SpaceX Internet Satellite Network: Starlink

[bxr140]

What are the spherical objects? And what size are they? Can’t get a sense of scale from the photo.

They're the gateway antennas, and IIRC they're 1m diameter. The spheres are just radomes so the antenna stays clean and its range of motion is clear (they can swing at least a full 90 degrees elevation and 360 degrees azimuth) as the antenna tracks a satellite overhead. Depending on the dynamic geometry between the spinning earth and the orbiting satellite, the satellite is only in view of the gateway for typically many minutes, maybe up to usually 15-20 for the high altitude shells--so you can imagine the antenna has to move reasonably fast. Each radome only tracks one satellite, so each gateway site will have a ton of antennas to track the many sats in view at any one time.

Here's a really simple diagram of the way the links work--you can imagine the satellite moving overhead of the fixed feeder and UTs and the angles constantly changing among them. Without ISLs a satellite needs covisibility with the gateway and the user(s), above whatever elevation threshold is set--often feeder elevation angles can go really low (like 10 and even below), and I think the minimum angle for the Starlink UTs is like 50 or something (?). That's a pretty high number for the UTs, but its ok because there's a gazillion satellites so they don't have to look super low like one might with, say, Iridium. That also makes it easier to have a phased array for the UT since they're, for the most part, less efficient the farther from "up" the signal is. A phased array is a good thing because it is just (more or less) a circuit board with some antenna elements on the top, rather than some complicated motorized mechanism (= cheaper and more reliable), and not going way down in elevation for the UT makes it easier to maintain a strong link--especially at higher frequencies, the more atmosphere the signal has to pass through the weaker it gets (there's also more ranging loss, higher doppler, more chance of occlusion, etc). And bottom line, the weaker the signal is, the worse your internet service is.

 

[dpeilow]

Familiar looking diagram in that post

 

[bxr140]

Familiar looking diagram in that post

Both are 100% random Googs, if it matters. I figured radomes and two way satellite communications aren’t exactly unique or cutting edge concepts, so I wasn’t picky about which ones I screenshot.

 

[dpeilow]

RCST and NCC are specifically DVB-RCS terms.

 

[bxr140]

Lol, fair play.

They’re also terms that are like 25 years old.

 

[dhanson865]

I think the cells shown on this tool are arbitrary but it's a decent tool to give you a rough idea how much of the time you have coverage now.

https://sebsebmc.github.io/starlink-coverage/index.html


For example the cell that includes my home says it is covered on average 994 minutes of 1440. This is approximately 69% of the day.

Seattle area cell says 1348/1440 93.6% of the day. Probably a good bet that Washington ST will be in the first wave of beta.

Here is a partially annotated screenshot of North America. Might be interesting to compare vs current numbers each week to see how this improves.

 

[dhanson865]

using launch dates compared to

It looks like it takes forever for these sats to get into position.

Launch 1 Nov 11 2019, first 1/3 in place after 50 days, mostly in place by April 2020 (about 150 days), still not in final positions after (190 days).

Launch 2 Jan 7 2020, first 1/3 in place after 75 days, mostly in place after 135 days.

Launch 3 Jan 17 2020, first 1/3 in place after 100 days,

Launch 4 Feb 26 2020, first 1/3 in place after 50 days,

Launch 5 Mar 18 2020, first 1/3 in place after 75 days,

then the next 3 launches haven't been up long enough to settle down so I'm assuming the pattern continues

Launch 6 Apr 22 2020, first 1/3 in place after ?100?
Launch 7 Jun 4 2020, first 1/3 in place after ?50?
Launch 8 Jun 13 2020, first 1/3 in place after ?75?

Sats that launched over 5 months ago still seem to be working towards a final position. So even if they stopped launching today the coverage would improve over the next 6 months. But every launch that is less than 2 months apart accelerates the pace of coverage.

 

[bxr140]

I think the cells shown on this tool are arbitrary but it's a decent tool to give you a rough idea how much of the time you have coverage now.

Yeah, they're arbitrary--they're just the cell size/shape/location that the person who set up the tool came up with. Its a decently small cell size and should be sufficient for the intended purpose. SpaceX is no doubt modeling to a finer resolution, folding in population density, sat capacity, etc.

