Black magic. Maybe a little red magic, too.

January 22, 2021 8:01 AM

An in depth RF focused teardown and analysis of a Starlink base station antennae.
posted by loquacious (18 comments total)

As soon as I saw the hyperlink description I hoped it was The Signal Path doing the analysis. The channel is great and provides great in depth reviews. Excited to finish watching this!
posted by Quack at 8:23 AM on January 22, 2021 [1 favorite]

Seconding. The Signal Path is such a great channel!
posted by wordless reply at 8:29 AM on January 22, 2021

This is also a fine example of someone who knows RF engineering demonstrating how much of antennae and RF design is trial and error and esoteric handwavium magic.

Many years ago I had a friend who was an advanced RF engineer. Like so advanced that Qualcomm kept trying to hire him and keep him as one of their own pet RF engineers and they kept making increasingly ridiculous and eye-watering salaries and signing bonuses and he wasn't interested in being trapped into working on cell phones and endless NDAs and the lack of independence.

He also told me something he didn't tell Qualcomm, which was that he actually personally hated them as a company because they kept buying up all the stock of really good specialty RF chips he used for independent contract projects, some of which were very large scale independent projects like inventing an entirely new custom digital/data/telemetry radio system that including everything from custom RF silicon engineering on up through the boards.

He was one of those rare real world Real Genius nerds that would show up at high level meetings at the DoD presenting to multi-starred generals while he was wearing thrashed jeans and a ratty old tie-dye shirt and no one would blink an eye - likely because they would be briefed and told in advance that they weren't allowed to say a damn thing otherwise he might walk.

It was like hanging out with a wizard. We mostly liked to hang out and drink beer and smoke weed, go figure, and he liked how I asked interesting questions when we were stoned, so go figure.

Of course I asked him to explain how antennae and RF radiators actually work and he just laughed and said if anyone claimed they could actually tell me they were either lying or were about to win a Nobel Prize and change our fundamental understanding of quantum physics and photons.

I remember him admitting and laughing that most of his actual work involved knowing how to run simulations and doing a lot of trial and error and just waiting for his personal office server farm to crunch lots of numbers.
posted by loquacious at 8:58 AM on January 22, 2021 [27 favorites]

The intersection of "trial and error" and computer modeling leads to some really weird stuff, too. At a former job we had one RF Genius who vetted board designs. One of ours was behaving strangely and the fix (proposed by Genius' modeling software) turned out to be rotating one antenna by just two and a half degrees. Then, on the weirder end of the spectrum, you have things like this evolved antenna (wikipedia).
posted by runehog at 9:38 AM on January 22, 2021 [6 favorites]

I am super excited about StarLink. I live in rural California where we have very few options for Internet and 12Mbps is considered "really fast!". Being able to point a small dish up and have high bandwidth, low latency access will be life changing. Not just for me but for a lot of the world. I'm eagerly anticipating beta access; it's still not available in my region. Their big problem is going to be having enough capacity in each cell to meet demand.

I've been following the subreddit, it's pretty good. FWIW, the colloquial name for the "Starlink Dish Phased Array" is "Dishy". Often personified, apparently he/him, folks seem to love their Dishy. Particularly the magic way he sloughs off snow. The service is up and working now with good bandwidth, but until more satellites are deployed folks report a few minutes of downtime a day.

Some useful resources gleaned from the subreddit... List of beta invite locations, in the US it's mostly folks north of 43°N. Also an animation of the satellite constellation as it's been deployed. And a map of ground stations.
posted by Nelson at 9:52 AM on January 22, 2021 [3 favorites]

Beamforming and phased arrays are just black magic fuckery. They work on differences in timing based on the speed of light and even though we've been able to crudely work with it for over a century, it's taken that century to be able to fabricate the double digit gigahertz transistors to truly have the mastery over them that we have today.

