Hacker News Reader

Comment by bri3d | Hacker News Reader

I really wonder why they don't introduce new hardware.

I know the system is fairly complex (tons of TI frontends building a big steerable phased array antenna), but the current hardware is awful for RV and really any use. There's a motor-oriented rectangular "dish" with a mandatory 1+ foot mounting arm, which stows in only one vertical orientation. Then, there are proprietary micro-HDMI esque connectors on both ends of a vendor provided fixed-length CAT5e from the dish to the router. For RVs, all of these are awful - the dish can't lie flat and must be brought inside for transit, and wiring is an unnecessary hassle.

As a Starlink user, my take is pretty much that the core technology is revolutionary in specific areas, but the product itself is mediocre in both cost and implementation. I wish they'd "unchain" the system with less bizarre hardware.

View on HN

> but the product itself is mediocre in both cost and implementation.

Go compare it to what you will get for $150 per month from consumer grade geostationary based highly oversubscribed services. Viasat/Hughesnet/wild blue/etc.

I could pull out my checkbook right now and start a contract for $6,000 per month on a 36mo term of dedicated SCPC geostationary capacity (requiring about $8500 of equipment at my site with a 1.8 meter dish, moderately powerful BUC and new modem) and not be able to come anywhere near matching the speeds and capability of my starlink terminal.

At least I can get service from the other providers. 2+ years on the waiting list and addresses in my area are now targeted for 2023.

I can see how that would be very frustrating, do you have any local options for WISPs?

WISPs come with their own challenges; namely time and line of sight to another node on the network.

Heck I about bought a schoolhouse in rural Michigan until the WISP got back to me and told me they couldn’t provide me service because of the trees in the yard.

I've seen people erect light duty 60' towers suitable for the wind load of one 60cm ptp microwave dish for a link to a local wisp, can be $2500-4500 all in including shipping, foundation, concrete work, etc.

That’s all fine and dandy; I’m happily on the west coast again instead. Happy it didn’t work out! That amount of uncertainty when buying a home and working from is not for me.

It's pretty sad that people put that much money into towers when the cost of hooking a home up to fibre (even rural) is in the $1800-2500 ballpark.

Only if the fiber provider is in the same area doing a build to serve at minimum a few hundred other homes at roughly the same time. Some people who live in rural areas where they would need a 60-80 ft tower for wisp cpe or starlink may be in a place so far from existing ISP fiber that a new build purely aerial fiber construction project could run $300,000.

It's perfectly possible to do FTTH builds for 30-40 homes at a cost of $30-40k. Just make sure you know what you're doing and avoid getting fleeced by engineering consulting firms and high margin construction companies.

I didn't want to put up a 100+ foot tower, so I started a rural FTTH ISP instead and ran my own fibre rather than pay the existing local WISP to give me crappy 5-10Mbps service.

Still, there are lots of cottages and other rural homes we can't get to and are trying to get Starlink at. The entire region I'm in has been stop-sell for a long time, yet there are zero uplinks anywhere nearby. I'd be more than happy to host an uplink for where we have dark fibre back to a data center to help increase capacity in the region, but nobody ever answers emails at Starlink. It's kinda awful how bad they are at customer service.

Starlink is only interested in colocation of their earth stations along major longhaul fiber routes, typically next to regen huts with existing Telco infrastructure. Just having dark fiber back to a medium or larger sized city isn't enough.

The level3 Prosser, WA hut site and starlink equipment there is a good example.

> Go compare it to what you will get for $150 per month from consumer grade geostationary based highly oversubscribed services. Viasat/Hughesnet/wild blue/etc.

Aside from "it's not a good deal just because it's cheaper than a bad deal" - your point is completely irrelevant to terrible design choices that have little to do with cost.

A permanently-attached cable on a very expensive outdoor device is mind-boggling levels of asinine, to the point that it seems done purposefully to help subsidize the service by making money off returned units with damaged cables.

