A new 650 satellite constelation

[Shpaget]

Of course there is no satellite production line, because there has never been a mass production of one type of satellite. However, modern assembly lines are all robots and CNC machines. They don't care nor know what they are making, its just another mill pass here another part placed there.

I won't pretend to know everything that goes into a communications satellite, but I imagine there are plenty of components that can be taken "off the shelf" or for which some standard manufacturing process already exists.

[Nibb31]

Of course there is no satellite production line, because there has never been a mass production of one type of satellite. However, modern assembly lines are all robots and CNC machines. They don't care nor know what they are making, its just another mill pass here another part placed there.

That's not how satellites (or even mass-produced aircraft) are built. Individual parts can be produced by machines, but assembly, wiring, and testing is a very involved manual process. In addition, spacecraft are built in a clean room, which is expensive.

Making a production run for 900 satellites is going to require a paradigm change for Airbus (or any other aerospace company). It also means that standards are going to be lower than what is in practice in the industry. This means lots of engineers working on new processes and standards, finding and certifying new suppliers, building and laying out production lines, etc... It also means that the sats have to be designed for cheap assembly, which has never been done before.

Basically, this is like creating a new industry from scratch. There is no way it's going to cost just a few million dollars.

I won't pretend to know everything that goes into a communications satellite, but I imagine there are plenty of components that can be taken "off the shelf" or for which some standard manufacturing process already exists.

Unfortunately, no. Typical production runs in the satellite industry range from 1 to 10 or 20 units. For typical comsats, Boeing or Airbus offer standard satellite busses that carry the propulsion, power, and avionics, and the customer adds antennas and transponders as required. However, those are still small volumes of large and expensive satellites (several tons and several $10 million).

As an example, for small runs of metal parts, it makes sense to mill them (or even 3D print in some cases). When it comes to larger production runs, casting those parts becomes more economical (you make a mold, which is expensive, but the unit cost is much lower than CNC milling). The problem is that your usual suppliers might be tooled for milling. Buying larger batches means that either your certified aerospace supplier has to invest in a new production line for casting, or you need to certify another supplier to aerospace standards who is already capable of meeting your volume.

[Shpaget]

It is a very involved and costly process because usually the company that wants a satellite put in orbit can only afford one (or however they plan to have) satellite. That is why they want to make sure that one individual satellite works. However, when you are making 900 of them, the production process changes. You can significantly reduce the manufacturing cost at the expense of reliability and manufacturing faults. You just test them and discard/repair the faulty ones.

Clean rooms are in place to provide as much of reliability as possible. They are not essential for the operation of a coms sat.

All that being said, 1000 units is not a large production run. At these sort of numbers, casting parts is still not economical. A lot cheaper is to take sheet metal and laser/plasma/water cut whatever you need cut out, then mill for the finish quality and for mating surfaces. With a bit of smart engineering you can, for example, design the satellite to use same side panels on all sides.

[magnemoe]

And the rest?

Falcon 9 could deliver 80+ of them in one go, but happens to be Branson's competitor.

How would that work anyway? You have 80 satellites clumped up in parking orbit, and need to push them each in its own specific orbit. Would each sat do its own burs, or would the upper stage do multiple different close orbits from which the sats would fine tune themselves?

They will be distributed out in equal distance around multiple orbits. My guess is that you launch to lower orbit then do an burn with satellite to put it in correct position then it passes.

[Nibb31]

It is a very involved and costly process because usually the company that wants a satellite put in orbit can only afford one (or however they plan to have) satellite. That is why they want to make sure that one individual satellite works. However, when you are making 900 of them, the production process changes. You can significantly reduce the manufacturing cost at the expense of reliability and manufacturing faults. You just test them and discard/repair the faulty ones.

Clean rooms are in place to provide as much of reliability as possible. They are not essential for the operation of a coms sat.

Absolutely. The fact they plan 200 spares for a constellation of 700 sats indicates that they expect a very high failure rate (nearly 30%!), therefore they are willing to take cuts in quality control and reliability. My point is that it is a huge shift that completely changes the way a company like Airbus does business.

The 25% failure rate is scary, because these sats are designed to be at an orbit of approx 700 to 1000km, where there isn't much drag to deorbit them if they fail. Any sat that doesn't deorbit will become a debris hazard.

All that being said, 1000 units is not a large production run.

As a comparison, it's equivalent to the number of A330 aircraft produced or it's twice the number or Eurofighters. It's a pretty large production run for the aerospace industry, but fare from anything the space industry has ever done.

- - - Updated - - -

They will be distributed out in equal distance around multiple orbits. My guess is that you launch to lower orbit then do an burn with satellite to put it in correct position then it passes.

You need at least one launch per inclination. A decent spread for ~700 sats would be something like ~26 satellites on ~26 different inclinations or a similar combination. That would mean ~26 launches of 26 sats that weight 150Kg each, meaning that if you add the weight of a dispenser, a launcher in the 6-tons-to-LEO class like Soyuz or Angara could do the trick. But then you start hitting schedule bottlenecks. How many Angaras or Soyuz can be produced every year?

