MechJeb 2 - Patch test bed release (October 10)

[Ratzap]

FAR complicates things for the ascent too since the gravity turn too low causes a ton of drag on one side of the payload and over she goes.

I can also confirm the hovering bug with the landing autopilot. I'm trying to land on Pol, it has the 2nd lowest gravity in KSP and even with very small engines the TWR is 12 or so. The latest MJ build just hovers at the final descent stage trying to get to 0m. At one point it was down to 16mm but it just kept on hovering and trying to get closer - only succeeding in pushing itself further away. I ended up repeating the landing then switching MJ off at the final descent and putting it down myself.

[johnsonwax]

As it says in the quote, if I close KSP, delete the .cfg file, and start it up again it seems to "fix it" but anyone with more than a few mods and more than a couple flights in the air at a time knows this is a tedious process.

This fixed by problem as well. Trying to dock 2 station components together since the update, it wanted to bring them together at a 90 deg angle. Deleting the files solved that problem.

[viperwolf]

Sarbian

I had to re install the old MJ unit craft file and just use your plugin. So the rover guidiance would work properly. because for some reason the MJ will not do this correctly unless it has control. without MJ being in control of the craft, the speed will not work , it just keeps increasing in speed. I needed the rover guidance for speed control and direction

[Revencher]

not sure what i am suppose to do with this https://raw.github.com/sarbian/MechJ...AR202/part.cfg? im i suppose to copy and paste to the old cfg file? or do i make a new one? if so how do i make a cfg? file?

[johnsonwax]

Ascent rendezvous doesn't seem to work properly any longer. Not sure when it stopped working right - I think it was working with sarbian's last patch before he started this thread. Basically it times the launch to when the target craft is directly overhead - it doesn't factor in the time to achieve orbit.

[Starwaster]

not sure what i am suppose to do with this https://raw.github.com/sarbian/MechJ...AR202/part.cfg? im i suppose to copy and paste to the old cfg file? or do i make a new one? if so how do i make a cfg? file?

Really don't know how to make a new file on your computer? Huh...

Just copy the new one into the KSP_win\GameData\MechJeb2\Parts\MechJeb2_AR202 folder, if it asks if you want to overwrite, you DO.

Freddy's your uncle, Fannie's your aunt.

[Starwaster]

Ascent rendezvous doesn't seem to work properly any longer. Not sure when it stopped working right - I think it was working with sarbian's last patch before he started this thread. Basically it times the launch to when the target craft is directly overhead - it doesn't factor in the time to achieve orbit.

Honestly I've never had that work properly for me. It puts my launch in the proper ballpark but I've always had to play with my ascent parameters to actually get close to it and I've never really done a proper rendezvous with just a launch...

Edit: And I've been using MJ2 since pre-Aprilish...?

[johnsonwax]

It worked pretty well before. It'd rendezvous between a few hundred meters and maybe 5km - which is fine as a range. I'd say 1/3-1/2 the time it'd rendezvous within physics range. Once it came in at 200m - and that was with a 150t payload. Now it's off by 80-100km which requires quite a bit of rendezvous work.

Another request is to change the KSP landing location from the launch pad to the middle of the runway. Deorbited one of my engines on the roof of the VAB, which was actually quite cool, but unlikely to be what the engineers would have preferred.

[Virtualgenius]

Woo Hoo docking fixed and landings 15m from target instead of 200 - 500m thank you very much

[Radam]

Woo Hoo docking fixed and landings 15m from target instead of 200 - 500m thank you very much

Yes it is absolutely wonderfull that it avtomatically lands precisely on the pad.

But in my oppinion that precise landing should be toggable. As it tends to start burning at a few kilometers, hover, go sideways to the correct position and then just get stuck. If one disables it and enables "land anywhere" it lands as usuall.

Also when under autopilot it doesnt want to go more than warp 1 for the whole descent (kerbin) and if one does toggle higher warp, it imediately turns it down and starts adjusting descent.

Unrelated to recent release, Mechjeb doesnt read thrust of the turbojets right. It just takes max power and goes with it. That would probably require it reading current thrust from the part itself... And subsequently it also shows current acceleration and max acceleration of the craft incorrectly when using jets or turbojets.

[Mihara]

...and here's another problem. Apparently MJ had it for months, but only today I finally discovered the specific bug trigger I can report.

If a custom window has an item "Time to impact", flying very close to the ground, bouncing on the ground, driving over the ground can cause the window contents to suddenly disappear entirely and only reappear when you are no longer landed. The blanked-out-window results in significant extra framerate losses while it is blanked out.

[sarbian]

In the current version, the "Use SAS when available" option is gone, but node autopilot will still lock up attempting to turn when SAS is enabled -- i.e. it won't disable SAS by itself, and it still won't kick SmartASS into auto mode either.

I went overboard when removing the ASAS control. MJ need to disable it when it take control, but I am not sure it should enable it back after.

For the ASAS not switching in AUTO mode I can't duplicate this each time, but I saw it append. Do you have a way to trigger it each time ?

The "Time to impact" problem is most likely some NaN. Is there anything in the log ?

For those who fly with FAR MJ2 won't be as precise while in the atmo. I looked into adding some specific code, but FAR current code does not allow to get some value easily. It's on my list since I like flying with FAR, but I'd rather fix bugs before.

For the "MJ can't handle a launch without spinning out of control sideways within meters in spite of the control part being correctly aligned" I'll need a ship file of save file to look into it.

@Payload : About the ascent autopilot. Do you have info on "optimal" ascent profile ? What would you change (keeping it simple) ?

johnsonwax & sojourner : So after it has enough velocity to get to the periapsis it should stay prograde until the circularisation burn ?

johnsonwax if you have a ship I can test ...

