Question about turn rate

[Dutch]

A question for you FDE gurus out there....

Which aircraft.cfg or air file parameters most affect turn rate of a prop aircraft (or any aircraft, but I am concentrating on single-engine fighters)?

I'm finding many of the ones out there have unrealistically low turn rates compared to their real-world counterparts.

Thanks
Dutch

 

[WarpD]

Rate of turn is calculated thus: 1092.39 * tan(B) / V where B is the bank angle and V is the velocity in knots. Please note that in no way does the geometry or aerodynamics of the aircraft get represented in the formula. If your aircraft can not maintain level flight at a given bank angle, I'd be looking at lift and drag components in the .air file that relate to pitch and roll.

 

[Roy Holmes]

You get maximum turn rate when flying at the speed where you can apply the limiting g of the airframe at the angle of attack for maximum lift coefficient.
At high angle of attack you get high induced drag. This frequently is such that you cannot maintain the speed.
Given the above you need:
High thrust, jet or prop.
Good sized wing area
High lift coefficient
High limiting G.
Controls capable of generating the pitch rate needed.
Not too much pitch stability.

To name but a few.
Roy

 

[Milton_Shupe]

Not sure the previous technical responses answered your question. I'm not a guru in these matters so they did not help me.

For most of us, we need to use available general help tools to get a somewhat reasonable flight model for our builds.
Once we have something in the ballpark, we can tweak the files a bit to get a response we feel better about (even though it still may not be accurate).

So, if you are like me, I use a tool to get started, then tweak afterwords.

When I want to tweak turn rates, I look at the air file table dealing with Cn_beta, Cn_da, Cn_dr dealing with side slip, aileron drag and rudder control.
Of course, other tables are available to tweak but these are my usual.

I test a two minute turn at 30 degree bank angle while holding altitude.

Its not what the gurus would do but its what we less knowledgeable folks can do.

Maybe others can more accurately address your specific question.

 

[Dutch]

Thanks all, for helping my understanding; and particularly Milton for pointing me to the air file records to tweak.

As a non-guru I'm looking for where the numbers are stored that I can play around with until I get the results I need.

This is all in support of putting together a library of realistic AI opponents for FSCAI.

Dutch

 

[Roy Holmes]

If the question had mentioned the fact that it was to do with AI aircraft rather than pilot controlled fighters, my answer would have been different.
Generally, AI aircraft do not do hard turns so limiting g and maximum lift coefficient are not relevant. However, in a steady turn, neither are side slip because there is no slip in a balanced turn, nor aileron drag because there should be no applied aileron in a balanced turn, nor rudder because the turn is generated by the inclined lift vector.
It has nothing to do with gurus and everything to do with knowing some flight dynamics and their effects.
Dutch, if you let me know what airplanes you are interested in I’ll help you with the solution you are looking for. Just specify a typical situation and what you want your models to do.
Roy

 

[Dutch]

Here's a fuller reply of what's going on.

My interest in enhancing the turn rate is experimental and it applies to both AI and user piloted aircraft. FSCAI is an acronym for FS Combat AI, and among other things it gives AI aircraft the ability to engage the user (or other AI) in aerial combat. The objective is not to make a game but to make a simulation with maximum authenticity. We are aiming for as much realism as FSX/P3D can give us. That's why I'm posting here.

A critical part of close-in guns combat ("dogfighting") is the ability to make hard turns. The aircraft that can turn tighter has a huge advantage on the one that can't. So to make AI fight realistically they must be able to turn as well as the real-world counterpart of the aircraft they are flying. Ideally we want it to be able to sustain the real-world turn rate but no more (we don't want gimped planes or super planes we want real planes.)

Much work has gone into this turn problem and most of the issues have been "solved" in the sense that AI pilots can be forced to bank up to 90 degrees and several SimConnect techniques are used to force even harder turns from the reluctant "airline-trained" AI pilots. This generally works well and they are formidable opponents - so now we are in another phase - tuning the turning capability to at least somewhat match real-world numbers.

Of course the problem arises that simulated aircraft both user and AI were not designed to be pushed to these limits and their ability to actually perform realistically in the sim varies widely, and in no particular relation to the capabilities of the actual airplanes they simulate.

