mav-jp

i know that all Falcon4 versions use HFFM , that is for sure... the question is did FF5 change the Falcon FM core ? answer is no therefore Falcon4 FM remains shit :)

ho yeah i have already investigated that possibility, i.e creating a CL/CD curve for each surface and compute locally the speed, alpha, beta to get the force from each surface and then integrating the whole thing It works..., but : 1) is extremly difficult to create the data (you need to know the CL/CD of elevator, rudder , etc etc.. which is difficult to have). 2) you dont simulate interaction of the surfaces with the others, so you need aditionnal modeling for interaction. all in all..it works but it's difficult to get it work right ...whereas globalmodeling can be much more accurate (all effects are in already)

reading the NASA implentation is a good start :)

correct , and after 40 years of testing, some people have concluded that for aerodynamic effects, most of the dependances are of the second, third of more orders which allows us to simplify / reduce the problem. Why the hell do you want to re do the work already done ?

you have never seen NS modeling ? you need to add some turbulence equations modeling, (like kolmogorov) plus boundary limits equations blah blah blah ... what about FF5.5 ???

what you are saying ha no meaning to me: *Whether you have an interpolated linear model and you know perfectly what are the number of parameters you want to use and the order of importance of it, in that case, i think the NASA model is quite correct and as explained you dont need 8D data... * whether you have a non linear model and the last known is Navier stokes... I dont deny you could propose a linear 8D model but it can NOT be unknown , you need to know it to measure it - or compute it with non linear model.

the fact that you would chose 8 parameter is already a model ;)... you can not say 'i have no model', even NS is a model ... and by the way, I dont have "My" model, NASA had a model ;)

Anyway , to get back on the original subject. FF5.5 has the same falcon4 FM ? ;)

In that case the 8D assumption is already incorrect, because it comes from a simplified model. you then need a more complex model rather than an interpolated model. but i'm saying complex model is NS and is not usable for real time computation. 8D has no interest because you use dat or order of importance too low. so there is no "in between" like 8D table there is 1) NS 2) Physical interpolation which needs ~ 2 Mb to get accurate modeling, adding data would be useless i dont believe in the 3) simplified NS , which is what X plane does... :)

this is what i call math masturbation and has absolutly no interest for engineering purpose ;) ;)

Well what people tend to forget is that EVERYTHING IS A MODEL, even the Navier Stokes Equations are a model ; the Turbulence equations are a model....so saying you can find the analytic solution of the problem is pure Math mind masturbation. The reality : * Navier Stokes equations + Turbu + boundary limit equations are a model and can not be used at the moment in real time. * For Real time simulation you can use * Simplified real time computation models * Physics model with tabular data (pre calculated by NS or testing) I'm saying when you are able to get sufficient data, tabular model is far far better than simplified real time computation. however, getting those tabular data is very time consuming and can be costly. NOW this is only for aero forces, of course after that you need to use a real EOM solver to compute the trajectory Falcon4 has very poor Physics modeling and tabular data Falcon4 have no EOM solver. End of story cheers

This assumption is very incorrect, saying that the function is unknown is saying that we have absolutly no knowledge of what aerodynamics is. OF course we know what kind of function is interpolating the best the reality (see equation i wrote above) and THANK GOD people have worked during hte 20th Century to understand the order of importance of PArameters !!!! You way of thinking remind me myself while i was studying purely math at University (before studying enginnering ). Think ENGINEERING, think PHYSICS :)

