New Glenn 9x4 discussion

[sstli2]

For GEO I definitely agree. For LEO you're likely right but I'm not yet convinced we've proven it. I don't see how to rule out the possibility that the LEO figure is with two stages and GEO is with a third stage that's not optimized.

The main assumption that moves the needle is dry mass. I start at 25 ton GS2 dry mass and with the 2 extra engines and ~8m of extra tank I end up at a 36 ton dry mass. Have to get rid of about 17 tons somewhere there. Maybe, but I'd need some kind of more precise color on GS2 dry mass to try that. Everything I've seen and heard so far point to it being an inherently heavy stage.

But what really intrigues me is this anecdote:

I was surprised because when they told me about 9x4 back at Space Symposium, they were targeting more like 45mT with first stage reuse.

And one fitting explanation for this is that they originally were targeting a quasi-50 ton vehicle with two stages - which jives just fine with the current estimates - and that the introduction of a GS3 is what changed the math. Besides, if you're Blue Origin, and you're marketing high-energy payload capability contingent on a GS3, why wouldn't you make the numbers look nice and do it on LEO too? After all, Falcon Heavy did something similar with their advertised 63 tons to LEO.

[hkultala]

This discussion about capabilities is hard to do in the abstract. You really need to dig down into the details.

I've been working on a spreadsheet, that I first debuted on the New Armstrong thread, to try to put together an estimate of what the delta-V capabilities of New Glenn actually are. Indeed, there are plenty of assumptions baked into this spreadsheet, and changing any of them by even a bit moves the numbers substantially.

I've linked the latest version of my spreadsheet here. I don't want to put too much emphasis in the numbers I came up with, because there are huge error bounds around them. Instead, I'd highlight some general conclusions from this exercise:

- The current 7x2 is dominated by gravity losses, and it's hard to get the TWR or the TWR/gravity loss curve into a place where that wouldn't be the case¹. I think the current performance is probably severely limited by this.
- The 7x2 with engine upgrades reverses these gravity losses and brings the performance closer to nominal.
- No amount of tinkering with assumptions gets the 9x4 performance close to 70 tons. Best I can do is 50 tons. Therefore, I believe that the target capability of 70 tons to LEO is based on performance characteristics (engine thrust improvements, dry mass optimizations) that are beyond what has been publicly shared to date.

Feel free to play around with this spreadsheet or the assumptions, if you think you can do a better job than I did.

¹ Interestingly, underfilling the propellant tanks also eliminates much of these gravity losses, and also brings the performance closer to nominal.

Couple of things noticed:

1) There is clearly something wrong with your mass numbers if you think current 7x2 variant has T/W ratio of only 1.05 on liftoff. Based on the video footage of the launch it was somewhere close 1.1 or slightly over 1.1, but much more than 1.05.

Maybe your stage 1 empty mass is too high?

2)
Your specific impulse for the updated BE-3U is too low.
Current BE-3U is already supposed to be 445s instead of 440s and I don't expect to see the upgraded variant to have smaller isp

with 445s for the updated BE-3U, LEO payload of 9x4 rises by 4 tonnes.

3) Assuming that uprated BE-4 would have smaller isp than current version is a bad assumption. Typically, when the pressure increases, also the isp increases slightly.

So, keep the updated BE-4 isp the same as the original and LEO payload of 9x4 rises by another 4 tonnes, now it is at 66 tonnes.

4) Your propellant mass for second stage of 9x4 is clearly not enough.

Increase the length of second stage to 36 meters and LEO payload goes up up 7 tonnes.
Make it 40 meters and LEO payload goes up by another 3 tonnes.

Now we are at 76 tonnes LEO for the 2-stage 9x4 and 48 tonnes for the uprated 7x2.

[sstli2]

Couple of things noticed:

1) There is clearly something wrong with your mass numbers if you think current 7x2 variant has T/W ratio of only 1.05 on liftoff. Based on the video footage of the launch it was somewhere close 1.1 or slightly over 1.1, but much more than 1.05.

Maybe your stage 1 empty mass is too high?

2)
Your specific impulse for the updates BE-3U is too low.
Current BE-3U is already supposed to be 445s instead of 440s and I don't expect to see the upgraded variant to have smaller isp

with 445s for the updated BE-3U, LEO payload of 9x4 rises by 4 tonnes.

3) Assuming that uprated BE-4 would have smaller isp than current version is a bad assumption. Typically, when the pressure increases, also the isp increases slightly.

So, keep the updated BE-4 isp the same as the original and LEO payload of 9x4 rises by another 4 tonnes, now it is at 66 tonnes.

