Starlink as a Real-Time Earth Observation platform.

[aero]

"What size camera equipment could fit on a Starlink satellite and what benefit could be derived from that real-time Earth observation capability?"

I could imagine a natural or unnatural disaster where real-time area observations/images were available as the emergency responders were being dispatched. TV news would love that! They could retire their helicopters and use Starlink instead.

[speedevil]

"What size camera equipment could fit on a Starlink satellite and what benefit could be derived from that real-time Earth observation capability?"

I could imagine a natural or unnatural disaster where real-time area observations/images were available as the emergency responders were being dispatched. TV news would love that! They could retire their helicopters and use starling instead.

Starlink currently is ~550km up.
Neglecting concerns about atmosphere, as one example, the Nikon P1000 camera has 125* zoom, and weighs about a kilo, neglecting the user interface parts.


For example - this would produce an image of the ground with pixels of the order of 20m, with a camera package probably in the several kilo class.

The Mars HIRISE camera weighs in at 65kg. https://en.wikipedia.org/wiki/HiRISE
This would be more like half a meter pixels.

65kg is in the range where you probably can't put it on starlink, so that is an upper limit.

[Ludus]

Planet labs Dove cubesats are 4 kilograms, 10 cm × 10 cm × 30 cm, orbit at about 400 km and provide imagery with a resolution of 3–5 m. That’s the whole cubesat including power and communications so just the imaging elements obviously are less. It’s also a more than 5 year old design but it’s proven for satellite operations and intended for production runs of satellites.

[meekGee]

Planet labs Dove cubesats are 4 kilograms, 10 cm × 10 cm × 30 cm, orbit at about 400 km and provide imagery with a resolution of 3–5 m. That’s the whole cubesat including power and communications so just the imaging elements obviously are less. It’s also a more than 5 year old design but it’s proven for satellite operations and intended for production runs of satellites.

The key here is "entire satellite".

Once you already have power (PV and storage), communications, attitude control, thermal control..  The imaging subsystem itself is only a fraction of an entire imaging satellite.

How small a fraction? I think it really depends on the optics design, and the hardest thing about that may be to conform to the tight packaging guidelines.

I can imagine maybe a main optic in-plane with the pizza box, and an imager that pops out of plane as part of deployment.

[aero]

Planet labs Dove cubesats are 4 kilograms, 10 cm × 10 cm × 30 cm, orbit at about 400 km and provide imagery with a resolution of 3–5 m. That’s the whole cubesat including power and communications so just the imaging elements obviously are less. It’s also a more than 5 year old design but it’s proven for satellite operations and intended for production runs of satellites.

That is getting there, but the resolution is still about an order of magnitude low. On the other hand, the dimensions seem to me to make it a lot smaller than a Starlink satellite. How far has camera resolution advanced in the last 5 years?

[Comga]

Planet labs Dove cubesats are 4 kilograms, 10 cm × 10 cm × 30 cm, orbit at about 400 km and provide imagery with a resolution of 3–5 m. That’s the whole cubesat including power and communications so just the imaging elements obviously are less. It’s also a more than 5 year old design but it’s proven for satellite operations and intended for production runs of satellites.

The key here is "entire satellite".

Once you already have power (PV and storage), communications, attitude control, thermal control..  The imaging subsystem itself is only a fraction of an entire imaging satellite.

How small a fraction? I think it really depends on the optics design, and the hardest thing about that may be to conform to the tight packaging guidelines.

I can imagine maybe a main optic in-plane with the pizza box, and an imager that pops out of plane as part of deployment.

My emphasis
Even if 100% of the Dove's were the optical system, it would be small compared to the Starlink satellite.
However, the issue is the attitude control, as highlighted above. 
Owning the satellite, Planet can point the Dove at targets.
A camera fixed to a Starlink satellite would not do this.
A pointing mechanism could be as expensive and large as the imager.
It might even have limits placed on it for torque imparted to the bus, making it more complex.
Not simple

PS Discussions of satellite resolution are also not simple.  There is the diffraction limit for any particular size optic.  There is the oft quoted Ground Sampling Distance (GSD).  Then there is the sharpness, generally quantified as the Modulation Transfer Function (MTF) at the highest resolvable spatial frequency, the Nyquist frequency, corresponding to two GSDs.  There are also issues of dynamic range and stray light, but Planet's Doves are commercially useful as is.

