To gain a better understanding of why this needs to be totally unique/custom/proprietary, it is helpful to first have a thorough understanding of current high capacity dedicated geostationary orbit based systems (1:1 SCPC/MCPC with dedicated kHz) and various shared-bandwidth VSAT type systems (TDMA timesliced architecture between one large earth station, one piece of satellite transponder kHz, and a number of N fixed CPE terminals within the satellite's spot beam.
After understanding current geostationary architectures, dig into the "How" and "why" o3b was created and has been such a success, and its general system architecture which is proprietary.
Satellite engineer here:
The OSI layer 1/2 needs to be totally custom because we're dealing with a unique architecture of, just off the top of my head:
a) dual satellite LEO architecture
b) possible satellite-to-earth station trunk links, and satellite-to-satellite
c) CPE terminals that have no moving parts and use phased array antenna systems to talk to two satellites at the same time. From the stationary point of view of a rooftop CPE, the satellite that is currently "rising" from the horizon and will be soon overhead, and the satellite that is currently overhead and will soon pass out of sight.
d) Densely packed high frequency spot beams on a moving LEO satellite. The closest thing that's ever been built to this before is again the o3b satellites, but there are a great deal fewer of them, they orbit much higher, and have much larger spot beams than these small LEO high-Ka/V-band satellites will have.
e) Custom indoor modem RF tech to talk to the rooftop CPE and provide a standard 100/1000 copper ethernet handoff (and possibly integrated 802.11ac dual band wifi). TDMA timeslicing per CPE and bandwidth allocation - there is no way that an individual CPE will get 1:1 dedicated bandwidth 24x7x365, the amount of capacity in an individual spot beam sized area will be oversold based on standard network architecture principles that most people don't try to max out the capacity of their circuit 24x7.
- Does it make sense to make a phone that acts as a receiver (assuming the phone is outdoors), because it's one fewer device to buy and power, or does it bring down the cost significantly to have a fixed unit installed on your terrace?
- If we want to provide remote regions with Internet access, would you go with Starlink or O3b? Would your answer change if the requirement was affordable internet access, no matter the speed?
- Under what conditions can Starlink or O3b compete with terrestrial internet in cities?
Basically, I'm trying to understand the impact these programs might have on the world, not the technicalities.
For your second question, both are suitable, but in different markets. O3B is intended to replace high cost Ku and C-band transponder capacity for telecoms and ISPs that are in a place economically impossible to reach with PTP microwave or fiber. The smallest terrestrial o3b terminal is a pair of 1.8 meter motorized tracking antennas semi-permanently installed on concrete pads. O3B gives an ISP one big fat pipe back to the terrestrial internet in somewhere with good connectivity, and then that ISP can distribute service around their region using whatever technology they want to use, or have access to (PTP microwave, point to multipoint, fixed LTE, DOCIS3.0, various types of FTTH, etc).
o3b only functions at latitude +/-45. Starlink and similar systems will be in polar orbits, so with network architecture for satellite-to-satellite relay, there is the possibility of truly global coverage (as Iridium service works equally well at the north pole, in remote parts of Nunavut, at the south pole, in Tierra del Fuego, etc).
It will not be competitive in cities against terrestrial based infrastructure. Given advancements in how much data can be shoved through old copper (DOCSIS3/3.1, VDSL2 30a, g.fast) and various active ethernet FTTH and GPON FTTH, those will have much greater throughput. A dozen starlink satellites will have less aggregate data throughput than what it is possible to push down two strands of singlemode fiber.