How cheap could space hardware potentially be made?

[InterestedEngineer]

Because the further you get from your source of supply and support, the more important it is to have hardware that is fault tolerant beyond consumer level here on Earth.

That doesn't follow from cheap launch costs.

You could spend 100x on the project making hardware that is "fault tolerant beyond consumer level here on Earth".

That's what we do today.  project mgmt maxim:  "scope, schedule, resources. Pick any 2, preferably 1"

Or you could launch 10-20 times the number of probes by making them in bulk from cheaper and heavier hardware.

The reason we don't do that today is launch costs would soar by 10-20x, and the launch costs are already over 25% of any deep space budget, so it'd blow up the budget by 300 percent or more.

But if launch costs DROP by 30x, then it make sense to launch 30 times more probes, or 15 times as many probes that weigh twice as much. Half of them fail, so what?

[Robotbeat]

And the reliability would increase with the number you make. Starlinks were falling out of the sky left and right at first and now are probably MORE reliable than the average satellite.

[Vultur]

Because the further you get from your source of supply and support, the more important it is to have hardware that is fault tolerant beyond consumer level here on Earth.

That doesn't follow from cheap launch costs.

You could spend 100x on the project making hardware that is "fault tolerant beyond consumer level here on Earth".

That's what we do today.  project mgmt maxim:  "scope, schedule, resources. Pick any 2, preferably 1"

Or you could launch 10-20 times the number of probes by making them in bulk from cheaper and heavier hardware.

The reason we don't do that today is launch costs would soar by 10-20x, and the launch costs are already over 25% of any deep space budget, so it'd blow up the budget by 300 percent or more.

But if launch costs DROP by 30x, then it make sense to launch 30 times more probes, or 15 times as many probes that weigh twice as much. Half of them fail, so what?

Yeah.

Custom flagship type stuff is always going to be really expensive. But how much can we do without using that style of design/development/manufacturing?

Most of the outer system probably isn't available with current tech, since you can't do RTGs in Starlink style mass production. Solar works at Jupiter, but the radiation belts are a big limit: you could study Jupiter and Callisto but not much more.

Some stuff at Saturn might maybe be doable ... I could see a swarm of probes that were mostly solar panel, perhaps.

One thing I'd love to see, but would probably never get funded by any space agency: an "asteroid explorer swarm". Hundreds or thousands of mass produced identical, small, cheap probes with just a couple basic instruments, sent out to lots and lots of different asteroids (send say 5 to each of the really high interest asteroids, assume a high failure rate). Get a really broad sample of different types of asteroids, look at ones that are potentially interesting from a resource perspective as well as a scientific one, etc.

[DanClemmensen]

And the reliability would increase with the number you make. Starlinks were falling out of the sky left and right at first and now are probably MORE reliable than the average satellite.

Starlinks have an operational lifetime of five years, and are supposed to be deliberately de-orbited at the end of that time. At least that's what all of the initial descriptions said. Do we have any reason to think that these satellites are not being deliberately de-orbited? If this is indeed what is happening, we should expect that in steady state a constellation will be de-orbiting satellites on average at the same rate that they are being launched. A constellation with 40,000 satellites will be launching 8000/yr or about 22 satellites a day, and de-orbiting the same number.

[Coastal Ron]

And the reliability would increase with the number you make. Starlinks were falling out of the sky left and right at first and now are probably MORE reliable than the average satellite.

Well, maybe for the period time they operate, but they are only designed to operate for, what, 5 years? GEO satellites are more like 15-20 years, and can last far longer.

And do we really know what the reliability is of orbiting Starlinks? If one fails they don't immediately fall out of orbit, so just because a Starlink is in orbit doesn't mean that it is operating, or operating as designed.

Do we have any hard numbers regarding how many of the current orbiting Starlink satellites are operational, and how many are dead?

[Vultur]

If this is indeed what is happening, we should expect that in steady state a constellation will be de-orbiting satellites on average at the same rate that they are being launched. A constellation with 40,000 satellites will be launching 8000/yr or about 22 satellites a day, and de-orbiting the same number.

Absolutely... But Starlink is nowhere near steady state yet.

Well, maybe for the period time they operate, but they are only designed to operate for, what, 5 years? GEO satellites are more like 15-20 years, and can last far longer.

Sure, but so what? As long as the 5 year satellite is at least 4x cheaper than the 20 year one, it still ends up cheaper per satellite life year. And Starlinks are cheaper than that.

[DanClemmensen]

If this is indeed what is happening, we should expect that in steady state a constellation will be de-orbiting satellites on average at the same rate that they are being launched. A constellation with 40,000 satellites will be launching 8000/yr or about 22 satellites a day, and de-orbiting the same number.

Absolutely... But Starlink is nowhere near steady state yet.