[meekGee]

Planet labs Dove cubesats are 4 kilograms, 10 cm × 10 cm × 30 cm, orbit at about 400 km and provide imagery with a resolution of 3–5 m. That’s the whole cubesat including power and communications so just the imaging elements obviously are less. It’s also a more than 5 year old design but it’s proven for satellite operations and intended for production runs of satellites.

The key here is "entire satellite".

Once you already have power (PV and storage), communications, attitude control, thermal control..  The imaging subsystem itself is only a fraction of an entire imaging satellite.

How small a fraction? I think it really depends on the optics design, and the hardest thing about that may be to conform to the tight packaging guidelines.

I can imagine maybe a main optic in-plane with the pizza box, and an imager that pops out of plane as part of deployment.

My emphasis
Even if 100% of the Dove's were the optical system, it would be small compared to the Starlink satellite.
However, the issue is the attitude control, as highlighted above. 
Owning the satellite, Planet can point the Dove at targets.
A camera fixed to a Starlink satellite would not do this.
A pointing mechanism could be as expensive and large as the imager.
It might even have limits placed on it for torque imparted to the bus, making it more complex.
Not simple

PS Discussions of satellite resolution are also not simple.  There is the diffraction limit for any particular size optic.  There is the oft quoted Ground Sampling Distance (GSD).  Then there is the sharpness, generally quantified as the Modulation Transfer Function (MTF) at the highest resolvable spatial frequency, the Nyquist frequency, corresponding to two GSDs.  There are also issues of dynamic range and stray light, but Planet's Doves are commercially useful as is.

Attitude control means either propellant or reaction wheels.  That's a big subsystem you get for free.
You still need pointing control of the optic.  But pointing is meaningless without a controlled platform to start off from.

Fine pointing can be done with minute motion of the sensor within the imaging plane.  (subject to optical design of course).

You can still stow the main optic within the main plane of the satellite, and gimbal it from there.
I consider pointing control a part of the optical subsystem that has to be added..

[whitelancer64]

Planet labs Dove cubesats are 4 kilograms, 10 cm × 10 cm × 30 cm, orbit at about 400 km and provide imagery with a resolution of 3–5 m. That’s the whole cubesat including power and communications so just the imaging elements obviously are less. It’s also a more than 5 year old design but it’s proven for satellite operations and intended for production runs of satellites.

That is getting there, but the resolution is still about an order of magnitude low. On the other hand, the dimensions seem to me to make it a lot smaller than a Starlink satellite. How far has camera resolution advanced in the last 5 years?

There are already satellite constellations with very close to real-time imagery available, currently in use by disaster management agencies and first responders.

Part of the problem (as you have already noticed!) with "retiring helicopters" is the image resolution. With a close-in helicopter (or drone, etc), your resolution is always going to be better than what you can get from orbit.

[Semmel]

Discussions of satellite resolution are also not simple.  There is the diffraction limit for any particular size optic.  There is the oft quoted Ground Sampling Distance (GSD).  Then there is the sharpness, generally quantified as the Modulation Transfer Function (MTF) at the highest resolvable spatial frequency, the Nyquist frequency, corresponding to two GSDs.  There are also issues of dynamic range and stray light, but Planet's Doves are commercially useful as is.

To fix the optical parameters, one best stars with the intended ground resolving power. Lets say 0.5m as minimal resolvable element, which would allow for identifying the presence of cars on a street or parking lot for instance. This could be very useful for traffic control, and autonomous driving, for instance to identify traffic jams far in advance and to find a suitable parking spot near a destination (obviously not the driving it self).

