Investigating significant drag and heating changes introduced in 1.8.0

[OHara]

3 hours ago, Kergarin said:

But squad DOES absolutely know all the time if they intended to change aero and heat in 1.8

I am convinced the changes were unintentional, because the are so inconsistent and make little physical sense.
The Big-S delta-wing, from the example at the top of this thread, had a drag coefficient for flow from below of C_d = 0.96 in versions 1.2--1.7, which makes sense for a nearly flat plate, but C_d = 0.02 in version 1.80 so when we try to belly-flop it cuts through the air as well as a needle.  The Terrier (v2) was modeled as 0.8-meters long in version 1.7, but 21.0 meters long in KSP 1.8.0

Even though it is a clear error, and easy for players to fix by pulling over the correct file from previous versions, Squad might be nervous about promising to have a solution soon. 

KSP has a system to automatically estimate, from the model of each part, for each direction of airflow, the facing area, degree of pointiness for C_d, and protrusion of the shape for shock heating.  Squad could put the correct values in the configuration files for each Squad part, but will want to repair the automated system for mods.  Part modders do not want to figure areas and drag coefficients.

Rather harmless errors started to creep in to the automatic drag-estimator with the _v2 parts around version 1.5,  as seen in the PartDatabase.cfg.  Then in 1.7.0 the basic nose-cone was given C_d = 0.04 for flow from the left, but a reasonable C_d = 0.54 for flow from the right.  So there might be a subtle bug in the shape-estimator.

[qzgy]

1 hour ago, OHara said:

The Big-S delta-wing, from the example at the top of this thread, had a drag coefficient for flow from below of C_d = 0.96 in versions 1.2--1.7, which makes sense for a nearly flat plate, but C_d = 0.02 in version 1.80 so when we try to belly-flop it cuts through the air as well as a needle.  The Terrier (v2) was modeled as 0.8-meters long in version 1.7, but 21.0 meters long in KSP 1.8.0

Well that doesn't sound at all right.

I will not weigh in much further, mostly since I've been staying in a 1.3.0/1.3.1 game for a long while now. Though it seems rather annoying for craft behavior to change this suddenly.

[vossiewulf]

Somebody checked in the wrong code that they'd been noodling around with and since the changes are the kind that would require careful A/B testing of lots of parts, it makes it through QA. Regression suites can only have so many cases, and their QA group is probably pretty small. 

[R-T-B]

Replacing the file is ok...  until ModuleManager wipes it out on a modded install.

I created my own solution to that debacle.  Basically, I updated module manager to reference a PartDatabase.cfg "template" in the game root named "StaticPartDatabase.dbk" (if it exists)and use it rather than blanking the file completely as it feels like it.  The format of the template file is the same as the original, it simply allows us to "stick" some entries at our own static settings.  You can drop the 1.7 PartDatabase.cfg in a 1.8 install as the StaticPartDatabase.dbk as an example to restore stock physics and still have working mods.

The pull request is very simple, but ModuleManager team will need to approve it.  If they don't, I'll be forced to maintain a patchset build for this useful feature (obviously, support for that would fall to me, even though it's a really simple patch).  I'm giving them a day or three to act though.  In the meantime if you'd like to test this for me, PM me.  I can make a test build provided distribution is controlled (you expllcityly agree not to bug sarbian about it).

Surprisingly simple PR link follows:

https://github.com/sarbian/ModuleManager/pull/153

EDIT:  Wow, that was closed quick.  And after discussion with Sabian, it seems a patch is incoming soon and he has no desire to see this patchset in the wild really.  His reasoning is pretty valid too.

 

Nevermind about the above then.

Edited October 27, 2019 by R-T-B

[BeanThruster]

Apparently, the drag / heat bug has been dealt with in 1.8.1. The drag coefficients in the PartDatabase.cfg file reverted back to values that are similar to 1.7.3. Didn't check them all, however the most significant variations (that is, the drag coefficients on z faces of the drag cube in fuselages (had been reduced by a factor of around 4), delta wings and shuttle wings (reduced by more than 10 times) I had found in 1.8.0 are no longer there. Which, together with the fact that the annoying not-saved-settings bug was corrected, allows me to happily step to the 1.8 universe.