No, but Starlink has been active since 2019. We would expect that satellites launched in 2019 began deorbiting last year and satellites launched in October 2020 will be deorbiting this month. They launched 180 satellites in October 2020, so we should expect (on rough average) they they would deorbit 180 satellites this month. That's roughly six a day. Yes, all the details are squishy, but the underlying reality is still there: the number to be deorbited roughly equals the number that were launched five years ago.

It appears that Chicken Little just woke up to this fact, but it's implicit in all of the information we have had from SpaceX for the last six or more years.

[Robotbeat]

And the reliability would increase with the number you make. Starlinks were falling out of the sky left and right at first and now are probably MORE reliable than the average satellite.

Well, maybe for the period time they operate, but they are only designed to operate for, what, 5 years? GEO satellites are more like 15-20 years, and can last far longer.

And do we really know what the reliability is of orbiting Starlinks? If one fails they don't immediately fall out of orbit, so just because a Starlink is in orbit doesn't mean that it is operating, or operating as designed.

Do we have any hard numbers regarding how many of the current orbiting Starlink satellites are operational, and how many are dead?

We actually do. Jonathan McDowell keeps track of this.

https://planet4589.org/space/con/star/stats.html

Look at Geosats like those used by Viasat, and the reliability doesn’t look great. Plenty of infant mortality, regardless of design intent for 10-15 year lifetime.

[thespacecow]

I have not yet listened to it, but the current episode Aviation Week's Check 6 podcast is entitled "Why do satellites still cost so much?".

I listened to this a few days ago, one thing they mentioned is that each layer of subcontractor will add a markup, and this accumulates to very significant amount after a few layers. The alternative is of course vertical integration, but that would require a large upfront investment. Maybe one stop shop like Redwire can reduce this somewhat, but maybe one just has to accept non-vertically integrated satellite remains relatively expensive.

Another thing they mentioned is that unlike the components contracted out, Starlink doesn't test every components produced by themselves, they have a large batch and only test a few of them to save cost. They accept the risk of some components failing due to not being tested, but subcontractor won't be able to take this risk.

[DanClemmensen]

Another thing they mentioned is that unlike the components contracted out, Starlink doesn't test every components produced by themselves, they have a large batch and only test a few of them to save cost. They accept the risk of some components failing due to not being tested, but subcontractor won't be able to take this risk.

This is a consequence of vertical integration at the system level, but it only works for constellations, where the cost of losing a small percentage of the satellites is relatively small. A typical one-off satellite needs to be much more reliable.

[Coastal Ron]

I have not yet listened to it, but the current episode Aviation Week's Check 6 podcast is entitled "Why do satellites still cost so much?".

I listened to this a few days ago, one thing they mentioned is that each layer of subcontractor will add a markup, and this accumulates to very significant amount after a few layers. The alternative is of course vertical integration, but that would require a large upfront investment.

Not necessarily so, but the real key is whether there is enough demand for what you are bringing in-house. Does you no good to hire a bunch of people if you are only going to use them 3 months out of the year. THAT is why companies use subcontractors, the lack of overall demand in-house for doing it themselves.

Another thing they mentioned is that unlike the components contracted out, Starlink doesn't test every components produced by themselves, they have a large batch and only test a few of them to save cost. They accept the risk of some components failing due to not being tested, but subcontractor won't be able to take this risk.

This is related to the earlier point that Robotbeat made about the auto industry, and it is related. However with cars if there is a failure you can have your local service center fix the problem, but in space you just write off that capability or that particular satellite.

The other thing too is that in order for this to truly pay off - the risk/reward calculation - you have your QA and engineering teams focus on your suppliers so that THEY build a quality product that doesn't need to be 100% tested or inspected when received at SpaceX. This is actually what the auto industry does, so that they can bring in car components and not have to inspect 100%. Plenty of other industries do this too, so not unique.

This topic is really about volume though. If you have enough volume, then you can implement initiatives that can significantly lower the per unit price. Prior to Starlink there were satellite busses that were in serial production, but certainly not what anyone would call "high volume" production. But then again those legacy satellites usually went to MEO or GEO and were required to have long lives, whereas Starlink has a service life of about 5 years, and there is plenty of service overlap in case of failures.

[Proponent]

I have not yet listened to it, but the current episode Aviation Week's Check 6 podcast is entitled "Why do satellites still cost so much?".

Now I've listened to the podcast. Two key factors raising the costs of satellites are identified. The first is a lack a vertical integration, which causes profit margins to accumulate exponentially. If a fully vertically-integrated supplier makes a 10% margin, then the customer pays 1,10 the actual cost of production. If, on the other hand, a chain of 10 suppliers each making 10% is involved, then the final customer pays 1.10¹⁰ = 2.60 times the cost. The problem with vertical integration is the huge capital outlay it requires, but it does appear to be a big part of SpaceX's success.

