[1.3.1] Ferram Aerospace Research: v0.15.9.1 "Liepmann" 4/2/18

[zzz]

I have last Kethane, modulemanager and even remotetech and all works fine(seems like).

[ferram4]

@Black-Talon: The panels will add a bit of drag, although I think that I'll have to increase the drag for them. You are correct in your assessment of how things should function there.

@foamyesque: IIRC the 3-man pod is only ~4 tonnes, so adding an extra tonne of mass to that (without much increase in surface area) will cause nasty things to happen. Also, the fact that the shield is very blunted will cause it to add quite a bit of drag (if I coded it properly) so that shouldn't affect things.

How recent is the control surface issue? I've never had this issue occur on my end.

@Clamps: I can't keep track of compatibility with other mods. If you manage to cause a problem with another mod, you must provide the output_log or I can't do anything about it. I have no idea where in the FAR / Kethane code the problem is occurring.

[foamyesque]

@foamyesque: IIRC the 3-man pod is only ~4 tonnes, so adding an extra tonne of mass to that (without much increase in surface area) will cause nasty things to happen. Also, the fact that the shield is very blunted will cause it to add quite a bit of drag (if I coded it properly) so that shouldn't affect things.

I've tested it with additional weight-- e.g., Rockomax decouplers-- and the same behavior does not occur.

How recent is the control surface issue? I've never had this issue occur on my end.

I figured it out. When I updated from the old version I forgot to revert the Squad parts to their base (because it's now overloaded via mod manager). Re-extracted the Squad parts folder, fixed the problem.

[zzz]

Some "science" here(parts are same but in different configuration..

Actually not, nothing to see here.

Edited July 4, 2013 by zzz

[camlost]

Q is dynamic pressure. Q=0.5*air density*velocity^2

You should be looking at Cd which is drag coefficient.

And IMO, the question asked here should be "how does it compare to physical laws", not "how does it compare to stock model", because the stock drag model is absurd.

[Autochton]

I'd be interested in how that same experiment looked when looking at 300 m/s. The stock drag model doesn't have a huge effect until ~200-250 m/s at lower altitudes, so closer to Mach=1 might yield more interesting results.

[zzz]

Q is dynamic pressure. Q=0.5*air density*velocity^2

My bad. By some reason(even help say I'm wrong but anyway) I thought there also drag in this equation, not sure why but for me "dynamic pressure" has more sense this way. I imagine like, depending on drag, force of air stream push different and this is mesure of it - is this have any sence? cd is independent from speed and q independent from drag so what is "drag force"? I'm beginning to understand that I do not understand anything.

So: cd of

"hammerhead" is 1.2

"Ksp-style" is 2.2

and "rocket" is 0.5

Have some sence for me. Roket seems to wins as expected.

I'm not compare to physical laws, I never mention word "realism". I'm compare(without reason, just from curiosity) two different drag models (no matter how absurd stock is, and both are models, we not deal with the "reality" in the game - this models are our reality. Even if I want I can't launch same things in reality so it will be only "unfair" assumption ) and what difference FAR made (if any).

I'd be interested in how that same experiment looked when looking at 300 m/s. The stock drag model doesn't have a huge effect until ~200-250 m/s at lower altitudes, so closer to Mach=1 might yield more interesting results.

Well, I can - but for stock all 3 will be same anyway isn't it? It does not take into account cofiguration, just mass, as far as I know.

Even more "science" (It not without reason in quotation marks): Now better - I should give more time for it, now difference more noticeable.

So in stock - no difference. In FAR, seem like, "rocket" wins and "ksp style" design is worst. But this is without nosecones, and no one build rockets in FAR without nosecones so results are useles but who cares.

Edited July 4, 2013 by zzz

[MR4Y]

My bad. By some reason(even help say I'm wrong but anyway) I thought there also drag in this equation, not sure why but for me "dynamic pressure" has more sense this way. I imagine like, depending on drag, force of air stream push different and this is mesure of it - is this have any sence? cd is independent from speed and q independent from drag so what is "drag force"? I'm beginning to understand that I do not understand anything.