Here is a partially annotated screenshot of North America. Might be interesting to compare vs current numbers each week to see how this improves.

Yeah. We'll see that the %'s start pushing down in latitude. At a given latitude the %'s are going to be more or less the same, where the major variable is gateway coverage, so the %'s should push down pretty equally.

We also see the strongest band of coverage is at ~50deg or so, which is a function of the 53deg inclination. The way multi-plane constellation geometry works is that the planes are closest to each other at the high latitudes, which means the satellites all converge in those regions. (Check any image/animation of the constellation for a visual) That's why the strongest band of coverage is and will always be in the high 40s-low 50s...and since constellation geometry is symmetric about the equator, the same strong band is in the Southern Hemisphere too.

Conversely, we see the equator is the worst coverage zone. That's because the planes are farthest apart at 0deg and so there's a ton of ground coverage gaps between each plane. On the global view of the tool the various color bands that ebb and flow between stronger and weaker coverage (as opposed to what might intuitively be a constant degradation in coverage from high to low latitudes) are a function of the specific constellation geometry and where there are secondary nodes of convergence, but in general coverage gets worse as you go from high to low latitudes.

The rollout strategy to release service in these high latitude regions (in North America, where the gateways are) makes a lot of sense, because even if there's only a few planes operational (say, every third or fourth plane of the final number of planes) the satellites still all converge close enough in those high latitude regions to provide useful continuous coverage. As further launches fill in the 'tweener' planes, those overlapping coverage bands will, as noted above, start to push down in latitude.[/QUOTE]

 

[bxr140]

Starlink-123 GIF | Gfycat

You can clearly see the convergence of the sat ground coverages at the latitude extremes. Looks like they're fixin' for something like 20 planes to start.

 

[dhanson865]

Starlink-123 GIF | Gfycat

You can clearly see the convergence of the sat ground coverages at the latitude extremes. Looks like they're fixin' for something like 20 planes to start.

That model is horribly misleading if you don't know what it is. It shows where the sats are visible from the ground, it is not a coverage map and only shows the sats from the first 3 launches.


> The "circles" around each satellite are not internet coverage areas - but rather visibility areas from the ground. If you are inside a circle, then you will be able to see that satellite >30˚ above the horizon.

> The color corresponds to illumination of the satellite - green is illuminated, red is not and different shades of yellow around the terminators correspond to partial eclipses


CelesTrak Orbit Visualization shows real positions of the sats from all 8 launches and they aren't anywhere near as neatly lined up in real life as they are in that gif.

From the youtube.com/watch?v=857UM4ErX9A I mentioned previously there are 3 planes per launch but it takes months for the sats to get into position. We have 8 launches so far but only 11 planes somewhat in position, 7 more waiting to be filled from the first 6 launches and 6 more planes for the sats from launch 7 and 8 (from earlier in June, just a few weeks ago).

 

[Cosmacelf]

CelesTrak Orbit Visualization shows real positions of the sats from all 8 launches and they aren't anywhere near as neatly lined up in real life as they are in that gif.

From the youtube.com/watch?v=857UM4ErX9A I mentioned previously there are 3 planes per launch but it takes months for the sats to get into position. We have 8 launches so far but only 11 planes somewhat in position, 7 more waiting to be filled from the first 6 launches and 6 more planes for the sats from launch 7 and 8 (from earlier in June, just a few weeks ago).

That link you provide gives a good idea of where we are at (click on the green dots to see orbits). You can see that it's going to take many more months before we get anywhere near continuous coverage anywhere.

SpaceX usually does things so quickly, we are conditioned to think that the sats are ready to go within days or weeks of launch, but that isn't the case. No doubt in this instance, SpaceX is being constrained by economics. It is cheaper to take many months to get into the correct orbits, and as Elon says, his goal is for Starlink to be the first LEO constellation that doesn't go bankrupt.

 

[dpeilow]

RCST and NCC are specifically DVB-RCS terms.