Edited June 17, 2015 by Nibb31

[Shpaget]

The 200 replacement units (I assume) are there to compensate mostly for launch losses and unexpected in-orbit failures. I would expect no more than 10-20 to have serious manufacturing defects, even under much less strict manufacturing standards.

Look at a modern mobile phone. It's a complex piece of engineering, yet they very rarely get to the customer broken.

I wouldn't compare these satellites to commercial jets, to me they are more like industrial equipment.

[Rune]

I wouldn't compare these satellites to commercial jets, to me they are more like industrial equipment.

I think you hit the mark there. These birds will be built with care... but not that much more care than any other piece of high tech electronic equipment with peculiar design requirements like, say, an MRI machine, or a cryocooled superconducting magnet, or even a plain old electro-optical hard drive (the tolerances on those are actually quite scary)... or you know, something equally complicated, but that is a far cry from the millions/kg space harware has historically cost. And maybe the MRI is not that good of an example, because the market for those is pretty small, actually. A satellite is dirt simple by comparison, requiring only a few ultra-high precision parts (other than the control gyros, nothing actually comes to mind right now). The rest is sophisticated electronics, manufactured in bulk, and if you don't use space-certified parts, you can probably provide much more redundancy and robustness through more advanced, smaller and powerful chips... A system-on-chip these days is a few grams, and it carries all you could ever need to run a sophisticated satellite (and more!)... not space rated? Encase them in aluminium, and put 5 watching each other's moves, one will survive the radiation long enough and the whole thing is still lighter than the outdated clunky rad-hardened electronics. Maybe you can have crazy tricks like doing away with a couple of space-rated cables by using wifi chips for the different parts of the sat to talk between them. That's probably also a very bad example, since these sats will be small, but you get the idea: when you go form building at most a dozen big satellites at a time, to building hundreds or even thousands of small ones, you can change everything about the way they are built.

Rune. Only in engineering anlaysis, you must be saving like, huge amounts, just on account on the design team/production team jobs ratio.

[Nibb31]

Well, since they are basically reinventing the wheel in terms of how a satellite is built, it will actually require more design team/production team people to not only design the new satellites using new methods, but also to source the new parts from new suppliers, layout a new manufacturing chain, and come up with new and innovative procedures to reduce man hours.

Instead of relying on established infrastructure and practices that are used in the industry, Airbus DS is going to have to throw away 50 years of experience and rethink vertically and horizontally the entire satellite industry from scratch. That is not going to be cheap or easy.

And since this is going to be pretty big production run (by industry standards), they can't simply try completely unproven solutions with certifying them through rigorous testing. They would need to be sure that your 5 system-on-chip CPUs or your on-board wifi works in space before committing the design to 900 satellites, because if one of those cheap systems fail on the first batch, then you have to rework the whole design, which will cost more in the end. So at the end of the day, those cheaper systems still need to have some sort of certification process before you can include them in the design.

Edited June 17, 2015 by Nibb31

[Rune]

Well, since they are basically reinventing the wheel in terms of how a satellite is built, it will actually require more design team/production team people to not only design the new satellites using new methods, but also to source the new parts from new suppliers, layout a new manufacturing chain, and come up with new and innovative procedures to reduce man hours.

Instead of relying on established infrastructure and practices that are used in the industry, Airbus DS is going to have to throw away 50 years of experience and rethink vertically and horizontally the entire satellite industry from scratch. That is not going to be cheap or easy.

And since this is going to be pretty big production run (by industry standards), they can't simply try completely unproven solutions with certifying them through rigorous testing. They would need to be sure that your 5 system-on-chip CPUs or your on-board wifi works in space before committing the design to 900 satellites, because if one of those cheap systems fail on the first batch, then you have to rework the whole design, which will cost more in the end. So at the end of the day, those cheaper systems still need to have some sort of certification process before you can include them in the design.

Oh sure! The opportunity can be botched just as much as it can be leveraged. For example, they would be pretty stupid not to prototype a few dozen sats first and subject them to hell (they have the facilities for that, shake tables and vacuum chambers and everything), see what breaks and what doesn't. Build true reliability numbers with production models. Then build another dozen, until a batch survives. You can't do that with a 5 ton GEO bird, because you will build one, six at most. So you have to simulate everything, and that takes more effort because you have to design you simulation and validate it in the first place, test it at the component level to death and have incredibly harsh quality controls during final integration.

And insurance? When you plan to waste ~200 sats in development, all you need is the launch insurance for a reflight, and that is up to the manufacturer of the launch vehicle. These birds should go mostly uninsured themselves, IMO. I did some order-of-magnitude numbers, BTW: At 50million each launch, 20 orbital planes are about a billion to fill. That's not quite a dedicated F9, unless F9R drops out of the sky with a hefty price reduction real soon, but some flights they will share I expect, so doable as an average and rough approximation, especially when you are buying a block of twenty launches. That means launch cost can be the traditional third of the "up to 3 billion" total cost of the system. But the operations segment has to enjoy some economies of scale, never mind the production of the satellites. You don't need a hundred times more ground stations and personnel to manage a hundred times more satellites! Hell, this might be on the low end of what they are projecting, for all we know now. But the numbers do seems to hold up to some (casual) scrutiny.

