TWR not TWR?

[Red Shirt]

I started asking questions in the MechJeb thread. I'm moving things over here rather than clog up Sarbians thread with clutter.

Background: I saw Cupcakes tiny Ion dropships and it inspired to to build my own Ion lander as an anomaly hunter. The first one worked perfectly on Minmus. I was able to take off and land with MJ multiple times without needing to dock with the mothership for refueling. Then I tried to use it on Mun and crashed. Rebuilt. Crashed. Rebuilt Crashed. At first I was running out of power, so naturally I kept adding batteries and solar panels. By the time I had sufficient power my TWR was pretty low - around 1.07. So I decided to go crazy and built this nearly 3 ton ship with 8177 dv. It should work. The engines have full power during the entire descent. My TWR is 1.69 starting at 15k orbit. When the ship slams into the ground at abt 50 m/s the TWR is 1.71. 

I am using MJ's landing guidance. Originally I thought MJ couldn't handle Ion engines. @Foxster slapped together a tiny ship and landed with MJ proving it can be done. So besides my particular beasty being over engineered, why won't it work? I can build a ship with any other engine (including the ant) with a similar TWR and it works. What is different here?

(Oh, and Foxster, I am apparently using a different version KSP than you as I cannot open your craft file. I'm still on 1.2.1) 

[Superfluous J]

I don't use ions or Mechjeb much, but I can tell you that 1.7 TWR is lower than most people are normally used to. It's a little more than the TWR of most of my Tylo landers and I almost always crash them on the first try. And Tylo's a bit flatter than Mun. It could be simply that you're smacking into a mountain due to having to come in so low and horizontal at such a high velocity to nail the landing.

MechJeb is telling you your acceleration is 2.8, which is about 1.7 TWR for the surface. So it should work. In theory.

Note: In theory, the difference between theory and practice is negligible. In practice, though...

Edited December 9, 2016 by 5thHorseman

[Starman4308]

14 minutes ago, 5thHorseman said:

I don't use ions or Mechjeb much, but I can tell you that 1.7 TWR is lower than most people are normally used to. It's a little more than the TWR of most of my Tylo landers and I almost always crash them on the first try. And Tylo's a bit flatter than Mun. It could be simply that you're smacking into a mountain due to having to come in so low and horizontal at such a high velocity to nail the landing.

MechJeb is telling you your acceleration is 2.8, which is about 1.7 TWR for the surface. So it should work. In theory.

I can't say much for MJ landings (I basically never use it), but people can and will use lower TWRs; I've landed on the Mun (6.4x resize, 3.2x rescale) multiple times with starting TWR of 0.95-ish. I think somebody once ran a bunch of plots showing that most moons, the optimal TWR is close to or less than 1, it just requires a lot of patience and judgement on how fast to descend without making a too-strong first impression on the Mun.

16 minutes ago, 5thHorseman said:

Note: In theory, the difference between theory and practice is negligible. In practice, though...

I laughed. Hard.

[foamyesque]

1.7 TWR is in theory enough to land, but the thing is, 1.7 TWR on the Mun is going to feel really painful trying to slow down. Your time to kill your original orbital velocity is going to be somewhere between three and a half to four minutes of burn time, plus more to account for the gravity losses during the tail end of the burn and to keep you aloft. You've got piles and piles of deltaV; take your time and just keep your vertical velocity cancelled out, at an altitude of around 5.5km, until your horizontal is also gone, then just drop it gently to the surface.

Edited December 9, 2016 by foamyesque

[Starman4308]

1 minute ago, foamyesque said:

1.7 TWR is in theory enough to land, but the thing is, 1.7 TWR on the Mun is going to feel really painful trying to slow down. Your time to kill your original orbital velocity is going to be somewhere between three and a half to four minutes of burn time, plus more to account for the gravity losses during the tail end of the burn and to keep you aloft. You've got piles and piles of deltaV; take your time and just keep your vertical velocity cancelled out until your horizontal is also gone, then just drop it gently to the surface.

I think I'm just going to continue laughing at this point. And maybe a little bit second-guessing my decision to play at 6.4x scale with my compulsive "must land/ascend efficiently with bare minimum of TWR" habits.

[foamyesque]

6 minutes ago, Starman4308 said:

I think I'm just going to continue laughing at this point. And maybe a little bit second-guessing my decision to play at 6.4x scale with my compulsive "must land/ascend efficiently with bare minimum of TWR" habits.

