Operations Summary – Week of 9/23/19

Written on 9/19/19

It’s been a little over a week and a half since I wrote the last Desk Notes and normally that would be cause for dismay in lack of productivity but with the new way I’m working things are not so bad. I have the entire next week already outlined, no rocket launches to work soon, and my last update said the game progression was in early November well now it’s in early December. So overall things are moving along well and this is the last time I’m going to talk about this because I’ve proven to myself this new way of working is the way to go moving forward.

Progeny Mk7-A launch

I actually wanted to do more in the way of launch delays but I couldn’t really come up with a compelling reason that would be worth all the extra work it would entail to write up the additional tweets and input the additional database records and reorganize the schedule of events during and following the launch. The most I decided to do was delay to the end of the day cycle for a different time of day than usual and be able to capture some new unique launch photos. I pushed the game clock forward to find a nice photogenic launch time and decided it would look cool to see the rocket rising out of the VAB’s shadow.

I did a bit of “simming” for this mission, which is not something I usually indulge in but for this case I wanted to see how much the rocket would pitch over initially depending on how much deflection was used in the guidance fins. I figured that was a reasonable thing to “model” prior to launch. The ascent code was written using this to try a new way of guiding the rocket off the pad.

Going to launch I ran into issues on startup of the mission code that the VSCode editor can’t really check for yet and there was also a weird issue where the program wouldn’t detect whether the support legs had retracted properly even though that code was proven in the previous launch. Thankfully the easy solution was just to delete the whole process of checking if the arms were down (it is of course kept in the published code) since that’s just superfluous.

Finally I got the mission launched but my first attempt was thwarted when the rocket reached space, I enabled SAS and saw that I only had Level 1 guidance control despite the fact that the tweakable in the VAB was set to the max of 3. This is due to an issue I already knew about but didn’t check for my probe core – if the guidance level is not specified in the CFG file, then it will only ever be able to set to Level 1, despite what the tweakable says. I continued the mission to see if any more problems arose and yea, when decoupling the fuel tanks & engine everything screwed up when my vessel got renamed. Turns out I forgot to check if the root part was properly set before launch. That’s now a new checklist item.

So after restarting the game my second attempt was also thwarted in the worst possible way – flew the whole damn mission fine then when the rocket crashed into the surface I had focus off the game window so I clicked to bring it back and at the time I made the click the screen area was empty but it just so fucking happened to be where the after-crash dialog box pops up with end mission info and options – one of which is Revert Flight that was under the cursor when I clicked and reset the mission before I had a chance to record and save vital post-mission data. Somehow I did not completely flip out and just sighed and reloaded for a third attempt.

There was one silver lining in that before I made my third attempt I checked the telemetry data to see if there were any mistakes and noticed stage fuel was only being recorded for the first stage booster. I quickly realized this was because I don’t actually use the stage command except for at launch to trigger the VOID logging data – I use kOS‘s ability to directly access part actions to do stuff like activate engines and decouple things, so the ability to track stage fuel was not working. Easy fix, just add staging commands to the code I was using for the actual flight (not the published code). The third time was the charm.

I had to go back to get the trajectory of the RTG, since after it was decoupled kOS automatically made a save file but for some reason, likely because it was a single part without any command capability, the game did not also save its trajectory information and when I reloaded the save I got a nullref error and no RTG to switch to from the rocket. So whatever, I just edited the save file to have the RTG attached on load and then had an instruction file ready for my boot script to execute and immediately decouple the RTG. So maybe it was let go like a second or so later that actual, but made little difference.

I also want to make perfectly clear that this mission was not written or designed to succeed. I had no idea if igniting the engine at 25km would be high enough for its gimbal to be effective, and I had not foreseen that it would almost drift too far off prograde during the coast to third stage ignition. That was a legit oh shit moment for me during the ascent! If it had failed for either of these reasons, or any others equally as plausible, I would have let if fail. Not knowing if a mission will fail or succeed makes things interesting for me as well since it’s a possible narrative branch I can’t plan beyond.

Ascension Mk1 video

What was said about waiting for a good mission to make a video for was pretty much exactly how I’ve felt since the rocket began missions back in mid-2018. Of course I added some well-reasoned “in universe” explanations for the delay as well, based on actual past lore, but basically I just didn’t feel like any prior missions deserved the full treatment I wanted to give the video. Most importantly though, I was waiting for development on Rocket Sound Enhancement to improve further, but it’s been stalled for a while now and although the mod author says he’s coming back that hasn’t happened yet. I’ve always deplored the rocket sounds in KSP but I felt they finally reached a usable point with this mod.

The video took about 5 days to complete, working in between sessions of advancing the game clock as I would have to reload the game anyways for various shots – the establishing, tower and tracking shots were all done with minimal graphics mods for better frame rates and process clock but I of course had to turn that all back up for the booster camera. I captured the audio with no graphics mods at all to ensure it was synced to in-game actions properly. The final booster cam footage of MECO and capsule release were also done with Ambient Light Adjustment set to 0 since it was is direct sunlight when looking at the capsule. KSP has issues with darkening the sky when looking towards the sun so I had to use Distant Object Enhancement to force the skybox to hide as well.

I also had to spend time debugging an issue with Hullcam VDS where the booster camera view was clipping into the rocket. Turned out to somehow be a result of using Physics Range Extender, which thankfully was not needed for the booster cam footage.

Video edited with Sony Vegas 9.0

After getting all the footage, then came the task of syncing it properly which meant I also had to record some segments twice, once with the GUI active so I could see the altitude of the rocket along with the mission elapsed time. This is important because due to the graphics mods the game does not run at the same speed for the entire ascent. Where not much background rendering is needed the game clock will be green, but as more of the planet comes into view the game clock turns yellow and slows down to keep the physics processing accurate. This means that, for example, the booster cam footage of the beginning of the launch in 5 seconds the rocket moves further than in 5 seconds once it’s at like 20km. Thankfully knowing the timing of events and being able to compare the GUI footage to the non-GUI footage to place the rocket at certain points in the ascent I can “time compress” the video to make the length of time it should have taken for that segment of the flight to occur.

Dunno if any of that was described well but suffice to say I can’t just record something and be done – it needs a lot of editing still, including taking out freeze-frames where the game pauses for garbage collection. Even Memgraph doesn’t do away with them entirely, just makes them shorter and less noticeable but still present.

Then came adding the sound, which as you can see above was manipulated still so the initial ignition would be softer with a beefier throttle up to launch thrust. Finally I could render it all out and upload it to YouTube, which still took three tries to get right because I don’t do this much and make dumb mistakes.

Overall I could cut this down to maybe 2-3 days since I had to rediscover solutions for a lot of mistakes I made and forgot about, so with experience this would become easier. It’s still a bit too much of a time sink to do still for every launch. Regardless I am pleased with the results and hope everyone else is as well.