sevenperforce

Ah, yep, that makes sense. I was going to be pretty shocked if the fineness was THAT low. I could see it for a hydrolox rocket but not for methalox.

IIRC that was basically the same failure mechanism as (presumptively) IFT-2 experienced: slosh in the tanks during a rapid flip that dumped vapor into the turbopumps.

I understand the point you’re making but perhaps it’s not the best example to give in regards to a rocket engine. For rocket engines the turbo pumps are part of the engine. If you understand the point being made, take the point being made instead of taking some other point that isn't being made. Suppose a truck manufacturer builds a semi-truck, and early tests of the semi by the manufacturer show that the rear wheel assemblies have a tendency to seize up, stop spinning, and frag the axles at highway speed. The truck manufacturer promptly redesigns the wheel assemblies, and they no longer exhibit this problem. Much later, in acceptance testing, the truck manufacturer learns that the transmission is not downshifting properly and is transmitting ten times too much torque to the axles, which causes the axles to fail. It would be very silly to suggest that this transmission failure in acceptance testing is the same wheel assembly failure exhibited in early testing, simply because both involve the axles.

That is a CHONKY boi.

It's so cool to watch with the audio on, because the plasma stream starts to heat up visibly before you get any significant wind noise. It's really something. All those burning bits from the ablatives are crazy to watch too.

I went and glanced through the last couple years of launches to try and get some benchmarks for the largest launches to given destinations. Falcon Heavy All Cores Expended ViaSat-3, 6.722 tonnes to GEO (5/1/2023) Center Expended, Boosters RTLS Psyche, 2.608 tonnes to heliocentric orbit (8/4/2022) Echostar-24 (Jupiter-2), 9.2 tonnes to GTO (7/29/2023) USSF-67, 3.75 tonnes to GEO (1/15/2023) Center ASDS, Boosters RTLS ArabSat-6A, 6.465 tonnes to supersynch GTO at 90,000 km apogee and 23° (4/11/2019) Falcon 9 Expended Galaxy 31 & 32, 6.6 tonnes to supersynch GTO at 283x58,433 and 24.2° (11/12/2022) Eutelsat 10B, 5.5 tonnes to supersynch GTO at 261x59,831 and 22.8° (11/23/2022) ASDS SES-18 & 19, 7 tonnes to GTO (3/17/2023) Intelsat 40e TEMPO, 5.59 tonnes to GTO (4/7/2023) Inmarsat-6 F2, 5.47 tonnes to supersynch GTO at 387x41,592 and 27° (2/18/2023) Starlink, 18.4 tonnes to LEO (12/7/2023) USA-343 GPS III-06, 4.352 tonnes to MEO (1/18/2023) Galaxy 33 & 34, 7.35 tonnes to subsynch GTO with apogee 19,800 km (10/8/2022) Danuri, 0.679 tonnes to ballistic lunar transfer (8/4/2022) RTLS OneWeb #17, 6 tonnes to LEO (3/9/2023) SARah 1, 4 tonnes to SSO (6/18/2022) Hakuto-R, 1 tonne to ballistic lunar transfer (12/11/2022) EROS-C3, 0.4 tonnes to retrograde LEO (12/30/2022) I feel like the supersynch launches are particularly helpful because they represent a maximization of the capability of the stage with a given mission profile. May try to math around with this a little.