Your assumption of equal dimension is incorrect: As the parameters have diffferent order of importance, the dimensionnal table to decribe their effect is not identical. stop thinking Math or Code, think Physics, you need to know the order of importance. To give you an exemple : if you take Cm (alpha,Beta, deltah, lef, sb,q ,ds) Cm can be expressed (read page 38 in Appendix B): Cm (alpha,Beta, deltah, lef, sb,q ,ds) = Cm(alpha,Beta,deltah) * Nu,deltah(deltah) + DELTACm,Lef(1-deltalef/25) + DELTACm,sb(alpha) *(deltasb/60) + Cbar*q/2V*[ CM,q(alpha) + DELTA Cm q, lef (alpha) *( 1 - deltalef/25)] + DELTACm(alpha) + DELTA Cm,ds(alpha,deltah) * Cm(alpha, Beta,deltah) has 20 alpha Bkpts, 19 beta Bkpts, 5 deltah Bkpts. this means a table of 1900 float values * Cm,lef(alpha, Beta) has 20 alpha Bkpts, 19 beta Bkpts, this means a table of 380 float values * DELTACm,sb(alpha) has 20 alpha Bkpts, this means a table of 20 float values *CM,q(alpha) has 20 alpha Bkpts, this means a table of 20 float values * deltah(deltah) has 5 deltah Bkpts, this means a table of 5 float values * DELTA Cm q, lef (alpha) has 14 alpha Bkpts, this means a table of 14 float values * DELTACm(alpha) has 20 alpha Bkpts, this means a table of 20 float values Finally to accuratly describe Cm , you need here 2359 float values. Size in the RAM : 2359 * 4 = 9436 bytes = 9,436 Kb !! Assumin you need more or less the same size for the 5 other force and moment adimensionnal coefficient (which is wrong because you dont need as much for Cn or Cl for instance)... you would get : 9,436 Kb * 6 = 56,616 Kb. Now , in the extreme case, you need all those coefficients for lets say 20 mach values (reynolds), Overall 1,132 Mb is needed.... NOT A BIG DEAL for the RAM, Believe me, the problem is not the storage in the RAM of those coefficient, the problem is the calculation of those coefficients and to WRITE THEM in a file LOL ;)

Your assumption is incorrect , In m'y previous post i was stating the dependancies but i should add the order of it becuse in order to correctly and accuratly compute a variable dependant of a second order parAmeter you dont need N but most likely N/50 variables. Read the Nasa document i have provided the simulation used in there has been used to develop thé fLCS of the real. As you CAN see there are indeed thousands of variable here but nothing close to 1 Gb :) and Bélieve me this was accurate enough to test and develop the deep stall situations which means a lot ( extrême enveloppe ) cheers

Aditionnaly some parameters have a importance of a second order , for example Cl vs deltar , therefore a simple table like 20 coefficients is only required to handle the coupling... second order parameters dont need as much values as first order :) i can GARANTEE you that our computers today have FAR enough RAM to handle the necessary tabular data to go donw to second order influence which is by far too much already of course, getting those coefficients is quite tricky and time/cost consuming. For the F16, hopefully, Wind data tunnel exist and of course experiment has been done for a reynolds that covers most of the flight regime... for Cl,Cm,Cr (mach 0.6) And HFFM covers other Reynolds for CL/CD because it is retro engineered :) the time when real time aero parameters computation is better than tabular is not here yet, maybe in a dozens of years ?

your assumptions are incorrect for all elevator positions, 5 points are very sufficient to get correct interpolation for some alpha cases 20 are sufficient for beta 19 are okay like this : # deltah "m" BREAKPOINTS # 5 # Number of deltah "m" Breakpoints -25 -10 0 10 25 # # deltah "m2" BREAKPOINTS # 7 # Number of delta "hm" Breakpoints -25 -10 0 10 15 20 25 # # deltah "m3" BREAKPOINTS # 3 # Number of delta "hm" Breakpoints -25 0 25 # # # alpha "m" BREAKPOINTS # 20 # Number of alpha "m" Breakpoints # -20 -15 -10 -5 0 5 10 15 20 25 30 35 40 45 50 55 60 70 80 90 # # alpha "m2 - LEF" BREAKPOINTS # 14 # Number of alpha "m2" Breakpoints # -20.0 -15.0 -10.0 -5.0 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 # # beta "m" BREAKPOINTS # 19 # Number of beta "m" Breakpoints # -30 -25 -20 -15 -10 -8 -6 -4 -2 0 2 4 6 8 10 15 20 25 30 # of course for CL/CD , more breakpoints are necessary like : # MACH BREAKPOINTS # 22 # Num MACH 0 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.95 1.0 1.05 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 # # ALPHA BREAKPOINTS # 46 # Num Alpha -20 -15 -10 -5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 14.5 15 15.5 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 40 55 70 90 # # BETA BREAKPOINTS # 19 # Num Beta -30 -25 -20 -15 -10 -8 -6 -4 -2 0 2 4 6 8 10 15 20 25 30 # all in all, i can guaantee you that you dont need 1GB ram to simulate a very accurate tabular model :) if you want more explanations about tabular data : read : http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/19800005879_1980005879.pdf and feel free to count the number of values :)

meuh non LOL!! you dont need 1000 steps per dimesion, you FOOL !! ;) for alpha, a step of 1 degree is largely sufficient in the -20 / °35 range, step of 5 or more after that, beta : no need more than -30/+30 Deltah : between -20 / +20 ; step 1+ largely sufficient etc etc ..it is doable with sufficient precision :)