So: cd of

"hammerhead" is 1.2

"Ksp-style" is 2.2

and "rocket" is 0.5

Have some sence for me. Roket seems to wins as expected.

I'm not compare to physical laws, I never mention word "realism". I'm compare(without reason, just from curiosity) two different drag models (no matter how absurd stock is, and both are models, we not deal with the "reality" in the game - this models are our reality. Even if I want I can't launch same things in reality so it will be only "unfair" assumption ) and what difference FAR made (if any).

Well, I can - but for stock all 3 will be same anyway isn't it? It does not take into account cofiguration, just mass, as far as I know.

Even more "science" (It not without reason in quotation marks): Now better - I should give more time for it, now difference more noticeable.

So in stock - no difference. In FAR, seem like, "rocket" wins and "ksp style" design is worst. But this is without nosecones, and no one build rockets in FAR without nosecones so results are useles but who cares.

Variances in your piloting for both ships and KSP's wonky physics itself makes this test not valid. Since ferram disaproves the use of Mechjeb for more consistent data results, we are left to wonder what the issue really is.

[zzz]

Variances in your piloting

Well, not too much, only "hammerhead" has some course deviations but not more then 5 degrees(durning most path is less) and in both models it was similar. With other two not more then one degree and I did 3 launches of eachh with use of dinamic warp at 1\2, 1\4 speed and results was similar- within the last digit-two - you can see it in stock model, ideally all numbers must be same, I think. But difference between crafts in FAR and acceleration in comparsion with stock is much more so some science still here, but not take it serious. And I think you use some needless overquoting here

Edited July 5, 2013 by zzz

[The Lone Wolfling]

Variances in your piloting for both ships and KSP's wonky physics itself makes this test not valid.

What about multiple tests?

[MR4Y]

What about multiple tests?

That only increases the chance of bigger differences. That's why equipment testing is done by machines.

[taniwha]

Since ferram disaproves the use of Mechjeb for more consistent data results, we are left to wonder what the issue really is.

Eh? I've been reading this thread for a while, and I've seen no such thing. What I have seen is a very true comment that MechJeb's drag loss figures are bogus and completely unreliable (and why).

While MechJeb's drag loss figures are bogus, its delta-v expected figures are correct, so one way of comparing is to use MJ's ascent guidance to get into orbit and compare the delta-v expended figures. I haven't not used FAR for a very long time, so I don't know the usual delta-v to get into 100km orbit with stock, but with FAR I usually see 3400-3600m/s.

[Autochton]

What about multiple tests?

That only increases the chance of bigger differences. That's why equipment testing is done by machines.

I beg to differ. Multiple tests with control input close to the same profile will tend to eliminate random error (i.e. deviations from variance in control input). You will tend to get a bell curve (presumably Gaussian) around the average, which is as good as any control for circumstances gets, anyway. If your results differ by more than the standard deviation, you've found a significant result.

Science!!!

[taniwha]

I've run into a problem where doing an AoA sweep out to 90 degrees causes KSP to lockup (it used to work, but I don't remember when I last did it). 89 degrees is fine, 90 degrees is bad. I do 90 degree sweeps to check the recoverability of my planes.

[edit]Sorry, just to be very specific: AoA sweep in the static analysis tab. I first noticed the problem with a pWings plane, then I tried it with another plane on which it used to work.

[foamyesque]

Eh? I've been reading this thread for a while, and I've seen no such thing. What I have seen is a very true comment that MechJeb's drag loss figures are bogus and completely unreliable (and why).

While MechJeb's drag loss figures are bogus, its delta-v expected figures are correct, so one way of comparing is to use MJ's ascent guidance to get into orbit and compare the delta-v expended figures. I haven't not used FAR for a very long time, so I don't know the usual delta-v to get into 100km orbit with stock, but with FAR I usually see 3400-3600m/s.