0.5m at a distance of 550km is arctan(0.5/550e3) = 9.1e-7 rad = 0.19 arc seconds. The diffraction limit for a circular aperture with diameter D is its angular resolution a =1.220*lamda/D (in radiants), with lambda is the wavelength of the light. Or, given our 0.5m, and say green light of about 530nm requirement:
D = 1.220*530e-9/arctan(0.5/550e3) = 0.7m. A sizable telescope. If you add all the other components and limitations like optical PSF and so on, its comes down to about a meter. Thats a full blown spy satellite type of optic. You can say, ok.. maybe one doesnt need 0.5m to identify empty parking lots, 1m is enough since a car is about 2m wide. So then its a 0.5m telescope. You can play a bit with the wavelength, say one goes to 450nm, which gets you a bit further and you are in the order of optically 0.3m but realistically more like 0.35m.

There are relatively compact amateur Schmidt–Cassegrain telescopes of that size in the 50kg range including a camera. You probably need motors for mirror alignment and other stuff for this to work, but its not impossible. I would say this is sort of the practical limit of what can be done.

In terms of pointing, I would not worry about that. With that many starlink satellites, choose the field of view such that successive satellites slightly overlap their ground coverage. That way you get a full coverage of earth every time earth rotates by the pitch of one orbital plane. In case of starlink, that is initially 24 orbital planes (if I remember correctly), so roughly one global image per hour.

[armchairfan]

1m [per pixel] is enough since a car is about 2m wide. [...]

In terms of pointing, I would not worry about that. With that many starlink satellites, choose the field of view such that successive satellites slightly overlap their ground coverage. That way you get a full coverage of earth every time earth rotates by the pitch of one orbital plane. In case of starlink, that is initially 24 orbital planes (if I remember correctly), so roughly one global image per hour.

my emphasis
Doing some first order calculations at the equator:

40e6 m earth circumference / 24 / 1 m/pix = 1.7 million pixels across. That's a heck of a sensor!

Similarly, 40,000 km / 24 = 700 km per satellite at the equator. Hand-waving instead of calculating, 550 km up with a 700km footprint requires about a 60 degree field-of-view. That's a huge field-of-view for a high res camera.

And now for something completely different ... I'm not a scientist but an atmospheric sounding sensor might be handy for weather prediction and other atmospheric modeling. No doubt we have better sensors deployed now but presumably not on a scale like Starlink. I'm wondering if this might be a case where "quantity has a quality all it's own."

[Asteroza]

Starlink already will be accommodating telescopes, in the form of lasercomm terminals. Which means the terminal is notified of spacecraft orientation and terminal basic stabilization is the spacecraft itself. Fine guidance is up to the terminal, telling the host spacecraft it's mass movement for bulk stabilization control. So, your basic starting point would be a crosslink lasercomm terminal. Your next limiter is packing space limitations, which puts a limit on the size of non-deployed components.

If you want a telescope bigger than the lasercomm terminal, you need to deploy optics from a stowed configuration. Something similar to Falconsat-7 with deployable membranes on triangular pantograph structures seems like a possible path.

[Comga]

Discussions of satellite resolution are also not simple.  There is the diffraction limit for any particular size optic.  There is the oft quoted Ground Sampling Distance (GSD).  Then there is the sharpness, generally quantified as the Modulation Transfer Function (MTF) at the highest resolvable spatial frequency, the Nyquist frequency, corresponding to two GSDs.  There are also issues of dynamic range and stray light, but Planet's Doves are commercially useful as is.

To fix the optical parameters, one best stars with the intended ground resolving power. Lets say 0.5m as minimal resolvable element, which would allow for identifying the presence of cars on a street or parking lot for instance. This could be very useful for traffic control, and autonomous driving, for instance to identify traffic jams far in advance and to find a suitable parking spot near a destination (obviously not the driving it self).

0.5m at a distance of 550km is arctan(0.5/550e3) = 9.1e-7 rad = 0.19 arc seconds. The diffraction limit for a circular aperture with diameter D is its angular resolution a =1.220*lamda/D (in radiants), with lambda is the wavelength of the light. Or, given our 0.5m, and say green light of about 530nm requirement:
D = 1.220*530e-9/arctan(0.5/550e3) = 0.7m. A sizable telescope. If you add all the other components and limitations like optical PSF and so on, its comes down to about a meter. Thats a full blown spy satellite type of optic. You can say, ok.. maybe one doesnt need 0.5m to identify empty parking lots, 1m is enough since a car is about 2m wide. So then its a 0.5m telescope. You can play a bit with the wavelength, say one goes to 450nm, which gets you a bit further and you are in the order of optically 0.3m but realistically more like 0.35m.

