egg changed the topic of #principia to: Logs: https://esper.irclog.whitequark.org/principia | <scott_manley> anyone that doubts the wisdom of retrograde bop needs to get the hell out | https://xkcd.com/323/ | <egg> calculating the influence of lamont on Pluto is a bit silly…
_whitelogger has joined #principia
<queqiao->
⟨thunkii⟩ yeah, that works well for me on the express install
<queqiao->
⟨the rocket science kid⟩ Is it kind of cheese to make flight plans before maneuver nodes are unlocked?
<queqiao->
⟨night_beast⟩ ⟪the rocket science kid⟫ Is it kind of cheese to make flight […] ⮪ You can, but you can't track the maneuver node, mechJeb won't auto execute it, so you have to just kinda point and do it the old fashioned way.
<queqiao->
⟨night_beast⟩ Is this a direct ascent? and if so, what's the ascent profile? I've always used launch to target LAN and then made my transfer node in orbit
<queqiao->
⟨thunkii⟩ ⟪night_beast⟫ Is this a direct ascent? and if so, […] ⮪ no, this isn't a direct ascent. Wait until the 'first window' time counts down to zero, then launch to the 'launch now incliniation'.
<queqiao->
⟨thunkii⟩ Then use principia or whatever to do a TLI from the orbit plane, which will be precisely positioned to allow you to get to the Moon after the specified number of days.
<queqiao->
⟨night_beast⟩ yea, I'm starting to get it. Will be interesting when I'm doing more than just impactors.
<queqiao->
⟨thunkii⟩ yi you got it!
<queqiao->
⟨night_beast⟩ I find it a little funny the M/S on the navball doesn't go down. And I had to figure out you had to click "Show on Navball" for MechJeb to see your flight plan
<queqiao->
⟨night_beast⟩ should come out with a principia RP-1 mod that adds LaGrange point and FRT programs so I can really learn this thing.
<paculino>
I think #σκοποσ has programs which might lead you to choosing satellites in those positions. It doesn't really force positioning though
<queqiao->
⟨night_beast⟩ you think who da what now has programs?
<paculino>
Egg's other mod, skopos
<paculino>
Iirc, it replaces the «put x satellites in y orbit, because that satisfies z» with «get a satellite (constellation) satisfying z».
<queqiao->
⟨thunkii⟩ yeah, the delta v not going down is a deliberate choice because it's less useful than in stock which always does instant-impulse maneuver nodes
<raptop>
[insert joke here about stock assuming you're always using project orion-style spacecraft]
<queqiao->
⟨patrikcath⟩ Stock is really stupid and lets you make SSTOs with the first liquid engines but showing the dV seems a lot more useful to me than having to calculate how much I'll have after the burn and having to use MJs dV stats to keep track of the dV
<queqiao->
⟨egg⟩ Mostly the Δv is not what you care about, you are not fundamentally trying to expend this much Δv in some direction, you are trying to get somewhere. If the burn does not go exactly according to plan, going on to expend the planned amount of Δv in the planned direction isn’t going to get you where you want to be—instead you want to be looking at map view to see what is going on.
<queqiao->
Alternatively if you want to just burn according to plan and correct later, you can watch the countdown and cut off when it reaches 0.
<queqiao->
⟨patrikcath⟩ ...there's a countdown?
<queqiao->
⟨patrikcath⟩ I don't think any tutorial mentioned that, I always just made a plan to estimate what I need to do
<queqiao->
⟨egg⟩ What tutorial?
<queqiao->
⟨patrikcath⟩ There definitely was one but I have no idea where, it was like a year ago
<queqiao->
> In order to execute a manœuvre you have to either take a look at the times in the flight plan editor (a countdown to ignition is provided, followed by a countdown to cut-off), or take a look at the shape of the prediction (which should change to become similar to the flight plan) and control your burn accordingly. Once the burn has been executed, you may want to revise your flight plan with trajectory correction manœuvres to account for errors,...
<queqiao->
... as in real life.