Rune. What would be the really crazy is to use aerospace project managers to do this.

[Nibb31]

Oh sure! The opportunity can be botched just as much as it can be leveraged. For example, they would be pretty stupid not to prototype a few dozen sats first and subject them to hell (they have the facilities for that, shake tables and vacuum chambers and everything), see what breaks and what doesn't. Build true reliability numbers with production models. Then build another dozen, until a batch survives. You can't do that with a 5 ton GEO bird, because you will build one, six at most. So you have to simulate everything, and that takes more effort because you have to design you simulation and validate it in the first place, test it at the component level to death and have incredibly harsh quality controls during final integration.

Yes, but really what matters in the picture is the cost of the entire system, not the cost of the individual satellite. Development through trial and error is a bad process that typically entails budget overruns and waste.

And insurance? When you plan to waste ~200 sats in development, all you need is the launch insurance for a reflight, and that is up to the manufacturer of the launch vehicle. These birds should go mostly uninsured themselves, IMO. I did some order-of-magnitude numbers, BTW: At 50million each launch, 20 orbital planes are about a billion to fill. That's not quite a dedicated F9, unless F9R drops out of the sky with a hefty price reduction real soon, but some flights they will share I expect, so doable as an average and rough approximation, especially when you are buying a block of twenty launches.

I really can't see SpaceX launching these birds. OneWeb is a direct competitor to SpaceX's own constellation and they wouldn't delay their own project when being the first-to-market is going to be crucial.

They are going to have to go with a more traditional launch provider or find a really really cheap alternative (good luck with that!).

That means launch cost can be the traditional third of the "up to 3 billion" total cost of the system. But the operations segment has to enjoy some economies of scale, never mind the production of the satellites. You don't need a hundred times more ground stations and personnel to manage a hundred times more satellites! Hell, this might be on the low end of what they are projecting, for all we know now. But the numbers do seems to hold up to some (casual) scrutiny.

Not the numbers I've seen. OneWeb claims the total cost of the project to be between 1.5 to 2 billion. I can't see how they could get 25 launches for less than a billion. If they are paying 1.4 billion for the satellite contract, they are already over budget. And there is still the commercial service to set up.

Rune. What would be the really crazy is to use aerospace project managers to do this.

Given that we are talking about Airbus Defense and Space (ex-EADS Astrium), don't get your hopes up too high...

[Nibb31]

So we have our answers. Oneweb has signed contracts with Arianespace and Virgin Galactic for:

- 21 Soyuz flights

- 39 LauncherOne flights

plus options for additional

- 5 Soyuz flights

- 3 Ariane 6 flights

- 100 LauncherOne flights

At nominal prices, that puts the price of launching the constellation at approximately 21x100 million+ 39x10 million= $2.5 billion. However they are probably get a large discount for ordering such a large volume. Something like 1.5 to $2 billion is probably closer to the truth.

Those must be the biggest bulk purchases price and volume in commercial spaceflight history.

Edited June 25, 2015 by Nibb31

[Kryten]

Purchasing dozens of flights on an unflown vehicle from a company with as checkered a history as VG is a very brave move.

[Nibb31]

Those are only replenishment flights, so there is less pressure. Virgin Group is a major investor in OneWeb, so I guess they had no choice. OneWeb justifies the existence of LauncherOne and vice-versa.

It's all a very bold project, with a multi-billion dollar investments in a strongly competitive market with predictably low margins, an unproven business model, and a reinvented industrial base. Taking risks is part of what makes an entrepreneur I guess...

Edited June 25, 2015 by Nibb31

[Kryten]

OneWeb have received $1.2 billion in a new round of investment. This should be the last round they need, with the remainder of the money coming from debt financing.

http://spacenews.com/oneweb-gets-1-2-billion-in-softbank-led-investment/

[Jonfliesgoats]

Satellite internet also has some big implications regarding the ability to disrupt and censor internet access in repressive  areas.  Granted signals can be jammed, but more information access seems to be on the horizon, making ti more difficult for your local strong man to filter your search results.  There are some unique surveillance opportunities with this too.

Edited December 20, 2016 by Jonfliesgoats

[wumpus]

Any mentions of the downlink device (for this or Musk's system)?  I can only imagine a direct to customer system (like Irridium) but other's (in spacex threads) have suggested that it would be to something like a wifi/wimax link (which would make it harder to justify customers in the developing world).  I'd think that getting a receiver (especially fixed) would't be *too* hard, but I've never designed much in the way of RF.

[Kryten]

Two bits of news;

-OneWeb are seriously considering adding another ~2,000 sats to the plan (representing about sixty more Soyuz launches)

-OneWeb and established GSO operator Intelsat have agreed to a merger, though there are several conditions that could cause it to fall through. If it does happen, OneWeb plans can be expected to be considerably accelerated.