 

Oh, I've done sub-1 TWR landings too, but I hate 'em and it's trivially easy to get better TWRs than that for the Mun in stock.

 

Besides which additional TWR makes better use of your deltaV

[Superfluous J]

15 minutes ago, Starman4308 said:

I can't say much for MJ landings (I basically never use it), but people can and will use lower TWRs

Oh for sure. I'm not saying it's not doable. I'm saying it's a chore and takes quite a deft hand.

[Starman4308]

4 minutes ago, foamyesque said:

Besides which additional TWR makes better use of your deltaV

Thing is, high TWR is more delta-V efficient, but not always more mass efficient. You pay for that higher TWR with a heavier engine, meaning more propellant to get the same delta-V. Thus, at constant mass, higher TWR uses less dV but has less dV to start with, and the optimal point is often pretty close to 1.0, assuming an efficient descent profile. It doesn't help matters that fuel tends to be cheaper than engines.

I mean, you may already know that, and just sanely choose "hey, I could save 10 kg on the lander, or I could save my sanity. Which do I choose?".

Edited December 9, 2016 by Starman4308

[Streetwind]

8 hours ago, Red Shirt said:

I am using MJ's landing guidance. Originally I thought MJ couldn't handle Ion engines. @Foxster slapped together a tiny ship and landed with MJ proving it can be done. So besides my particular beasty being over engineered, why won't it work? I can build a ship with any other engine (including the ant) with a similar TWR and it works. What is different here?

I can only speculate, ultimately, but I can think of a few things. For starters, it's probably the high Isp of the ion drives that's screwing you over.

Sounds weird? Well, consider this. Your ion craft has a whole ton worth of engines. Additionally, it has two large solar arrays. Then, it has a tank with a mass ratio of only 2.27. And finally there's a large battery. Meanwhile, a similar craft with an Ant engine has only a small battery, much smaller solar panels, a tank with a mass ratio of 9, and engines massing 0.01 tons per Ant. All in all, even if the initial TWR is the same, the ion craft carries much more dry mass and much less fuel mass than the liquid fuel variant. When combined with the much poorer Isp of the Ant engine, the liquid fuel variant will lose fuel mass much much faster than the ion craft during its descent, resulting in the total mass of the craft dropping rapidly. This results in its TWR growing much faster during the descent, and reaching a much higher number at the end of it.

Ergo: precisely because the ion engines are so efficient, they put you in a situation where your TWR at the end of the burn is hardly any different than the TWR you started with. A liquid fuel craft however benefits from shedding a lot of mass on the way down and can boast a significantly higher deceleration in the critical last few moments.

 

The second thing is, quite simply, distance. Take a car, driving on the road at a set speed. If the driver wants to come to a dead stop as soon as possible, there is an absolute minimum distance the car needs to brake to decelerate from this speed, no matter how hard the driver slams the brakes. It is much the same with your spacecraft. It is traveling at a certain speed, and can decelerate only so much per second. If you want to land successfully, then your distance to the surface must be larger than the minimum distance you must travel in order to remove all your speed. If your TWR is really low, that minimum distance is really high - in fact, in your case I'm pretty sure that it is higher than the distance your spacecraft travels from orbit down to the surface. Hence, it impacts the surface while still going much too fast to survive.

Now, in contrast to a car, a spacecraft has six degrees of freedom, and therefore can 'cheat'. You have the ability to sacrifice a part of your deceleration and divert it towards pushing you away from the surface, thereby lengthening the distance to impact. This is entirely a (manual!) piloting exercise. Instead of burning directly at the horizon, you burn slightly towards the surface, waching and controlling your vertical speed. The slower you go, the more you have to divert towards the surface, and therefore, the longer it takes to remove the remaining speed. You are essentially performing a fluid transition from an unpowered orbit to a powered stationary hovering.

Reading this, you might be thinking: isn't this awfully fuel-inefficent? And the answer is: yes, yes it is. Hovering is a bad thing, which you ideally want your lander to never do. The lower your TWR, the more dV you need to throw at gravity in order to bribe it into not slamming your spacecraft into the ground at speed. The growth in cost is exponential with lowering TWR.

