Physics and Aerodynamic Comparison between Real Flight Test and MFS2020 - Arcade Feeling

The entire paradigm for the flight model in MSFS is supposed to be “emergent behavior” - ditch the table based flight model form FSX and get the actual basic aerodynamics and physics right, and things like ground handling, stall, spins, etc should “emerge” correctly from that.

Response to a sudden disturbance is one of the simplest tests of that basic emergent behavior in terms of inertia and damping. If this isn’t correct, then many of the other “emergent” behaviors will not be correct either.

If the emergent behavior isn’t correctly modeled then aircraft makers have to put tables of fudge factors on top of it (tables…like FSX). Only now there is a lot more complexity going on with the detailed aerodynamics model, so it actually becomes much harder to properly tune an aircraft with the complex model+correction factors for edge cases.

What this test is showing is that they didn’t even get the basics correct that this new flight model is supposed to be based on.

The twitchy handling in yaw when lifting off the ground or touching down in a crosswind for example, could be a result of the lack of inertia and damping shown in this test.

The ‘response lag’ they added to the controls settings is just a hack to hide thins behavior. No other flight sim I have ever seen or used has proposed the idea of intentionally delaying control inputs. It does not make physical sense to do so. The fact that they added that option was a huge red flag to me.

7 Likes

I hope you post this on the dev Q&A thread so that it can be gather enough votes and be acknowledged on next weeks live stream by the devs.

5 Likes

Ive been here all along, what you talking about?

I hope they fix this

Kudos to the OP for approaching this as an engineering problem and using appropriate tools to test the MSFS 172. In the same spirit, I ran a similar test on both the default X Plane 172, as well as the Simcoder add-on, which is a payware product in which, to quote the developer, “The flight dynamics are replaced to match the real world data by Cessna”.

I used the elegant data output capabilities of XP (latest update, but the classic flight model) to produce the following graph:

I should note that the response of both the default and the mod seemed essentially identical, so only one is shown. The rudder deflection is shown in yellow (this is the physical rudder, not the pedals), while the sideslip is displayed in magenta. I set both the control sensitivity and artificial stability to zero. I’ll try with other settings and let you know if that changes anything.

To the extent that the MSFS handling might be considered arcade-like, it seems that the XP handling belongs in a totally different arcade. The nose swings back and forth in a poorly damped oscillation. Neither it nor the MSFS model seems to faithfully display the characteristics of the real airplane flying in the real atmosphere.

Best wishes to all.

18 Likes

Something I can add and which in my eyes also shows that something is not correct with sideslip etc. is that the beta dot is not correct. while it should be the derivative of beta and being zero when sideslip is constant, it seems to be more like r in the body axis.

Wobble Wobble

This is my DA62 at around 120 indicated and full fuel in the wings.
yaw stability 0.2 and yaw_moment_yaw_damping -1.91730
Around 7 oscilactions

I noticed yaw_moment_yaw_damping doesnt do anything on the modern flightmodel.
EDIT: Nah, they are connected. Yaw_stab seems to be a multipler
Here with Yaw_stab 1. Not much difference, 1 oscillation less.

Here with Yaw_Stab 5


2

2.5

3

4

My DA40ng with a yaw stability 0.1 and yaw_moment_yaw_damping -11.91730
I will keep that

9 Likes

@MrTommymxr that was an excellent contribution.

With a yaw stability coef. of 0.2 it looks more acceptable, but Alexis mentioned testing the TBM with some rather low value there too… unless the TBM has some kind of an hidden yaw damper ?

really, 0.2?
Personally i think 2 looks and felt the best.

I will set 1 in my next update. I will investigate further when the thrust bug is fix and look at the flightmodel into more detail later

1 Like

This is my first time doing this stuff in a flightsim.

My personal opinion on the physics and flighmodel in FS20:

The physics are not perfect, and neither are the flightmodels.

Bad physics can be compensated for with workarounds on the flightmodel, but good physics will not fix a bad flightmodel.

A lack of adverse yaw is not a bad physics engine, but bad settings in the flightmodel.
A da20 doing wheelie with the parking break on is both.
The tail should not really be affected by the propwash and the elevator shouldn’t be that strong.
The flightmodel’s tail can be moved above the propwash and the effectiveness would need to be turned down.

What does need a change is the CL table. Im unable to get a steep enough increase in lift and have the correct stall AoA. These things have been well documented already however

3 Likes

@MrTommymxr what is interesting about your plots is they also show a lack of any inertia. As soon as the rudder starts returning to center, the sideslip does the same. In the real world flight test data it is clear that the sideslip continues to increase after the rudder returns to center, due to the inertia of the aircraft.

3 Likes

1 Like

Also my applications are not perfect and i possibly was to slow…

All 3 i would say

1 Like

It would be interesting to make a script using Joystick Gremlin that generates a consistent input deflection every single time. Would allow for more controlled testing.

1 Like

Have you flown planes in DCS that have AFM like the A10?

You can clearly feel the différence between planes that have weight & inertia in them vs those in msfs2020

1 Like

I haven’t flown DCS but I did spend four years studying aeronautical engineering. So I’ve spent a bit of time staring at the differential equations for aircraft stability. In my career I have ended up doing mechanical stuff for space applications, but I remember enough of the aircraft stuff to see when things are grossly wrong. And I kept all my textbooks and class notes and might dig through them if I have time.

4 Likes

I was not putting you skill/knowledge in question.

Just pointing out that the planes in DCS feels more “Alive” / less on rail

Sometimes you can play for free, you should try it to compare

1 Like

I’ve been meaning to give it a try now that I have a VR headset.

I do think having more inertia would indeed translate into the aircraft feeling more ‘alive’ which was something Asobo wanted to do anyway.

@Alec246 Thanks for putting in the time and effort to describe how an important part of the sim experience could be fixed for everyone’s benefit.

In addition to your investigations, I’d like to see the flight model changed to see adverse yaw, p-factor and torque effects that don’t have me off the side of the runway put in place.

I don’t expect a Level-D full motion simulator experience for a home entertainment product. Though I do expect a few key flight characteristics to be changed to better reflect real world operations.

A little more effort on the flight modelling to address these issues would, IMO, make MSFS a lot more enjoyable and a lot less open to criticism.

2 Likes

Nice but not entirely true.

  • First of all this is an old problem, as all predecessors of the present sim were actually incorrect in this regard.

  • One can still do an acceptably working model related to rudder, spring-roll or dutch-roll, the problem is that it will make the aircraft artificially unstable in normal situations, where it needs to be much more stable.

  • It is a common misunderstanding regarding the “table-based” flight model implementation, because:
    a./ MSFS has much more tables compared to its predecessors (the rest is simple marketing)
    b./ tables (being the industry standard now and a long time to come) are not something bad or avoidable.

  • All, I repeat ALL similar problems can be more or less solved in LEGACY mode, by a rightly designed flight-model, whereas the default “Modern” mode is nothing else but an experimental implementation of what Asobo feels to be the future - with the slight disadvantage that a C-172 will never be right while using the same equations or algorithms compared to a B738.

That is a futile experiment, because the actual characteristics can only be mimicked or repeated if one uses the real world test-flight-data, that simply cannot be interpreted by any other form, apart from detailed tables that represent the actual airplane, which will be always different by maker to maker, model by model.
There is NO generic physics solution available for both butterflies and commercial jets.

5 Likes