Likewise. MechJeb mostly works fine with FAR and is perfectly reliable for its dV expended, steering loss, and gravity loss numbers (from which a drag loss number can be computed). It simply does not report drag correctly nor does the autoland feature work with aerobraking.

Oh, and usual 100x100 dV without FAR is 4.5km/s. Less if you nail the ascent profile, but not by much.

[The Lone Wolfling]

That only increases the chance of bigger differences. That's why equipment testing is done by machines.

Exactly. So if he repeats the test several times, and the with FAR results are clustered and different from the without FAR results, he can be sure that it isn't just variance.

[MR4Y]

Exactly. So if he repeats the test several times, and the with FAR results are clustered and different from the without FAR results, he can be sure that it isn't just variance.

Nope. We can't be sure of anything. Human beings can't reproduce any task exatcly the same way twice. And when we talk about computer games, that's worse either.

[Autochton]

Nope. We can't be sure of anything. Human beings can't reproduce any task exatcly the same way twice. And when we talk about computer games, that's worse either.

He doesn't need to. It just needs to be close enough that the error introduced is not sufficiently large, and flying the ascent with MechJeb will do that adequately. Hell, for the purpose of this test, setting the Smart A.S.S. to Surface, 90 degrees off horizon will do. And thus, he can fly very nearly the same ascent many times, record results for each, and then calculate the average and variance of them. And if the results differ by more than the variance (especially if they differ by multiples of the variance), he has significant results.

This is the way it's done by real scientists in the real world. Ever hear of an 'error bar'?

[Resender]

Anyone know of a tutorial how to calculate the surfaces ect...

[Frederf]

Maddening isn't it? It's as if the laws of statistics are invalid because "people not perfect video games."

[MR4Y]

He doesn't need to. It just needs to be close enough that the error introduced is not sufficiently large, and flying the ascent with MechJeb will do that adequately. Hell, for the purpose of this test, setting the Smart A.S.S. to Surface, 90 degrees off horizon will do. And thus, he can fly very nearly the same ascent many times, record results for each, and then calculate the average and variance of them. And if the results differ by more than the variance (especially if they differ by multiples of the variance), he has significant results.

This is the way it's done by real scientists in the real world. Ever hear of an 'error bar'?

Have you ever seen any ABY and similar testing having it's critical steps performed by humans? Differences in response of even a milisecond can make all the difference in the end.

[taniwha]

I have no idea what ABY testing is, but ascent profiles do not need millisecond timing to get a reasonably good error bar, especially when getting to orbit around Kerbin takes about 4500m/s in stock and I got a ship into 100km orbit (using MechJeb) using 3294m/s. If I had done it by hand, I would expect between 3400 and 3600. That's still significantly different from stock.

[Autochton]

Have you ever seen any ABY and similar testing having it's critical steps performed by humans? Differences in response of even a milisecond can make all the difference in the end.

Aerodynamics is not a chaotic system, no matter what you may think. If it were, no human could ever fly a plane. And only chaotic systems fail in the way you describe there. Meanwhile, we still use statistics on large amounts of data to predict them, and it works. As Frederf says above, your disbelief does not actually disprove the laws of statistics, upon which most of modern science is built.

But go on, prove your hypothesis: After all, if, controlling for as many variables as possible, you find the standard deviation overshadows the differences between results, you'll be right. You lose nothing by trying it, and right now, yours is the more unlikely hypothesis. You stand to gain bvy proving it true. (In fact, there'd be some mathematicians whom I think would be interested in talking to you.)

[Cinocal]

So I have some control surfaces that were assigned to only Pitch but when in game I apply roll input and it also activates the surfaces that are only supposed to pitch. I've tried changing it but it still is doing it, KB or Joystick input same result?

[praise the suuun]

Ferram: I've identified a conflict of modulemanager.dll between FAR and B9 R3.1c (intended for .20 not .202). here's my write up. I don't know much about code conflict but I believe them to be connected.

http://forum.kerbalspaceprogram.com/showthread.php/38899-Stock-air-breathing-engine-s-values-remaining-frozen-at-a-different-value-than-stock?p=502877#post502877