RF engineers are wizards. Literal fucking wizards.
posted by Your Childhood Pet Rock at 10:15 AM on January 22, 2021 [7 favorites]

I know antenna design and related stuff is really complex, but I'm not sure that I buy that we just don't know how it works. If we can simulate it accurately, and those simulations translate to the real world when we build the thing, then I'd say it's understood, it's just really complicated. Very cool, though.
posted by Joakim Ziegler at 1:54 PM on January 22, 2021

but I'm not sure that I buy that we just don't know how it works

Sounds like, based on the comments above, too, that we don't understand it well enough to make good, reliable predictions, which is why people keep running a bunch of trial and error simulations and prototypes?
posted by danhon at 4:16 PM on January 22, 2021

Yup, RF engineering (or any HF analog engineering) is black magic .... or as one practitioner I know says - "it's more necromancy, you only become an adept when you know where all the bodies are buried"

Which is to say, you have to make a whole lot of mistakes to become a real expert
posted by mbo at 5:20 PM on January 22, 2021 [1 favorite]

Oh, we know how Electricity & Magnetism work, it's just too time-consuming (in both human effort and computing time) to model all the fiddly details. We know how semiconductors work but it's not realistic to build them exactly the same every time; the same applies to circuit boards. It's easier, at this point of technological progress, to build prototypes and test them, rather than making perfect mathematical models.
posted by Standard Orange at 6:06 PM on January 22, 2021

I understand like 90% of that just from the physics/math/ee/ce bits. What I'm left wondering is "is that really a Power Over Ethernet (POE) jack?"

Is there enough in that System On a Chip (SOC) to turn this into a full blown router? Or is there a second box that just uses that POE cable for cost/ease and is the router bit leaving the cable just as a means to connect to an antenna over a distance. If so, are they really using Ethernet? It would be a Point To Point (PTP) connection and half of the Ethernet protocol is not needed... they could just be using a Cat-6 cable because it's cheap and easy and non-confusing to end users but running some other protocol over the same wires.
posted by zengargoyle at 7:41 PM on January 22, 2021

Of course I asked him to explain how antennae and RF radiators actually work and he just laughed and said if anyone claimed they could actually tell me they were either lying

We understand very well how basic antennas work at more frequencies, but it's undeniably true that as frequency increases and more interacting elements are added our understanding rapidly decreases until it is eventually all handwavium and/or trial and error.

It's actually pretty amazing how some people were able to make the folded antennas that enabled the tiny candy bar tri-band cell phones of the late 90s and early 2000s. By all rights they shouldn't have worked at all, much less at several different non-integer multiple frequencies in fringe reception areas, but they somehow managed to pull it off despite having nowhere near the simulation capability that we have today.
posted by wierdo at 11:13 PM on January 22, 2021

I thought cell phones were some weird fractal antenna design that worked well enough for the power/distance bit of the equations. That's just a bit of... we can etch this out without all of the resonance sort of stuff and still get a good enough signal back and forth to the nearby cell-tower. It's balanced phase array in the structure to the point that it works.
posted by zengargoyle at 1:40 AM on January 23, 2021

There's a brilliant story from early experiments in 'hardware evolution' in the mid 90s.

The informatics researcher began his experiment by selecting a straightforward task for the chip to complete: he decided that it must reliably differentiate between two particular audio tones. A traditional sound processor with its hundreds of thousands of pre-programmed logic blocks would have no trouble filling such a request, but Thompson wanted to ensure that his hardware evolved a novel solution. To that end, he employed a chip only ten cells wide and ten cells across⁠— a mere 100 logic gates. He also strayed from convention by omitting the system clock, thereby stripping the chip of its ability to synchronize its digital resources in the traditional way.
...
Finally, after just over 4,000 generations, the test system settled upon the best program... As predicted, the principle of natural selection could successfully produce specialized circuits using a fraction of the resources a human would have required. And no one had the foggiest notion how it worked.

