Hacker News Reader

Comment by mNovak | Hacker News Reader

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.

View on HN

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.

Reply on news.ycombinator.com