unmodified LTE and 5G seem way too verbose to be able to handle the constraints of a much higher number of devices per base station
That's a matter of pricing, isn't it?
Also, won't this primarily be used where very few people roam? Otherwise they could just use the regular land-based LTE infra.
Remember: SpaceX can't just magically offer cheaper mobile coverage globally; they don't own any rights to use 4G/5G frequency bands. These are typically very expensive.
It makes a lot of sense for SpaceX to partner with mobile operators who own these rights but don't utilize them very well in sparsely inhabited areas.
Not necessarily: Users don't pay for signalling traffic.
I could imagine even just rejecting attachment requests of thousands of roaming clients could easily saturate a single satellite cell. Existing mobile-to-satellite protocols are much more optimized in that regard (and limited in that there are simply not that many clients out there looking for signal).
Sure, but ignoring doesn't solve the problem of signalling channel saturation – if there is something to ignore, it means that some scarce resource has already been consumed.
That's fine if you need to tell a few terrestrial clients to go away every few minutes or so (until they finally give up until the next reboot or so), but I could imagine it being much more tricky when you're talking about cell sizes with a diameter of hundreds of kilometers.
> wouldn't this primarily be activated in very desolate areas?
It would be enabled globally (to fill in coverage gaps etc), I'd assume – and even in very remote areas, you might easily have a sizeable number of devices per square kilometer. A typical cell size (for Iridum, as an example) is 250 miles of diameter – that's the area of New York state! And that's not even considering overlaps of remote and densely populated areas.
To summarize (please correct me if I'm wrong here):
You see this being deployed to service every customer of the partnering mobile networks, wherever they may be.
I see this being deployed to service those customers who aren't covered by existing and much cheaper land-based LTE-coverage from the partnering mobile networks.
My concern is really of a more technical nature: 2G-5G are somewhat "chatty" on their signalling channels, as far as I understand.
Even when a given network does not support your operator/SIM card, your device will attempt to connect to it at least once (and hopefully cache the resulting rejection permanently and not keep trying in a loop).
But in case your device does support the satellite-based service (e.g. as a T-Mobile customer in the US), it would then attach to the satellite network if no terrestrial service is available and send periodic location updates etc.
That's normally not a problem, since cell sizes on earth are limited (either by geometry/physics or intentionally by network capacity planning), and by extension also the ratio of signalling chatter per cell.
I'm just not sure if unmodified LTE clients can be convinced to be less chatty only from the network side.
That's a matter of pricing, isn't it?
Also, won't this primarily be used where very few people roam? Otherwise they could just use the regular land-based LTE infra.
Remember: SpaceX can't just magically offer cheaper mobile coverage globally; they don't own any rights to use 4G/5G frequency bands. These are typically very expensive.
It makes a lot of sense for SpaceX to partner with mobile operators who own these rights but don't utilize them very well in sparsely inhabited areas.