Starlink: Collision risks

[su27k]

A long twitter thread about close approaches between Starlink and other objects:

https://twitter.com/ProfHughLewis/status/1436321009242710022

The important part is this:

In August, @pbdes reported that @SpaceX had made 2219 collision avoidance manoeuvres in the six months to the end of May 2021 [13/n]

If we look at the distribution of the maximum collision probability predicted by SOCRATES for #Starlink conjunctions we can see an average of 77.4 conjunctions per week for the six months to the end of May 2021 (excludes Starlink-on-Starlink) [14/n]

This is equivalent to 2012.3 conjunctions over the  period, on average, which represents about one-third of all the conjunctions < 1 km involving one Starlink. [15/n]

Keeping in mind the limited accuracy of predictions made using public TLEs, it appears as though the number of conjunctions for which the maximum collision probability is > 1-in-100,000 provides an estimate of the number of collision avoidance manoeuvres (+/- 10% or so). [16/n]

Looking at the most recent 3-month period, there were 130.5 conjunctions per week, on average, where the maximum collision probability was predicted to be > 1-in-100,000 (again, about one-third of all conjunctions < 1 km), suggesting a further 1696 manoeuvres. [17/n]

The approach to orbital space safety for #Starlink appears to be quite cautious & is likely enabled by the high specific impulse of the low-thrust propulsion system, which mean that avoidance manoeuvres only have a small effect on overall propellant mass. [18/n]

Combined with some flexibility in the orbits needed to provide the communications service, Starlinks are able to avoid relatively low-risk encounters with debris & other operators. This approach has led to the safety of the system despite the growing system size. [19/n]

This confirms SpaceX is indeed using the very conservative 1 in 100,000 collision probability to determine if they'll perform collision avoidance maneuvers.

[eeergo]

This confirms SpaceX is indeed using the very conservative 1 in 100,000 collision probability to determine if they'll perform collision avoidance maneuvers.

Conservative?

1e-5 probability means that, with just over 1500 satellites currently and 2220 x 2 = 4440 conjunctions/year exceeding that threshold in the first half of this year (during which time more than half of the currently orbiting Starlinks were launched, which means the situation was not in a steady state and the actual number of conjunctions with the current amount of Starlinks should be around 10,000 conjunctions/year), you'd most likely get a catastrophic collision in 10 years, which would likely generate a rapid cascade effect considering the shells are very thin and debris couldn't be tracked fast enough to notify all satellites repeatedly passing through the debris cloud. If they set the bar any lower, they'd be all but guaranteed to initiate that process before 2030.

This doesn't take into account inherent inaccuracies in TLEs CDMs (<1km but unknown how much better) and other tracking of non-Starlink conjunctions. As the tweet shows there are 3000 conjunctions at less than 1 km a week involving at least a Starlink (156,000 a year), and TLEs often have that inaccuracy already built in. Let's remember Starlink "automated" approach relies on TLEs CDMs for their maneuvers, which have uncertainties even if lower than TLEs.

Of course, 1500 satellites is an eigth of the initial megaconstellation's proposed size of 12,000. Then, the 1e-5 collision probability threshold for maneuvering would mean a likely collision in slightly over a year if they didn't maneuver, again without uncertainties smudging the picture. With the full 42,000-strong system, not taking into account competitors or increased number of other satellites in Starlink's orbital regime, the 1e-5 threshold would imply a collision (followed swiftly by a myriad of others, let's remember - there's no room for error here) every trimester.

One wonders what happens with conjunctions with 4e-6 probability though, for which their "conservative" threshold wouldn't apply, and which imply collision likelihood frequencies twice rarer, yet unmitigated - i.e. every 2 years for a "basic" Starlink system, and every 6 months for a complete system.

[envy887]

This confirms SpaceX is indeed using the very conservative 1 in 100,000 collision probability to determine if they'll perform collision avoidance maneuvers.

Conservative?

1e-5 probability means that, with just over 1500 satellites currently and 2220 x 2 = 4440 conjunctions/year exceeding that threshold in the first half of this year (during which time more than half of the currently orbiting Starlinks were launched, which means the situation was not in a steady state and the actual number of conjunctions with the current amount of Starlinks should be around 10,000 conjunctions/year), you'd most likely get a catastrophic collision in 10 years, which would likely generate a rapid cascade effect considering the shells are very thin and debris couldn't be tracked fast enough to notify all satellites repeatedly passing through the debris cloud. If they set the bar any lower, they'd be all but guaranteed to initiate that process before 2030.