About the landing AP : I ll look into it more this week.

[MainSailor]

Is there a known issue with MJ mis-reporting DV with asparagus staging? After upgrading the .dll on Saturday, my standard multi-mission launcher gives me 0 DV for the stage after the asparagus stage drops off, even though there's a full tank of fuel. I've also noticed that SRBs cause this issue as well. I haven't tried Engineer so I don't know if it's a .craft issue...just wanted to check to see before I ripped apart my launcher to see if there's a staging bug coming out of the VAB (although it flies just fine.)

I'd post my craft file but I have a number of custom parts that I'd have to dump .cfgs out for so it would be messy.

[sarbian]

There is an issue with docking port. Someone made a patch and I need to include it. Check with Engineer anyway

[Hesh]

For the "MJ can't handle a launch without spinning out of control sideways within meters in spite of the control part being correctly aligned" I'll need a ship file of save file to look into it.

Deleting the mechjeb global cfg seems to have fixed it more permanently, if the problem returns I'll post a save file.

[MOARdV]

johnsonwax & sojourner : So after it has enough velocity to get to the periapsis it should stay prograde until the circularisation burn ?

The current behavior (I do not have the latest update, just your previous patch) is preferable - stay prograde until the ship clears atmo, and then plot the circularization burn and rotate towards the node, and *then* warp. The previous version of MechJeb (1.x?) stayed prograde until apoapsis, and then it failed to circularize because the craft takes too long to rotate to the correct direction. I don't want to have to spam monopropellant and RCS on launch stages just to handle that one maneuver. That's my opinion.

[MainSailor]

There is an issue with docking port. Someone made a patch and I need to include it. Check with Engineer anyway

Will do, thank you sir.

[synik4l]

Im having one issue with the ascent autopilot. At times on the launch pad.When I hit enable autopilot. Usually you would expect it to goto 100%. Instead it bounces like crazy around 40% throttle. Usually I just reset the launch a couple times and finally it wont do it....Or I launch it myself. Then enable it once in the air at full throttle. It works fine like that. Just makes it alot harder for me to get a timed launch to work. This may be a combination of my mods but im not sure. If needed I can get you a save file later.

[Payload]

The current behavior (I do not have the latest update, just your previous patch) is preferable - stay prograde until the ship clears atmo, and then plot the circularization burn and rotate towards the node, and *then* warp. The previous version of MechJeb (1.x?) stayed prograde until apoapsis, and then it failed to circularize because the craft takes too long to rotate to the correct direction. I don't want to have to spam monopropellant and RCS on launch stages just to handle that one maneuver. That's my opinion.

I agree with this assessment. This is the correct action as I see it. If your ship is fast/agile enough, then the old way is fine. If it isn't, then you are SOL.

@Payload : About the ascent autopilot. Do you have info on "optimal" ascent profile ? What would you change (keeping it simple) ?

Currently getting the very best out of the auto pilot requires supervision. It is also very difficult/impossible to do hands off space plane orbiting. The problem is the constant best fit angle and single acceleration target.

Consider the following as a crude example.

If we could click on the picture and add POIs at determined altitudes, we could then set a target angle and target acceleration and an action group or groups to be executed when we reach that target altitude.

Different systems have different flight parameters as can be seen through the differences in High TWR Rockets and low ones or Rockets VS Spaceplanes.

With a system similar to this we could literally fly anything, hands off, into space.

Edited August 26, 2013 by Payload

[Radam]

How is the path calculated anyway? Rough estimate at least, and I think it is altitude/downrange type right?

I used logomatic to dump data into excel and tried a few shallow acsents with mechjeb and tubojets. And in altitude/surface velocity graph looks like this:

All turns started at 8km ending at 70 and different turn grades.

Very qurious.

Is a gravity turn such a straight line in this type of representation? Or does mechjeb calculate how fast it should be going at ceirtain altitude?

Edited August 26, 2013 by Radam

[Payload]

Radam, did you have corrective steering enabled?

[Starwaster]

Is a gravity turn such a straight line in this type of representation? Or does mechjeb calculate how fast it should be going at ceirtain altitude?

I don't know that it checks speed but it does check altitude. If you're at a certain point in your flight path (if you have corrective flight enabled) and you're too low then it will pull up more steeply to compensate. Last night I realized I had not staged some empty boosters and the dead weight was holding me back. MJ2 tried to compensate by pulling into a steeper climb. As soon as I shed the dead tanks it pulled up until it had the ship back on track.

Basically if there's any deviation from the flight path it will try to fix the problem. If the option is enabled.

[Radam]

Radam, did you have corrective steering enabled?

I did, but it does go lower without it.

[johnsonwax]

The current behavior (I do not have the latest update, just your previous patch) is preferable - stay prograde until the ship clears atmo, and then plot the circularization burn and rotate towards the node, and *then* warp. The previous version of MechJeb (1.x?) stayed prograde until apoapsis, and then it failed to circularize because the craft takes too long to rotate to the correct direction. I don't want to have to spam monopropellant and RCS on launch stages just to handle that one maneuver. That's my opinion.

I agree. Though out of atmos it turns prograde relative to the maneuver node, while in atmos it should just point prograde relative to its current position - as if you hit the prograde button in SmartASS. In fact I would have it do precisely that.

[Starwaster]

I agree. Though out of atmos it turns prograde relative to the maneuver node, while in atmos it should just point prograde relative to its current position - as if you hit the prograde button in SmartASS. In fact I would have it do precisely that.

That's what it WOULD do unless there's a maneuver node. At which point it's already cut engines so it's time to start turning to face the node. The proposed behavior is just to keep it steady until the circularization has been plotted.