Delta-V woes

[Fraktal]

So. I have a Mun rocket I've already successfully ran several missions with - but when I fired up KSP today, I ended up spending six hours without a single mission actually reaching Kerbin orbit because it just doesn't seem to want to work the way I used it before. Maybe I'm messing up the ascent, but I'm not sure. Anyway, the rocket in question (couldn't make screenshots because KSP ran like excrement after six hours and took fifteen minutes to exit):

• Booster stage
  • Three Reliants on a tricoupler with four FL-T400 tanks each, plus three Thumpers and three winglets. Struts applied where necessary.
  • Originally the Thumpers were mounted on the side of the LF tanks, similar to how an S-shaped asparagus staging is set up, but I dropped that design because I wasn't sure the drag/flex wouldn't cause rotational torque. So instead I hung a number of girders off the tricoupler's exact center (offset tool until RCS Build Aid showed zero engine torque), onto which I attached three extended radial decouplers that clip through the LF tanks to the outside.
  • This stage is capable of outputting a max-throttle TWR of 1.7 up until 81.88 tons. My payload is lighter (60-something, below 70), so I throttled back all engines in the VAB for 1.7 TWR at both takeoff and SRB separation. In this configuration, the stage has just short of 3000 m/s of dV.
• Transfer stage
  • Single Terrier with fuel equivalent to two FL-T400 and one FL-T200 tanks.
  • Originally made up of a single FL-T400 and three radially-mounted FL-T200s, with nosecones on both ends of the radial tanks, to shorten the rocket's height for less flexing. Changed to two FL-T400s and a single FL-T200, all centrally mounted, to test if this configuration is more drag-friendly, but noted no difference aside from gaining over 100 m/s of dV from losing the nosecones' weight.
  • Packs just over 1900 m/s.
• Lander stage
  • Single Terrier throttled back to 60%, with an FL-T400 and an FL-T100, plus four landing legs and four searchlights.
  • Packs over 2100 m/s to have enough spare fuel for a non-equatorial landing (about 20° away from the equator max before cutting it dangerously close, even if I zero out my horizontal speed a few dozen kilometers above the target area and land with a suicide burn).
• Payload stage
  • Command pod with three Communotrons (the extendible, not the radial) and three solar panels on top.
  • One materials bay.
  • Two service bays. One with a science payload (1 exp. storage, one thermometer, one barometer, one goo canister), one with a guidance payload (OKTO and 200 power).
  • Heat shield with 80 ablator.

So then. My problems with this are twofold. First, if I deviate more than about 3-4° away from prograde, the rocket starts rolling. Not flipping due to lack of stability, rolling around its lengthwise axis in a clockwise direction, turning the pitch into yaw, which introduces an inclination change, which causes SAS to start yawing in the opposite direction, which makes the roll even worse until it eventually stabilizes. If I turn off SAS, it keeps rolling for about ten more seconds before stabilizing. I've noticed this on other rockets before, but they all started rolling at far bigger AoA, never this soon. Because of this, I'm having a hard time doing a proper gravity turn, only managing to turn 20° before hitting the target apoapse. I've spent hours trying to figure out why this is so. The rocket is stable (barely) with the winglets it has and adding more winglets doesn't help. Launching without the SRBs doesn't help either. RCS Build Aid indicated a ~5-6 kN torque which I fixed by nudging around the LF tanks on the booster (they weren't centered properly, for some reason). Now I still have a torque of 0.6 in the science payload, but that's not it.

Which leads me to my second issue. Even once I dropped the transfer stage's radial tanks to reduce drag, I'm still losing so much dV during takeoff that it costs me about 4000 m/s to circularize at my usual parking orbit of 120 km. While the transfer stage still has more than enough dV left for a munar encounter, it doesn't have enough left afterwards for me to deorbit it into the Mun like I've done with my previous launches of this design. Yet when I tried reducing my apoapse to 80 km as per the dV map, I hit apoapse so quickly that I ended up 30 seconds late at starting the circularization burn. I'm losing a colossal amount of dV somewhere, I just don't know where.