1.8.0 is dead, long live 1.8.1

Ill keep a game backup ready for reverting back to 1.7.3, though

[Kergarin]

Yes, this is fixed. Thanks squad!

[djr5899]

Curious as to anyone else's experience with 1.8.1 with this issue.  I just finished some game time, starting a new career save in 1.8.1, and, while I see an improvement over the 1.8.0 drag/heating issues, it's still not back to pre 1.8.0 levels.  Simple, early mission I did in the brand new career:

• MK1 Command Pod
• MK16 Chute attached to Command pod
• 4 Mystery Goo, 4 Barometers, and 4 2HOT Thermometers attached to the command pod
• A science junior
• 2 MK2 radial parachutes attached to the science junior
• A 1.25m Heat Shield

This was my entire craft re-entering Kerbin atmosphere.  I did not achieve full orbit.  My max speed under 70km was 2003m/s, yet, the heat shield failed, the science junior exploded, and all the external attached experiments blew up as well.  The only thing that didn't overheat and explode was the MK1 Pod and MK16 chute.

Pre-1.8, I could come in with a craft 2-3 times as heavy, as fast as 3000-3300 m/s hitting the atmosphere and still have all the parts of the craft survive.

[AHHans]

7 hours ago, djr5899 said:

Curious as to anyone else's experience with 1.8.1 with this issue.  I just finished some game time, starting a new career save in 1.8.1, and, while I see an improvement over the 1.8.0 drag/heating issues, it's still not back to pre 1.8.0 levels.  Simple, early mission I did in the brand new career:

• MK1 Command Pod
• MK16 Chute attached to Command pod
• 4 Mystery Goo, 4 Barometers, and 4 2HOT Thermometers attached to the command pod
• A science junior
• 2 MK2 radial parachutes attached to the science junior
• A 1.25m Heat Shield

This was my entire craft re-entering Kerbin atmosphere.  I did not achieve full orbit.  My max speed under 70km was 2003m/s, yet, the heat shield failed, the science junior exploded, and all the external attached experiments blew up as well.  The only thing that didn't overheat and explode was the MK1 Pod and MK16 chute.

Hmmmm...

I just re-created that vessel, added a decoupler and a kickback, and kickback-ed it into space. (SCNR) Once outside the atmosphere I decoupled and selected retrograde/surface. It entered the atmosphere at about 2100 m/s and hat no real problems. (A temperature-gauge on the science junior showed up for a few seconds, and at the end it couldn't keep itself fully retrograde anymore, but that's it.)

[djr5899]

1 hour ago, AHHans said:

Hmmmm...

I just re-created that vessel, added a decoupler and a kickback, and kickback-ed it into space. (SCNR) Once outside the atmosphere I decoupled and selected retrograde/surface. It entered the atmosphere at about 2100 m/s and hat no real problems. (A temperature-gauge on the science junior showed up for a few seconds, and at the end it couldn't keep itself fully retrograde anymore, but that's it.)

Thanks.  I'm going to re-test the same craft file in Sandbox and see if that makes any difference.  I also re-verified files and found 6 files that failed to verify.  Hoping that is it.  I'm not sure if the actual changes are published, like, in 1.8.0 look in X file and X setting was this value, and now it should be Y value in 1.8.1.

MOAR TESTING.....

[Streetwind]

If a heatshield fails on a Kerbin suborbital hop, you know something is significantly broken. Heatshields have so much passive temperature tolerance, you can remove all ablator and still do a direct reentry from a Duna return. Perhaps even Dres, though I haven't tried that yet.

Whenever you think something about a part is significantly broken, delete partdatabase.cfg in your KSP install directory. If you want, delete physics.cfg as well, to make sure there are no borked up values in there either. The next time you start KSP, these files will be automatically recreated at default values.

[djr5899]

To me, it seems like all science parts are more susceptible to overheating.  My heat shield isn't failing....the science junior it is attached to is.  Literally the slightest deviation from dead center retrograde and the science junior starts overheating.  I mean, it does make some sense that science experiments would overheat and would need to be put in a service module, but, a science junior....that won't work.