Yep - check who wrote the standard

 

[dhanson865]

Starlink-123 GIF | Gfycat

You can clearly see the convergence of the sat ground coverages at the latitude extremes. Looks like they're fixin' for something like 20 planes to start.

Here is another visualization that I think conveys the convergence at the extreme latitudes and also allows a good overall visualization.

SpaceX Starlink Orbit Constellation Series: Update 2020-06-14

 

[dhanson865]

Starlink https://spacex.moesalih.com/starlink is a continually updated page based on the animation seen in

 

[Cosmacelf]

Starlink https://spacex.moesalih.com/starlink is a continually updated page based on the animation seen in

OK, that's a cool and useful page! Easy to understand too.

 

[Wahbash]

These are some great links, and this is all fascinating the more I learn about it. Hopefull that beta will come to us here in northern NY relatively soon, looks like we may have about 85% coverage now.

I had no idea it took up to 75-100 days for each batch of birds to settle into position, but now that I think it through it makes sense.

 

[bxr140]

I had no idea it took up to 75-100 days for each batch of birds to settle into position, but now that I think it through it makes sense.

It’s pretty complicated geometry made even more complicated and lengthy by electric propulsion that does long slow burns around a significant portion of each obit (as opposed to an old school propulsion system that basically does most maneuvers in half an orbit or less and with essentially instantaneous impulses), but there’s basically three things going on that sum up to the ~3 month-to-station time:

1. Orbit raising. The sats all deploy into a lower orbit than the final altitude and then use onboard propulsion to make up the difference. This is because it is more energy efficient than being injected into the final orbit altitude, but also because it enables 2 and 3. This is a hard duration to calculate because it depends on a number of things (primarily actual thrust and duty cycle of the EP system) and presumably someone has an as-executed number reported from SpaceX, but generally these kinds of orbits/sats will be 1-2 months for orbit raising. This is where the spiraling over hundreds of orbits happens--there's ~15-16 orbits per day, times 30-60 days.

2. Orbit phasing. All of the sats are deployed at basically the same exact point on the 360 degrees of an orbit, but ultimately they need to be space evenly around those 360 degrees. While the EP (not to mention the non-circular injection orbit) makes it far more nuanced, this one is pretty easy to visualize and implement: If the plan is 20 sats evenly spaced around an orbit, as the just-deployed flock of sats went around in the injected orbit, each one would wait progressively 18 more degrees (because 360 deg / 20 = 18deg) than the last sat before starting its orbit raising sequence. There's really no discreet duration for this element of the constellation deployment as its all kind of wrapped into a mega EP burn strategy, but you first order visualize this as each sat has waiting 4.5 min longer than the previous sat (4.5 min = 90 min orbit / 20 evenly spaced sats) before starting its orbit phasing. So the last sat has to wait 83.5 more min than the first.

3. The most complicated element is the Plane Phasing. Each launch of 60 sats is currently filling 3 planes, but only the first plane of sats can immediately start #1 and #2 upon deployment (as described it reads more like #2 then #1, but I digress). The other 40 sats basically just have to wait it out in the low parking orbit before they can raise up to their final altitude because they would otherwise be misaligned with their destination planes that are 18 and 36 degrees away, respectively. The way orbits work the lower the altitude of an orbit the faster that orbit’s plane naturally rotates (‘precesses’) around the earth. So, kinda like in Frogger where you have to wait for the slower and faster moving things to line up and you can’t just force it, the sats have to hang out at the low orbit to allow the faster precessing plane that they're orbiting in to catch up to their slower precessing destination planes. Specifically, if we're talking about a 300km parking altitude and a 550km final altitude (actual start and finish altitudes may vary but I think that's pretty close for Starlink), that difference in precession between the two orbits is ~0.6 degrees/day. So for the current launches filling 3 planes, the first plane of sats will immediately orbit raise, the next plane of sats will hang out for ~29 days and then start orbit raising, and the third plane will hang out for ~58 days. Note that its not quite that straightforward (reality ends up with an optimal mash up of #1 and #3), but that conversation would take us way in the weeds...

Of course, all of those maths change if the launches are filling a different number of planes with a different number of sats (like, when Starship actually becomes operational), but the principals above still all hold true.