So then, what are my options? The booster stage still has a few tons of weight capacity left, should I use that to give the booster stage more fuel? That solves the dV budget issue, but what about the rolling?

Oh, and ever since I switched the SRB mounting from a regular decoupler to an extended one, the SRBs are pitching inwards after separation and hit the LF tanks if the rocket isn't aiming perfectly prograde, despite me having mounted them literally at the bottom of the decoupler to center the decoupler force on the CoM. I even offset the SRBs one notch lower, no difference. When I was using the regular decoupler, mounting the SRB on the very bottom caused it to separate without pitching, why isn't this the case with the extended decoupler?

[bewing]

For the rolling, the KSP control algorithms are kinda bad when it comes to 3-way symmetry. If you go to a quad-adapter and do everything 4-way, I bet your rolling problem goes away. Additionally, it sounds like you should be using tailfins instead of winglets. Maybe one of those LF tanks on top of the thumpers at high priority, so it will empty and be discarded early.

You don't want the decoupler pushing at the CoM of the spent SRB. You want it pushing above the CoM, so that the noses of the SRBs splay out on decoupling. Putting the tailfins at the bottom of the SRBs will help accomplish this by adding a torque.

edit: I built a stock version of your ship (with some small modifications) and I'm not having any trouble at all getting it to LKO with 1600 m/s remaining in the transfer stage.

 

Edited January 5, 2019 by bewing

[Aegolius13]

A screenshot of the craft would help diagnose.  But a couple thoughts:

-In terms of high delta-v to reach orbit, it could be an issue with your ascent profile.  Are you going more or less straight up for while, and then making a sharp term towards horizontal?  That's going to be less efficient than a true gravity turn that gets you building horizontal speed as quickly as possible.  You mention having to go a ways off prograde.  Ideally, after your initial pitchover maneuver, you leave SAS locked on prograde and let the gravity turn work itself out.  Some adjustments are usually necessary since it's hard to do an exact pitchover by hand, but the closer to prograde you can stay, the better.  

-Judging by the description of your core stage, it sounds like this rocket might have a lot of drag (again, a screenshot would be helpful here). That might be contributing to the high delta-v requirement.  

-If RCS Build Aid is showing an actual torque figure, that suggests maybe something isn't truly symmetrical.  Maybe try to rebuild the booster from scratch and see if it does the same thing?  Or take it apart one part at a time to see if there's a particular point at which the torque shows up / leaves?

 

 

[Fraktal]

I go vertical until 100 m/s, then start turning. I never turn at the same rate between two takeoffs because I'm not quite that good at paying attention to three things at once, so I just shrug and eyeball it as best as I can.

SAS locking is what I usually go with, yes, but not with this particular rocket (the OKTO can't do prograde, the command stage has only one seat and I'm sending a scientist).

I went through the booster tinkering with each part. If I just attach parts on the tricoupler's triplet attachment nodes, there's torque, but if I select them with an offset tool and click on any of the arrows (not drag, just click), the torque in this stage goes to zero. So KSP isn't placing them properly. The booster stage and the transfer stage are torque-free: rotating them with the rotation tool changes neither the strength nor the direction of the torque. If I rotate the service bay with the science payload, however, it rotates the torque's direction too. Rotating anything underneath that service bay doesn't change anything. So the torque is definitely at that bay. The interior of the bay has a thermometer on one wall (placed on the outside, then rotated and offset to be inside), a barometer on the opposite wall, an exp. storage bay in the center and a goo canister hanging off the side of the storage bay, towards one of the doors. That last one might point at the goo canister as the reason for the torque... except not only RCS Build Aid's torque indicator points in the barometer's direction, if I take off the goo canister, the torque is still there and it's still pointing at the barometer, even though it has the exact same weight as the thermometer and is in the exact same position, so the weight distribution should be symmetrical.

Oh, and the torque direction RCS Build Aid shows is linear, not rotational.

