Parachutes ineffective below 8m above sea level

[CaptainBalls]

Hi this is with KSP 1.3.1, RSS 13.1.0 and Principia εὐκλείδης. playing on windows 10 64-bit
When attempting to splash down with a probe or capsule on earth, the parachutes become ineffective (at about 8m sea level altitude) causing the vessel to impact the water at high speed.

the time stamp in the linked "INFO" file is "08:20:08 . 332936" line 614
https://gist.github.com/CaptainBalls/c8474656a9365311ab5c32589a522a96
there is no "FATAL" file.

[eggrobin]

From IRC logs:

<Mike`> egg|laptop|egg, so apparently "chutes cut" isn't what happens, but physics stop working. Chutes stop working, engines stop working as well ("they kept firing but had no effect"). The guy retested and apparently the issue was gone when principia was removed.

[eggrobin]

Some questions to help us determine where we should look:

• does it happen with all kinds of parachutes, or just the FASA mercury ones that you were using here?
• does this happen deterministically?
• is this new in Εὐκλείδης, or have you encountered it in previous versions of Principia?

[CaptainBalls]

Some questions to help us determine where we should look:

• does it happen with all kinds of parachutes, or just the FASA mercury ones that you were using here?

it happens with realchutes as well

• does this happen deterministically?

all attempts to land in the ocean end with exploding parts

• is this new in Εὐκλείδης, or have you encountered it in previous versions of Principia?

this is my first attempt to play with Principia

[MikeOnTea]

I can reproduce the problem. Also happens with the previous release.
I suspect there aren't many reports because in my test the capsule survives the splashdown, at around 15 m/s, while it had a speed of 6 m/s while the chute was working.

We also found out that RSS sets earth's mean radius 8.4 m lower than what principia uses here which seems to be roughly matching the distance where the chutes become ineffective. Increasing this value for RSS to the one principia uses seems to fix the issue (haven't personally verified it).

If the value principia uses is more accurate here we might want to change the RSS value.

[eggrobin]

8,4 m is indeed the difference between the mean radius that we use,

Principia/astronomy/sol_gravity_model.proto.txt

Lines 86 to 93 in 9e29425

	body {
	# Spherical harmonic coefficients and reference radius from the combined
	# GRACE/GEOS model GGM05C.ICGEM at
	# ftp://ftp.csr.utexas.edu/pub/grace/GGM05/.
	name : "Earth"
	gravitational_parameter : "3.9860043543609598e+05 km^3/s^2"
	reference_instant : "JD2451545.000000000"
	mean_radius : "6371.0084 km"

and the one that RSS uses:

KSP-RO/RealSolarSystem/GameData/RealSolarSystem/RSSKopernicus/Earth/Earth.cfg at 4c60999c, lines 30 to 34:

		Properties
		{
			displayName = Earth
			isHomeWorld = true
			radius = 6371000

Ours comes from the IAU WGCCRE 2009 report, which gives a value in km of 6371.0084 ± 0.0001, where the meaning of the uncertainty is unclear, since

The uncertainties in the values for the radii and axes in Tables 4 and 5 are generally those of the authors, and, as such, frequently have different meanings.

The IAU WGCCRE 2015 report gives still gives 6371.0084 ± 0.0001, with the following note:

(b) These values for the Earth are for comparison only. An IAU resolution (IAU 2015a) has adopted values for the equatorial and polar radii with less precision, giving values of
6378.1 and 6356.8 km, respectively.

However the IAU 2015 resolution A values are nominal values, and only include a value for the equatorial and polar radii, so that resolution is not very relevant if we want to keep using mean radii.

Principia/quantities/astronomy.hpp

Lines 25 to 28 in 708a01e

	// System of nominal solar and planetary conversion constants, Résolution B3
	// "Sur les valeurs recommandées de constantes de conversion pour une sélection
	// de propriétés solaires et planétaires" adopted at the XXIXth General Assembly
	// of the IAU in 2015.

Principia/quantities/astronomy.hpp

Lines 39 to 44 in 708a01e

	// Planetary conversion constants.
	// “If equatorial vs. polar radius is not explicitly specified, it should be
	// understood that nominal terrestrial [or jovian] radius refers specifically to
	// [the nominal equatorial radius], following common usage.”
	constexpr Length TerrestrialEquatorialRadius = 6.3781e6 * si::Metre;
	constexpr Length TerrestrialPolarRadius = 6.3568e6 * si::Metre;

If the value principia uses is more accurate here we might want to change the RSS value.

Possibly, but we should also try to figure out what is going on here, this is certainly unexpected.

There are some places in Principia that assume that being under the mean radius implies collision, which is broken in principle (although since KSP models bodies as spheres rather than ellipsoids it is not very visible); we should revisit that.

On the other hand, we would expect a collision detection from Principia to result in

1. the vessel blowing up;
2. a log message at the WARNING severity;

neither of which is happening. 8 m of free fall with ineffective parachutes is definitely not expected.

[zilti]

I have the distinct memory of having had that problem about two years ago, though not sure if with or without Principia.

[pleroy]

I was looking at this code recently for unrelated reasons and noticed that we were not adding the intrinsic accelerations after a collision was detected. I believe that this explains the bug.

Subsequent questions:

1. Should we use mean_radius or 0.9 * mean_radius or something for the collision detection? The pole on Earth is like 15 km below the mean radius.
2. The change in Propagate errors found during integration to kill the vessel #2064 should cause us to destroy the vessel when it collides with a celestial. Why is this not happening here? Should it? Do we have a lagging vessel?

[eggrobin]

Note: this issue was introduced by baa6eb9#diff-360562aedddbc765add7381b8a92a3f8R415, in #2020, released in Erdős. @zilti’s two-year old issue is likely separate.