Example. If in testing I fly a well-simulated airplane (like the A2A Spitfire) I can turn and fight realistically. But if for instance I fly the Just Flight Hurricane (a fighter with a turn rate better than a Spitfire) against the same AI they have their way with me because the JF Hurricane can't even sustain a 4-G turn. The AI 109s out-turn me and shoot me down relentlessly if I try to turn fight with them. This isn't realistic, the real Hurricane can out-turn a 109. I want to be able to fix this in the JF hurricane air file.

The same problem holds with AI aircraft but it's even worse because most AI were put together from cut-and-paste air files and cfg files so that they can fly from point A to point B and land (perform their AI role in FS) but if called upon to climb, dive, and turn like a combat airplane many are either unable to do it or - they do it too well. I want to be able to fix AI to perform more or less like their real world counterparts. I know this will never be anywhere near perfect but I want to be able to fixthe most ridiculous cases.

I'm not a newbie with these configurations I know how to adjust power and speed many other flight characteristics, but turn rate is something I don't have experience fixing.

I'm not sure why tinkering with turn rate ability would be different for AI versus a user aircraft. It seems like the numbers should be in the same place. But I may not understand fully how it works.

Examples on the user side are the A2A Spitfire vs the JF Hurricane as mentioned above, and on the AI side I've been testing with the MAIW Spitfire, Hurricane, and BF-109. All three of these have the same turn capability in the sim - whereas in the real world the Mk.I Spitfire could do a sustained no-flaps 360 degree turn in 20.8 seconds, the BF-109E4 could do the same in 24.1 seconds, and the Hurricane Mk.I in 22.1 seconds (source is RAF flight testing from 1940.)

I want to be able to tweak some numbers to give these AI something reasonably those approximate rates of turn, if possible.

Dutch

Link to the FSCAI program for anyone interested:
32 bits - http://fscaptain.net/downloads/FSCAI_Alpha_SC32_200226.zip
64 bits - http://fscaptain.net/downloads/FSCAI_Alpha_SC64_200226.zip

Forums where FSCAI is discussed by testers: https://fscaptain.proboards.com/board/29/fscai

 

[bstolle]

Examples on the user side are the A2A Spitfire vs the JF Hurricane as mentioned above, and on the AI side I've been testing with the MAIW Spitfire, Hurricane, and BF-109. All three of these have the same turn capability in the sim - whereas in the real world the Mk.I Spitfire could do a sustained no-flaps 360 degree turn in 20.8 seconds, the BF-109E4 could do the same in 24.1 seconds, and the Hurricane Mk.I in 22.1 seconds (source is RAF flight testing from 1940.)

Without the associated IAS the time for a 360° turn doesn't say a lot about the actual turning performance.
As has been mentioned before, the important values for the turn performance with a perfectly flown turn are CL at the turn AoA and drag/power. All this can be tweaked in the cfg file but much finer and more precise control can be achieved in the air file.
The problem is, if the turn performance is off, the rest of the FDE will be most likely off as well in other areas and tweaking only the turning performance can easily result in the rest of the FDE being even worse than before.

 

[Dutch]

Without the associated IAS the time for a 360° turn doesn't say a lot about the actual turning performance.
As has been mentioned before, the important values for the turn performance with a perfectly flown turn are CL at the turn AoA and drag/power. All this can be tweaked in the cfg file but much finer and more precise control can be achieved in the air file.
The problem is, if the turn performance is off, the rest of the FDE will be most likely off as well in other areas and tweaking only the turning performance can easily result in the rest of the FDE being even worse than before.

My bad in leaving off the speeds; the IAS for the Spitfire is 120mph, 115mph for the 109, and 100mph for the Hurricane - all best rate of turn speeds.

I get you, and it may well turn out that trying to adapt the FDE in various AI and non-AI airplanes isn't practical. The fallback position is to force a turn rate which we control outside the FDE. I'd rather not do that, but it may be that it's the least painful option. The downside is that it's going to mean a custom-made configuration file for each aircraft for FSCAI purposes, and that kind of thing is difficult, time-consuming, and requires constant maintenance as new airplanes are announced. But having to review and change air files would probably be worse.

But based on some of the replies here I'm going to see if some number tweaking will do the trick, can't hurt to try.

Dutch

 

[bstolle]

If you are lucky you only need to modify the CL at the required AoA to achieve the required turn rate. Maybe this is already sufficient.

I'm surprised about the low speeds. Just checked, VY for the Hurricane is 140mph and VS clean at high weight is 90mph.