Dr. Thompson peered inside his perfect offspring to gain insight into its methods, but what he found inside was baffling. The plucky chip was utilizing only thirty-seven of its one hundred logic gates, and most of them were arranged in a curious collection of feedback loops. Five individual logic cells were functionally disconnected from the rest⁠— with no pathways that would allow them to influence the output⁠— yet when the researcher disabled any one of them the chip lost its ability to discriminate the tones. Furthermore, the final program did not work reliably when it was loaded onto other FPGAs of the same type.

It seems that evolution had not merely selected the best code for the task, it had also advocated those programs which took advantage of the electromagnetic quirks of that specific microchip environment. The five separate logic cells were clearly crucial to the chip’s operation, but they were interacting with the main circuitry through some unorthodox method⁠— most likely via the subtle magnetic fields that are created when electrons flow through circuitry, an effect known as magnetic flux. There was also evidence that the circuit was not relying solely on the transistors’ absolute ON and OFF positions like a typical chip; it was capitalizing upon analogue shades of gray along with the digital black and white.

Back when I was learning about conventional hardware chip design in the early 2000s, this blew my mind. Transistors that don't work as switches, but only work because of 'incidental' EMF effects?

Our ability to build conventional designs kept improving fast enough that we were able keep increasing chip capability by basic brute force - by making ever smaller transistors, so higher density - but we're finally starting to appear to hit the limits of what immensely complicated lithography advances can deliver. Intel, the 'big beast' in lithography improvements, was really stuck getting down to 10nm for several years, and still can't do it at large scale. TSMC can't make their 7nm chips (equivalent to intel's 10nm) anywhere near fast enough to meet demand, which is why it's still really hard to buy the latest gen graphics cards and AMD processors and consoles.

I do wonder if in the future we're going to return to looking at what we can do by tapping into the black magic of EMF interference on transistors beyond RF applications, instead of trying to minimise and avoid them.
posted by Absolutely No You-Know-What at 2:38 AM on January 23, 2021 [7 favorites]

I thought cell phones were some weird fractal antenna design that worked well enough for the power/distance bit of the equations

The ones in the free phones the carriers were giving away were basically just barely good enough, but a couple of manufacturers managed to come up with designs that worked better than the previous external antennas.

Ironically, most smartphones these days are big enough that they use relatively simple antenna designs where the only real cleverness is how they're hidden in plain sight. (And the way they keep you from accidentally shorting them out with your hand) I have yet to find a smartphone that gets reception anything like my Nokia 6230 did despite UMTS and LTE having an order of magnitude or more coding gain to help them out. The last phone I had that supported analog (a Nokia 6340i) was similarly impressive. I could be out in the middle of the woods where a car phone had trouble and get a perfectly usable signal.

I stuck with Nokia to the bitter fucking end because of my experience living in a place where other people's phones would literally not get a signal anywhere in the neighborhood on any carrier while mine was still showing -90dBm and worked fine in a pocket, when I was sitting on it, or even when I was wrapping my hands around it trying to get it to drop the damn signal. Even Motorola's phones would occasionally have issues there, and they were themselves considered wizards of RF engineering for good reason.

Antenna design matters, kids. Better reception means better battery life, after all.
posted by wierdo at 3:46 AM on January 23, 2021 [4 favorites]

What I'm left wondering is "is that really a Power Over Ethernet (POE) jack?"

It almost certainly is. SoCs with built in Ethernet interfaces are readily available. The required data bandwidth to/from the customer's router isn't overly taxing Ethernet bandwidth, so there's no need to spend time and money improving the connection. It's just plain vanilla.
posted by monotreme at 4:10 PM on January 23, 2021

Yeah, just wondering if you could plug it into a computer and use your standard Ethernet stuff and reverse engineer the protocol. There's plenty of room to not use Ethernet and do the hard stuff in the attached box instead of putting it in the expensive and hard to make flat dish under limited power and subject to extreme conditions.

Next up... can I install Linux on that dish?
posted by zengargoyle at 6:52 PM on January 23, 2021

It already runs Linux. ;)
posted by wierdo at 9:03 PM on January 23, 2021 [1 favorite]