 

My personal recommendation is that you should perform a powered landing with no less than an initial TWR of 2 and/or a terminal TWR of 3. Anything less than that, while still possible, starts becoming a pain in the butt. And even if you aim to save on money and launch mass and fuel... look at your ion craft. It probably saves on none of these things. It needs so much heavy and expensive stuff, just to achieve a deeply suboptimal deceleration force. All your dV isn't going to be worth anything if you cannot expend it fast enough before plowing into the ground, or if you have to shovel it out the window to appease gravity. Ion drives make poor lander engines in most places with more gravity than Minmus. (And IRL, they can't be used for landing at all.)

Edited December 9, 2016 by Streetwind

[Starman4308]

3 hours ago, Streetwind said:

Reading this, you might be thinking: isn't this awfully fuel-inefficent? And the answer is: yes, yes it is. Hovering is a bad thing, which you ideally want your lander to never do. The lower your TWR, the more dV you need to throw at gravity in order to bribe it into not slamming your spacecraft into the ground at speed. The growth in cost is exponential with lowering TWR.

Thing there is that exponential growth is only significant when your final TWR is very, very close to 1.0.

Trigonometry is your friend for landing. In the final moments of an ideal suicide burn, the vertical component of your velocity must be 1.0, thus TWR * sin(angle) = 1. The fraction of your thrust lost to gravity is not (1.0 / TWR), but rather 1 - cos(angle) = 1 - cos(sin^-1(1/TWR)). The effects during the rest of the burn are a bit more complicated to calculate; while you'll have less TWR (having not burned through all your fuel), effective gravity will be less than 1, because of orbital effects.

Anywhere save very near 1.0, increasing TWR doesn't help gravity losses much, and does require significant increases in engine mass. Now, the full story is more complicated than that; you have to integrate these losses over the entire descent profile, and you can be more aggressive with high TWR, as low TWR burns often entail planning to go over not only the hills you see, but also the hills hidden behind the curvature of the Mun. I'll try to get some simulations going, but I do rather specifically remember that somebody already did a lot of these calculations, and came up with the conclusion that for many moons, particularly the heavier ones, optimal starting TWR is pretty close to 1, and sometimes less than 1.

[Streetwind]

I'd argue your definition of "not significant". Dropping from 2.0 to 1.5 increases your gravity losses by 150%, according to your chart. That's pretty significant to me! But, let's do an example:

Take a 20 ton lander with a mass fraction of 50%, fitted with a Terrier. That's a Mun TWR of 1.875 and a dV of 2345 m/s roughly. Now add a second Terrier. The vessel now weighs 20.5 tons and has a Mun TWR of 3.66. Its dV dropped to 2263.6 m/s. That's 96.5% of what it had before, 81.5 m/s less. The vessel lost 3.5% of dV. But the gravity losses went from being ~15% of the whole descent to being, I dunno, that looks like less than 5% in the chart. If your Mun landing took, let's say 600 m/s to reduce velocity to 0, then roughly 106 m/s gravity losses went on top of that for a 15% share. But if we assume just 4% gravity losses, then you'd add only 25 m/s. The difference? 81 m/s.

Result: by adding another engine and doubling your lander's initial TWR from 1.875 to 3.66, you managed to make your descent that much more comfortable while essentially breaking even on the dV front. And 1.875 isn't all that far from my own rule-of-the-thumb minimum of 2.0. I'm pretty sure that as you drop below 1.8 initial TWR, and/or use engines with a higher base TWR than the Terrier's awful <12, you only stand to gain by adding more engine. So for my personal opinion's purposes, the losses start becoming "significant enough" below 2.0.

 

Also keep in mind the OP's actual scenario, with the ion engines. He manages to combine very high engine mass with very low craft TWR. He definitely suffers a lot from various inefficiencies here.

Edited December 9, 2016 by Streetwind

[Norcalplanner]

I'm going to be a bit heretical here - don't use MJ's auto-landing feature.

The way I use MJ when landing is a bit different, and you can still use the auto-landing feature at the very end. The critical thing is to add one of the data points hidden in the custom window editor under "misc", in this case the suicide burn timer.  Do a slight initial deorbit burn so that your orbit line intersects a bit past where you want to land, but you're still coming in at a fairly shallow angle. Then lock your heading on surface retrograde using SmartASS. Once the suicide burn timer gets down to 3 or 4 seconds, go to max throttle and start watching the suicide burn timer like a hawk. As long as it doesn't go more than a second or two negative, you're fine. If it goes more negative, dial in a few degrees of positive pitch using SmartASS until the timer gets back up to zero. The trigonometry works out such that burning 5 or 10 degrees off retrograde results in minimal losses. As you approach the surface and the timer starts to increase, remove any positive pitch and reduce throttle slightly. You should end up going less than 10 m/s only 100 or 200 meters above the surface, then complete the landing using whatever method you choose (including hitting the auto land button at this point). I tend to use SMARTASS all the way in, switching to "up" orientation at the end to make sure the lander doesn't flip accidentally.

