Spit/109 sea level speed comparisons in 1.08 beta patch

[swift]

Originally Posted by ACE-OF-ACES

Correction..

The 6DOF math (what you call thumb rules) is more than adequate to simulate flight!

And the 'data' that the 6DOF math uses has nothing to do with any of the real world performance data (ROC, TSPA, etc). The 'data' the 6DOF math uses is coefficients only. That is to say the 6DOF math for a P51 is the same as that for a Bf109, what makes a P51 a P51 is the coefficients loaded into the 6DOF equation. That is to say, no where do you 'load' say the ROC or TSPA values from a WWII performance test.

The only time you make use of the WWII performance data is in the validation of the 'outputs' of the 6DOF math and the corsponding coefficients selected.

That is to say the math never changes, only the coefficients.

Basically they can get a good set of coefficients to use based off the geometry of the plane (CL, CD, mass, wing loading, etc). Than they 'tweak' the coefficients until the outputs of the equations match the real world data. As part of all this the power plant (engine) is also simulated and is one of the inputs to the 6DOF (thrust) equation.


The more complicated versions were an issue back in the early 90s.. Where games like AOTP made use of fixed point math, in that the floating point processors were just not fast enough to do the complex calculations in real time. Mater of fact back then they were even limited to a 3DOF flight model, but than around 1995 a flight sim called Pacific Air War 1942 came out, that was one of if not the first PC flight sim to implement a 6DOF flight model, it still used fixed point math. These days there is no need for fixed point math and thus no need to use the simplified versions of the 6DOF flight model equations.


Who wouldn't?


Depends on which data your referring too.. As noted above, a good estimate of the 6DOF coefficients can be derived from the planes geometry. Actually the hard part to simulate is the engine! In that many of those records do not exist and no good way to derive them from looking at the dementions of the engine.


As noted above, if they have enough info to draw the plane in 3D, then they have enough info to derive many if not all the coefficients for the 6DOF FM.. What is lacking in the power plant info (thrust)


Disagree 100%


To each his own than

My point is that you will have difficulties to derive reliable coefficients and you seem to agree on this. I do know a little about flight mechanics and trajectory computation (where the 6dof equations intervene) and about coefficient determinations (it is my daily business).

At work we frequently use a simplified tool to calculate the aerodynamic coefficients for subsonic and transsonic flight conditions and I can tell you I would not trust them for applications such as CoD. We use them for different applications where the impact is minor so we can live with it. But CoD would rely heavily on these coefficients and I'd say to obtain something that is halfway close to reality such a tool is not sufficient. And from experts working for years in the aerospace business using modern and highly sophisticated cfd tools I know that using these methods for subsonic regions is far from trivial. And it would take hours to days to calculate just one flight point for one configuration and probably would take longer if one would take into account the viscious terms instead of relying on simplified Euler calculations.

But what we need in CoD does not stop at the determination of lift, drag and lateral force coefficients and the moment coefficients about the three axis. We also need the derivate coefficients to obtain a believable flight model. Up to now the means with which these are "determined" is more than crude and very little reliable.

This gets even more complicated when one considers that each flap, rudder and aileron movement will have an impact on the aerodynamic coefficients (the 6 static coefficients and the derivates). You'd need a database set for several flap, rudder and aileron deflection combination. Then we have the trimmed and untrimmed flight conditions and other aerodynamic control surfaces such as flaps and airbrakes. Now let's talk about canopy open or closed and radiator and oil cooler openings ...

And these are just the coefficients for the airframe. We'd also need reliable data for the propulsion set.

I really do not believe in being able to obtain a full AEDB that will result in a flight performance that will be close to the real thing anyway, provided we even know where the real thing was. I think it is smarter to take the bottom up approach by tweaking the used coefficients in such a way that they fit to the experienced behaviour including test results and, where values are missing, to anecdotical evidence as long as there is a bunch of anecdotes saying the same.

BTW: CoD is definitely using 6dof. What we are disputing is how they come up with the forces and moments they inject into the 6dof equations.

[David198502]

Originally Posted by swift

My point is that you will have difficulties to derive reliable coefficients and you seem to agree on this. I do know a little about flight mechanics and trajectory computation (where the 6dof equations intervene) and about coefficient determinations (it is my daily business).

At work we frequently use a simplified tool to calculate the aerodynamic coefficients for subsonic and transsonic flight conditions and I can tell you I would not trust them for applications such as CoD. We use them for different applications where the impact is minor so we can live with it. But CoD would rely heavily on these coefficients and I'd say to obtain something that is halfway close to reality such a tool is not sufficient. And from experts working for years in the aerospace business using modern and highly sophisticated cfd tools I know that using these methods for subsonic regions is far from trivial. And it would take hours to days to calculate just one flight point for one configuration and probably would take longer if one would take into account the viscious terms instead of relying on simplified Euler calculations.