The other point made was the cumulative cost of repeated tests, where a component is vibration-tested and thermally cycled by its manufacturer before being sold to the builder of subsystem, who vibes and cycles again, before delivering to a system builder, etc. Not only do costs accumulate, but the repeated testing is potentially a hazard to reliability.

[Coastal Ron]

I have not yet listened to it, but the current episode Aviation Week's Check 6 podcast is entitled "Why do satellites still cost so much?".

Now I've listened to the podcast. ...
The problem with vertical integration is the huge capital outlay it requires, but it does appear to be a big part of SpaceX's success.

I've mentioned this before in that vertical integration only works if you have full time need for the things you are going to make. Or at least full time employment for the team that you brought on to build the components for vertical integration.

If you were building 10 units per year before vertical integration, and 10 units per year afterward, then it probably didn't make sense to do that because of the vast amount of skills and equipment you had to bring onboard to do vertical integration.

You NEED additional demand to support the cost and overhead of vertical integration.

The other point made was the cumulative cost of repeated tests, where a component is vibration-tested and thermally cycled by its manufacturer before being sold to the builder of subsystem, who vibes and cycles again, before delivering to a system builder, etc. Not only do costs accumulate, but the repeated testing is potentially a hazard to reliability.

I'm a little leery about this claim, since doing component testing once you have already integrated the components into their final assembly (like a circuit board) risks having to perform rework of the board to R&R the failed component. Plus, the test environment of the component may be harsher than the test environment of the assembly.

As to vibration, won't this be mitigated by moving away from launchers like Atlas V and Vulcan, which use solid rocket motor boosters?

For instance, I would think the ride on a Starship to space would be far more benign from a vibration standpoint than on a Vulcan. Any data on this?

[InterestedEngineer]

Hear from a CEO of as company who is working to make stuff cheap.

Lots of interesting things to say why things are not currently cheap.

https://x.com/ti_morse/status/1980736116727972120

[Proponent]

I have not yet listened to it, but the current episode Aviation Week's Check 6 podcast is entitled "Why do satellites still cost so much?".

Now I've listened to the podcast. ...
The problem with vertical integration is the huge capital outlay it requires, but it does appear to be a big part of SpaceX's success.

I've mentioned this before in that vertical integration only works if you have full time need for the things you are going to make. Or at least full time employment for the team that you brought on to build the components for vertical integration.

If you were building 10 units per year before vertical integration, and 10 units per year afterward, then it probably didn't make sense to do that because of the vast amount of skills and equipment you had to bring onboard to do vertical integration.

You NEED additional demand to support the cost and overhead of vertical integration.

I think we implicitly agree -- it won't be worth the capital investment unless you have the scale to make efficient use of it (and can raise the capital in the first place).

The other point made was the cumulative cost of repeated tests, where a component is vibration-tested and thermally cycled by its manufacturer before being sold to the builder of subsystem, who vibes and cycles again, before delivering to a system builder, etc. Not only do costs accumulate, but the repeated testing is potentially a hazard to reliability.

I'm a little leery about this claim, since doing component testing once you have already integrated the components into their final assembly (like a circuit board) risks having to perform rework of the board to R&R the failed component. Plus, the test environment of the component may be harsher than the test environment of the assembly.

I'm sure you know more about it than I, but I would think cost and reliability would be optimized by finding a happy optimal testing rate short of testing every component. Balance the high cost of failed systems tests against the cost of redundant tests of components and the cost of operational failures due to excessive cycling.

As to vibration, won't this be mitigated by moving away from launchers like Atlas V and Vulcan, which use solid rocket motor boosters?

For instance, I would think the ride on a Starship to space would be far more benign from a vibration standpoint than on a Vulcan. Any data on this?

All-liquid launch vehicles tend to vibrate less, but every payload user's guide exhibits the vibration environment in detail. A couple of the recent Starship failures were due in significant part to vibration.

[Coastal Ron]

The other point made was the cumulative cost of repeated tests, where a component is vibration-tested and thermally cycled by its manufacturer before being sold to the builder of subsystem, who vibes and cycles again, before delivering to a system builder, etc. Not only do costs accumulate, but the repeated testing is potentially a hazard to reliability.

I'm a little leery about this claim, since doing component testing once you have already integrated the components into their final assembly (like a circuit board) risks having to perform rework of the board to R&R the failed component. Plus, the test environment of the component may be harsher than the test environment of the assembly.

I'm sure you know more about it than I, but I would think cost and reliability would be optimized by finding a happy optimal testing rate short of testing every component.

Quite a while ago, when I was still early in my career as a Operations Program Manager, to speed up getting product through the various tests that military electronics of the time had to go through, I would drive production units to the local vibration test facility (usually in town, but sometimes 2 hours away). Sometimes I had time to chat up the test guys and watch the vibration tests. They were used to listening for loose stuff and stopping tests before too much damage happened. 