This doesn't take into account inherent inaccuracies in TLEs and other tracking of non-Starlink conjunctions. As the tweet shows there are 3000 conjunctions at less than 1 km a week involving at least a Starlink (156,000 a year), and TLEs often have that inaccuracy already built in. Let's remember Starlink "automated" approach relies on TLEs for their maneuvers. So there can be many other approaches which are not registered.

Of course, 1500 satellites is an eigth of the initial megaconstellation's proposed size of 12,000. Then, the 1e-5 collision probability threshold for maneuvering would mean a likely collision in slightly over a year if they didn't maneuver, again without uncertainties smudging the picture. With the full 42,000-strong system, not taking into account competitors or increased number of other satellites in Starlink's orbital regime, the 1e-5 threshold would imply a collision (followed swiftly by a myriad of others, let's remember - there's no room for error here) every trimester.

One wonders what happens with conjunctions with 4e-6 probability though, for which their "conservative" threshold wouldn't apply, and which imply collision likelihood frequencies twice rarer, yet unmitigated - i.e. every 2 years for a "basic" Starlink system, and every 6 months for a complete system.

I think "a rapid cascade effect" is exceptionally unlikely in the event of a collision, for a number of reasons.

Also, the probability for each event is different. You can't assume that events with a probability below the threshold of 1/100,000 are all actually 1/100,000 when they are really in some distribution ranging from 1/100,000 all the way to exactly 0.

[dondar]

This confirms SpaceX is indeed using the very conservative 1 in 100,000 collision probability to determine if they'll perform collision avoidance maneuvers.

Conservative?

1e-5 probability means that, with just over 1500 satellites currently and 2220 x 2 = 4440 conjunctions/year exceeding that threshold in the first half of this year (during which time more than half of the currently orbiting Starlinks were launched, which means the situation was not in a steady state and the actual number of conjunctions with the current amount of Starlinks should be around 10,000 conjunctions/year), you'd most likely get a catastrophic collision in 10 years, which would likely generate a rapid cascade effect considering the shells are very thin and debris couldn't be tracked fast enough to notify all satellites repeatedly passing through the debris cloud. If they set the bar any lower, they'd be all but guaranteed to initiate that process before 2030.

This doesn't take into account inherent inaccuracies in TLEs and other tracking of non-Starlink conjunctions. As the tweet shows there are 3000 conjunctions at less than 1 km a week involving at least a Starlink (156,000 a year), and TLEs often have that inaccuracy already built in. Let's remember Starlink "automated" approach relies on TLEs for their maneuvers. So there can be many other approaches which are not registered.

Of course, 1500 satellites is an eigth of the initial megaconstellation's proposed size of 12,000. Then, the 1e-5 collision probability threshold for maneuvering would mean a likely collision in slightly over a year if they didn't maneuver, again without uncertainties smudging the picture. With the full 42,000-strong system, not taking into account competitors or increased number of other satellites in Starlink's orbital regime, the 1e-5 threshold would imply a collision (followed swiftly by a myriad of others, let's remember - there's no room for error here) every trimester.

One wonders what happens with conjunctions with 4e-6 probability though, for which their "conservative" threshold wouldn't apply, and which imply collision likelihood frequencies twice rarer, yet unmitigated - i.e. every 2 years for a "basic" Starlink system, and every 6 months for a complete system.

so much emotions here.

 "conjunctions" are not a thing within starlink system.
 They are events of possible orbit intersection with other satelites (or for now between rising starlinks from different batches.)
The "fall" in conjuctions reported by Lewis indeed directly correlates with the number of rising starlink in intersecting planes.
SOCRAT counts also satellites in formation as a collision candidates (Because the system is simplified, read STUPID).

The probabilities even for the conjuctions with min 0.1km can be still in the range off 1e-7. And there are many of those.

I will not go into "estimated numbers". Or what even SOCRAT calculates.

Right now Starlink sit strongly under 1% per year for their system. Totally. Please don't comment, especially "doom" without proper numbers or ability to defend scientifically your opinion.

[eeergo]

I think "a rapid cascade effect" is exceptionally unlikely in the event of a collision, for a number of reasons.

Also, the probability for each event is different. You can't assume that events with a probability below the threshold of 1/100,000 are all actually 1/100,000 when they are really in some distribution ranging from 1/100,000 all the way to exactly 0.

Could you list some of those reasons? Common sense would indicate hundreds or thousands of satellites at roughly the same orbital height but mostly differing velocity vectors (except those in the same plane) would be exceptionally vulnerable to a debris field squarely intersecting that height - especially if centered on it.