Or guarantee that units will eventually die; it sounds like Musk wants to shift to being a satellite backbone, not providing interwebs for plebs - and one great way to shift to that is to simply stop selling dishes to consumers and wait for the existing units to die of cable failure.

(And before you jump all over me: yeeeeees, I'm aware that implementing a connector for the cable has costs. Starlink chose to cut those costs.)

It's completely up Musk's alley to pull this sort of blatantly anti-consumer nonsense. His entire life he's been profiting off exploiting his customers. For example, in the case of paypal, blatantly stealing money from people's accounts for "suspicious" activity, and getting away with it because "paypal isn't a bank, neener neener."

Far from me to defend Musk but assholes have a place in society -- they just need to be monitored.

Whatever the means he has used, he kickstarted (in the minds of consumers) at least three industries.

Would I willingly give him money? No. Unless I had an RV :)

> he kickstarted (in the minds of consumers) at least three industries

In reality he didn't.

He didn't kickstart the stagnant commercial space launch industry? Really?

Kickstarted is a strong word here, but revolutionized is perfectly adequate. If starship works out, he’ll monopolize it.

No, he's not an engineer or a scientist. He paid others to do the work. They did the work.

> "at least three industries"

He bought other companies that were already started. Bought, not founded and not invented technology himself.

Somehow, when it comes to billionaires, "I paid a writer" becomes "I wrote a book".

> it sounds like Musk wants to shift to being a satellite backbone,

Musk and his network team are not that dumb. They know about the Shannon limit in rf microwave and millimeter wave and the multi terahertz channel width capacity of basic 9/125 SM fiber.

Anything RF based has incredibly tiny capacity compared to modern 100/200/400GbE 40/80 channel dwdm systems. The capacity of two strands of fiber on a long haul path is incredible.

Internet backbone links carry far too much traffic to handle through even the most optimistic starlink sat to sat laser links.

I 100% concur with you on the foolishness of the cable. It should be some kind of ip68 rated twist lock Ethernet connector.

Why do you think 100gbps+ is not feasible for sat to sat lasers? The vacuum of space is a better environment than 500km of glass with a bunch of inline repeaters.

Because a few discrete OOK lasers in a single frequency aiming at another moving target, while they will have a lot of capacity compared to an rf link, will have a minuscule amount of capacity in Gbps compared to a whole DWDM based, coherent 100GbE+ optical transport system operating on singlemode fiber.

Like, literally, twenty or thirty individual full duplex 400GbE circuits in two strands of fiber, if you have enough money to throw at the problem.

The only way you could approach matching the same capacity on a sat to sat link would be if you had a massive array of 30-40 separate laser tx and corresponding massive array of 30-40 rx receptors on the other side. It's easier to understand if you've seen a DWDM mux and demux in person in a telecom facility.

I'd like to be proven wrong if somebody can do multi Tbps of capacity on a sat to sat laser link, that would be awesome, but the challenges are very high.

> The only way you could approach matching the same capacity on a sat to sat link would be if you had a massive array of 30-40 separate laser tx and corresponding massive array of 30-40 rx receptors on the other side.

This does sound like the sort of thing Musk would try, especially post-Starship, if the rest of the setup was physically possible.

One of the other problems is that the ultimate last hop for data would still have to be an rf link satellite to earth station through atmosphere, where lasers don't work well at all (there's a rain and junk in the air reason the telecom industry has given up on free space optics lasers for 1-2km, 1 to 20Gbps data links roof to roof and uses millimeter wave fdd radios instead)

I think it might be more about latency then pure capacity.

> Because a few discrete OOK lasers in a single frequency aiming at another moving target

Two sats in the same shell are not moving relative to each other.

Two satellites from the same launch, in the same orbital inclination, following each other in a conga line of starlink satellites are still slightly moving relative to each other, considering the narrow beam width of a laser shot at distances of 50 to 80 km or more between satellites.