<queqiao->
⟨patrikcath⟩ Yeah I think I did the latter and just used the flight plan to know where to point and roughly how long to expect to burn
<queqiao->
⟨patrikcath⟩ Mechjeb unfortunately kept doing whatever it wanted so I didn't get to try that
<queqiao->
⟨thunkii⟩ I personally think it makes more sense to keep the delta v countdown as a rough guide tbh, it threw me for a loop for a bit but 🤷♀️
<queqiao->
⟨Neamesis⟩ in principia the planets are simulated using a high order symplectic integrator but the rockets and stuff are simulated using RK4 (which I pressume PhysX in ksp uses)
<queqiao->
⟨Neamesis⟩ so basically you're having two different integrators? for the planets you want high precision and not much energy loss so symplectic and for physics obejcts you want it to be fast so RK4
<queqiao->
⟨Neamesis⟩ in principia the planets are simulated using a high order symplectic integrator but the rockets and stuff are simulated using RK4 (which I pressume PhysX in ksp uses)?
<queqiao->
⟨eggrobin⟩ I don’t think we use RK4 except in tests.
<queqiao->
⟨Neamesis⟩ so you had to rewrite the entire physics system to run using symplectic integrator
<queqiao->
⟨eggrobin⟩ We use a conjugate-symplectic method for the vessels as well, just one with a different step.
<queqiao->
⟨Neamesis⟩ * integrator?
<paculino>
The bridge has renamed you, eggrobin
<queqiao->
⟨eggrobin⟩ ⟪Neamesis⟫ so you had to rewrite the entire […] ⮪ What do you mean by the physics system?
<queqiao->
⟨Neamesis⟩ well the part of the game that handles physics i guess
<queqiao->
⟨eggrobin⟩ ⟪paculino⟫ The bridge has renamed you, eggrobin ⮪ blargh. That is because of the new discord username thing, egg was taken.
<paculino>
It normally handles nicknames fine
<queqiao->
⟨eggrobin⟩ ⟪Neamesis⟫ well the part of the game that handles […] ⮪ That is distressingly nonspecific.
<queqiao->
⟨Neamesis⟩ ⟪egg⟫ That is distressingly nonspecific. ⮪ I meant the normal physics engine used in ksp is PhysX it handles everything like collisions and it uses RK4 but you had to change it to use a "conjugate symplectic" integrator m
<queqiao->
⟨egg⟩ That does not use RK4, this is good old Euler.
<queqiao->
⟨Neamesis⟩ wait really?
<queqiao->
⟨Neamesis⟩ that's even worse
<queqiao->
⟨egg⟩ Anything that has _AddForce_ as an API is going to be stuck in there; you can only specify a fixed force over a given step, which you compute only based on the present state, so no room for stages there.
<queqiao->
⟨egg⟩ (Also, PhysX doesn’t even try integrating Euler’s equations so rigid body rotation is nonsense.)
<queqiao->
⟨Neamesis⟩ also where did you learn about integrators in standard numerical analysis books i really cant seem to find anything except euler and RK4. But when I look up stuff there seems to be a whole world of integrators, like "RK89" where you run both RK8 and RK9 and use it to correct errors or something and of course there's a whole bunch of symplectic integrators and there's stuff for solving non-linear ODEs it's wild!
<queqiao->
⟨Neamesis⟩ * integrators?
<queqiao->
⟨Neamesis⟩ also where did you learn about integrators? in standard numerical analysis books i really cant seem to find anything except euler and RK4. But when I look up stuff there seems to be a whole world of integrators, like "RK89" where you run both RK8 and RK9 and use it to correct errors or something and of course there's a whole bunch of symplectic integrators and there's stuff for solving non-linear ODEs it's wild!
<queqiao->
⟨egg⟩ Numerik II at ETHZ, and then reading a bunch of papers when I started running into it in Principia
<queqiao->
⟨egg⟩ I did learn about embedded methods in Numerik II, probably heard about symplectic methods too though I don’t recall being taught the specifics of SPRKs. Certainly did not see conjugate-symplectic linear multistep stuff.
<queqiao->
⟨Neamesis⟩ ah so basically reading papers in that field makes sense, I was looking through the change log of principia like all the different versions to see what kind of integrators it used and I was surprised to see 3 or 4 different ones that you mentioned and I was like "damn this guy knows a lot about numerical integration" lol