So, basically, SLS? OOOOOOH that burns like an pad without a deluge. I dunno why but that gives me the ick. Kind of figured something along these lines. A real solid workable EVA suit -- one they actually want to build on rather than use as a tethered one-off -- is tough. But doable. There are a lot of unknowns about internal operational cost of ASDS vs RTLS and so forth. In theory, there are a total of 17 different flight options for the Falcon family: One core RTLS ASDS Expended Two core Both RTLS Booster RTLS and center ASDS Both ASDS Booster RTLS and center expended Booster ASDS and center expended Both expended Three core All RTLS Boosters RTLS, center ASDS Boosters RTLS, center expended One booster RTLS, one booster ASDS, center ASDS One booster RTLS, one booster ASDS, center expended Both boosters ASDS, center ASDS Both boosters ASDS, center expended All expended I really have no idea what the correct ordering is in terms of payload capability. I suppose the idea is that if a tile comes loose, a smaller area will be exposed. Am I missing other possible advantages? That's one advantage, but another one probably comes before that: smaller tiles have a lower torque arm moment and thus are less sensitive to flexion of the underlying substrate surface. This would make sense if this is a problem area where they keep losing tiles no matter what.

The header tanks contain 6.25 tonnes of liquid methane and 23.75 tonnes of liquid oxygen.

I don't know how else to tell you, man -- if you put garbage in, you're going to get garbage out. If we could get magic, weightless nozzle extensions for free, that don't take up any volume or mass, then rocket engineers everywhere would be adding them to every engine everywhere. And no, it's still not over 130 tonnes to LEO, because not only are your numbers all wrong, but you're also not applying the overestimation factor. It's a matter of adding a nozzle extension to an already intrinsically-high-performing engine cycle. An expander cycle engine is fundamentally different than a gas generator engine. Look at the RL10A-4-2. It has an expansion ratio of 84:1, lower than the expansion ratio of the J-2X, but it achieves 451 seconds Isp because it is an intrinsically higher-performing engine. Gas generator engines simply do not have the same efficiency potential as expander cycle engines. The highest-efficiency GG engine flying today, the HM-7B, has an expansion ratio matching the RL10A-4-2 but it only pulls 446 seconds. The record-holding LE-5 gas generator engine had an expansion ratio of 140:1, more than 50% higher than the J-2X, and it was only able to reach 450 seconds Isp. And this is before you factor in issues like changes in O:F ratios, chamber pressure, chamber temperature, and so forth. Rocket science is a complex tradeoff balancing many different factors to decrease weight and size while increasing thrust, efficiency, and reliability. You can't just slap an expander-cycle nozzle extension onto a gas generator engine and expect to get expander-cycle efficiency any more than you can slap the straight pipes from a Mustang onto a Prius and expect the Prius to go 0-60 in 4 seconds. This just isn't true. The largest-diameter nozzle extension for an RL10 comes in at 2.15 meters, while the interstage of the Delta IV has a diameter of 5 meters. A 4.66-meter nozzle extension for the RL10 (the same size, relative to the Delta IV diameter, as your notional J-2Z) would bring its expansion ratio to 1,315:1. So why not do that? It should be obvious: the efficiency savings wouldn't outweigh the mass bloat. As I discussed upthread, impulse scales approximately with the square root of (1 - RP(k-1)/k), where RP is the inverse of the expansion ratio and k is a specific heat ratio. Let's pause and take a closer look. If the expansion ratio is infinite, the RP term drops away, and the specific impulse scales to the square root of 1, which is just . . . 