Project Kuiper's custom-built antenna architecture | Hacker News Reader

Project Kuiper's custom-built antenna architecture

aboutamazon.com

48 pointsboolean52 comments

As an antenna engineer working on phased arrays, I find the claim that reducing the physical size directly reduces manufacturing cost by an order of magnitude laughable. As if the plastic radome or PCB factors into the cost at all.

The cost of these systems is 100% in the electronics. What I presume is going on here, is that Amazon (like SpaceX) has developed custom RFICs, which handle both uplink and downlink in a single package, thus requiring the apertures to be overlapping. The advancement is in having developed a better IC, the size reduction is an indirect outcome.

If Kuiper is taking a very different approach, more akin to Kymeta or ThinKom, the original statement might be credible, but the aperture looks too thin for that to be the case.

The enclosure costs will be negligible, but the PCB will not be. These phased array antenna designs from SpaceX and Amazon are some of the most complex PCBs ever theorized, with many dozens of layers and requiring manufacturing processes that do not currently exist at scale.

Not really. I just built a 256 element array covering 2 octaves BW. The feed board is seven layers. Arrays are lots of repetitive circuitry. I think the impressive part are getting microwave RFIC for pennies.

Now this PCB is crazy:

https://youtu.be/nurL3N1Etuc

Two quick questions that may or may not make any sense.

If you've seen the Starlink dish teardown, would you suspect that it's built for just a single beam or could it support multiple simultaneous beams?

And what would you expect are the primary performance/capability differences between the 'dish' antenna and the one they fly in the satellite (space-hardening aside)?

Not the OP.

We have to clarify what you mean by "support multiple simultaneous beams". The SL receiver is half-duplex, so it can only send or receive at any given time, and, in theory anyway, it uses beamforming to communicate with just one bird at a time (if it couldn't beamform to within sufficient tolerance, then that would be an FCC no-no.) There's no theoretical reason, AFAIK, that it couldn't rapidly switch between birds, but the actual feasibility of that depends on terminal firmware, the ground network being setup to handle that, the arc between two birds, and whether or not the antenna's scan angle can encompass that arc.

The primary difference between the ground terminals and the antennas in space is the amount of power passing through them, but also their sophistication (primarily the beamforming capability.)

Ah, was actually wondering about the duplex situation as well, thanks for answering a bonus question :)

Re: multiple beams I'm thinking about the ability to split the array up into 'subarrays' with independent beamforming capability. Probably doesn't work for reasons above my head but I was just trying to think about the mechanics of the transition between satellites. Having a chameleon mode with one eye pointed at one satellite and another pointed at the next seems like one way to ensure a seamless transfer. But realistically the dish probably knows where to steer next within an approximation of its beamwidth, so it probably just flips like a channel.

Thanks for the answers!

Ah, OK, I understand better.

Yes, in theory, a terminal with a phased array antenna can be built to emit more than one beam simultaneously and steer them independently. Doing so is currently still costly in terms of research, electronics, and power, so it's unlikely that the SL terminal can do it. It would also require there to be a second PHY chip somewhere in the terminal, which would also drive up cost.

Also, it's not really necessary. Satcom already requires nanosecond-level timing accuracy, so I would think their terminal can probably steer its beams with at most ms lag.

Do you have a reference for the half-duplex note? That would seem a waste given that they already have FDMA. Also most satcom applications are full duplex.

https://twitter.com/TMFAssociates/status/1300438995843776514

Based on the teardown, it's impossible to tell; again largely comes down to what their ICs are capable of. But based on the patents, and what I've heard unofficially, it does support 2 simultaneous beams. The point of that isn't to double throughput or anything like that, it's to accomplish "make before break" handoff as one satellite moves towards the edge of view, and another is better.

Don't know much about how they've set up their satellite terminals--for a typical GEO station, they'd support a huge number of static beams simultaneously to multiplex as much through the station as possible. With many birds in LEO, it's unclear how extreme they have to get about that.

Previous threads on this project:

https://news.ycombinator.com/item?id=24209940

https://news.ycombinator.com/item?id=24142135

https://news.ycombinator.com/item?id=24010633

(If anyone's curious. Not dupes.)

Good for them, and good for all of us. If there was only one provider of satellite internet, it’d be priced out of range for all of us.

But both Amazon and SpaceX look like they’ll be able to ship viable products. This really is a game changer though, because it makes rural areas much more attractive places to live in. With access to stable, fast internet anywhere, remote workers could live wherever they wanted to.

But more than that, once this becomes affordable I’m excited to see how getting nearly everyone internet access changes the internet.

I wouldn't hold your hopes for blue origin being able to produce anything any time soon. remember they started before spacex, and have accomplished very little during that time.

Nobody mentioned Blue Origin, what do they have to do with Kuiper?

Many people are thinking tribally and assuming that Bezos's Kuiper must launch on Bezos's Blue Origin.