The cable isn't permanently attached on the new square dish, iirc

The cable isn't permanent on the rectangular dish. It's got weird proprietary micro-HDMI esque connectors on both ends (and each end is different!) This at least isn't completely out there, since the cable can be replaced, but certainly isn't as good as an off the shelf cable that can be replaced in almost any town.

I think this must have been done due to either weatherproofing (RJ45 is awful for this, you can make it waterproof with weird enclosures but it's painful) or some kind of issue with the excessive power requirements from the dish. The cable is just CAT5e. I cut it and put RJ45 in the middle (through a gas discharge surge arrestor) to avoid needing to drill another hole in my house, and my system works fine even in snow melt mode, so... shrug.

The weirdest part about the rectangular Dishy system is that the router doesn't have an Ethernet port, and, to add one you have to buy a proprietary dongle because there's no Ethernet PHY for an additional port in the router itself. Plus, there's no official way to eliminate the router hardware - you can put it in passthrough mode and eliminate the actual routing part, but the proprietary PoE injector is fully integrated into the router hardware.

Things are improving slowly, though - at least there's now a "remote management" function in the Starlink app, so passthrough mode users have (proprietary) access to their connection's vitals. Previously, enabling passthrough mode eliminated all telemetry from the connection, so you couldn't see if the link was obstructed, for example.

I worked for a while with PoP-in-a-Box container systems you could drop in anywhere and run. The ammount of effort tuning each partner satellite provider to work with voice & video with ultra low codecs etc was huge overhead in addition to the cost for SCPC service/equipment. People complaining about ease of use are totally missing how hard the equivalent was not so long ago. Yes, there were simple point plug and play suitcase mobile solutions like BGAN but only at a fraction of the bandwidth of a StarLink system.

That and a bgan terminal data has pricing from multiple dollars per megabyte transferred.

>the product itself is mediocre in both cost and implementation

Getting a phased array antenna with a gain enough to get range of more than 500km for $500 is mediocre in cost? I've got curious and googled Starlink antenna teardown - man, that looks like a fighter radar hardware, may be not of a 5G fighter, yet definitely not every 4G fighter has such a radar. There is no way it costs $500 even in BOM. The BOM itself must be several thousands at least. Starlink is basically giving away its hardware in order to get customers.

Supposedly $1500 for a V1 unit, the new rectangular dish hasn't been discussed afaik. https://www.businessinsider.in/tech/news/spacex-is-spending-...

1) I was mostly referring to the monthly cost, although the upfront is steep as well.

2) As a customer I don't care if the product has a $5000 COGS to the provider or has unicorn blood in it, I care what it does. Compared to most WISPs, LTE/5G, or using the Internet access provided by a commercial campground or RV park, it's extremely expensive and really inconvenient. For example, my parents chose not to redeem their Starlink invitation as at home, they'd get similar performance to their existing WISP for over 2x the price and a really inconvenient installation process, and for the road, the cost just didn't justify the rare occasion they were at a campsite without LTE or commercially available Internet nearby.

Compared to geostationary and existing constellation satellite solutions, Starlink is a massive winner, and competes well. But, those markets were already shrinking and IMO Starlink will eventually have to compete with terrestrial wireless solutions as well. To do that, they'll need a better cost and CPE model.

There is already an aviation and Maritime mobile antenna unit in beta test. I've seen photos of it. You can also see the prototypes of the flush/conformal mount aviation terminal stuck to the side of some of the starship prototypes.

The units that exist right now are meant to be relatively low cost and compete with consumer geostationary vsat terminals.

And v1 supposedly had a COGS of $1500 https://www.businessinsider.in/tech/news/spacex-is-spending-...

That is okay for a business

The rate at which SpaceX does capital raises suggests otherwise.

See: https://www.cnbc.com/2022/05/22/elon-musks-spacex-looks-to-r...

It's going to be interesting to see in what shape Space X survives the end of the "free money" era.

Considering their current position as a launch provider (and independent of Russian rockets/engines), I'd say they're going to do just fine.