1. That's because every engine has an inherent maximum specific impulse set by the chemistry of the propellants and the thermodynamic constraints of the universe. Increasing the expansion ratio decreases the loss associated with underexpansion at the nozzle outlet, but it can never exceed that theoretical maximum. However, that equation for impulse is only part of the picture. Imagine a rocket engine with a nozzle of zero length. If the expansion ratio is simply 1, then 1 raised to the power of anything is still 1, and so impulse would scale with the square root of (1-1) which is zero. Does a rocket engine with a zero-length nozzle have no thrust? Of course not -- intuitively we know that it would still have thrust: The thrust produced by this engine is simple: it's the chamber pressure times the throat area. Of course this is highly inefficient, because all of the heat from the exhaust is being lost to vacuum and all of the gas molecules are flying in every conceivable direction instead of lending their momentum properly. But it's still thrust. If you imagine chopping the nozzle off of an engine (ignoring that many engines would stop working because they use regenerative cooling in their cycle), this is the pressure thrust (FPressure) you would get from various engines: Engine Chamber Pressure (kPa) Throat Area (m2) FPressure (kN) RL10B-2 4412 0.01297 57.2 RL10A-4-2 4200 0.01280 53.8 RL10C-1 4400 0.01270 55.9 J-2X 9515 0.07931 754.6 J-2 5260 0.10994 578.3 RS-25D 20640 0.05327 1099.4 RS-68A 10260 0.21571 2213.2 Note that because even similar engines have slightly different mass flows, FPressure is not going to line up exactly the same. In a pressure-thrust regime, the only way to increase thrust is to either increase chamber pressure or use a wider nozzle. Using a wider nozzle will of course require you to have higher mass flow to maintain chamber pressure. In order to better utilize some of that heat and pressure, we add a nozzle to our engine: We still have some hot, underexpanded gases escaping around the edges, but now a bunch of our exhaust has cooled and is now pointing downrange, which helps us out tremendously. Now, our total thrust becomes the sum of two components: the original pressure thrust component and a new "momentum thrust" (or "dynamic pressure") component: FTotal = FPressure + FMomentum The difficulty is that the contributing proportions of these two components change nonlinearly with chamber pressure, expansion ratio, combustion temperature, mixture ratio, and more -- certainly beyond the scope of this thread. But this should illustrate why you can't just keep imagining a longer and longer nozzle (and apparently a weightless one at that?) and conclude that arbitrarily high specific impulses are possible. The J-2X already has a nozzle extension. A J-2X with a different nozzle extension would be a different engine. Even if we imagined that the J-2X was not a gas generator and was actually just a really big uprated RL10, getting up to 465.5 seconds of specific impulse would make it prohibitively heavy. Since we keep running into the same issues with Silverbird, I went ahead and did a little tweaking. As I said in my last post, setting up the actual values for SLS with a TLI trajectory gives about 11.2% more than what Block 1B can actually send to TLI: This is primarily from overestimation due to how Silverbird handles altitude-compensating engines like the RS-25. To fix this, just downgrade the efficiency of the RS-25s to 95.95%, setting total core thrust to 8,747 kN and vacuum specific impulse to 434 seconds: This allows Silverbird to better characterize the performance of the RS-25s and yields 42.2 tonnes to TLI, right about where you want it to be. Note that setting this to a 185 x 185 km nominal parking orbit with direct ascent yields 111.1 tonnes to LEO, again right about where NASA says the max is for Block 1B.