Regarding the distribution: sure, that's why my numbers are a lower limit simplification, because 1e-5 is the threshold for action. There will be much higher probability conjunctions im the distribution too, but even taking a "monochromatic" probability at 1e-5 you can see the results of not maneuvering, which is why the threshold is not "conservative" at all.

The issue here aren't conjunctions below 1e-5 (hence needing no avoidance maneuvers, even if as shown they will need to lower the threshold if the constellation reaches anywhere close to its planned total size). The issue is the distribution over 1e-5 (and well below 1, to be clear) if left alone, or the part of the distribution just under 1e-5 (and hence necessitating no maneuvers by the current standard). There's also the question of TLE CDM uncertainty which SpaceX uses to determine the probabilities (<1 km but unknown how much better).


EDIT: By the way, realizing people might not have eead the referenced thread and where the 2200 CAMs figure might come from: https://mobile.twitter.com/pbdes/status/1429778141506060289

[Redclaws]

I think "a rapid cascade effect" is exceptionally unlikely in the event of a collision, for a number of reasons.

Also, the probability for each event is different. You can't assume that events with a probability below the threshold of 1/100,000 are all actually 1/100,000 when they are really in some distribution ranging from 1/100,000 all the way to exactly 0.

Could you list some of those reasons? Common sense would indicate hundreds or thousands of satellites at roughly the same orbital height but mostly differing velocity vectors (except those in the same plane) would be exceptionally vulnerable to a debris field squarely intersecting that height - especially if centered on it.

Regarding the distribution: sure, that's why my numbers are a lower limit simplification, because 1e-5 is the threshold for action. There will be much higher probability conjunctions im the distribution too, but even taking a "monochromatic" probability at 1e-5 you can see the results of not maneuvering, which is why the threshold is not "conservative" at all.

The issue here aren't conjunctions below 1e-5 (hence needing no avoidance maneuvers, even if as shown they will need to lower the threshold if the constellation reaches anywhere close to its planned total size). The issue is the distribution over 1e-5 (and well below 1, to be clear) if left alone, or the part of the distribution just under 1e-5 (and hence necessitating no maneuvers by the current standard). There's also the question of TLE uncertainty which SpaceX uses to determine the probabilities (~1 km).

I’m not in a position to dispute the numbers you’re using, but I’d ask you a question.  The relatively simple numbers you’re putting up seem to clearly show doom for this constellation - totally unworkable levels of collisions, etc - so why do the regulators think it’s ok?  It seems reasonably clear the real numbers do not work out as you’re describing - the regulators are competent and can do basic math.  So…

[eeergo]

so much emotions here.

Many.

"conjunctions" are not a thing within starlink system.They are events of possible orbit intersection with other satelites (or for now between rising starlinks from different batches.)

They are as in any satellite. Between different heights and at the same height alike. They are reportedly minimized automatically based on SpaceX tracking -of which we know little or nothing about, but let's take it at face value and say they're not an issue (even if they clearly exist as shown by the SOCRATES data). There are still plenty of conjunctions <1 km with other objects as shown in the Twitter thread, and within those many that SpaceX judges worthy of executing a CAM for (~10000/year currently).

SOCRAT counts also satellites in formation as a collision candidates (Because the system is simplified, read STUPID).

What are you talking about? You're calling SOCRATES stupid?

The probabilities even for the conjuctions with min 0.1km can be still in the range off 1e-7. And there are many of those.

They can. I did not conflate both. I did take however SpaceX's CAM numbers, and they state only events with probability >1e-5 require action, so all those CAMs are over 1e-5. All 2200 of them. Some much higher probably (continuous distribution from 1e-5 all the way to 1). I did some elementary calculations for longer times and/or satellite numbers.

Right now Starlink sit strongly under 1% per year for their system. Totally. Please don't comment, especially "doom" without proper numbers or ability to defend scientifically your opinion.

Excuse me, where are your proper numbers? You just said you'll not go into SOCRATES data? Neither basic multiplication because it's "estimated"? I explicitly showed my work, you're pulling a 1% out of your hat (seriously, where in the world do you deduce that from? By their own admission they did 2200 CAMs in 6 months when Starlink had half the satellites it has now, and now it has 1500 birds...). Where exactly are my extrapolations of their own CAM numbers and probability thresholds wrong?

Talk about emotional responses...