Basically There are 6 “global” aerodynamic coefficients to completely describe a model: The Forces: ·CL : Lift Coefficient (reference airflow) ·CD : Drag Coefficient (reference airflow) ·CY : Y-Axis Force Coefficient (reference body) The Moments: ·Cm : pitching moment ·Cn : yawing moment ·Cl : rolling moment List of Variables: α = Angle of Attack β = Side Slip Angle p = Roll rate q = Pitch rate r = yaw rate δsb = Speed Brake deflection δlef = Leading Edge flap deflection δh = Elevator deflection δr = Rudder deflection δa = Aileron deflection If you can get the tabular data for the dependances as below , you will have a VERY accurate description of forces and torques. Those data can be calculated via a 3rd party NS solver (not in real time) or from wind tunnel testing Variable dependence CL / CD ................................................ Retro Engineered CY........................................................ α, β, r, p, δa, δlef, δr Cm........................................................α, β, q, δh, δlef, δsb, Cl........................................................α, β, r, p, δh, δa, δr, δlef Cn.......................................................α, β, r, p, δh, δa, δr, δlef

no hard feeling , but it seems to me that you have misunderstood what the problem of falcon4 FM is. Let me try to explain. First thing to understand is that a fligth model (and generally any motion model) is 1) a model that computes forces and torques applied on the model 2) A routine that solves the 6 dof motion equations taking as input the forces and torques and giving as output the speeds, rot speeds, accel and rot accels. In the specific case of Flight model, the forces and torques are generally dependant on motion itself (i.e Speed, AOA, BEta etc...) which make the differential equation of motions beeing of the second order (or more) and coupled. Aditionnaly, the aero forces are very often dependant on several parameters, for instance , CL (adimensional coefficient for Lift) is dependant on AOA and Speed...(approximation). From that you have two options: 1) you decide to have a linearisation of the forces and torques with some of the parameters , CL (mach,alpha) , CD (mach, alpha) etc...which means you have TABLES of coefficients. 2) you try to calculate the forces and torques by using complete resolution of Navier stokes equations (not applicable for simulation) or using simplified aero equations (Xplane for instance). The solution (1) is in 99% of the case the one chosen in fligth simulation (including pro simulation)because it allows the best precision and is very fast. Aditionnaly in some extreme cases, all the coefficients static and dynamic are available wether from real wind tunnel testing, wether from complete NS calculation (not done in RL time). SO when you have a complete desciption of the forces and torques, then you can solve the equations of motions, which is not a big deal because it is hopefully simplified in coupled equations of the 1st order and can be solves by algo like Runge Kutta... Now what is the problem of falcon4 ? well falcon4 has two major flaws : 1) The only aero data that are available for the model are CL/CD and thrust (partial CY as well) ..nothing about Cl,Cd, Cn (torque coefficients) ,and the dozens of other aero coefficients for specific cases (LEF/ TEF, Sb, etc...). Aditionnaly those aero coefficients in F4 are NOt dependant on the position of the controls and they are not dependant either on Beta. TO summarize: the problem of the data does not come from the fact they are tabular , it comes from the fact that they are not exhaustive and not dependant on vital parameters. 2) there is not resolution of the equations of motions, in fact the position of the GC is computed by Euler integration, but the rotation in the x,y,z axis are not computed but are generated from a FLCS code routine... It means that the position of the AC is somehow independant from the rotation motion. In fact , AC is calculated as if it was a dot and rotation is calculated by scripts routines corresponding more or less of what the FLCS is supposed to do. As a conclusion: * dont believe using tabular data is wrong, because it is the best thing to do for real time simulation. It is very bad when you have only 1000 values, but when you have 1 000 000 of them (that can come from Wind tunnel testing or NS computation), then it is much better than any approx aero Real Time computation * the "FM on rail sensation" comes from those two flaws * Of course, real equation of motion solver would be the first step to make falcon4 FM react as a FM....but that would demand hell of a work and probably a complete rewrite of the FM code and integration in F4 code, which would be IMHO very diffcult. NOTA : As far as edge pass is concerned for the F16 , it is possible to hold it under very specific circonstances which is a very low fuel amount (<1000 lbs) . Indeed, in that case, most of the "lift" is not produced by the tail itself , but by the Engine thrust that is angled at around 8 / 10 degrees. If the speed is sufficient when entering the manoeuver, the combination of thrust/ low weight and "Lift" generated by tail + fuselage is sugfficient to allow the AC to hold it. In that specific case, the FLCS does not restrain the Yaw imput of the pilot because speed is >250 kt and AOA is low