Is it realy? The volumes they want to transfer at some point are huge, hundrets of cubic meters. Even with some big fuel connectors this will take some time, requiring lots of delta-v to provide the acceleration. I could only imagine a hybrid solution, ullage to provide e.g. 0.01g so the propelant is settled during the transfer while the actual transfer happens with strong pumps, keeping the time under acceleration to a minimum. To be clear, you would definitely need continuous acceleration for settling regardless of whether you used a pump. The weight of a pump that can move tens of tonnes of liquid quickly is non-negligible, and neither is the weight of a sufficient power source. In contrast, if you're transferring by constant ullage acceleration then you're only losing gas -- gas that would have to be displaced anyway as the destination tank is filled.

Wait, what? I've not been following along... The plan is to use inertia to shift the fuel? Isn't slosh a problem? Well that kind of gets at the root of the issue: there is no straightforward way to transfer propellants in microgravity. Think about it. You've got two tanks, and you've got them connected somehow, and you've got a mix of gases and liquid in one tank but only gas in the other tank, and you need to transfer the liquid. How do you do it without using gravity? Even if you have a way to increase the pressure in one tank and decrease the pressure in the other tank, you're just going to be pushing gas, because gas pressure waves move at the speed of sound while liquids move...well, slowly. Pumps won't work unless you can get the liquids into the inlets to begin with. Prior mechanisms to transfer propellant have been extremely volume- and mass-limited and have typically used stable hypergolics. Russian vehicle prop transfers typically use a feeder tank in a disposable refueling vehicle, where the propellant is held inside a bladder that can be compressed by gas pressure to physically push the propellant into the destination tank: But that's a small-scale solution. You can't very well fit every Starship tanker with inflatable bladders holding all of their propellant -- it would be so heavy that the damn thing would never get to orbit. Propellant settling is one big problem here. That's been tackled in a number of ways; most upper stages use solid ullage motors or cold-gas thrusters for settling, while the Space Shuttle and most capsules used a complex engine feed inlet with a structure that holds fluid inside it by surface tension. Once settled, the subsequent propellant burn provides the acceleration needed to keep the propellant flowing into the inlets. Starship currently uses propulsive vents (basically cold-gas thrusters) for settling before restart, although it may switch to hot-gas thrusters eventually. So if they provide settling with gas thrusters, they'll create an impulse that will get the propellant flowing (or, technically, an impulse that will move the tanks in one direction while the propellant is stationary relative to the original trajectory). At that point they could try to use some sort of pump, but it's a more mass-efficient solution to just keep the gas thrusters thrusting and let the propellant drain "down" relative to the thrust vector, just as it would under the influence of gravity. In particular, the flow from the donor tank in to the recipient tank is going to lower the pressure in the donor tank and increase the pressure in the recipient tank, so the pressure increase in the recipient tank can be used to continuously feed gas thrusters on that tank, keeping the whole stack settled: Of course you'd need COPVs or maybe a resistance heater in the donor tank to keep it at equal or higher pressure than the recipient tank, but that's really no big deal.

This was the original plan for HLS as an accompaniment to SLS, but then Starship came along and obviated that.

From what I’ve gleaned it’s just a transfer between main and header tanks. Little to no change to vehicle, plumbing’s already there (to fill header in the first place), zero additional risk to mission. I believe I had read that they would have a separate actual tank in the payload bay and would test transfer between it and the vehicle tank, rather than transfer in and out of headers. The connections would already be in place and it would simply be a matter of opening the valves and using a small ullage burn to nudge the props from one tank to the other.

What did SN8, SN9, SN10, SN11, and SN15 use for their landing attempts, then? If you review the Starship landing tests, it has happened multiple times a Raptor engine has leaked fuel and caught fire. This was only with one or at most 3 engines. Imagine this with 33 engines. I’m positing the Raptor is no more reliable now than on those landing tests on relights. You said "restarted in flight". The high-altitude hop tests demonstrated in-flight restarts. You may certainly opine that SpaceX cannot reliably restart the Raptor in flight (although all of those were Raptor 1 and there have been no demonstrated refiring/restart issues with Raptor 2), but you cannot deny that those Raptors did in fact restart in flight. And you still have no evidence beyond speculation that Raptor 2 has any startup issues at all, given that (a) the lack of completed Superheavy startups for OFT-1 was likely associated with sensor excursions and no actual engine failures, (b) the subsequent shutoffs on OFT-1 were associated with leaks from a hydraulic TVC system that no longer exists, and (c) IFT-2 had no difficulty keeping all 33 engines firing from ground to staging. I don't know why you're on about the engines. Engines are an area where SpaceX seems to have the most success. There are more issues with propellant slosh, maneuvers, the survivability of tile failures on re-entry, and so forth.