[eeergo]

I think "a rapid cascade effect" is exceptionally unlikely in the event of a collision, for a number of reasons.

Also, the probability for each event is different. You can't assume that events with a probability below the threshold of 1/100,000 are all actually 1/100,000 when they are really in some distribution ranging from 1/100,000 all the way to exactly 0.

Could you list some of those reasons? Common sense would indicate hundreds or thousands of satellites at roughly the same orbital height but mostly differing velocity vectors (except those in the same plane) would be exceptionally vulnerable to a debris field squarely intersecting that height - especially if centered on it.

Regarding the distribution: sure, that's why my numbers are a lower limit simplification, because 1e-5 is the threshold for action. There will be much higher probability conjunctions im the distribution too, but even taking a "monochromatic" probability at 1e-5 you can see the results of not maneuvering, which is why the threshold is not "conservative" at all.

The issue here aren't conjunctions below 1e-5 (hence needing no avoidance maneuvers, even if as shown they will need to lower the threshold if the constellation reaches anywhere close to its planned total size). The issue is the distribution over 1e-5 (and well below 1, to be clear) if left alone, or the part of the distribution just under 1e-5 (and hence necessitating no maneuvers by the current standard). There's also the question of TLE uncertainty which SpaceX uses to determine the probabilities (~1 km).

I’m not in a position to dispute the numbers you’re using, but I’d ask you a question.  The relatively simple numbers you’re putting up seem to clearly show doom for this constellation - totally unworkable levels of collisions, etc - so why do the regulators think it’s ok?  It seems reasonably clear the real numbers do not work out as you’re describing - the regulators are competent and can do basic math.  So…

As you say, it's basic math, and there's no way around it unless you believe in impeccable tracking and constellation control over many decades. Now you want me to read into politicians' minds, and that my friend, is just asking for miracles. But I do know it's not the first time gold rushes happened, either in space or on the ground - rarely because they made long-term sense.


On the other hand, they may be expecting SpaceX to reduce tracking errors with their own 'Space Fences', or lowering the threshold an order of magnitude as Starlink grows, or some software wizardry that hopefully will solve space traffic control congestion in a few years' time because it'll become highly error-free and automated, or any number of 'let's wait and see's.


I was taking issue at calling 1e-5 "conservative" with the numbers of conjunctions, CAMs and satellites already at hand. Not claiming some law of Nature forces this to become doomsday - but current numbers just aren't flattering, and already leave little room for error: human, systematic or otherwise.

[envy887]

I think "a rapid cascade effect" is exceptionally unlikely in the event of a collision, for a number of reasons.

Also, the probability for each event is different. You can't assume that events with a probability below the threshold of 1/100,000 are all actually 1/100,000 when they are really in some distribution ranging from 1/100,000 all the way to exactly 0.

Could you list some of those reasons? Common sense would indicate hundreds or thousands of satellites at roughly the same orbital height but mostly differing velocity vectors (except those in the same plane) would be exceptionally vulnerable to a debris field squarely intersecting that height - especially if centered on it.

After a collision, approximately none of the fragments have exactly the same velocity as the original satellite. The impact is impulsive, which means they are now in new orbits that intersect the original only at most twice per orbit. That means the opportunity to re-collide with another satellite in the shell happens at most twice per 90 minutes.

If they got an infinite number of orbits, they would eventually collide... but they won't have that chance. A few things start happening. For one, most of the fragment will lose energy and velocity, and end up with a lower perigee. That exposes them to more drag, which quickly lowers apogee out of the shell, so they no longer intersect at all. Fragments kicked higher also start to drag. They stick around a little longer, but at the 550 km level drag is pretty high and only the tail end of the distribution gets enough energy to stick around a while.

Also, in a few days, the trackable objects will get added to the catalog, and the rest of the constellation will start maneuvering to avoid them. So the big (most dangerous) fragments only get a few tens of shots at causing more damage. And Starlinks aren't big enough to make a lot of big fragments, perhaps a few tens as well. All these factors reduce the number of debris conjunctions after a collision, which reduces the probability of a cascade.

Regarding the distribution: sure, that's why my numbers are a lower limit simplification, because 1e-5 is the threshold for action. There will be much higher probability conjunctions im the distribution too, but even taking a "monochromatic" probability at 1e-5 you can see the results of not maneuvering, which is why the threshold is not "conservative" at all.