4) Your propellant mass for second stage of 9x4 is clearly not enough.

Increase the length of second stage to 36 meters and LEO payload goes up up 7 tonnes.
Make it 40 meters and LEO payload goes up by another 3 tonnes.

Now we are at 76 tonnes LEO for the 2-stage 9x4 and 48 tonnes for the uprated 7x2.

Thanks for all the feedback. I'm trying to improve this spreadsheet as more info comes out.

1 - I agree it's a questionable number, but there aren't many ways to fix this. One way would be reduce dry mass as you said, but how would you anchor this number to something objective? It's definitely a heavy stage, but how heavy? I went back and forth on this, compared to some other stages, and concluded that it's probably quite heavy. The strakes, fins, BE4s, high number of landing legs, HTP thrusters, structure, etc. Happy to change this from 125 tons but I don't have a great sense of what it should be instead and why another value would be better.

Another consideration is that there's 5.6 seconds of burn before the hold downs release. By the time NG clears the tower it's burned off quite a bit of propellant and that very low TWR probably creeps closer to 1.10 by then. Napkin math is almost 50 tons of propellant or almost +.04 TWR by then.

The possibility that I like better than an overestimated dry mass is underfilling the tanks, which is why I added an input for GS1 propellant level percent. If you underfill the tanks you do improve TWR/grav loss and payload a lot. I wanted to avoid this because there's not much evidence for it, but of the options to get to a 1.10 TWR, this seems like the best bet.

2/3 -  I think that's fair. I wanted to haircut the specific impulse a bit because I figured that what's being marketed is under optimal conditions at the test stand, but perhaps that's too conservative.

4 - I think I'm going to have to do some pixel counting on the renders, as I would like to anchor the estimate of GS2 tank stretch to something real. But I agree it's worth a look.

Edit: Here are the changes I just made:
- Updated the specific impulse to the nominal level as stated above.
- Set the minimum TWR to 1.10 and set GS1 propellant level to 95% to meet that threshold.
- Did a pixel analysis of the 7x2 / 9x4 render:
  - I have the GS1 being 311 pixels tall previously, and 343 after, or 10% taller.
  - I have the GS2 being 74 pixels tall previously, and 136 after, or 83% taller.
- Reduced GS1 tank volumes slightly to match Dave Limp's tweet.
- Added notes for citing volume/density inputs and rationalizations for other assumptions like dry mass.

After applying these, indeed the 9x4 looks a lot better. As was previously pointed out, the high-energy numbers are still too far from feasibility, but the LEO numbers are more plausible.

[deltaV]

If you underfill the tanks you do improve TWR/grav loss and payload a lot.

Ignoring practical issues like damage to the pad from a slow liftoff, as long as the initial thrust-to-weight ratio is at least 1 I don’t think underfilling the tanks can ever improve performance. To see why, consider the following thought experiment. Launch a rocket fully fueled, let it run for a minute or so until its thrust-to-weight ratio is better, and then magically teleport it back to the launch pad with zero velocity (but its propellant loading is unchanged) and launch it a second time. This magical teleportation step is clearly counter-productive since it removes valuable velocity and altitude. But the rocket after the magical teleportation is in exactly the same state as a rocket that was launched underfilled with no magic and hence will perform as well. So underfilling doesn’t help performance (assuming T/W > 1).

The fact that your spreadsheet predicts that underfilling the tanks substantially improves performance suggests that your gravity loss model could be improved. I unfortunately don’t know the right fix. There’s a loss model at https://silverbirdastronautics.com/LaunchMethodology.pdf but I have no idea how good it is. The nuclear option would be writing a simple trajectory simulation and optimization tool.

[meekGee]

If you underfill the tanks you do improve TWR/grav loss and payload a lot.

Ignoring practical issues like damage to the pad from a slow liftoff, as long as the initial thrust-to-weight ratio is at least 1 I don’t think underfilling the tanks can ever improve performance. To see why, consider the following thought experiment. Launch a rocket fully fueled, let it run for a minute or so until its thrust-to-weight ratio is better, and then magically teleport it back to the launch pad with zero velocity (but its propellant loading is unchanged) and launch it a second time. This magical teleportation step is clearly counter-productive since it removes valuable velocity and altitude. But the rocket after the magical teleportation is in exactly the same state as a rocket that was launched underfilled with no magic and hence will perform as well. So underfilling doesn’t help performance (assuming T/W > 1).

The fact that your spreadsheet predicts that underfilling the tanks substantially improves performance suggests that your gravity loss model could be improved. I unfortunately don’t know the right fix. There’s a loss model at https://silverbirdastronautics.com/LaunchMethodology.pdf but I have no idea how good it is. The nuclear option would be writing a simple trajectory simulation and optimization tool.