 

[Bipo]

Do we know in which plane will the next batch be launched?

Will those new satellites still shift into three different planes? Maybe since the full 360° are already covered (in 20° slices) the follow on batches will fill up those planes or even start to complete the polar part of the constellation...

 

[ggies07]

It’s pretty complicated geometry made even more complicated and lengthy by electric propulsion that does long slow burns around a significant portion of each obit (as opposed to an old school propulsion system that basically does most maneuvers in half an orbit or less and with essentially instantaneous impulses), but there’s basically three things going on that sum up to the ~3 month-to-station time:

1. Orbit raising. The sats all deploy into a lower orbit than the final altitude and then use onboard propulsion to make up the difference. This is because it is more energy efficient than being injected into the final orbit altitude, but also because it enables 2 and 3. This is a hard duration to calculate because it depends on a number of things (primarily actual thrust and duty cycle of the EP system) and presumably someone has an as-executed number reported from SpaceX, but generally these kinds of orbits/sats will be 1-2 months for orbit raising. This is where the spiraling over hundreds of orbits happens--there's ~15-16 orbits per day, times 30-60 days.

2. Orbit phasing. All of the sats are deployed at basically the same exact point on the 360 degrees of an orbit, but ultimately they need to be space evenly around those 360 degrees. While the EP (not to mention the non-circular injection orbit) makes it far more nuanced, this one is pretty easy to visualize and implement: If the plan is 20 sats evenly spaced around an orbit, as the just-deployed flock of sats went around in the injected orbit, each one would wait progressively 18 more degrees (because 360 deg / 20 = 18deg) than the last sat before starting its orbit raising sequence. There's really no discreet duration for this element of the constellation deployment as its all kind of wrapped into a mega EP burn strategy, but you first order visualize this as each sat has waiting 4.5 min longer than the previous sat (4.5 min = 90 min orbit / 20 evenly spaced sats) before starting its orbit phasing. So the last sat has to wait 83.5 more min than the first.

3. The most complicated element is the Plane Phasing. Each launch of 60 sats is currently filling 3 planes, but only the first plane of sats can immediately start #1 and #2 upon deployment (as described it reads more like #2 then #1, but I digress). The other 40 sats basically just have to wait it out in the low parking orbit before they can raise up to their final altitude because they would otherwise be misaligned with their destination planes that are 18 and 36 degrees away, respectively. The way orbits work the lower the altitude of an orbit the faster that orbit’s plane naturally rotates (‘precesses’) around the earth. So, kinda like in Frogger where you have to wait for the slower and faster moving things to line up and you can’t just force it, the sats have to hang out at the low orbit to allow the faster precessing plane that they're orbiting in to catch up to their slower precessing destination planes. Specifically, if we're talking about a 300km parking altitude and a 550km final altitude (actual start and finish altitudes may vary but I think that's pretty close for Starlink), that difference in precession between the two orbits is ~0.6 degrees/day. So for the current launches filling 3 planes, the first plane of sats will immediately orbit raise, the next plane of sats will hang out for ~29 days and then start orbit raising, and the third plane will hang out for ~58 days. Note that its not quite that straightforward (reality ends up with an optimal mash up of #1 and #3), but that conversation would take us way in the weeds...

Of course, all of those maths change if the launches are filling a different number of planes with a different number of sats (like, when Starship actually becomes operational), but the principals above still all hold true.

Great explanation. thanks!

 

[mongo]

Do we know in which plane will the next batch be launched?

Will those new satellites still shift into three different planes? Maybe since the full 360° are already covered (in 20° slices) the follow on batches will fill up those planes or even start to complete the polar part of the constellation...

They'll start new planes, perhaps on the 10 degree offsets phased half a sat off the current ones. (Like a 5 on a die).
Evaluating SpaceX's Starlink Push - NASASpaceFlight.com

This article mentions a request to go from 24 planes to 72 SpaceX says more Starlink orbits will speed service, reduce launch needs - SpaceNews.com
which was approved
SpaceX modifies Starlink network design as another 60 satellites gear up for launch – Spaceflight Now
So starting at 20 planes and increasing from there.

I wouldn't expect polar for a while.