1e-5 is the ceiling. There will be no conjunctions above that, because those always result in maneuvers to place them below the threshold. So there will be some at 1e-6, but none at 1e-4. There will also probably be some at 1e-7, and maybe a few at 1e-8 or 1e-9 (passes within an arbitrary distance aren't necessarily anywhere near the 1e-5 threshold, depending on uncertainty in the orbits and the geometry of the pass). You need the distribution of event probabilities (or at least the mean event probability?) to compute the overall probability of at least 1 collision in a given timeframe.

[envy887]

On the other hand, they may be expecting SpaceX to reduce tracking errors with their own 'Space Fences', or lowering the threshold an order of magnitude as Starlink grows, or some software wizardry that hopefully will solve space traffic control congestion in a few years' time because it'll become highly error-free and automated, or any number of 'let's wait and see's.

SpaceX doesn't need a "Space Fence" for their own satellites. They have highly accurate data from the INS/GPS/star trackers on their satellites and can use this to avoid Starlink fratricide. Starlink on other satellite and Starlink on debris conjunctions are the main concerns, but I'm not sure that anyone realistically expects SpaceX to start tracking everything else in orbit.

[su27k]

This confirms SpaceX is indeed using the very conservative 1 in 100,000 collision probability to determine if they'll perform collision avoidance maneuvers.

Conservative?

1e-5 probability means that, with just over 1500 satellites currently and 2220 x 2 = 4440 conjunctions/year exceeding that threshold in the first half of this year (during which time more than half of the currently orbiting Starlinks were launched, which means the situation was not in a steady state and the actual number of conjunctions with the current amount of Starlinks should be around 10,000 conjunctions/year), you'd most likely get a catastrophic collision in 10 years, which would likely generate a rapid cascade effect considering the shells are very thin and debris couldn't be tracked fast enough to notify all satellites repeatedly passing through the debris cloud. If they set the bar any lower, they'd be all but guaranteed to initiate that process before 2030.

This doesn't take into account inherent inaccuracies in TLEs and other tracking of non-Starlink conjunctions. As the tweet shows there are 3000 conjunctions at less than 1 km a week involving at least a Starlink (156,000 a year), and TLEs often have that inaccuracy already built in. Let's remember Starlink "automated" approach relies on TLEs for their maneuvers. So there can be many other approaches which are not registered.

Of course, 1500 satellites is an eigth of the initial megaconstellation's proposed size of 12,000. Then, the 1e-5 collision probability threshold for maneuvering would mean a likely collision in slightly over a year if they didn't maneuver, again without uncertainties smudging the picture. With the full 42,000-strong system, not taking into account competitors or increased number of other satellites in Starlink's orbital regime, the 1e-5 threshold would imply a collision (followed swiftly by a myriad of others, let's remember - there's no room for error here) every trimester.

One wonders what happens with conjunctions with 4e-6 probability though, for which their "conservative" threshold wouldn't apply, and which imply collision likelihood frequencies twice rarer, yet unmitigated - i.e. every 2 years for a "basic" Starlink system, and every 6 months for a complete system.

1. Yes, conservative, because SpaceX said so, and now it was independently confirmed by a space debris expert.

2. You're under the mistaken impression that a collision avoidance threshold of 1e-5 means after the collision avoidance maneuver the probability of collision (Pc) is only reduced to 1e-5, that's not the case, as SpaceX explained in their FCC filing which is quoted here. In most cases, after the collision avoidance maneuver Pc is reduced to be lower than 1e-6. That filing also explained other reasons why what they're doing is conservative, for example they use 10m radius in their modeling of Starlink satellite, which is much larger than the actual satellite.

3. Starlink does not use TLEs for their automated collision avoidance system, this was explained in a FCC filing just a few days ago, see here, what they use is Conjunction Data Messages (CDM) from 18th SPCS.

4. And when 18th SPCS calculates conjunctions and issues CDMs, they most definitely does take inaccuracies in position and speed of tracked satellites into account. This is explained in the Spaceflight Safety Handbook for Satellite Operator, the term you're looking for is "covariance".

5. Dr. Hugh Lewis has stated multiple times that mega constellation like Starlink won't cause collision cascade reaction like you suggested (aka Kessler syndrome), see this tweet for example: https://twitter.com/ProfHughLewis/status/1387532062446456835

So unless you can back up your claims with some real math and simulations, I'm going to insist that "conservative" is the right word to use for the current constellation. Will the threshold need to be adjusted for bigger constellations? I don't know, maybe it will, but given on average each satellite is only doing 3 collision avoidance maneuvers every *year* right now, I don't see a lower threshold will present any difficulties.