Yup! With the exact caveat that you specified, but noting that that's what bit them on the first launch.

[sstli2]

If you underfill the tanks you do improve TWR/grav loss and payload a lot.

Ignoring practical issues like damage to the pad from a slow liftoff, as long as the initial thrust-to-weight ratio is at least 1 I don’t think underfilling the tanks can ever improve performance. To see why, consider the following thought experiment. Launch a rocket fully fueled, let it run for a minute or so until its thrust-to-weight ratio is better, and then magically teleport it back to the launch pad with zero velocity (but its propellant loading is unchanged) and launch it a second time. This magical teleportation step is clearly counter-productive since it removes valuable velocity and altitude. But the rocket after the magical teleportation is in exactly the same state as a rocket that was launched underfilled with no magic and hence will perform as well. So underfilling doesn’t help performance (assuming T/W > 1).

The fact that your spreadsheet predicts that underfilling the tanks substantially improves performance suggests that your gravity loss model could be improved. I unfortunately don’t know the right fix. There’s a loss model at https://silverbirdastronautics.com/LaunchMethodology.pdf but I have no idea how good it is. The nuclear option would be writing a simple trajectory simulation and optimization tool.

Good point, and it's a byproduct of my non-linear gravity loss estimate that grows more quickly as TWR falls than the corresponding improvement in delta-V due to better PMF. At the same time, I want to make sure that the model embeds a non-trivial credit to performance for improvements in thrust. The model you linked looks simple enough; I'll explore using it. A full-blown trajectory analysis would be a bit beyond the scope of a spreadsheet.

Edit: I decided to remove gravity losses from my model entirely. It's very hard to come up with something that's defensible and well-grounded, and likewise hard to come up with a way to quantify what the impact of future improvements in thrust will be on that value. Instead, I'll anchor the required delta-V to particular orbits as inclusive of gravity losses, setting LEO to a conservative 9,700 m/s.

Edit 2: I re-introduced delta-V losses into the model. I first derived burn times for the other variants based on the NG-2 burn times (notes on how in the spreadsheet) and used the above delta-V penalty estimation to calculate losses.

[deltaV]

The model you linked looks simple enough; I'll explore using it.

Cool. See also https://silverbirdastronautics.com/LVperform.html for a tool using that model.

A full-blown trajectory analysis would be a bit beyond the scope of a spreadsheet.

Agreed, a spreadsheet is not the right tool for non-trivial programming tasks like trajectory optimization. We'd need either a mainstream programming language like python or C++ or something like MATLAB.

[seb21051]

The model you linked looks simple enough; I'll explore using it.

Cool. See also https://silverbirdastronautics.com/LVperform.html for a tool using that model.

A full-blown trajectory analysis would be a bit beyond the scope of a spreadsheet.

Agreed, a spreadsheet is not the right tool for non-trivial programming tasks like trajectory optimization. We'd need either a mainstream programming language like python or C++ or something like MATLAB.

Might be interesting to try your hand(s) at one of the AI Coder programs. You just tell it what you want in the language of your choice. I have tried a few simple ones and been impressed with the results.

[deltaV]

According to https://www.blueorigin.com/news/new-glenn-upgraded-engines-subcooled-components-drive-enhanced-performance 9x4’s first stage thrust is 640*9/(550*7)-1=50% more than old 7x2’s. For the second stage the increase is 400*4/(320*2)-1=150%. So why are they increasing the second stage thrust so much, i.e. why not have 2-3 engines instead of 4? I think the answer is with a third stage (at least for high energy orbits) the second stage will always be pushing a relatively heavy mass so its burnout mass matters less so a larger second stage optimizes better.

[Brigantine]

why are they increasing the second stage thrust so much

2nd stage is where all the ISP advantage is - and the faster your 1st stage goes the more prop you need to reserve for the re-entry burn (by another name) or possibly a boostback burn.

A high TWR is important for a hydrogen 2nd stage to LEO, and yes the CLT/GS3 has removed (de-emphasized?) high energy destinations from GS2's requirements so makes sense that it becomes a more LEO-focussed design.

I wish we had numbers about how much the GS2 tank stretch is

[sstli2]

I wish we had numbers about how much the GS2 tank stretch is

If the render is to scale, and you count pixels, the stretch is roughly 80%, or going from 23.4m to ~42m (as tall a Falcon 9 booster).