There are relatively compact amateur Schmidt–Cassegrain telescopes of that size in the 50kg range including a camera. You probably need motors for mirror alignment and other stuff for this to work, but its not impossible. I would say this is sort of the practical limit of what can be done.

In terms of pointing, I would not worry about that. With that many starlink satellites, choose the field of view such that successive satellites slightly overlap their ground coverage. That way you get a full coverage of earth every time earth rotates by the pitch of one orbital plane. In case of starlink, that is initially 24 orbital planes (if I remember correctly), so roughly one global image per hour.

0.5m GSD is not needed to count cars
That's the Rayleigh criterion, which is only loosely applicable to ground resolution.  Certainly not to three significant figures.  It's more complicated than that.
It's not the pointing of the satellite.  It's the pointing of the line of sight.
One global image per hour is enormous bandwidth

[Lar]

One global image per hour is enormous bandwidth

Conveniently, the system has enormous bandwidth capacity...

[matthewkantar]

One global image per hour is enormous bandwidth

Conveniently, the system has enormous bandwidth capacity...

I guess the question is weather it is better to sell the bandwidth or use it for Earth observations. Seems like it might be a different trade depending on what the sat is located above at any given time.

[Lar]

One global image per hour is enormous bandwidth

Conveniently, the system has enormous bandwidth capacity...

I guess the question is weather it is better to sell the bandwidth or use it for Earth observations. Seems like it might be a different trade depending on what the sat is located above at any given time.

Sure. But I'm not seeing this straining the system at all. And if it does, just put up more birds.

Also "weather" ...  LOL ... good segue to

If one could stop tornados, there would be an annual benefit of $10+ billion in saved expenses from damages.  Saved costs due to preventing damage from hurricanes could be an order of magnitude larger.  It doesn’t seem likely the damage costs equate to what is spent to recover (in other words, this money is going somewhere else otherwise everyone in Florida should be a millionaire by now).

An earth observation system like what's being mooted can't stop tornadoes. It probably isn't even high frequency enough to do good at warning against them. They come up fast and move fast.

An earth observation like what's being mooted can help in tracking hurricanes, but we are already pretty good at tracking them with our weather satellites. Improving tracking and intensity prediction can help some.

But in both cases building things to resist the effects better (and convincing people to evacuate, or not evacuate, as appropriate) has a higher payoff I think.

[whitelancer64]

If one could stop tornados, there would be an annual benefit of $10+ billion in saved expenses from damages.  Saved costs due to preventing damage from hurricanes could be an order of magnitude larger.  It doesn’t seem likely the damage costs equate to what is spent to recover (in other words, this money is going somewhere else otherwise everyone in Florida should be a millionaire by now).

$10 billion divided by 21 million (the population of Florida) is $476.19

Let's say a town with a population of 50,000 is completely demolished by a hurricane. $10 billion divided by 50,000 is $200,000 each.

Ten billion sounds like a lot of money, but spread out over tens of thousands of claims, it's not very much.

[Ludus]

One global image per hour is enormous bandwidth

Conveniently, the system has enormous bandwidth capacity...

I guess the question is weather it is better to sell the bandwidth or use it for Earth observations. Seems like it might be a different trade depending on what the sat is located above at any given time.

It might be considered a lower priority use of capacity rather like deploying Starlink itself can soak up unused launch capacity that SpaceX would have anyway as a consequence of rapid reusability. Clients like financial institutions willing to pay a big premium for guaranteed lowest latency (after intersat laser links are up) would get top priority, other data customers next and internal uses like this last. Large areas of the network wouldn’t have much priority traffic at any given time.

[speedevil]

Large areas of the network wouldn’t have much priority traffic at any given time.

Storing the image for even a few minutes dramatically improves the likelyhood of actually unused bandwidth being available.
As does sending the data in 'silly' directions - hopping between twenty satellites that are unused as they're all over ocean before getting to an underutilised ground-site, rather than fighting with priority data.