[eeergo]

Could you list some of those reasons? Common sense would indicate hundreds or thousands of satellites at roughly the same orbital height but mostly differing velocity vectors (except those in the same plane) would be exceptionally vulnerable to a debris field squarely intersecting that height - especially if centered on it.

After a collision, approximately none of the fragments have exactly the same velocity as the original satellite. The impact is impulsive, which means they are now in new orbits that intersect the original only at most twice per orbit. That means the opportunity to re-collide with another satellite in the shell happens at most twice per 90 minutes.

That isn't the general case. It will depend on the geometry of the collision. Even for forward-scattered collisions, the distributions end up having a strong double-lobed distribution of resulting objects, with some reaching higher apogees, some lower, a scatter of inclinations, but most staying right around the original orbital altitude and inclination (see first attached image from FY-1C's ASAT test, where the collision was highly deboosting since the warhead was extremely slow compared to the satellite's orbital velocity). Then you get precession and orbital perturbations (solar, atmospheric, n-body) that will spread the debris cloud around in a few months' time, which is within the resulting debris' lifetimes in orbit (at the initial height, of course there will be those higher): an example is the spread of the Iridium-Kosmos debris clouds from tight distributions at their initial approximately 90-degree-apart respective orbits, and the homogeneous distribution some months later - see second and third attached images.

If they got an infinite number of orbits, they would eventually collide... but they won't have that chance. A few things start happening. For one, most of the fragment will lose energy and velocity, and end up with a lower perigee. That exposes them to more drag, which quickly lowers apogee out of the shell, so they no longer intersect at all. Fragments kicked higher also start to drag. They stick around a little longer, but at the 550 km level drag is pretty high and only the tail end of the distribution gets enough energy to stick around a while.

Orbital lifetimes for most objects at 550 km are on the order of months to years. Even decaying to lower orbits than the main shell's will take such a cloud uncomfortably close to other lower orbiting satellites, like incidentally the ISS.

Also, in a few days, the trackable objects will get added to the catalog, and the rest of the constellation will start maneuvering to avoid them. So the big (most dangerous) fragments only get a few tens of shots at causing more damage. And Starlinks aren't big enough to make a lot of big fragments, perhaps a few tens as well. All these factors reduce the number of debris conjunctions after a collision, which reduces the probability of a cascade.

I'm afraid this is just completely anumerical. They will take weeks if not months to catalog resulting debris, based on recent experiences with the Indian ASAT or several on-orbit collisions. Satellites won't be able to maneuver clear of the cloud, which will span a large range of altitudes (centered at the operational altitude) and soon start spreading in inclination (which the satellites cannot significantly impulsively change). As you say, then there's the issue of trackability: a small untrackable object under about 10 cm but larger than 1 cm can still cause a lot of damage, yet they are untrackable - and an orbital collision has the peculiarity of creating lots of those. See here: https://conference.sdo.esoc.esa.int/proceedings/neosst1/paper/399 (10% of potentially catastrophically-damaging debris is trackable).
All these issues are valid currently. With an order of magnitude more active satellites *at the same height* this is a substantially non-linearly worsening challenge.

Regarding the distribution: sure, that's why my numbers are a lower limit simplification, because 1e-5 is the threshold for action. There will be much higher probability conjunctions im the distribution too, but even taking a "monochromatic" probability at 1e-5 you can see the results of not maneuvering, which is why the threshold is not "conservative" at all.

1e-5 is the ceiling. There will be no conjunctions above that, because those always result in maneuvers to place them below the threshold. So there will be some at 1e-6, but none at 1e-4. There will also probably be some at 1e-7, and maybe a few at 1e-8 or 1e-9 (passes within an arbitrary distance aren't necessarily anywhere near the 1e-5 threshold, depending on uncertainty in the orbits and the geometry of the pass). You need the distribution of event probabilities (or at least the mean event probability?) to compute the overall probability of at least 1 collision in a given timeframe.

Finally I understand what you're referring to, I think - and I'm afraid you're misunderstanding the dynamics here.
You're assuming SpaceX has kind of a color-coded board with all satellites in green (probability of collision close to 0), and slowly some are becoming progressively redder, until reaching bright red (probability of 1). Once they become orange though, they have the ability to slightly maneuver them so that none gets beyond a certain tone of orange (1e-5) which they set as threshold. That indeed would be the ideal situation. It *may* be for inter-Starlink conjunctions, because they have their own GNSS packages, as you mention in your following post, and that *may* be enough to approach that idealization - we'll have to take their word for it, which is not ideal anyway.

