Hacker News Reader

Comment by miked85 | Hacker News Reader

BMW recently sent out an email of impending loss of 3G connectivity.

> The decision to phase out 3G network technology was made at the discretion of the respective cellular carriers and lies beyond the control of BMW. As a result of the sunset of 3G service by wireless carrier partners, by February 2022 your vehicle will no longer be able to receive any ConnectedDrive/BMW Assist services, such as BMW Assist eCall, Advanced Real‑Time Traffic Information, Remote Services and BMW Online, depending on your BMW model.

View on HN

The loss of 3G is going to be an unmitigated disaster it seems. Might even be appropriate for the FCC to step in and force them to keep it up.

A lesser loss is that 3G often works in a big crowd where LTE is overloaded. Maybe that'll at least change to 5G vs LTE over time.

That's not the case everywhere. Depending on your network's configuration, it's entirely possible to have areas of high traffic where 3G service is so poor even 4G is miles ahead. This is because networks may decide to provides lots of spectrum to 4G whereas the spectrum available to 3G has be reduced to the bare minimum they need to operate. This is the case for all three networks operating here in New Zealand. One carrier has gone from providing 30 MHz of 3G spectrum down to only 15 MHz (basically halving the capacity), meanwhile the same network provides up to 80 MHz of 4G spectrum with plans to reduce 3G usage further so it can be reallocated to 4G or even 5G. Also 4G is so much better at handling poor signal conditions—I know of several places where the rollout of 4G saw marginal singal locations improve from sporadic data (coming in bursts randomly every now and then) to actually being able to slowly browse the web anytime with data coming in at a steady stream. The difference was like night and day. I'd be glad to see 3G go but am disappointed about all the issues with VoLTE as its making the migration to 4G more difficult.

It’s insane to keep a massive portion of spectrum on this old technology to support a very small amount of people with outdated hardware.

Not just that – it complicates cell planning a lot.

As far as I know, UMTS and LTE behave very differently under load, and at least in the early versions of UMTS there was the phenomenon of "cell breathing", where the serviceable cell radius varies with network load (due to the CDMA modulation/multiple access scheme used).

Having to provide denser cells in sparsely populated areas just to avoid dropped calls during the odd traffic spike sounds like a real pain.

In my old house, on UMTS, I would have perfect signal except for 6am-4pm when the local school was there. It was fine in the summer. The local school soaked up the entire “cell” and my house was a dead spot.

It's more about these devices that aren't so easy to replace like car modems. If 3G won't be it then we need to settle on something else that can be promised to be there in the foreseeable future for the longevity of IoT/cars/whatever else.

My 2013 Ford was delivered with a 2G modem, and had a recall to upgrade it to 3G. I guess when that goes offline, the online features will be totally abandoned. The final (presumably) radio firmware already significantly reduced utility of them anyway, the system no longer wakes up for network to car requests, so might as well not make them.

Old Teslas (like 5+ years) will lose net when 3g goes. You can do a computer upgrade but I don’t think it’s cheap ($500-$1500 range offhand)

It’s really annoying this isn’t software defined for future proofing. As long as your RF hardware is robust onboard, the rest can be done in software.

Are you suggesting some 3G RF hardware can be retrofitted with software patches to be compatible with 4G networks ?

This is theoretically possible (SDR) but it's very expensive.

(Before anyone replies about $20 SDRs, those aren't even close to 4G/5G capable.)

Indeeed. Last time I looked these chips were $1300 each in quantities of 1000. Too expensive for even a luxury car.

I think the answer here is an internal USB modem or something. Easy to upgrade. Just up to the car manufacturer to commit to writing new drivers. (I assume that they don't want to do this, so are probably happy when technology changes out from under them. Everyone has to buy a new car!)

What makes this silicon so expensive?

Future-proofing puts you on the wrong side of Moore's Law. Let's say 4G requires 10x the transistors of 3G. In reality that's not a problem because 4G came out 10 years after 3G so transistor costs came down 10x or more. But if you wanted to build a 4G-upgradeable radio years early you'd have to pay for that 10x performance upfront using older and more expensive technology. This same argument applies to future-proofing other tech like video codecs as well.

- Car stuff has to be certified in order to be road legal (in Europe, ECE marked)

- Car stuff has to deal with a way wider temperature range than normal consumer stuff (-40 to +125 °C). That is the thing that makes silicon so expensive.

- Car stuff has to deal with vibrations and mechanical stress, meaning more expensive PCBs, less options for solder (remember the infamous NVIDIA solder issue of ~2010?) and connectors

- Car voltage regulation is a hot mess, at least for ICE vehicles - IIRC during engine start, voltages can drop to something like 6-ish volts and after that (or during battery disconnects/other load dumps) into 120-ish volts, and the voltage has an awful lot of ripple because usual generators are AC that's driven through a diode rectifier, with the battery acting as stabilizer.

- Most stuff in cars requires hours of work to replace it - Tesla were to my knowledge the first ones who did software updates outside of shop visits at all.

Regarding the temperature range; this procides a bit more detail and nuance:

https://electronics.stackexchange.com/questions/301826/how-t...