But that was a military hardware that had to survive in battle conditions, and I would imagine other than launch that space electronics don't have to worry about vibration, just temperature and radiation. This is a good example too of where we didn't have the volume of product to merit doing such testing in-house.

To your point, I think yes, that you are looking for happy mediums, and...

Balance the high cost of failed systems tests against the cost of redundant tests of components and the cost of operational failures due to excessive cycling.

There are two good examples that SpaceX has provided us about using commodity electronic components:

1. Dragon spacecraft - they use commodity processors to run the Dragon spacecraft, and they handle radiation related concerns with redundancy. Great article about this from 2012.

2. Plan for loss of vehicle with redundancy - which for satellites is what Starlink does (amongst other things too). If a Starlink satellite fails, for whatever reason, the constellation is able to continue to operate at a degraded level that still meets their customer service goals.

Building things in higher volume is certainly one way to lower costs, so commoditizing space hardware can help with that.

I guess one way to look at what SpaceX has done with Starlink is that they have redesigned the "system" of how electronic signals are bounced around the world. Instead of sending signals from Earth out to GEO, they only send them to LEO, where transportation costs are far less, the operating environment is more moderate, and the customer gets the advantage of quicker response rates. So the combination of more moderate operating environment and lower launch costs allows for a higher tolerance of failure from less optimized components.

[spacenut]

The new Starlinks are larger and probably can last longer than 5 years with more propellant, solar panels and thrusters.  I just bought a Starlink mobile.  Easy to set up and speeds over 200-250 downloading.  Streams movies without a hickup as well as high def TV.  Starlink has increased it's speed by 50% since this past January with more satellites in orbit.  Even more are coming.  Well worth it for traveling and in remote areas where no cable exists.  Very small notebook size and very light weight. 

Now, with mass production, 3D printing, and more robotics yes any space hardware would be cheaper. 

[DanClemmensen]

The new Starlinks are larger and probably can last longer than 5 years with more propellant, solar panels and thrusters.  I just bought a Starlink mobile.  Easy to set up and speeds over 200-250 downloading.  Streams movies without a hickup as well as high def TV.  Starlink has increased it's speed by 50% since this past January with more satellites in orbit.  Even more are coming.  Well worth it for traveling and in remote areas where no cable exists.  Very small notebook size and very light weight. 

Now, with mass production, 3D printing, and more robotics yes any space hardware would be cheaper.

If satellite technology continues to advance, satellites will continue to become technologically obsolete after a five-year lifetime. My personal guesstimate is that this will continue for at least another ten years (two satellite generations) and maybe more. If the cost of launch continues to drop, it makes sense to replace old satellites with much higher-bandwidth satellites on a regular schedule. It becomes a continuous process instead of a periodic cycle of launch campaigns.

[DanClemmensen]

Now, with mass production, 3D printing, and more robotics yes any space hardware would be cheaper.

In the 1990's, various visionaries (or nutcases, or whatever) such as Ray Kurzweil and Eric Drexler thought that three synergistic advancements would lead to a Singularity. They were fusion, superintelligence, and nanotechnology. We are still waiting. Of the three, nanotech is no longer getting much attention, but it may get a new kick from AGI, which is on the "superintelligence" leg of the trilogy. Among other things, full-up nanotech is the ultimate 3D printer and it drives the cost of hardware down to the cost of dirt plus the cost of the energy it needs.

Incidentally, these ideas finally came to the attention of Bill Joy in 2000 (Why the Future Doesn't Need Us), which may in turn have caught the attention of Elon and driven his Mars Sanctuary efforts.

[Vultur]

I am skeptical whether the really optimistic versions of nanotechnology are really physically possible. When you get down to the real nanometer scale, you're working with chemical systems rather than 'mechanical' systems. Viruses are tens of nanometers, and they can't really fit in enough stuff to be independently self-replicating. Ribosomes by themselves are on the same scale (~20 - 30 nm).

So you'd probably have the same limits as biochemistry - needing a liquid medium and therefore a relatively limited temperature range (between freezing and boiling of your liquid medium), relatively limited energy use (to avoid damaging the molecules), etc. - but maybe with extremely advanced SF tech you could use a supercritical fluid medium and relax the temperature limits somewhat?

Anyway, even if there are ways around those limits, I was more thinking of using current-tech mass manufacturing practices to make space hardware much cheaper, enabled by cheaper launch costs. If you assume near-Singularity nanotech manufacturing that makes anything from dirt+ energy*, the whole nature of the world's economy is completely upended anyway. So I'd rather stick to more or less current tech.

*Another reason I'm skeptical of that is that random dirt may not have the right elements to make many high tech components.