This approach makes no sense. If you think that you can simply arbitrarily increase the specific impulse of any hydrolox engine, then why do it with the J-2X but not with the RL-10s? Also, your numbers were wrong to begin with: the RL10C-3 for the EUS develops 460.1 seconds Isp, not the 448 seconds you gave it. In reality, you've got it upside down. That particular report is featuring an expander cycle engine, which has an intrinsically high specific impulse because it doesn't burn any of the propellant to run the pumps. The J-2X, like the J-2 before it, is a gas generator engine which (as I've told you before) cannot achieve anything near the maximum attainable specific impulse of an expander cycle. Comparing the J-2X to the J-2 will illustrate that pretty cleanly. The J-2X has 3.35x the expansion ratio of the J-2 and 1.57x the chamber pressure of the J-2, but only increased the vacuum specific impulse by a paltry 6.4%. Specific impulse is roughly proportional to the square root of (1 - RP0.26), where RP is the pressure ratio between the exit and the chamber, so if you increase the expansion ratio by from 27.5:1 to 92:1, you're going to get a MAXIMUM of 9.4% increase in vacuum specific impulse...obviously the actual number is slightly lower due to built-in inefficiencies and other factors. If you use a massively oversized nozzle to go ahead and increase the expansion ratio from 92:1 to 625:1, you're going to get a probable increase of ~5.8% which is going to get you no higher than 474 seconds. And that would be for a nozzle diameter 2.61 times greater than the already massive 3-meter nozzle of the J-2X: nearly 8 meters. Adding a nozzle extension onto the J-2 to create the J-2X already increased its mass by 680 kg (actually more, because the other J-2X upgrades decreased the mass of the powerplant and chamber). Approximating the J-2X nozzle extension as a truncated cone with a lower diameter of 3 meters, a height of 1.3 meters, and an upper diameter of 2.1 meters, we get a lateral surface area of 11.02 square meters. Adding an additional nozzle extension to increase the exit diameter to 7.83 meters would require the new extension to be 6.98 meters in length and have a lateral surface area of 125.65 square meters, adding something like 7.98 tonnes. Which brings us to additional problems. All four RL10C-3 engines together have a mass of 920 kg and occupy 2.17 meters of vertical space with the nozzle extension stowed. The J-2X has a mass of 2.47 tonnes and a fixed length of 4.7 meters, while your super-nozzle-extension J-2 (I'll call it the J-2Z) would have a minimum stowed length of 5.84 meters. So you'd need to remove 3.67 meters of length from the EUS, reducing its propellant capacity by something like 24%. Remember that you can't just make the EUS longer because this single-launch architecture of yours ALREADY doesn't have enough vertical space to fit both a capsule and a sortie lander. Your math didn't account for increased mass OR increased length. Finally, there were other errors in what you plugged into Silverbird. For example, the dry mass of an SLS booster is 102 tonnes, not 110 tonnes. The dry mass of the SLS core is 85.3 tonnes but you have to add 7.4 tonnes for the interstage (Block 1's LVSA interstage massed 4.5 tonnes with a lateral surface area of 177 square meters so scale up for Block 1B's longer interstage with its cylindrical surface area of 292 square meters). Similar errors have been corrected throughout. Here are the correct calculations (spoilered for length). First, we create a baseline using the Silverbird calculator with the existing SLS Block 1B to gauge overestimation: It gives just under 120 tonnes, which is significantly more than the 105-110 tonnes claimed by NASA. As I've explained before, Silverbird tends to overestimate both T/W ratio and sea level specific impulse when dealing with altitude-compensating sustainer engines like the RS-25. So we know we need to reduce whatever Silverbird tells us by ~11.4%. Next, let's use the J-2X. Note that I have added the increased mass of the bigger engine but also reduced the length, propellant capacity, and dry mass of the EUS to accommodate the greater length of the engine. With the overestimation reduction, that's 119.6 tonnes to LEO: certainly an improvement, but not enough to get us the kind of performance we'd need for a single-launch architecture. And now finally let's use your imagined J-2Z, adding 7.98 tonnes for the nozzle extension and reducing the length and propellant capacity of the EUS accordingly: And applying the same reduction, that gives us 116.3 tonnes to LEO. So this super-high-expansion-ratio idea is sending us in the wrong direction. That makes intuitive sense, after all: if it made sense to get higher specific impulse by attaching a bigger nozzle extension, that's what rocket engineers would do! But it doesn't, because nozzle extensions are really quite heavy, so there's always a balancing test. You need around 3.15 km/s for TLI. Assuming a 185x185 km parking orbit, your parking orbit velocity is 7.793 km/s from the vis-a-vis equation. Adding 3.1 km/s at perigee raises your "apogee" to 451,517 km (which makes sense given that the moon is a little under 400,000 km away and you need to swing past it for an efficient lunar orbit insertion). So if you plug my baseline SLS figures above into Silverbird just as I did before but setting a desired apogee of 451,000 km, you get an estimated TLI payload of 46.7 tonnes, 11.2% more than the 42 tonnes that NASA claims can be sent to TLI by SLS Block 1B Cargo. (Note: this should give us some confidence in our method because the overestimate remains consistent.) Using the J-2X figures from above, we get a TLI payload of 48 tonnes, which comes to 43.2 tonnes when you apply the overestimate. So it's really not a significant increase once you are looking at TLI deliveries. Could it do an Apollo-style mission? Sure: if you gave it the exact same capsule, service module, and lunar module of Apollo. But so could ordinary SLS Block 1B.