But what we need in CoD does not stop at the determination of lift, drag and lateral force coefficients and the moment coefficients about the three axis. We also need the derivate coefficients to obtain a believable flight model. Up to now the means with which these are "determined" is more than crude and very little reliable.

This gets even more complicated when one considers that each flap, rudder and aileron movement will have an impact on the aerodynamic coefficients (the 6 static coefficients and the derivates). You'd need a database set for several flap, rudder and aileron deflection combination. Then we have the trimmed and untrimmed flight conditions and other aerodynamic control surfaces such as flaps and airbrakes. Now let's talk about canopy open or closed and radiator and oil cooler openings ...

And these are just the coefficients for the airframe. We'd also need reliable data for the propulsion set.

I really do not believe in being able to obtain a full AEDB that will result in a flight performance that will be close to the real thing anyway, provided we even know where the real thing was. I think it is smarter to take the bottom up approach by tweaking the used coefficients in such a way that they fit to the experienced behaviour including test results and, where values are missing, to anecdotical evidence as long as there is a bunch of anecdotes saying the same.

BTW: CoD is definitely using 6dof. What we are disputing is how they come up with the forces and moments they inject into the 6dof equations.

really good thread so far, and very good and interesting post!

[ACE-OF-ACES]

Originally Posted by swift

My point is that you will have difficulties to derive reliable coefficients and you seem to agree on this.

Key word being reliable..

Now ask yourself..

How do you prove to yourself they are reliable?

Answer is you compare the results (outputs, such as tas, roc, roll-rates, turn rates, etc) of the 6DOF using these coefficients to the real world data..

Which is pretty easy to do when you have the real world data!

But what do you do when you don't have any real world data?

You guess it.. You rely on, aka trust, the math!

That is my point

That being we do NOT have real world data on each aspect of the WWII plane.. So unless we come up with a time machine, we are going to have to rely on calculated results (the math) for not only simulation but validation.

Originally Posted by swift

I do know a little about flight mechanics and trajectory computation (where the 6dof equations intervene) and about coefficient determinations (it is my daily business).

You too? Here as WSMR we use a lot of trajectory math to calculate the launch to impact site of the missiles we test here. We also have our own flight simulation software that we call RAGE

http://www.csc.com/public_sector/suc...tual_landscape

I work with the guy who wrote that software on a daily bases.. I also write plug-ins for RAGE but he is the true gu-roo of the software. All in all a great job, on my way to work I may see anything from a F22 to a UAV fly by at tree top level (landing or taking off from holloman).. It can be dangerous sometimes.. In that I am always looking up in the blue instead of looking forward at the road!

Originally Posted by swift

At work we frequently use a simplified tool to calculate the aerodynamic coefficients for subsonic and transsonic flight conditions and I can tell you I would not trust them for applications such as CoD. We use them for different applications where the impact is minor so we can live with it. But CoD would rely heavily on these coefficients and I'd say to obtain something that is halfway close to reality such a tool is not sufficient.

Well than I guess we will have to agree to disagree on that point

Originally Posted by swift

And from experts working for years in the aerospace business using modern and highly sophisticated cfd tools I know that using these methods for subsonic regions is far from trivial. And it would take hours to days to calculate just one flight point for one configuration

Note I never said it was easy or trivial.. My only point is it is doable! Many have for many years now.

Originally Posted by swift

and probably would take longer if one would take into account the viscious terms instead of relying on simplified Euler calculations.

I know of only one PC flight sim that implemented a real-time computational fluid dynamics flight model (what you call 'viscious') and that was FLIGHT Unlimited

http://en.wikipedia.org/wiki/Flight_Unlimited

It was done back in 1995.. And was not only overkill IMHO but too much for the PCs of that time. Maybe even today, I don't know in that not many make use of it in that the 6DOF (what you call simple euler) has proven itself to be more than adequate for military applications, thus more than adequate for PC games IMHO.

Originally Posted by swift

But what we need in CoD does not stop at the determination of lift, drag and lateral force coefficients and the moment coefficients about the three axis. We also need the derivate coefficients to obtain a believable flight model.

Key word being believable..

Now ask yourself..

How do you prove to yourself they are believable?

Answer is you compare the results (outputs, such as tas, roc, roll-rates, turn rates, etc) of the 6DOF using these coefficients to the real world data..

Which is pretty easy to do when you have the real world data!

But what do you do when you don't have any real world data?

You guess it.. You rely on, aka trust, the math!

That is my point

That being we do NOT have real world data on each aspect of the WWII plane.. So unless we come up with a time machine, we are going to have to rely on calculated results (the math) for not only simulation but validation.

