lbsft

At full throttle, brakes applied, with the weather conditions given above, we have:

n-aero-lbsft = -102.95
n-prop-lbsft = 443.59

After releasing the brakes the absolute values of both moments decrease until n-prop-lbsft eventually become negative at which point the aircraft starts to turn left as expected.

This can also be easily checked by setting the p_factor parameter to 0 in fgdata/Aircraft/c172p/c172p.xml line 212.

Now the strength of the P-factor can possibly be questioned but it is a matter of tuning.

Status: WontFix

[Anonymous]

Originally posted by: bcoco... (code.google.com)@gmail.com

After some more thoughts and especially after having read JSBSim bug http://sourceforge.net/tracker/?func=detail&aid=3171743&group_id=19399&atid=119399 (which is actually unrelated to our issue), I think the P-factor formula is currently not handling well high values of alpha and beta and this must explain the "hard tendency" to turn right.

Summary: c172: P-factor with high alpha and beta values
Status: Accepted

[Anonymous]

Originally posted by: g...@av8n.com

It appears that comment 8 must be considered a retraction of comment 7.

In any case, let me follow up on comment 7.  Everything in that comment
is wrong, as one can easily discern in multiple ways.  First of all, in
the case of a crosswind from the right, I observe a pronounced tendency
to turn downwind -- to the left -- in the flagship c172p model.  This
simply cannot be explained by P-factor or anything else having to do with
the propwash.  Secondly, if we build up a little bit of speed and then
retard the throttle to idle, I observe a pronounced tendency to turn
downwind (no matter whether the wind is from the left or right), which
cannot be explained by anything having to do with propwash, since the
engine is at idle.  Thirdly, looking at the code in c172p.xml, there is
no way this code could produce the correct result.

I made these observations using a fresh pull of c172p.xml.  The last commit
to it was the notorious quick-and-dirty "sin" patch: [rb7fb005052cefb31f83fbf73e787187747cf93fb]
Author: Stuart Buchanan <stuart_d_buchanan@yahoo.co.uk>
Date:   Fri Feb 11 20:16:49 2011 +0000

    Stop c172p pitching backwards in a tailwind. Patch from Ron JENSEN.

As a very very minor tangential point, please note that P-factor properly
speaking refers to asymmetric disk loading.  There is no asymmetric disk
loading during the initial takeoff roll in a tricycle gear airplane, not
of the kind that could produce the left-turning tendency that we observe
in the real aircraft.  The actual effect is properly explained by propwash
hitting the side of the vertical fin and rudder ... not by P-factor. 

Misconceptions about this are very widespread, but they remain misconceptions.
For details, see
  http://www.av8n.com/how/htm/yaw.html#sec-helical-propwash

I have no idea who changed the "summary" line of this bug or why, but it
was not an improvement.  I am changing it back.  Also note that the issue
is not confined to the c172.

Additional observations and root-cause analysis and suggestions for how to
proceed have been discussed on the flightgear-devel list e.g. 06/19/2011
09:50 AM,

Summary: crosswind conniptions

[Anonymous]

Originally posted by: bcoco... (code.google.com)@gmail.com

Mr. Denker,

Everything I said in comment #7 is correct and can be checked by anyone simply by running FG and observing the properties I mentioned. Have you tried to re-run your own test with the P-factor parameter set to zero ?

Comment #8 is an addendum in that I initially implied that P-factor calculations just needed tuning but they actually need some more considerations.

Now you are free to defecate your arrogance on the elements I brought but I would prefer to see facts that can be checked.

[Anonymous]

Originally posted by: cumuluni... (code.google.com)@gmail.com

Hmm - two sides fighting, both claiming to be right. We seem to have no judge here and I tend to apply "in dubio pro reo" here - but who is "reo"?

My solution: Status=Stalled for now.

Status: Stalled

[Anonymous]

Originally posted by: bre... (code.google.com)@gmail.com

Issue 434 has been merged into this issue.

[Anonymous]

Originally posted by: digi.furry9@yahoo.com

May I add my tests & observations to this conversation? 

I assume we are all agreed that single engine, puller prop airplanes turn left when the nose/tail wheel leaves the ground during takeoff roll.  I accept too that planes tend to "weathervane" into a crosswind.  With that in mind, I made some "reality-check" tests with my default 2.4 release on WinXP SP3, using KSFO in its default state. 

Wind at ground is 3 kts, from bearing 150 (from terminal across runways).  I used runway 10L/28R, so the crosswind is at a slight angle.

C172 on 10L: Wind from right.  Turns slightly right on takeoff roll, then very slight tending left at liftoff.  With crosswind only 3 kts, and 172's natural left turning tendency, I would expect that at best the turning effects would cancel each other out.  But at least the right turn is mild.  This might be acceptable.

C172 on 28R: Crosswind now from the left.  Plane turns hard left on takeoff roll.  Combination of natural left turn plus weathervaning: OK, this one is believable.

DHC2 (Beaver) on 10L: Wind from right.  This powerful radial-engined 4-seat tail dragger would be expected to have a strong torque and strong left-turn tendency.  Here, as the tail wheel lifts, it pulls hard right.  Much harder than the C172.  Further, with no correction it goes across the grass toward the terminal and never turns left even a little.  With only 3 kts of crosswind, I submit that this cannot be supported by the weathervaning concept.  No matter what the internal spec's of FG claim, this is not believable.  I submit the simulation is simply wrong here.  Too much weathervaning effect.

DHC2 (Beaver) on 28L: Here, this plane pulls hard left, a behavior to be expected.

Since this turning-into-crosswinds phenomenon did not happen in FG 2.0, it seems to be a new addition in 2.4.  I respectfully submit, gentlemen (& ladies, if any), that it seems a good idea, but is a little overdone right now and needs to be toned down some.

[Anonymous]

Originally posted by: digi.furry9@yahoo.com

Testing update.  Discovered how to adjust simulated wind. 

1) C172, KSFO, 10L.  Set wind to 0kts (no wind).  Slight left turn during takeoff roll.  Interestingly, it turned left while nose wheel was still on pavement.  No turn when nose wheel left pavement.  Real planes exhibit their turning effect when their steering wheel leaves the pavement and loses its effect.  Maybe this needs some adjustment too.

2) C172, KSFO, 10L.  Set wind from 100 (3 kt headwind, straight down runway).  Same as above.

3) DHC2 (Beaver), KSFO, 10L, wind from 100.  Strong left turn AFTER tail wheel left ground.  Not as strong as previous test (Comment 13) on 28L.  Very believable behavior.  (Very nice model!)

This validates idea that right-turning behavior that caused this thread is caused by simulated cross-wind weather-vaning effect.  Simulated effect does seem too strong.

[Anonymous]

Originally posted by: p...@freeflightsim.org

Is this DEAD? its an olde bug

[Anonymous]

Originally posted by: cumuluni... (code.google.com)@gmail.com

assuming, it's dead.

Status: WontFix