Yeah, but also firing an RVac on a modern launch vehicle was definitely something. I don't believe we've clustered more than two engines of any appreciable size on an upper stage since the Saturn V, and the whole concept of a fixed-mount (that is, non-gimbaling) vacuum engine on an upper stage with separate full-size engines providing attitude control is pretty unique. We also haven't used vacuum-expanded regeneratively-cooled-nozzle engines in a vacuum...like, ever.

Obviated to some degree, yes, but if we are going to contemplate any expendability to Starship+Superheavy then we should just go with an all-up expendable architecture like the Saturn V.

Elon had previously said that Raptor V3 would bump a "better optimized" Starship V1's expendable payload up to 300 tonnes LEO. I'm assuming he's accounting for a vehicle that is specifically being launched without heat shielding and so forth. A while back he had suggested 40 tonnes dry for a minimal-mass expendable BLEO Starship without SL Raptors, so if we keep all six Raptors then we're looking at something like 45 tonnes dry. We also know that the main tanks on Starship V1 carry 1170 tonnes of propellant. Plugging these values into the rocket equation (with an average ~371.75 seconds Isp, assuming an even split between SL and Vac Raptors for the first two-thirds of the burn and only one SL Raptor accompanying the Vacs for the last third of the burn) gives us 5,396 m/s of Δv from staging off an expendable Superheavy to get to LEO. But Starship V2 has an expected tank stretch and bump up to nine engines. The main tanks of Starship are about 23 meters in (vertical) length, so we could expect a notional 7-meter tank stretch to add around 350 tonnes of propellant. Crudely subtracting off 9.6 tonnes for engines from the 45-tonne Starship V1 got us tankage and plumbing mass of ~35 tonnes, so let's bump this up to 42 tonnes to account for the stretch, then add in nine engines to get to around 57 tonnes dry. Staging velocity will be just slightly lower here due to the increased mass for Superheavy to lift, so let's subtract 100 m/s from staging velocity (increasing the corresponding Δv requirement to 5,496 km/s). We bump up the average specific impulse to 374.9 seconds, though (assuming all nine engines firing for the first third of the burn and only one SL raptor accompanying the Vacs for the last two-thirds of the burn). Load that up with payload, and you find that the expendable Starship V2 can presumably put 383 tonnes into LEO. If it can put 376 tonnes into LEO, then it can send ~101 tonnes to TLI monolithic. But we can do better. Let us imagine a "Starship Saturn" with two expendable upper stages that together have the same total expected propellant capacity of Starship V2. All nine engines on the second stage and two Raptor Vacuums on the third stage: Doing it this way allows us to stay within known variables for the stack (since we are only increasing propellant and tank weight by what is already expected for Starship V2), so this shouldn't require any significant changes to Superheavy. It gets taller, of course, but only by the height of the new second interstage. With these constraints, figuring out where to "divide" Starship V2 into two separate stages becomes a simple optimization problem. Maximum payload is reached by placing 461.5 tonnes of propellant onto the third stage, which allows us to send just over 132 tonnes to TLI, and the third stage has a staging T/W ratio of 0.98:1, which is fine since it will be mostly firing in orbit. It will stage just shy of LEO and burn for 101 seconds to circularize, and you'll only need a 126-second burn for TLI. You'll notice that the propellant mass on this third stage is only a bit more than the total amount of propellant that gets added from the tank stretch, so it may be much easier (for this purpose) to just keep the original Starship V1 tank tooling but strengthen and plumb the bottom for nine Raptors instead of six. Won't reduce your TLI delivery by much. Another possibility, depending on what your lunar stack design is like, is to use the terminal stage as the braking stage to get into cislunar orbit. Things get a little trickier here because we assume some boiloff will occur in transit. Let's assume a four-day coast and 0.1% boiloff per day, measured against total original propellant mass on the third stage. Now the total required Δv is much higher, which changes up how you balance the stages: here the optimal propellant load on the third stage is 368 tonnes, and the sorties stack that is delivered all the way to low lunar orbit is a whopping 95.5 tonnes. If you want to squeeze out just a little more from the whole affair, you can even drop down to a single vacuum Raptor, although that will double the length of your burns which could cause either Oberth or cosine losses depending on the trajectory optimization you use. If you can deliver 95.5 tonnes all the way to low lunar orbit in a single launch, then you can deliver Orion AND a 69-tonne lander. That's enough mass budget for pretty much any approach you want to use.