 

[Roy Holmes]

Dutch,
Thanks for the detailed explanation, sounds like a very interesting project.
AI airplanes in general are built to be pretty docile and, for example, they all fly an approach at 5 degrees AOA. That gives them a 2 degree nose up attitude on a 3 degree approach. Most jet fighters use around 10 degrees AOA on the approach and when used as AI and being tied to 5 degrees they approach very fast and have difficulties landing and stopping. That problem should not exist for piston fighters but I just use it as an example that the default AI behaviour is based around light airplanes. I do not know how your system copes with that. You may be able to get them to turn and pull more g but are they AOA limited even then?. If they are you have your explanation why they do not turn as well as you would expect.
I 'm also of the opinion you need a lot more than small tweaks. The first thing I would suggest is some fairly extensive parameter recording on the airplanes as you have them. In particular to see if there are artificial limits placed on such things as AOA. You can do that through listing the parameters of each AI type one at a time , something like <Script>(C:ITrafficInfo:SLANE ALTITUDE, feet) (&gt;L:TargetAltitude, feet)</Script>, which gives you the altitude. Similarly you can record other values like g and AOA. Through this process you could find out if there are some built-in performance limits to AI airplanes in the sim.. If there are it would be possible to say double the CL at 5 AOA.

As regards speed, from WarpD's equation you can see that the slower an airplane can achieve a given Angle of bank the bigger its rate of turn. Compare a sailplane circling in a thermal to a bigger heavier airplane with a smaller wing, which has to fly faster.

So I think the point here is that before doing much in terms of rebuilding the air file I would gather as much info as I can to find out the rules of the game.
Roy

 

[Heretic]

Mind that AI aircraft are controlled by PID controllers for heading and airspeed.
So in order to increase turn rate, you'll have to increase the controller's proportional gain so it chases new a headings more aggressively. On the other hand, a large P gain means more instability, which has to be counteracted by tweaked I and D gains. AI aircraft therefor may become jerky in all phases of flight.

See here:

Flight Models

docs.microsoft.com

https://docs.microsoft.com/en-us/previous-versions/cc707067(v=msdn.10)?redirectedfrom=MSDN
("PID Controllers" sections)

Did anybody ever analyze AI flight dynamics?
I honestly doubt that they are using the same set of tables and parameters as user aircraft as this would just cost way too many CPU cycles if applied to a lot of traffic.

 

[Dutch]

Mind that AI aircraft are controlled by PID controllers for heading and airspeed.
So in order to increase turn rate, you'll have to increase the controller's proportional gain so it chases new a headings more aggressively. On the other hand, a large P gain means more instability, which has to be counteracted by tweaked I and D gains. AI aircraft therefor may become jerky in all phases of flight.

See here:

Flight Models

docs.microsoft.com

https://docs.microsoft.com/en-us/previous-versions/cc707067(v=msdn.10)?redirectedfrom=MSDN
("PID Controllers" sections)

Did anybody ever analyze AI flight dynamics?
I honestly doubt that they are using the same set of tables and parameters as user aircraft as this would just cost way too many CPU cycles if applied to a lot of traffic.

My own feeling from interacting with AI is that they are using a simplified flight model but I haven't yet done any careful scientific measurements.

At this point increasing the turn rate isn't the objective, that's been done. It's customizing the turn rate by aircraft type to be more like the real airplane. We maybe can do that internally (the airplane itself) or externally (via FSCAI) and I'm trying the internal route first.

Increasing the turn rate was done by using SimConnect to forcibly make the pilot bank much more than he wanted to, up to 90 degrees of bank. That worked but it didn't get us all the way there so it was necessary to selectively compel some small heading changes during the turn to force it around a bit faster. We index the degree of those changes (we call it "turbo boost") by the AI skill rating (recruit to ace). It was also necessary to adjust pitch to maintain AoA in very hard turns. But all that works well. The ace level AI; well, I've never beaten one yet; they are just too good. (A better combat pilot than me could, but only by tricks to get them to overshoot.)

It looks and feels seamless when in actual combat with them.