[sardia]

19 minutes ago, Norcalplanner said:

I'm going to be a bit heretical here - don't use MJ's auto-landing feature.

The way I use MJ when landing is a bit different, and you can still use the auto-landing feature at the very end. The critical thing is to add one of the data points hidden in the custom window editor under "misc", in this case the suicide burn timer.  Do a slight initial deorbit burn so that your orbit line intersects a bit past where you want to land, but you're still coming in at a fairly shallow angle. Then lock your heading on surface retrograde using SmartASS. Once the suicide burn timer gets down to 3 or 4 seconds, go to max throttle and start watching the suicide burn timer like a hawk. As long as it doesn't go more than a second or two negative, you're fine. If it goes more negative, dial in a few degrees of positive pitch using SmartASS until the timer gets back up to zero. The trigonometry works out such that burning 5 or 10 degrees off retrograde results in minimal losses. As you approach the surface and the timer starts to increase, remove any positive pitch and reduce throttle slightly. You should end up going less than 10 m/s only 100 or 200 meters above the surface, then complete the landing using whatever method you choose (including hitting the auto land button at this point). I tend to use SMARTASS all the way in, switching to "up" orientation at the end to make sure the lander doesn't flip accidentally.

I agree, there's something wrong with the mechjeb landing functions. You can use it to de orbit and to land at the end perfectly. But if you use it to set up the whole landing, it misjudges your burn, and crashes you into the planet. 

[Red Shirt]

OK, I stripped my lander down to no frills. With a surface TWR of 1.84 I landed with MJ Landing Guidance

. That extra 0.1 made the difference between a 0.5 m/s safe landing and a 50 m/s crash and burn.

With the much lower initial dv, the use of an ion multiple anomaly hopper craft isn't so appealing. I remain confident that I could land the 4 engine lander but not with MJ Landing Guidance. Without it I won't land anywhere near my target but I may try just for giggles.

1 hour ago, Norcalplanner said:

I tend to use SMARTASS all the way in

I have never understood this feature. 

[Foxster]

This is just some weird space science stuff we poor mortals may never understand. I build everything by guesswork and "feel" because the maths just whooshes over my head.

Someone mentioned something about dV/s being too low for your craft. They might as well have said the faeries per sponge was too ebullient.  

 

Edited December 9, 2016 by Foxster

[wumpus]

1 hour ago, Norcalplanner said:

I'm going to be a bit heretical here - don't use MJ's auto-landing feature.

I'd simply use the "hold retrograde" SAS feature, set a timer for five minutes, switch to windowed mode (which is unfortunately quite cumbersome in KSP) and then chat on the forums for a few minutes (you can find out how long by creating a maneuver node in orbit that goes to a complete stop (vertical fall) and checking how long that will take).  When you are more or less falling vertically, take over and land it.

Can't help you with the orbit you should start at, but I'd go high in the belief that you don't need to care about an ion engine's fuel.  Too low and crashing while reading the forum is possible.

[Mastikator]

UIm wat, why are you using ion thrusters to to to the moon?  @Red Shirt

If you're using a seat as a pilot then surely a FL-100 (or FL-200) and a spark would do the trick much easier?

[Red Shirt]

6 minutes ago, Mastikator said:

UIm wat, why are you using ion thrusters to to to the moon?  @Red Shirt

If you're using a seat as a pilot then surely a FL-100 (or FL-200) and a spark would do the trick much easier?

I blame Cupcake. The Owl Ion dropship was just too cute. Worked great on Minmus. Wanted my own design. Also worked great on Minmus. The reason I attempted this is with a small FL-T or even the Oscar tank and a small engine (the ant works well) I can land and take off. I might be able to hit two anomalies before needing to refuel. My thought with the ion was I might hit 4 or 5 before needing to refuel. I liked the challenge. The math suggested it works but I couldn't get the job done.. 

[Sharpy]

I have no clue about MJ but I've been landing on Mun with lousy TWR before. I landed a rover with TWR of 1.3 or less.

