Also seems likely that if they can find a way to make small sats with the ability to accurately point a laser at another sat, that would have pretty obvious implications to the defense dept. SpaceX is a military contractor after all.

Here's the design 101 from base principles:

In practice, you need a big-ish telescope on both ends to control beamspread and to collect light from a big aperture. Aperture is a given based on link budget (and you can trade off power to make the aperture smaller, but halving the aperture diameter on both sides means you'll need 16x the power); if you target F/2 then it'll be twice as long as the aperture. Maybe think about a 6cm aperture and 12cm long telescope for a starting point. This will get you a 6 arcsecond wide beam @ 1000nm.

Then, you need to slew this at pretty fast rates-- perhaps 15 degrees per second for acquisition, and control the pointing within 3 arcseconds while tracking at peak rates of a few degrees per second. Yeek! This pretty quickly takes you towards some kind of direct drive fork mount that is very gimbally-looking.

One bit of fun is that you need to have a lot of bandwidth on your reaction control system on the spacecraft, too-- because when you snap one of these telescopes around, the whole craft is going to want to counterrotate, so the reaction control wheels (and/or other telescopes for links in other directions) will need to react. Feedforward is advised.

You can't get the same degree of directivity. As wavelength decreases, you get more directivity for a given aperture. Light has 1/5000th the wavelength of plausible radio links, so both the sender and the receiver can have much higher gains. You also can have much more bandwidth, and thus you obtain many orders of magnitude higher data rates per unit of power used.

E.g. a 6cm telescope has 82dB of gain on each side for 1000nm light.

A 1 meter aperture (about what a 3.2x1.6x0.2m Starlink satellite can likely present to another satellite) has 53dB of gain on each side.

So for equivalent power, you have 6 orders of magnitude more signal strength, and you can occupy 10x the bandwidth, too, even if you have a very large phased array.

> The only reason you'd go for lasers here is thin civilian cover for developing a weapons platform.

This kind of system has very little in common with how I would build an anti-satellite laser system.

?

> if it wasn't there would be existing examples.

Let's see. NASA downlinked from the moon to the ground, through the atmosphere at OC-12 rates back in 2013-- so about 100x the distance, with the added penalty of traversing the atmosphere. NFIRE did 5.6 gigabit/sec LEO to ground (again through the atmosphere) in 2011-- shorter distances but higher angular rates which is the "hard part". And EDRS does 1.8gbit/sec over longer distances in geostationary orbit. Both flown and proven.

I'm comparing to LEO to ground, which has a higher rate of angular movement (e.g. harder to point at) than LEO-to-LEO in the same shell, among other things.

I've built systems that point to sub-arcsecond precision at satellites in LEO. It's not quite an off-the-shelf controls problem (e.g. good luck getting a COTS motion controller to hit-a-fast-moving-target-at-a-chosen-time, rather than follow a track and not care about time) but it's not super hard, either.

Given that there's systems that have successfully flown doing links from GEO to LEO (e.g. high angular rates again), using several year old conservative technology, it's not so bad.

There's a million little details, of course. Just conduction cooling for fast optical transceivers is going to be annoying in space, for instance.

I don't understand what this means. If you mean Facebook-- Facebook didn't drop billions of dollars in to free space optical comms.

> GEO to LEO is not LEO to LEO

Yes, GEO-to-LEO is worse in every way (assuming the LEO satellites are in the same inclination and have the approximate same orbital period):

* Longer link distances. (More path loss, worse link budgets)

* Higher peak angular rates for pointing.

LEO-to-Earth is mostly worse:

* Shorter link path, but atmospheric dispersion (More path loss, worse link budgets, plus things like multipath).

* Higher peak angular rates for pointing.

* Less demanding pointing precision due to shorter path, though.

The hard part isn't the optical comms in space. The hard part is fitting multiple precision-pointed transceivers into a tiny volume and mass budget.