Hmm. Sea level thrust of a single Merlin 1D is 845 kN, and only one row of three engines is restartable. Max thrust for liftoff is therefore 2.54 MN. They could add more TEA-TEB to accomplish more restarts, at a minor weight penalty. Assuming the need for a T/W ratio of at least 1.3, the landed stage could have a liftoff mass no greater than 199 tonnes. Assuming 24 tonnes for an empty booster, that allows 175 tonnes of propellant. However, thrust will be lower since the fuel barge won't be able to effectively subcool the propellants, so let's say it gets 150 tonnes. That's a total Δv of 5.48 km/s, so definitely doable. What would be more difficult would be strengthening the legs to increase their load-bearing capability by a factor of more than seven and redesigning them so they can fold back up on their own at liftoff. Kinda pointless when you have to send the landing ship out and bring it back between launches anyway.

Cascading engine failure due to fuel starvation followed by commanded AFTS.

The phased arrays on Starlink satellites can target 12" x 19" receiving dishes (although obviously the signal is significantly overlapping the receiver). I wonder if it would someday be possible to safely beam power to moving vehicles on the ground. It just occurred to me that if you can place a satellite in a Molniya-style orbit to keep it in the sun's shadow as much as possible, you can also place it in an opposite orbit that keeps it in sunlight as much as possible. If they were able to solve linear losses in phased-array beaming (or at least partially solve them) then the extra distance wouldn't be a problem. I'm reminded of the deep space telescopes that use lasers to create a guide star to adjust for atmospheric diffraction: What if the phased array power transmitter and the corresponding receiver could communicate information about the shape of the power beam in order to use the phased array to adjust for atmospheric diffraction and reduce transmission losses?