Realistically, isn't that the only path to a competitively priced product? What other launch provider are they going to use? Surely they can't utilize the falcon / starship as that directly funds their competitor

A. Go out of business (e.g. because Blue Origin is not ready and the FCC is about to revoke the spectrum license)

B. Fund your competitor

If it came down to it, would you really choose A?

I would think B would quickly lead to A as Kuiper would not be able to compete with Starlinks pricing advantages

> Instead of placing antenna arrays adjacent to one another, we used tiny antenna element structures to overlay one over the other. This has never been accomplished in the Ka-band

Starlink does this but in the Ku band. Starlink is 2 feet in diameter vs. 1 foot for Kuiper. Much more portable, although I imagine still far too power hungry to run off a laptop and of course still far from cell phone sized.

Seems like they both have a lot of potential on the roof of vehicles (airplanes included). Or even as a 'ground station' for a neighborhood or village (but that depends on the latitudes they cover, and their billing structure).

Does Amazon have a way of putting the satellites in orbit yet? Is that going to be up to Blue Origin?

I sincerely hope not. The lawsuits will fly if Bezos awards Blue Origin launch contracts.

Really? Who would sue?

It's a joke

Reference: Amazon lawsuit against Pentagon for Azure contract

Reference: Amazon lawsuit against Future Retail

etc

Or Blue Origin trying to patent rockets landing on floating platforms and then suing SpaceX. At least we are soon coming up on the 5 year anniversary of the first successful landing!

https://spacenews.com/spacex-gets-partial-win-in-patent-ruli...

Amazon investors if there was preferential treatment given to the owner of the company's other, unrelated company. I don't know why the above comment is so downvoted - Elon Musk recently had to field a similar set of lawsuits for Tesla's solar city acquisition.

Jeff Bezos would probably just subsidize the launch cost by dumping another couple billion into BO, making it the cheapest option to Amazon.

Blue Origin is also talking up their moon lander, but we still haven't seen them actually put a payload in orbit. The BE-4 engine has been delayed and they've still only delivered test stand versions to ULA. If you read the Blue Origin Glassdoor and subreddit where a number of anonymous employees talk, they're not doing well. New Glenn looks unlikely to fly until 2022 at the earliest, and presumably they'd need at least 6-10 launches even being optimistic about fitting hundreds of sats in the large fairing.

I can't see them being able to offer service until 2024 at the earliest, and there's a non-nil chance Blue Origin implodes trying to leapfrog from suborbital test vehicle to reusable heavy-lift rocket with their own (possibly flawed) engine and Kuiper would need to launch with SpaceX or fold.

Is Kuiper completely dependent on Blue Origin? I assumed they would start by buying a few launches from other providers, though I didn't expect them to try SpaceX. I guess SpaceX would sell to them although it would be a little weird.

Per this tweet, they're planning on using multiple launch providers:

https://twitter.com/jeff_foust/status/1339306611488288768?s=...

> Interesting comment at TC Sessions: Space by Amazon’s David Limp about launching its Project Kuiper constellation: we hope that Blue Origin can provide launch capacity, but hope others will, too. Expecting to use multiple launch providers.

FCC approval only gives Amazon until July 30, 2026 to launch at least 50% of their 3,200 sat constellation into orbit. I don't know of any existing provider other than SpaceX that could plausibly put that much payload into orbit in 5 years. It's not like ULA has a ton of spare launch capacity sitting around.

Blue Origin may move slowly but it would still be surprising if New Glenn couldn't easily handle that by 2026. It should only take 14 launches for 1600 satellites assuming 120 per launch (double Starlink/Falcon 9), and New Glenn is supposed to be reusable and operational next year.

I can’t find the article/tweet but I believe SpaceX said they’d launch competitors’ satellites.

Of course they won’t get the same at-cost deal SpaceX gives themselves.

>If you read the Blue Origin Glassdoor and subreddit where a number of anonymous employees talk, they're not doing well.

Got any specific examples? Glassdoor is practically useless for reading into a company's actual culture, and after checking out the BO subreddit I only saw one single negative post about BO, and even on that post most of the comments were disagreeing with the premise that BO is failing.

The comments are interspersed in random threads, but there's a bunch in last month's about the COO being replaced, one striking example:

https://www.reddit.com/r/BlueOrigin/comments/k5icgf/coo_of_j...

> I work for Blue Origin. blue is not telling the public..

> 1. They built 4 different rockets..none are built the same.

> 2. They never documented the builds. The allowed legacy engineers transfer to New Glenn.

> 3. The rockets are marginal at best. tail 4 just came back from Kent..They had to cut it open to replace helium tanks. the tanks have a life cycle.Instead of replacing with new..they ground on the welds till the cracks went away. ATP test was need to know only.

> 4. Software between tails is different..they cant talk to one another.

> 5. BE3 is a single engine dud..The last flights have all been done with the same engine cracks and all.

> 6. All BE3 engine assembly attempts have produced an Engine

While that's certainly not a stellar take on things, none of those points seem to be indicating that "Blue Origin is not doing well". Those seem like pretty standard occurrences for anyone who's worked on a large iterative engineering team. I remember reading similar stuff about SpaceX back in the Falcon9 1.0 days.