Oh well, it's playable now so I'll just go with it.

[Leszek]

Even after the 1.81 update I notice that basic planes don't want to fly as the control surfaces do not have any authority.  That is unless there is something I am missing.    Anyone else having this problem?

Screenshot: https://www.dropbox.com/s/2buc8ve4bsk2j77/screenshot0.png?dl=0

[pmborg]

1 hour ago, Leszek said:

Even after the 1.81 update I notice that basic planes don't want to fly as the control surfaces do not have any authority.  That is unless there is something I am missing.    Anyone else having this problem?

Screenshot: https://www.dropbox.com/s/2buc8ve4bsk2j77/screenshot0.png?dl=0

There is nothing wrong with airplanes, they fly like in 1.7.3

Hello  @Leszek, try to make the center of gravity a bit further from the center of lift.

This one for example takeoff at 80kts and weights 363tones.

When you press CTRL + F12: you will enable the forces overlay:

BLUE: Lift

YELLOW: Control

RED: Thrust

 

Edited November 2, 2019 by pmborg

[Shaun]

There's definitely something strange about atmospheric drag: before a Mk1 cabin within a fuselage had between 0.04 and 0.06 drag at 245m/s. It now has around 0.64 drag and I can't even get to 200m/s.

[AVaughan]

7 hours ago, djr5899 said:

My heat shield isn't failing....the science junior it is attached to is.  Literally the slightest deviation from dead center retrograde and the science junior starts overheating.

I stopped trying to build such designs back around KSP 1.3.  As you mentioned, even the slightest deviation from retrograde will cause the science junior to overheat, and that design isn't aerodynamically stable, so it is likely to deviate at least slightly.

 

6 hours ago, Leszek said:

that basic planes don't want to fly as the control surfaces do not have any authority.

From the screenshot that you posted it looks like most of your control surfaces are close to the center of mass (and the center of lift).  Either add a tail or some canards all the way forward.  Control surfaces are more effective if they have more leverage.

[Leszek]

22 hours ago, pmborg said:

There is nothing wrong with airplanes, they fly like in 1.7.3

Hello  @Leszek, try to make the center of gravity a bit further from the center of lift.

This one for example takeoff at 80kts and weights 363tones.

When you press CTRL + F12: you will enable the forces overlay:

BLUE: Lift

YELLOW: Control

RED: Thrust

 

Crap well something is wrong and perhaps it is a mod or something.  The thing is that the normal planes I make and fly don't work anymore.  The one in the screenshot has lots of troubleshooting to still fail.  But for the life of me I can't figure out why if it isn't 1.8.1.

 

17 hours ago, AVaughan said:

I stopped trying to build such designs back around KSP 1.3.  As you mentioned, even the slightest deviation from retrograde will cause the science junior to overheat, and that design isn't aerodynamically stable, so it is likely to deviate at least slightly.

 

From the screenshot that you posted it looks like most of your control surfaces are close to the center of mass (and the center of lift).  Either add a tail or some canards all the way forward.  Control surfaces are more effective if they have more leverage.

Sorry I should have made it more clear but I spent a lot of effort.  That plane is heavily modified when the ones I normally make didn't work.  Those are canards on the front and yes the elevators are close to the center of lift on this one, they weren't originally.  The closer your center of lift to the center of mass the less leverage you need to control the plane and one of the last things I did was move the wings forward to make the difference between the two smaller.  This does weaken those elevators but the canards are still good and this plane should be more than able to fly.  Something is definitely wrong and if it isn't 1.8.1 then it must be a mod.  I am just not sure which one would make this kind of difference as I am using stock aerodynamics.

Thanks anyway for your input guys.  At least I can figure it isn't KSP and I may have been looking in the wrong place.

 

Update:  Now that I know it was a mod the only thing I could think of is the airplane autopilot mod.  Turns out it that was it.  The newest 1.8 version isn't out yet but the build the modder pointed out seems to resolve the issue.  I didn't think this mod made those kind of changes.

Edited November 3, 2019 by Leszek
Update