The secret is not to burn pure retrograde, but between retrograde and radial out.

As you lose horizontal speed, you're gaining vertical speed with growing gravity drag. With engines pointing retrograde - mostly horizontal - by the time you lost enough horizontal speed for the engines to point downwards and arrest your vertical speed, you're moving too fast to brake before hitting the ground. So you need to keep your vertical speed in check at all times, use as much of horizontal component as you can afford for arresting horizontal speed, while sacrificing most of thrust to keeping descent speed low.

My rover had one more ugly problem: fixed solar panels. As I was not landing near the 'noon' area, I'd have to tilt it to have my battery charge, reducing thrust not to accelerate sideways too much, then use up whatever I gained in arresting the fall, then repeat... uh, ugly work.

[Red Shirt]

I've been tinkering with ion landers for several days now. I finally have a 1.3 ton, 2 engine ion engine, lander with a TWR of 1.93 and 4198 dv after the first landing. And...

it's all pointless as the cfg I wrote thinking it would allow multiple flags without reentering an actual pod, doesn't work. Oh well, I can scratch this arch off my to do list.

[Enceos]

The lower TWR is the higher the gravity losses during the landing, sometimes it exceeds the required dV to land with a conventional engine by 30% and more. dV numbers are misleading when TWR is low on the lander.

Edited December 9, 2016 by Enceos

[FleshJeb]

I've done a lot of experimenting with low TWR (1.1-1.3) landings on the Mun.

The problem isn't the TWR, it's the absolute acceleration. Say for instance you were coming straight down at 50/ms: That leaves you only (1.7- 1.0) * 0.167 * 9.8m/s^2 = 1.15 m/s^2 to play with. It will take 43s to kill that off. This is out of bounds of what MJ can handle.

I'm pretty lazy, I like to automate/simplify as much as possible. What I started doing was modeling a pre-burn for my landings. The goal is to halt almost all your horizontal velocity, and minimize your vertical velocity over the landing point. Since I don't have my spreadsheet on me, I'll give the short version:

• Pick your landing point and turn on predictions with MJ's Landing Assist.
• Manually place a node, and edit it:
  • Retrograde is your orbital velocity (I think it's about 542m/s @ 8km).
  • Set Radial Out to 50m/s (100m/s for TWR 1.1-1.3). the resultant is 544, and I normalize* this to be my orbital velocity. So Retro becomes 537, Radial becomes 49 (533 and 98 for lower TWRs).
• Adjust the time on the node until the landing prediction hits the correct longitude
• Edit the node to add whatever Normal/Anti-Normal gets you close to your landing latitude.
• Adjust the time again to minimize the landing error. (It's better to fall short than it is to go long, because you want the correction burn facing up.)
• QUICKSAVE
• Have MJ execute that burn. Your craft will end up rising quite a bit (mine go from 8km-20km-12km), but you'll be almost over your landing point, with minimal horizontal velocity, and only a little vertical.
• Punch Autoland. The automatic correction burn should be pretty small. (My spreadsheet keeps it between 2-10m/s).
• MJ should be able to deal with your residual vertical velocity.
• If it fails, Quickload and add more Radial Out, reducing the Retrograde appropriately.

Obviously this isn't as pretty or efficient as a Constant Altitude Landing, but it's simple, gets pinpoint landings, and it's repeatable.

*normalizing the resultant (I forget why this works--it just does):

SQRT( 542^2 + 100^2 ) = 551

542 / 551 = 0.983

542 * 0.983 = 533 Retro

100 * 0.983 = 98 Radial

 

[Starman4308]

Let it not be said that I do not put mine money wherein be mine mouth. My overall conclusion is that for small bodies, I should be using a higher TWR than I am, but I'm in an excellent range for the colossal bodies like Tylo, where surface gravity is high and delta-V requirements are through the roof.

Note: I will probably revisit this shortly, because I completely forgot to simulate the ascent.

Assumptions:

Spoiler

 

Constant-altitude landing at 15 km (stock) or 25 km (64k-world, where I live)

Engine thrust-to-mass ratio: 200.0 (stock, that of a Spark engine), 750.0 (RealFuels, about a TL 7 Sparkler MMH/NTO engine)

Fuel full-to-dry ratio: 9.0 (stock, exact), 18.0 (RealFuels, approximate for default tank with MMH/NTO)

Useful payload: 1 ton

Timestep for landing simulations: 0.04 seconds (it's what KSP uses, so hey!)