If you're comparing their rocket to Falcon 9 1.0 from 7 - 10 years ago, that is concerning, especially when they were founded 2~ years earlier.

Right and a lot of SpaceX's scrappiness and scrambling was due to funding constraints in the early days - if you're running your rocket program like a broke startup when you get $1B/year from daddy Bezos, something's amiss. But broadly, they've been around for 20 years and haven't put anything into orbit yet, so it's hard to call it a serious program.

I know the goals and scale are different, but Astra's on the cusp of orbit after only 5 years and orders of magnitude less investment. Something at BlueOrigin is seriously broken: https://en.wikipedia.org/wiki/Astra_(aerospace)

It feels like Blue Origin is Jeff Bezo’s hobby. A very, very, very expensive hobby.

Woof. Cracks in welds is always a bad sign of poor engineering management and quality control. Getting welders that can consistently produce xray quality welds isn't hard or even expensive. You just need to get good people and treat them well. As for discovering them you can xray all the welds during fab to ensure nothing like that makes it out. Though depending on what surfaces you were joining you can often just grind the weld out and try again, though this is pain in the ass with full penetration welds.

My brother runs a high end fabrication company (the kind you might hire to fabricate a piece of your rocket) and such things would never leave the floor of his shop.

Yeah that was my gut reaction too - especially in a town like Seattle... I'm sure there's one or two shops up there that can competently weld critical components that are designed to fly.

Blue Origin doesn't need to be worried about lack of cash. As long as Bezos is willing to burn billions they will keep on going. But that also means they don't have fear of failing, so they keep on doing things slowly. Elon on the other hand was and still is relentlessly pushing his companies to a breaking point. Model 3 production hell was a great example - do or die for Tesla, with crazy production lines in tents, flying machines from Germany in matter of days etc. Starship program is also managed as if the world was ending tomorrow with every other prototype ending with a big boom. The big pivot from carbon fiber to steel is example of agility, not failing for sunken cost fallacy and optimizing for rapid development speed.

SpaceX is a company that understands, respects and capitalizes on failure. Their entire development process is based around failing early and often in order to prune the search space of viable solutions as quickly as possible.

Proper Agile development, but for rockets.

Not quite. SpaceX isn't just throwing random designs at the wall and using some gradient descent to find a good rocket design.

They are starting with competent designs and well thought out systems. The most interesting aspects to their rockets is how boring they are. The Falcon 9 basically uses a Beowulf cluster of Merlin engines. The second stage uses yet another Merlin engine. All use RP-1/LOX so there's minimal cryogenics involved in fueling and defueling the rocket on the launchpad. The Falcon Heavy is a Beowulf cluster of Falcon 9s.

They didn't happen into any of those designs. They've existed from the beginning. Where SpaceX accepts "failure" is in their integration tests, these are their more visible "failures". All of their subcomponents have been bench tested before their integration tests and are known to work correctly. Their integration tests are literally testing their large scale integration, the things that can't be bench tested.

Their process is waterfall AF. They might take iterative steps in the process but they're not stumbling around looking for good designs.

Dan Rasky, Senior Research Scientist at NASA gave great description of SpaceX processes and how they differ from Old Space: https://www.youtube.com/watch?v=MxIiiwD9C0E

Well they could buy space on Falcon9's :-) Internet Service is a known way to generate cash.

But that said, BO seems to be an "announcements" company and not a "doing things" company, which is sad. The former relates to research and good ideas, the latter relates to execution of said ideas. I've met one engineer who used to work there and their experience was not particularly positive but it was also anecdotal so I don't give it a lot of weight.

What I do give weight to are seeing results. This is something SpaceX has lately excelled at and BO has not. It seemed like they were "ahead" of Virgin Galactic on the "Suborbital Tourists / Minting Astronauts" thing but now VG seems to be a couple of proof flights away from their first customers. Then they seemed to be ahead of SpaceX's efforts to get the Raptor engine going but the BE-4 stalled (although it seemed to unstall[1] as well). Not that SpaceX's time lines vs announcements are that much better but SpaceX seems to be getting over the finish line more often. As a result I wonder sometimes what is going on over at BO.

[1] https://arstechnica.com/science/2020/10/ula-chief-says-the-b...

The picture shows UpLink at 28-28GHz and Downlink at 17-19GHz I'm curios why not the opposite?

Those are the bands designated by ITU / FCC for everyone else as well. To change them would cause big interference problems for existing systems.

Oh great, yet more space junk.

I wonder when it’ll be commercially available.

Blue Origin was founded two years before SpaceX.

Bezos also makes the same references that Elon Musk spearheaded in his interviews... like Bezos says that noone would use throwaway planes that would be only good for one flight. Musk said that long ago also.

So, the observation that a lot of copying is going on is more incorrect, but that's the case with most things in life.

Shows though how strong Musks influence is in the area

Reply on news.ycombinator.com