Flight model / Overspeed issue
-
… @Blu3 and Cloud9
Most of the time, we joke, we laugh, sometimes we sigh … then just after, we launch the soft to try to repro and report it on the Dev forum. What happens next … just look last Jp post.
I do not say that we are wizards … but in most cases, we take a look and see what can be done. Sometimes it is fixable … sometime it takes a lonnnnng time. Sometimes it is more problematic or just impossible.Remember … this is “Falcon4.0” … it could have been named “Compromises40”.
Let’s enjoy the “Funcky Chiken” and wing flex in cockpit view … some times ago, it wasn’t there. … meanwhile, hope for more. But let’s try to stay reasonable.
-
@Cloud:
What is wrong with ppl?? Blu3Wolf was simply responding to DJ’s post:
You are correct. I misread his post. My post has since been edited and I again apologize to Blu3Wolf.
-
ok and now the fun begins :
Which clearly indicated , as i said in my first post : this IS FPS related.
not saying this is because of too low FPS, but just the fps gives a timestep wich makes the RK4 hard to be precise at those speeds….
The bad news now
if i can not reproduce it on my rig, i can not fix it
Oky guys, so now, since it is FPS related, if you want Jp to be able to repro and fix it … you will have to give him a donation to purchase a “new” kicking ass computer.
Go figure.
-
@ Mav-JP
I am still getting the “jetwash bug” … or to be more precise
if i fly maximum speeds (or close to > 700kts) at low altitudes (300feet-1000feet) … my nose starts “vibrating” to left and rigth (like a “rudder flattering”).
For test i made the deadzone of rudders maximum to ensure, that it is not caused by false minute inputs, because this “flattering” can be induced by rapid rudder input changes (or maybe “spiking” stick inputs too?!?) …still same.
My FPS is over 100 and i dont change views.THE PROBLEM IS… this is very random, so i can not reproduce it either sometimes, meaning i have to find (create) more speficic regimes in order to reproduce it.
Another phenomena is, that if i pull high Gs and pass from sub- to super-sonic regimes (and vise versa) while under G-load, i can feel a tiny little “bump” in the pitch of the craft. This does not concern me much and feels quiete logical in fact.
-
Regarding the timestep issue, when I’m doing researches on flight controls, I recreated the F-16 FLCS according to a General Dynamics’ F-16 control block diagrams in another platform some time before. At that time I used 0.006 seconds for timestep and I found that the whole plane behaves laggy and have pitch oscillation above 700kts.
I know that the F-16 FLCS has a ton of filter and I’ve also modeled them line by line, but that’s too laggy compared to any F-16 HUD video. Then I adjusted the timestep to 0.009 seconds, and the pitch becomes more crispy, the aircraft no longer oscillates > 700kts.
-
Interesting.
-
Oky guys, so now, since it is FPS related, if you want Jp to be able to repro and fix it … you will have to give him a donation to purchase a “new” kicking ass computer.
Go figure.
If it is low FPS related, that can be tested for with his existing one I suspect. Just start running prime95 in the background.
@A.S:
Another phenomena is, that if i pull high Gs and pass from sub- to super-sonic regimes (and vise versa) while under G-load, i can feel a tiny little “bump” in the pitch of the craft. This does not concern me much and feels quiete logical in fact.
Other than how regular it feels, it seems not unusual to have that kind of kick in the transonic regime.
-
@A.S:
@ Mav-JP
I am still getting the “jetwash bug” … or to be more precise
if i fly maximum speeds (or close to > 700kts) at low altitudes (300feet-1000feet) … my nose starts “vibrating” to left and rigth (like a “rudder flattering”).
For test i made the deadzone of rudders maximum to ensure, that it is not caused by false minute inputs, because this “flattering” can be induced by rapid rudder input changes (or maybe “spiking” stick inputs too?!?) …still same.
My FPS is over 100 and i dont change views.THE PROBLEM IS… this is very random, so i can not reproduce it either sometimes, meaning i have to find (create) more speficic regimes in order to reproduce it.
Do you Guys even read what I wrote ?
-
Regarding the timestep issue, when I’m doing researches on flight controls, I recreated the F-16 FLCS according to a General Dynamics’ F-16 control block diagrams in another platform some time before. At that time I used 0.006 seconds for timestep and I found that the whole plane behaves laggy and have pitch oscillation above 700kts.
I know that the F-16 FLCS has a ton of filter and I’ve also modeled them line by line, but that’s too laggy compared to any F-16 HUD video. Then I adjusted the timestep to 0.009 seconds, and the pitch becomes more crispy, the aircraft no longer oscillates > 700kts.