Unfortunately, that's not the case for the rest of objects, simply because of the innacuracies of the TLEs and the non-deterministic refinement of tracking. You don't get continuous perfect tracking that allows you to steer clear as soon as the collision probability reaches 1e-5 and not beyond. You get more or less frequent updates with "jumps" in the parameters, and most of those will instantaneously change the collision probability by a lot. Most will stay very low or 0, but some will jump well above 1e-5 immediately, with a distribution that rarely reaches 1 (or 1e-1 for that matter), yet will include lots of 1e-4, 5e-3, 1e-3 and so forth. It's the fundamental difference between a continuous, well-sampled process with low uncertainties, and a discrete, undersampled process with high uncertainties.

Their proverbial color-coded board would be alternating all kinds of colors with each tracking update, and the "magic orange tone" of 1e-5 will be the threshold over which they'll do something about it - but won't be the reddest the board gets, with some points in the board suddenly turning dark orange just after being clear green.

Once you have a collision though, as I mentioned above, all bets are off regarding Starlink-on-Starlink: then Starlink on fragement-of-Starlink conjunctions are no longer GNSS-defined and will be subject to the same limitations.

[eeergo]

This confirms SpaceX is indeed using the very conservative 1 in 100,000 collision probability to determine if they'll perform collision avoidance maneuvers.

Conservative?

1e-5 probability means that, with just over 1500 satellites currently and 2220 x 2 = 4440 conjunctions/year exceeding that threshold in the first half of this year (during which time more than half of the currently orbiting Starlinks were launched, which means the situation was not in a steady state and the actual number of conjunctions with the current amount of Starlinks should be around 10,000 conjunctions/year), you'd most likely get a catastrophic collision in 10 years, which would likely generate a rapid cascade effect considering the shells are very thin and debris couldn't be tracked fast enough to notify all satellites repeatedly passing through the debris cloud. If they set the bar any lower, they'd be all but guaranteed to initiate that process before 2030.

This doesn't take into account inherent inaccuracies in TLEs and other tracking of non-Starlink conjunctions. As the tweet shows there are 3000 conjunctions at less than 1 km a week involving at least a Starlink (156,000 a year), and TLEs often have that inaccuracy already built in. Let's remember Starlink "automated" approach relies on TLEs for their maneuvers. So there can be many other approaches which are not registered.

Of course, 1500 satellites is an eigth of the initial megaconstellation's proposed size of 12,000. Then, the 1e-5 collision probability threshold for maneuvering would mean a likely collision in slightly over a year if they didn't maneuver, again without uncertainties smudging the picture. With the full 42,000-strong system, not taking into account competitors or increased number of other satellites in Starlink's orbital regime, the 1e-5 threshold would imply a collision (followed swiftly by a myriad of others, let's remember - there's no room for error here) every trimester.

One wonders what happens with conjunctions with 4e-6 probability though, for which their "conservative" threshold wouldn't apply, and which imply collision likelihood frequencies twice rarer, yet unmitigated - i.e. every 2 years for a "basic" Starlink system, and every 6 months for a complete system.

1. Yes, conservative, because SpaceX said so, and now it was independently confirmed by a space debris expert.

2. You're under the mistaken impression that a collision avoidance threshold of 1e-5 means after the collision avoidance maneuver the probability of collision (Pc) is only reduced to 1e-5, that's not the case, as SpaceX explained in their FCC filing which is quoted here. In most cases, after the collision avoidance maneuver Pc is reduced to be lower than 1e-6. That filing also explained other reasons why what they're doing is conservative, for example they use 10m radius in their modeling of Starlink satellite, which is much larger than the actual satellite.

3. Starlink does not use TLEs for their automated collision avoidance system, this was explained in a FCC filing just a few days ago, see here, what they use is Conjunction Data Messages (CDM) from 18th SPCS.

4. And when 18th SPCS calculates conjunctions and issues CDMs, they most definitely does take inaccuracies in position and speed of tracked satellites into account. This is explained in the Spaceflight Safety Handbook for Satellite Operator, the term you're looking for is "covariance".

5. Dr. Hugh Lewis has stated multiple times that mega constellation like Starlink won't cause collision cascade reaction like you suggested (aka Kessler syndrome), see this tweet for example: https://twitter.com/ProfHughLewis/status/1387532062446456835

So unless you can back up your claims with some real math and simulations, I'm going to insist that "conservative" is the right word to use for the current constellation. Will the threshold need to be adjusted for bigger constellations? I don't know, maybe it will, but given on average each satellite is only doing 3 collision avoidance maneuvers every *year* right now, I don't see a lower threshold will present any difficulties.