A killer app for maximizing the utility of low-cost heavy-lift launch services can come in three flavors: Do it better. There are things that we do on Earth, like silicon wafer manufacturing and power generation, that are constrained by being on Earth's surface (e.g. by gravity or by lack of access to peak solar wattage) and thus could potentially be done better in space. Do it more. There are commodities here on Earth that don't have a large market because they simply cannot be produced at scale. An example that definitely wouldn't lend itself well to going to space is seafood farming: the price of crab (and a number of other seafood products) is high because crab can't really farmed and has to be fished. There may be commodities on Earth that have high value regardless of supply and that we've never thought of producing in space, but we could. Do it first. The killer app may be something we've never thought of before. Not too long ago, the concept of streaming services was simply unimaginable. Nobody was coming up with the idea of offering streaming video on demand because the infrastructure didn't exist. We kind of need to be looking in different places here on Earth to fit these three different niches. I wonder what thought has been given to using specific orbits for specific purposes. Many of the concepts for space-based infrastructure focuses on what to do in an essentially frozen orbit, but orbits don't have to be frozen. What if you placed your infrastructure in an eccentric orbit with a periapsis on the daylit side (with some sort of approach that uses precession to keep it there) in order to maximize the amount of time spent in Earth's shadow? Would those "free" heating and cooling cycles be useful for anything? The phased arrays on Starlink satellites can target 12" x 19" receiving dishes (although obviously the signal is significantly overlapping the receiver). I wonder if it would someday be possible to safely beam power to moving vehicles on the ground. Even if you coated every inch of a Tesla with solar cells, it wouldn't pick up enough power to recharge as it drives. But if you had a large space-based power generation array that was collecting many thousands of square meters of sunlight and beaming it directly to individual vehicles, that might work. The United States sorely lacks interstate mass transit infrastructure, and bus tickets are terribly expensive in part due to the cost of fuel. If you could beam power directly to the roof of an electric bus so that it would never need to stop and recharge batteries, you could slash transit costs. This was literally a completely different Raptor iteration.

I would push back on this. If there was a flow failure or attitude deviation or some other anomaly shortly before SECO, that could easily result in flight path deviation and would trigger AFTS. If there was an overall thrust shortfall then it could have same same effect as OFT-1 on the booster, where it kept doing its best but ended up out of flight path right at the end of the burn. Is there any indication that the burn was taking longer than planned, or that the ship was moving slower than it should have been at that time? LOX leaks in an engine pretty rapidly lead to engine-rich combustion. Doubt that’s the case. Damage to the LOX tank, perhaps from spalling during hot staging, could be a more likely root cause. It wouldn’t even need it to be a leak alone; if you have a small leak, then you will have autogenous pressurization issues and end up consuming more LOX that way too. Yep. Hydraulic shock (the “water hammer” effect) can be difficult to model, especially when there is almost no way of accurately predicting the retrograde acceleration induced on the booster from Starship’s six engines, let alone the actual gradient and rate of change of that acceleration during the hot staging event. I don’t see that ever being possible. Raptor has two turbopumps, one directly in line with the engine thrust factor and one slightly offset. The turbopumps are slightly different sizes as well. Even if you could do a complete redesign and rotate the pumps by 90°, they likely still wouldn’t be on parallel axes, making any flip alignment impossible. Even if you manage to line up the pumps on the same axis, you would have to rotate all of the raptor engines onto the same alignment to make that work, which is a complete plumbing redesign as well.

Ooooooh. It would seem, perhaps, that six Raptors blasting Superheavy at close range produce more “negative thrust” than the prograde thrust of three Raptors running at the other end. To the point that the propellant would continue forward while Superheavy started to slow.

In the video @tater attached above, the Starship+Superheavy stack clears its own height in under eight seconds. Very sporty coming off the pad. I don't see any indication that the engines were being downthrottled. Remember that the engine bay leaks in OFT-1 were associated primarily with hydraulic fluid. Full replacement with ETVC means no hydraulic fluid to leak this time. I'm not sure where that particular dude on Twitter is getting the notion that the booster should have dumped 3.2 km/s at staging. A nominal Falcon 9 mission with boostback, like Crew-7, stages at 1.72 km/s airspeed; that's only slightly higher than Starship+Superheavy which is itself a completely different vehicle with a different launch profile and different boostback characteristics (not to mention that this was a test).