It is hard to make time step in f4 fixed which makes things more tricky
But as you can see the FM is very smooth
-
@A.S:
Another phenomena is, that if i pull high Gs and pass from sub- to super-sonic regimes (and vise versa) while under G-load, i can feel a tiny little “bump” in the pitch of the craft. This does not concern me much and feels quiete logical in fact.
Look in your EM charts , between 0.9 and 1.0 Mach ,the lift is very discontinuous .
This is the bump you feel
-
Do you Guys even read what I wrote ?
Ofc ! Just trying to help/add information. You had problems reproducing it, no?
I am not the only one who has expirienced this “random flatterering” and this also happens with constant and high FPS.
The ONLY way i can reproduce it “by force” is to fly fast and violently give right-left rudder inputs.
Then it resonates (continues) on its own, which makes me wonder, if there is maybe a “buffering false input” produced somehow, or if input devices of certain individuals are the cause. -
-
With All Respect Ninja, I was responding to DJ.
My apologise Blue I misread then I thought you was referring to Jp’s post
-
Yeah sorry I am a bit picky those days LOL
Nah, you are just pround about a phantastic FM - and you can be
-
the time steps are half the normal when speed is over 500 kts to help precision of the algorythm ….
i can triple the loops above 700 , that should help …but that will eat some fps (not that much , RK4 is light)
Would it change something for us that have very powerfull PC to have an option to have always tripple he speed(like you suggested to do for 700kts) or we wont see any differences?
-
The fact that fps changes between cockpit and external make the time step of the RK4 algorythm change , which can play a significant role in convergence at the edge of the model
The fact that the phenomenon happens in a certain view and not another is just a consequence of this time step change .
Could you do the aero calculations in another thread and put the results back into the sim that way? As a way to make that particular set of calculations time step be independent of the FPS? Obviously that would introduce the issues of multithreading though.
-
Could you do the aero calculations in another thread and put the results back into the sim that way? As a way to make that particular set of calculations time step be independent of the FPS? Obviously that would introduce the issues of multithreading though.
What happens when the display frame rate changes and the sim is operating at it’s fixed frame rate? You’ll get temporal aliasing, objects will jump across the screen as their aero calcs are feed into render. Not to mention when you start multi-threading there are a host of other issues. You’ll need to buffer and store your interpolations, then send the position updates to the render. Then there is fact that falcon 4 is 18 years old and would need to be rewritten entirely to make use of multi-threading. The increased number of instructions and buffer, eat up cycle time and memory, so there may not be a huge performance gain for the effort involved in an edge case.
There are some other solutions to this issue. One involves locking the frame rate and this is easy to do in game, Just turn on Vsync. People tend to not like it, as it has a habit of locking the frame at the lowest multiple of their refresh rate. The second is a bit more complicated, get your dt from the render and then intergrate using a step of that dt. You end up with accumulating time, which you store as a variable. This variable is used to interpolate the current physics state from the last, thus smoothing out any errors due to the variable frame rate. You send this lerped physics state to the render to be displayed.
I can see why you wouldn’t want to use this approach as you’re not really simulating the given variables at any moment. You’re lerping between last state and current. Which is why it appears that mav’s approach is to increase the time step in the integration. He’s actually simulating the forces at a given moment and sending it the rendered and advancing a head. Which is why you can end up with edge cases where the accelerations and frame variance produce positioning error.
Here’s one the basics of Integration and physics sim, if you want to understand rk4.
http://gafferongames.com/game-physics/integration-basics/And another that covers time step and render times. This provides a better explanation of the approach described above with easy to read code examples.
http://gafferongames.com/game-physics/fix-your-timestep/ -
There are some other solutions to this issue. One involves locking the frame rate and this is easy to do in game, Just turn on Vsync. People tend to not like it, as it has a habit of locking the frame at the lowest multiple of their refresh rate.
This is exactly what my buddy and I did to fix the problem. With a stout enough system and a good monitor it will lock in at 60 FPS no problem. I’d much rather have 100% stability in flight and 60 FPS than have 110+ and wing rock.
-
If we could confirm that actually too many fps are the problem for high speed I could increase the time step as well in those cases !!!
And if you wonder why BMS fm is so smooth compare to others sims , now you start to get it adaptative RK4 with full physics in is a bitch
-
If we could confirm that actually too many fps are the problem for high speed I could increase the time step as well in those cases !!!
And if you wonder why BMS fm is so smooth compare to others sims , now you start to get it adaptative RK4 with full physics in is a bitch
Well from my testing years ago the lowest FPS I had at the worst duplication of this problem was 50 FPS all the way up to 100+ so I’d say that is definitely part of the problem. That being said since you can simply turn on VSync and tame this I’m not sure if adjusting this would be a worthwhile venture if it has any possibility of causing issues with lower performance systems.