Originally Posted by swift

Up to now the means with which these are "determined" is more than crude and very little reliable.

Well than I guess we will have to agree to disagree on that point

Originally Posted by swift

This gets even more complicated when one considers that each flap, rudder and aileron movement will have an impact on the aerodynamic coefficients (the 6 static coefficients and the derivates). You'd need a database set for several flap, rudder and aileron deflection combination. Then we have the trimmed and untrimmed flight conditions and other aerodynamic control surfaces such as flaps and airbrakes.

Which has all been done before and done for years!

Originally Posted by swift

Now let's talk about canopy open or closed and radiator and oil cooler openings ...

Better yet..

To drive my point home..

Let talk about you going out and finding the real world test data for each plane in the game of the 'effects' of the canopy open vs. the canopy closed on each aspect of the plane (tas, roc, roll-rate, turn-rate, etc)

Allow me to spare you that effort!

In that you wont find such data!

Thus, back to square one of my point

You will have to trust the math and how it says the canopy open vs. closed will 'affect' the flight

Oh sure you may find some anecdotical evidence for some of the planes.. For example we have all read the stories.. Like the Me262 that was stuck in a high speed dive, until the pilot popped the canopy and started to bail out, at which point he noticed that popping the canopy 'changed something' such that he was able to regain control, and thus didn't bail out. What is not 'clear' about such stories is the parameters to re-create that scenario in the game to see (validate) the flight model. For example..

What was his altitude when he popped the canopy?
What was his speed when he popped the canopy?
What was his dive angel when he popped the canopy?
What was his flap setting when he popped the canopy?
What was his trim setting when he popped the canopy?

The list goes on and that is only for the point in time of when he popped the canopy..

So all we know from that story is that 'something' changed.. For all we know his plane was slowing down and the popping of the canopy had NOTHING to do with it.. It could have just been a coincidence that he popped the canopy at the same time the plane had slowed down enough that he was able to regain control

Originally Posted by swift

And these are just the coefficients for the airframe. We'd also need reliable data for the propulsion set.

Ah, glad to see you agree with what I said in my last post!

So not all is lost, in that we can agree on some things!

Originally Posted by swift

I really do not believe in being able to obtain a full AEDB that will result in a flight performance that will be close to the real thing anyway, provided we even know where the real thing was.

Bingo!

So do we give up?

Or do we trust the math and move on?

Originally Posted by swift

I think it is smarter to take the bottom up approach by tweaking the used coefficients in such a way that they fit to the experienced behaviour including test results and, where values are missing, to anecdotical evidence as long as there is a bunch of anecdotes saying the same.

Which is exactly what I was saying minus the anecdotical evidence (aka pilot combat reports)

Just too many variables involved to make most if not all anecdotical evidence useful

As I noted early on

Many 'feel' the can derive some sort of statistical average of the anecdotical evidence.. Many have tried, all have failed! Which is not surprising when you consider the fact that the anecdotical evidence is not something that varies a 'little' As in one reports says the top speed is 305, another say 307, and another says 302, and another say 310. If that was the case it would be a simple mater of taking the average and calling it good! But that is not the case for anecdotical evidence, what we have there is Spitfire pilots saying they could out turn 109s and 109 pilots saying they could out turn Spitfires. So based on that it is not surprising that those who have tried have failed.

Originally Posted by swift

BTW: CoD is definitely using 6dof.

Ah, glad to see you agree with what I said in my last post!

So not all is lost, in that we can agree on some things!

Originally Posted by swift

What we are disputing is how they come up with the forces and moments they inject into the 6dof equations.

No, we know how they come up with them.. What we are disputing IMHO is how they validate the flight model.. For the aspects of which we have no real world test data to use in the validation. In such situations I am saying we have to trust the math to fill in the blanks, where as your saying we should use anecdotical evidence to fill in the blanks..

In summary

I put more faith in the 6DOF math than you do
You put more faith in anecdotical evidence than I do

Other than that I think we agree

[swift]

You do not need to do this surgical argument deplugging (what I always feel to be a bit rude although I understand that you do not mean it this way). I understood perfectly where you want to go.

My point is that with the maths what you have in mind you will be likely as far off the reality as with what we have right now in the game if you cannot check it against reality.

So you apply maths but the chances are great that you will be far off the mark as with the current methods (which might by the way be based on some simplified maths I guess). You will never know if you will be below or above the mark. So I could as well trust in what we have now. I could as well trust in anecdotical evidence provided the sample is large enough to allow a statistically sound picture about the real thing. If 1000 pilots say the spit could outturn the 109 I'd tend to believe that 1000 pilots cannot be wrong even if I do not know the 1000 initial conditions. The number of pilot accounts however may suggest that the variety of initial conditions in which these guys made their observation was large enough to provide for a good hint about a qualitative not measurable behaviour.