Let's say they make 30 million profit from each Falcon 9 launch. Last year they did 31 launches, but half of those were Starlink launches. That's 15x30 = $450 million profit off Falcon 9, plus whatever they made from Dragon.

You're probably right that they would be fine as a company if they couldn't get any more money. SpaceX could operate at a decent profit and still do either Starlink or Starship at reduced pace. The "free money" just allows them to get to their goals a lot faster.

So what? At 100$/month, the initial 1000$ "subsidy" is paid off and the dish generates naked profit - internet contracts can run for extremely long, especially if there is no competition.

Cable and DSL networks renting you a shitty 50$ modem for 5-10$ a month or charging you 5$ for the privilege of using the shitty wifi AP are worse in that regard.

Absolute worst case scenario it goes bankrupt and ends up like the second corporate incarnation of Iridium under entirely new owners. Which is very much a successful and essential, mission critical business today.

A big difference is Starlink satellites are low enough that they will reenter the atmosphere in 5-10 years. So, a short shelf life limits its value as an investment.

Musk has said that for Starlink to be successful they need the launch capabilities of Starship.

Department of Defense has had a perhaps unexpected taste of Starlink’s capabilities in real warfare and they will not let them go. I expect a large infusion of cash from that direction if anything bad was on the horizon.

I mean, read what little official commentary is available from them. You can practically hear those guys salivating at what they’re seeing: unjammable, indestructible, untappable communications infrastructure with high bandwidth and 5 min setup time in the field, man portable if necessary.

You make your money on the annuity. This is like utilities 101 (see -> Verizon Wireless, AT&T Wireless, etc.)

Terrestrial fiber based ISPs doing GPON can regularly spend several thousand dollars in new build outside plant fiber construction costs to bring a single home on net, but they know they have a very good chance of getting at least ten years if not more of revenue at $85 a month from it.

Yes and that why the already not using that anymore and are now using something cheaper and very likely already have the next version in the pipeline.

>product itself is mediocre in both cost and implementation

they're not interested in competing on cost, because they have no real competition.

implementation will improve I'm sure.

I am surprised you have not raised a major issue others have highlighted, that the dish is a MASSIVE energy hog in RV terms, where electricity is a premium when you are not hooked up to services where internet is usually also available, even if it is spotty.

2nd gen dish not so bad, it's about a 65W load if not in ice melt mode. The first gen is 85W to 150W depending on if heaters are on or not. But even so , yes, the cumulative kWh from a 65W load powered up for 1 month is a lot. Probably best handled on an RV that can mount multiple 360W class big pv panels on the roof.

Hmm. Are PV cells transparent or opaque in this frequency?

Elon said a few times they're working hard to get the power draw down.

> Starlink for RVs is not designed for use while in motion.

This is really their current service, but not strictly limited to only one location cell. So, you can use it whenever you reach your destination, but it isn’t designed to be used while you’re in motion.

Though you absolutely can use it in motion, surprisingly, as long as you're moving in roughly a straight line (in rural areas).

Is the straight line caveat because the dish tries to predict motion by extrapolation?

To my best guess, yep that's exactly it. Lots of weaving turns in suburban areas seems to throw it's prediction off

I can see plenty of room for improvement to optimize it even further. They could use accelerometers/gyroscopes to detect & predict "regular" movement (boats, etc) as well as current position to predict turns for land-based vehicles (from map data) so that the antenna can essentially "lead" the predicted movement (and maybe even do some magic with TCP flows to prepare them for predicted dropouts so they pick back up faster once the link is reestablished).

Yeah, I asked it because if that was the reason then it seems fixable.

In practice I'm not so sure it's such a bad thing for it to not be permanently mounted on the roof because of the importance of avoiding obstructions. Imagine your campsite is under a tree, but there's a clearing 50 feet away.

Though I did see one clever installation under a cover on a spare tire on the roof (for the round Dishy). You could leave it up there and pull it down as needed to avoid obstructions when parked. The cable would even coil up nicely in the tire.

Reply on news.ycombinator.com