Dutch

 

[Roy Holmes]

About 8 years ago I did a series of tests of AI performance with 50 different sim airplane types where they started 8 miles from a runway at 1500 ft and began a descent at 5 miles for an approach and landing. I had an F-15 parked near the touchdown point with a display of about 16 flight parameters as they flew. The parameters were the normal ones you would look at as a pilot if you were flying the approach and came from a traffic display. I noted and recorded the most significant parameters and it became obvious that every airplane flew the approach at the same angle of attack, 5 degrees. It did not matter whether it was a 747 or an F-15. The AOA dictated the airspeed, so the F-15 which normally would have been at 10 AOA and about 150 kts approached at 5 AOA and 195 kts. The 5 AOA rule began at about 6 miles out when the gear and full flap were lowered. Before that the rule seemed to be speed and some of the airplanes were struggling at very high AOA, because they did not fit the norm.
So it looked like there was a presumption of the sort of airplane that can operate as AI in their sort of operations. That would imply limits on performance.
What is not yet clear to me is if Dutch has software that changes or overcomes those limits.
Roy

 

[Dutch]

We overcome those limits by using SimConnect to basically make them do what they need to do, like having your hand on the wheel and making him pitch, roll and yaw over and above his sim-designed "limit". I had serious doubts about whether this would work or not; but it does, as proven in the sim both FSX and P3DV4

Recent video of an expert alpha tester in a turn fight against an "average" skill level AI:

And here we have 2 on 2 jet action:

 

[Roy Holmes]

Without seeing recorded data it is not possible to make any conclusions on the amount of performance exhibited. My impression is that both videos show similar rates of turn whereas the props should have about twice the rate of the jets.
That tells me some aircraft type independent rules are in play, but It is simply an impression based on my experience in those situations.
Any testing done to evaluate improvements needs hard results. I can help you get the numbers, I have gauges that can display them and Logger can record them.
What you have is potentially a great training aid for virtual air forces and that’s what I fly with 4 days a week, hence my interest.
Roy

 

[Roy Holmes]

I’m thinking your turn methods probably should be closer to the way an airplane is actually flown. For example, putting the airplane in a particular bank angle only will cause it to turn at a specific rate if it does things like increasing elevator deflection to create a balanced turn. If this does not happen the airplane would behave as if a slow roll was stopped about halfway around. If the AI does not behave this way then it is using a different flight model. That could be something like 60 bank means turn with 2 g. I have no idea if this is what happens, but it needs to be investigated.
Roy

 

[Dutch]

We use Tacview for analysis. This is a video of the F-14 vs Mig-23 fight above, perhaps this might help a bit:

 

[Roy Holmes]

If you had a record of angle of attack, g loads and indicated airspeed you could see whether the airplanes are turning steadily or at their max rate. The more I think about it I believe when you compare the turn rate you are getting from the AI airplanes to those in the sim, there are two factors.
First it may be that the AI air files are not as amenable to high g turns as are the ones in the sim, or
Second they could be the same but the AI are not turning at the max rates they are capable of.
Hard numbers will give you the answer. Tacview telemetry has speeds, g, and AOA so the capability is there to record what you need.
Roy

 

[Ellover Dee]

A question for you FDE gurus out there....

Which aircraft.cfg or air file parameters most affect turn rate of a prop aircraft (or any aircraft, but I am concentrating on single-engine fighters)?

I'm finding many of the ones out there have unrealistically low turn rates compared to their real-world counterparts.

Thanks
Dutch

Let's be clear on roll rates and turn rates; rate of roll is distinct from turn rate. The plane turns due to the horizontal component of lift so the wing is what gets you turned. (The rudder is involved very little, just to compensate for adverse yaw, so forget about it, lol!

A quick roll rate will get you to the start of your directional change faster, of course, but you could sure have a plane with a moderate roll rate and high turn rate. Just keep that stuff in mind.

So you will want to be looking at making sure your weight and balance, and MOI values are reasonable as well as your area dimensions for ailerons and elevator.

Then you want to be investigating Section 1101 (or the corresponding newer section, whatever that is) for Cl_da, which is aileron authority; Cl_P, roll damping; and Cm_de/Cm_q, which are elevator control/damping values.

Next you will want to be modifying the tables in 517 and 518, which provide a simulation of the action of dynamic pressure on the ailerons and elevator, and vary control authority. I'd flatline those or just delete 'em from the .air file to start with.

Finally you have tables 341 and 342, which vary the control surface movement with the stick/yoke movement, imparting increased or decreased "leverage ratios", so to speak. I'd delete those or flatline 'em as well to start, just get what you're looking for without them and then fool about tuning them later.

Let us all know how you get along.