> well 3 degrees wide means at best ~28 square degrees on the sky (pi * 3 * 3).
If you use a diameter for a circle area, you need to divide by 4, to get like ~7. I was just going to do 3*3 = 9.
> It looks like Starlink has 7 GHz total downlink bandwidth from satellite to customers. A cell may be ~300 sq km, but no way you can pipe all that bandwidth in adjacent cells due to interference
Both the downlink and the uplink are directional. Yes, there's some power lost in side lobes, but..
So, if you have multiple satellites overhead, you can have a mosaic pattern where satellite A illuminates a hex, and then other satellites illuminate its neighbors. And even though some energy spills over, it's strongly rejected by the receivers because it's coming from the completely wrong direction.
> How many people live in places with fewer than a few people/km^2?
In addition to the incorrect pessimism about downlink bandwidth (also, one should expect ~5bits/Hz)... You're not taking into account any reasonable capacity factor, and you're assuming each service is one person. You're also assuming that you sell through to 100% of the populace and there's no one in the cell served by other technologies. If 95% of a suburban area can get FTTH, cable, or reasonable LTE, that still leaves 5% as possible customers for Starlink to fill in.
> Only Alaska averages below that in the US.
OK, but people are concentrated into cities, leaving a whole lot of land area with concentrations below this even in states with much higher average population density.