One more thing. I've read somewhere yesterday that clipping parts into each other causes phantom forces. I have a reaction wheel clipped between the top of the booster stage's tricoupler and the decoupler connecting the booster stage to the transfer stage, as the command stage's reaction wheels don't have enough torque to meaningfully steer the booster stage in the upper atmosphere. Could that be causing rotational torque?

Edited January 5, 2019 by Fraktal

[Aegolius13]

The lack of balance in your bay is probably because the barometer is a "physicsless" part (mass is just added to the parent part) whereas the goo container is not.

https://wiki.kerbalspaceprogram.com/wiki/Physicsless_part

I wonder if you're getting asymmetric drag on this stuff.   Aero shrouding can be wacky.  A couple ways to test: you can cheat the ship into orbit with the alt-f12 menu and see if the same problem happens in vacuum.   Or you can use the alt-f12 > physics >aero menu to enable drag info when you right click a part, and see what's generating drag. 

[Fraktal]

I moved the thermometer and barometer to the sides of the exp. storage bay and offset the goo canister as far into the exp. storage bay as I could. Torque is now minimal, though still non-zero.

Also, I decided to disregard the rolling and go aggressive with the gravity turn, flying nearly horizontally by the time my apoapse is beyond the atmosphere. That did the trick, so the problem was definitely with the ascent profile. Still, I decided to go back to the side-tank configuration for the transfer stage but with four FL-T200s instead of three. This nets me nearly 300 m/s surplus dV in the transfer stage by the time circularization at the Mun is complete at 50 km altitude.

Now I just need to figure out why in the world is the Mk1 lander can so damn draggy. According to the UI, it's got nearly 11 drag while the Mk1 command pod directly in front of it has less than 1 and during reentry from a munar flyby, the lander can heated to 79% despite being behind a heatshield with ablator. Do I need to use a 2.5m heatshield here, or is the heatshield dumping heat into the lander can due to the heatshield being attached directly to the lander can with no parts between?

[Aegolius13]

The lander can is intentionally designed to be draggy and bad at reentry - it's intended for vacuum.   That's the tradeoff for lightness.  To minimize drag on launch, best to put it inside a fairing.   

On reentry heat. I'd argue if it didn't blow up, you're doing ok.  79% is actually not that bad.   The heat you're seeing is probably from the incomplete coverage of the shield.  You could use a larger heat shield,  but those are heavy and draggy.  A second command pod might end up being easier.   

[bewing]

1 hour ago, Fraktal said:

or is the heatshield dumping heat into the lander can due to the heatshield being attached directly to the lander can with no parts between?

I'm going to disagree slightly with Aegolius. I think ^this is the correct reason, thermal conduction.

[Aegolius13]

2 hours ago, bewing said:

I'm going to disagree slightly with Aegolius. I think ^this is the correct reason, thermal conduction.

I did a quick test, dropping what I think this ship looks like from Munar altitude.  Here's a screenshot of reentry around point of max heating.  Skin temperature of heat shield is about 2000K, but internal temp is under 300K.  Lander can shows heat  coming in via convection, but conduction is net cooling (of course, this includes heat conducting to the command pod behind the lander).  So I do think most of the heat is getting to the can around the edges, rather than straight through the heat shield.  

But in any event, if the skin temp of the heat shield plateaus at 2000K (increased ablation seems to keep it from going any higher), conduction alone should not get the lander can to explodey-time.

 

[EDIT - just for kicks, I tried the same thing but without the lander can, so that the command pod is fully occluded by the heat shield.  The skin temp of the command pod topped out at only 400 and some degrees, showed zero convection flux and very small positive conduction flux.  Suggests the protruding shape of the lander can is the major difference.

Edited January 6, 2019 by Aegolius13

[pincushionman]

Is the tri-coupler still oddly non-symmetrical, or did that get fixed in the update?

[Fraktal]

Not fixed. I took a tricoupler and put a trio of Ants onto the triple nodes. RCS Build Aid showed non-zero torque. Then I grabbed one of the Ants with the offset tool and clicked one of the arrows. Not dragged anywhere, just clicked. Torque instantly became zero. So yes, the default part placement is NOT symmetrical.