High complexity, low volumes, I would imagine. A few of the answers below talk about it being a car part but SDRs that can do 3G/4G are targeted at people developing cheaper chips. Electronic test gear is much more expensive than cars (at least at the bleeding edge), so the pricing doesn't preclude it from being used there.

the radio cannot be software defined (and user modifiable) for type acceptance and regulatory reasons

In what country? The FCC certainly allows SDR to be certified. "User modifiable" is sufficiently broad as to be nearly meaningless.

https://softwarefreedom.org/resources/2007/fcc-sdr-whitepape...

Here's the actual text on restrictions for SDR: https://www.law.cornell.edu/cfr/text/47/2.944

The FCC defines SDR here extremely broadly, any radio which "frequency range, modulation type or maximum output power (either radiated or conducted)" can be changed.

This describes a substantial portion of radios - including many if not most WiFi access points. They are legal and many US companies design, manufacture, and/or sell these types of radios (I work for one of them).

Absolutely they are legal to sell, but you have to go through type acceptance under part 15, and then you need to make it so the user cannot modify them.

“ The rules require any manufacturer certifying a device under the new process to take steps to prevent “unauthorized” changes to the software on the device that might alter its radio frequency and power parameters in a way that takes it out of compliance with the regulations known as FCC Part 15 regulations.”

Test equipment is exempt because it is not licensed as an intentional emitter under part 15.

$500 to $1500 sounds relatively cheap for a Tesla repair since a battery issue can be in the tens of thousands (because they want to replace the entire battery) and even relatively minor body damage can be in the thousands or the tens of thousands.

They really should have used an easily replaceable mPCIe cellular module. $50 replacement cost. Interface has been around since 2005.

And $1450 labor.

I have actually found Tesla labor prices very reasonable.

Car manufacturers don't even try to pretend that their prices have anything to do with the actual costs of components. You want an "original X" part? That will be $$$$, thank you very much.

Just because the old modules were mPCIe doesn't mean a new one will be too.

I have an Thinkpad T430s, that came with mPCIe 3G modem; for LTE, it was possible to find a replacement, but for 5G, I don't expect there will be any (any whitelists in firmware aside).

Or similarly, I have an openwrt router with mPCIe 802.11-ac wifi (Turris Omnia). It is not possible to replace them with -ax adapters, they all have m.2 interface.

With cars and other long-used goods I would expect the same problems.

Sure, most have USB-only connections and some have actual PCIe connections (I suppose). And then there's the fragmentation at the software/AT command level.

Still, Tesla could have spent the software effort required to make it work with a few of the most common 3G/4G/(5G) modules from reputable manufacturers.

This is apparently a photo of the easily replaceable card that it's on: https://www.ebay.com/p/3035604932

I'm not sure how you plan on taking apart the dashboard for less than $50 of labor, though.

It doesn't have to be placed in the dashboard.

So you're saying that existing cars, where it's in the dashboard, can be modified for almost nothing to put this card someplace else?

Edit: I mean, hindsight is great when you're designing a new car, and Tesla did make it much easier to replace this card in newer models, but we're talking about old cars. 75 minutes of labor for an S, 90 for an X.

> So you're saying that existing cars, where it's in the dashboard, can be modified for almost nothing to put this card someplace else?

Of course I'm not saying that. Please re-read my post if necessary.

Edit: This is a design deficiency, is what I'm saying.

Isn't hindsight awesome? Even though your first bit of advice was implemented on day 1 (the modem is on a separate card), it's still expensive to replace it.

Tethering off your phone seems simpler, especially if you got their app running.

That's not accurate. You don't need to upgrade the MCU/infotainment computer and the retrofit is only $200.

"A source has informed me that due to AT&T shutting down its 3G network by Feb 2022, Model S's without an LTE-capable modem will lose connectivity features unless connected to WiFi. HOWEVER, Tesla will offer retrofits for $200. Affected vehicles are pre June 2015."

"Tesla will be offering retrofits for these vehicles for a bargain price of only $200, even though the vehicles affected by this could be up to nine years old."

Source:https://twitter.com/SawyerMerritt/status/1415851250055323648

See https://www.tesla.com/support/update-3g-services

They're going to upgrade the modems to LTE for $200 each.

I really appreciate my 2011 Civic's lack of a cellular modem.

Ditto with our 2003 Citroën. This always online paradigm will cause more and more problems down the road, as the first IoT generations lose Internet access.

Do you? Because when it's -10F here, I can remote start my car, even if outside remote start range. Rather nice to always have a toasty car (or cool in summer).

Okay that's good. Last I heard was that they stopped doing daughter board upgrades, and you had to swap your entire computer. Things like this tend to change pretty fast/often in Tesla land, so can't always keep up.

(Also don't have a 2014 Model S so doesn't impact me personally)

I'm getting bit by this too, so I've been keeping an eye on possible workarounds. People are finding ways doing upgrades that BWM is mostly saying are impossible: https://www.reddit.com/r/BMWi3/comments/of67m0/i_had_a_3g_to...

I was going to say, this is going to be awful for so many devices. There's still quite a few areas along I5 that only get 3g.

Subaru upgraded my telematics unit in my car for free when 3g was ending in my area.

Reply on news.ycombinator.com