[sstli2]

According to https://www.blueorigin.com/news/new-glenn-upgraded-engines-subcooled-components-drive-enhanced-performance 9x4’s first stage thrust is 640*9/(550*7)-1=50% more than old 7x2’s. For the second stage the increase is 400*4/(320*2)-1=150%. So why are they increasing the second stage thrust so much, i.e. why not have 2-3 engines instead of 4? I think the answer is with a third stage (at least for high energy orbits) the second stage will always be pushing a relatively heavy mass so its burnout mass matters less so a larger second stage optimizes better.

Since your last comment, I implemented the gravity loss / penalty delta-V model you had linked to, which is based on an actual ascent time and an adjustment to account for the fact the original Townsend model assumes 3 stages each with equal thrust and specific impulse. This also required an estimate of burn time on the other variants (have some notes on this derivation in the spreadsheet, but it's based on ratios of thrust / specific impulse / dry mass / wet mass).

Edit: See the latest spreadsheet, I had to make some fixes as my I had an error in my GS2 dry mass formula. Also reduced BE-4 sea level ISP to 310.

Excel version is located here and read-only web version is located here.

[ZachF]

2)
Your specific impulse for the updated BE-3U is too low.
Current BE-3U is already supposed to be 445s instead of 440s and I don't expect to see the upgraded variant to have smaller isp

with 445s for the updated BE-3U, LEO payload of 9x4 rises by 4 tonnes.

With an open expander cycle, increasing thrust probably will lower ISP some, as increasing thrust requires increasing turbine power, which requires increased flow to the turbine, which will decrease net isp as a larger percentage of the remass will exit the turbine instead of the nozzle.

[deltaV]

Since your last comment, I implemented the gravity loss / penalty delta-V model you had linked to, which is based on an actual ascent time and an adjustment to account for the fact the original Townsend model assumes 3 stages each with equal thrust and specific impulse. This also required an estimate of burn time on the other variants (have some notes on this derivation in the spreadsheet, but it's based on ratios of thrust / specific impulse / dry mass / wet mass).

...

Excel version is located here and read-only web version is located here.

Thanks! Two issues with your latest spreadsheet (after importing into my own Google sheets) from a quick (incomplete) look:
1. The formulas for GS1 delta vee for all 4 vehicles seem to be missing several masses inside the log. Payload mass, payload adapter, payload fairing at least are missing from both numerator and denominator, and GS2 initial mass is missing from the numerator.
2. There's apparently a circular reference involving "GS1 Sea-Level Ascent Duration". Switching the settings to resolve the circular dependency iteratively doesn't help - it results in divide by 0.

[sstli2]

Since your last comment, I implemented the gravity loss / penalty delta-V model you had linked to, which is based on an actual ascent time and an adjustment to account for the fact the original Townsend model assumes 3 stages each with equal thrust and specific impulse. This also required an estimate of burn time on the other variants (have some notes on this derivation in the spreadsheet, but it's based on ratios of thrust / specific impulse / dry mass / wet mass).

...

Excel version is located here and read-only web version is located here.

Thanks! Two issues with your latest spreadsheet (after importing into my own Google sheets) from a quick (incomplete) look:
1. The formulas for GS1 delta vee for all 4 vehicles seem to be missing several masses inside the log. Payload mass, payload adapter, payload fairing at least are missing from both numerator and denominator, and GS2 initial mass is missing from the numerator.
2. There's apparently a circular reference involving "GS1 Sea-Level Ascent Duration". Switching the settings to resolve the circular dependency iteratively doesn't help - it results in divide by 0.

1) GS1 Initial Mass and GS2 Initial Mass are poorly named, these mean wet mass of GS1 plus everything above it, and wet mass of GS2 plus everything above it. I used to have these named "Lift-Off Mass" and "Staging Mass", perhaps I should have stuck with that so it's less misleading. Given this, I believe the delta-V formula should be correct as is.

2) The circular reference is intentional, as re-entry and landing propellant are calculated off of re-entry and landing burn times, and the burn times are calculated off of the GS1 mass prior to re-entry and landing, which references re-entry and landing propellant. I've seen the issue you are describing where the solver gets stuck on a divide-by-zero before, but I thought it was gone. It usually can be solved by putting a dummy term (very small decimal) into the denominator, which I will look into doing. Excel behavior and Google Sheets behavior sometimes differs here.

P.S. Feel free to PM for anything else specific, as I don't want to crowd this thread too much with the quirks of my model.

[deltaV]

Cross post from New Glenn Thread 2:

Job posting from yesterday: Technical Designer III - GS3

A third stage is clearly more than just an idea.

The new GS3 may be available for 7x2 too so it's probably best to continue discussion of GS3 in that thread. I'm mentioning it here just to inform people who read about third stage speculations in this thread.