If a cell reads NaN (Not a Number), it exceeded 20 iterations on simulating the lander weight (after calculating the delta-V). This only occurred for 6.4x scale Tylo, and only for very high TWR and low Isp combinations.

The program uses Apache FastMath as a dependency; I've included the pom.xml I used to compile the program. If you can't be bothered, just replace FastMath.anything() with Math.everything() from the Java core libraries.

There is some commented-out code; rather than pass in any command-line arguments, I just edited the Java class file as needed. Bug me if you really need something more user-friendly.

The stuff past the Planet class is irrelevant and used for figuring out how effective Java 8 Streams were.

 

THE PROGRAM: https://www.dropbox.com/s/ot471gupq469e0c/Calcathing.java?dl=0

THE RESULTS: https://www.dropbox.com/s/uovrt0ctafn08pt/landerMassFracts-All.xlsx?dl=0

THE POM.XML: https://www.dropbox.com/s/963o55bs4plsnmo/pom.xml?dl=0

THE SHORT DIGEST (at isp 340):

Stock Mun: optimum is 1.95

Stock Minmus: optimum is 1.75

Stock Tylo: optimum is 1.35

64k Mun: optimum is 3.4

64k Minmus: optimum is 4.05

64k Tylo: optimum is 1.2

My conclusion from the stock-64k difference is that the much better thrust-to-mass ratios of RealFuels engines make it trivial to just slap on a giant engine and call it a day, because it doesn't mass much, except for Tylo which just hates you forever.

EDIT: Ladies and Gentlemen: when attempting to SSTO off 64k Tylo, the optimal TWR for 340 isp is 1.2, providing an amazing landing weight of negative 21.7139 tons. I'm sure this is a real result and not because the math returns a bogus result for situations like "not physically possible to SSTO at that Isp and wet/dry tank ratio even with infinitely light/powerful engine". I think I'm going to finish implementing land-and-ascend tomorrow, and actually play some dang KSP today.

Edited December 10, 2016 by Starman4308

[Starman4308]

So, added the ascent to the simulations in two ways. First, I just assumed a single stage for both descent and ascent, and second, I just assumed the ascent stage would be completely independent of the landing stage and not share any parts or other shenanigans.

Updated code: https://www.dropbox.com/s/c0mw713bvmn74zq/Calcathing-v2.java?dl=0

Results for single-stage: https://www.dropbox.com/s/e7wh0fsdvzcend8/singleStageFracts-all.xlsx?dl=0

Results for ascent-only: https://www.dropbox.com/s/gnuv094wlfj383p/ascentMassFracts-all.xlsx?dl=0

As an aside: the wonder-negative-masses from last night were mostly because, when updating the mass fraction (relative to the start) at the end of each ascent simulation timestep, I forgot to add in the delta-V consumed by the landing.

Overall, for ascent-only stages, you'll want a little bit more TWR than descent-only stages, in part because you've got the most gravity losses when your craft is the heaviest, unlike landing where gravity is strongest after shedding the most fuel.

Stock Mun: optimum is 2.1 (again, assuming engine with 200 kN/ton and 340 s Isp)

Stock Minmus: optimum is 1.8.

Stock Tylo: optimum is 1.7.

64k Mun: 3.75

64k Minmus: 4.2

64k Tylo: 2.0

For single-stage land/ascend vehicles, this trend is amplified for the smaller moons, but reversed for the big moons (Tylo, 64k Mun, see note on 64k Tylo).

Stock Mun: 2.2

Stock Minmus: 2.05

Stock Tylo: 1.15

64k Mun: 3.35

64k Minmus: 4.55

64k Tylo: heeheehaahaa, they're coming to take me away, hee-hee, ho-ho, haha, to the funny farm! Not even hydrolox cuts it here; the only solution is maybe a LOX-boosted nuclear thermal rocket.

64k Tylo with stock engines and tanks: there's a little corner in the lower-left which corresponds to "LV-N hacked to have way more thrust than it should".

[theend3r]

I've landed on mun with TWR of about 0.6 so that's not the problem. I use MJ, too, but definitely not for landings or docking, since I find those not only fun but also sometimes quite unpredictable.

(BTW this sound like a great challenge and you'd get some empirical data, too).

Edited December 10, 2016 by theend3r