It will be simply an impossible task to have fully viscious cfd simulations for each aircraft for a game that is basically just a niche product. It had perhaps been tried once. It has never been done again. This talks books. And cfd is again basically useless anyway if it cannot be checked against wind tunnel tests. Now this won't ever happen anyway.

And don't mix up the effort you and your company can put into a product for which your company will be payed a fortune with the possibilities of a small game developer company.

[5./JG27.Farber]

I suppose there could be a way to settle this. If AoA used his method on a spitIa and Swift used his method on a SpitIa and we had some results?

Aside from having different approches to the same problem, what about some results?

[Crumpp]

So it would appear at sea level the 109 is too slow by at least 30kmh.

There was another monster silly thread over this issue.

It appears all of the aircraft are slow by a similar margin of error.

So while actual performance is not correct, relative performance is not effected.

[Robo.]

Originally Posted by Crumpp

So while actual performance is not correct, relative performance is not effected.

Oh yes it is, unfortunately.

[ATAG_Snapper]

Originally Posted by Crumpp

There was another monster silly thread over this issue.

It appears all of the aircraft are slow by a similar margin of error.

So while actual performance is not correct, relative performance is not effected.

Right.

[ACE-OF-ACES]

Originally Posted by swift

You do not need to do this surgical argument deplugging (what I always feel to be a bit rude although I understand that you do not mean it this way).

Not a need as much as a style that I find useful and actually respectful in that I take the time to address each of your comments.

PS wrt my style, I have spoken with several mods and it does not break any forum rules!!

Originally Posted by swift

I understood perfectly where you want to go.

Really? Because I did not get that impression

Originally Posted by swift

My point is that with the maths what you have in mind you will be likely as far off the reality

Yes we have already established that you do not trust the math..

Got it!

So if not math.. Than what should we trust?

Originally Posted by swift

as with what we have right now in the game

And how far off is it?

Only way to tell is to apply some math and derive a percent error

Originally Posted by swift

if you cannot check it against reality.

Agreed 100%

The part we don't agree on is the definition of reality..

I for one am all for checking against reality when reality exists.. Like real world test data

But when real world test data does not exist, than my point is you have to trust the math to fill in the blanks

Where as you want to fill in the blanks with anecdotical evidence..

And I have already given several examples of why that is a bad idea

Originally Posted by swift

So you apply maths but the chances are great that you will be far off the mark as with the current methods (which might by the way be based on some simplified maths I guess). You will never know if you will be below or above the mark. So I could as well trust in what we have now. I could as well trust in anecdotical evidence provided the sample is large enough to allow a statistically sound picture about the real thing. If 1000 pilots say the spit could outturn the 109 I'd tend to believe that 1000 pilots cannot be wrong even if I do not know the 1000 initial conditions.

If large enough?

Again, as I noted, EASY to say, but until you try you don't know just how hard it is to try and do that..

Many have made that claim over the my past 20 years of simming

All have failed!

All in all I would still take my chances with 6DOF math over some sort of statistical conclusion drawn from anecdotical evidence..

Why?

Because there is something you are forgetting about anecdotical evidence

We only hear from (read the reports of) the pilots that made it home to write about it..

Thus a filtered set of data!

In short, we don't know how many Spitfire pilots were SHOT DOWN trying to turn with a 109 and thus never got a chance to write about it!

Originally Posted by swift

The number of pilot accounts however may suggest that the variety of initial conditions in which these guys made their observation was large enough to provide for a good hint about a qualitative not measurable behavior.

Chances are it will not..

Again, not trying to bum you out

All I am saying is that many like yourself have made such claims over my past 20 years of simming

All have failed

But who knows, maybe your different?

Maybe you are the one to do it?

On that note, don't take this wrong, but talk is cheap!

Do it and than lets talk about it!

Originally Posted by swift

It will be simply an impossible task to have fully viscious cfd simulations for each aircraft for a game that is basically just a niche product.

My guess is the main reason it has not been done is that it is over kill from the get go..

That is to say the benefits of it are so small that most would not even notice the difference..

At the human level that is!

Originally Posted by swift

It had perhaps been tried once. It has never been done again. This talks books. And cfd is again basically useless anyway if it cannot be checked against wind tunnel tests. Now this won't ever happen anyway.

Yes as I pointed out, FLIGHT Unlimited tried it back in 1995, and as I noted in my last post, I don't know of anyone else that has tried it since.

Originally Posted by swift

And don't mix up the effort you and your company can put into a product for which your company will be payed a fortune with the possibilities of a small game developer company.

LOL

Trust me, we are as small if not smaller than 1C when it comes to budgets

[swift]

Sh, I tell you a secret: CloD is already based on 6dof ... but don't tell anybody