1- Ok then, if they say so we must all shut up and revere them. Experts have been saying all sorts of things regarding Starlink and you've been consistently and openly dissing them. But suddenly one expresses a favorable opinion *for the current situation* (as my numbers showed, 1e-5 will give you 10 years of clean operations for a thousand satellites, even if no maneuvers took place) and they get all authoritative in your view.

2- I'm doing no such thing. I'm noting TLEs have uncertainties on the order of km, and most if not all the POCs>1e-5 need that kind of precision at minimum. Obviously if they perform a maneuver they're going to minimize the POC as much as they can, and with the available uncertainties 1e-6 is the minimum they can be confident they did something (positive) about it. The modelling of satellites as "spherical cows" with diameter equal to the satellite's largest axis is SOP in space debris management, as you can't be certain of the relative attitude at closest approach. Starlink v1.0's array is around 10 m long, so there you go.

3/4- Evidently they take uncertainties into account, through covariance matrices or any other method to do so - yet sizeable uncertainties exist in the first place, and you can't get around them, just mitigate their smearing effect to *statistically* avoid running into more trouble through them than you're running away from. One point I did learn with your remark about CDMs is that these Messages have **lower** uncertainties than run-of-the-mill TLEs, which is reasonable considering no meaningful prediction could be made if your minimum uncertainty is on the km scale. This means that the uncertainty smearing is smaller than I estimated before, and I have corrected that from my previous posts - yet it was an additional factor to the numbers I shared, not a basing assumption: those are still valid, since the hard number of CAMs are the end result of all the CDM-derived analysis.

5- There are plenty of experts that point out how we're already into Kessler dynamics - just that they aren't, and never were going to be, happening in a few hours' time like in Hollywood movies (think Gravity). You're grossly misrepresenting Prof Lewis' words by the way:

https://twitter.com/ProfHughLewis/status/1336584183783837696
https://twitter.com/ProfHughLewis/status/1304426186320347138
https://twitter.com/ProfHughLewis/status/1385535670094008326
https://twitter.com/ProfHughLewis/status/1385537020789989378

He's referring to immediate obliteration of the shell. His last tweet refers to the *current* environment below 600 km. But of course you're looking for slogans, not for nuanced scientific debate, for which you already showed your disdain plenty of times.

6- You're free to consider it as conservative as you want, the numbers are what they are. Plus: nyet, those numbers of 3 per satellite-year are for around 800 satellites. You don't have data to say how many they're performing now with 1500, although it's probably around 10. Still, that sounds like a small number until you realize that's a statistical process we're talking about, and the absolute numbers are big, ensuring probability will take advantage of any mistake: around ten thousand a year with a 10th of the initial constellation built up, and a non-linear increase Prof Lewis is showing in SOCRATES data. A complete minimal (12000) system will have millions of conjunctions a year and similar orders of magnitudes of CAMs. If we consider smooth nominal operations forever, everything is and will be fine obviously - here we're talking about uncertainties, mistakes or oversights, and their consequences.

As for the simulations: please do so yourself before asking others to develop dissertation-grade work for you. I'm taking real-world number disclosed by SpaceX themselves and showing what they mean mathematically for different degrees of extrapolation. You're misconstruing an expert's words stripping them of their nuances and twisting the concluding message in order to advance your agenda, as you've done before most notoriously with Dr McDowell's points

[su27k]

1- Ok then, if they say so we must all shut up and revere them. Experts have been saying all sorts of things regarding Starlink and you've been consistently and openly dissing them. But suddenly one expresses a favorable opinion *for the current situation* (as my numbers showed, 1e-5 will give you 10 years of clean operations for a thousand satellites, even if no maneuvers took place) and they get all authoritative in your view.

If an expert says things contradicting what SpaceX says, then there's a different of opinion and we will need further analysis to see who's right. But if both expert and SpaceX says the same thing, then there's much higher likelihood of them being right, high enough to override some opinion from someone who clearly doesn't know how collision avoidance works in the first place.

And yes, I very much like to see expert "expresses a favorable opinion" for Starlink, especially one who has less favorable opinion before, but has now changed his opinion based on fact, which is what everybody should be doing. Nobody is stopping you from quoting other expert who has less favorable opinion, if you can find them.

2- I'm doing no such thing. I'm noting TLEs have uncertainties on the order of km, and most if not all the POCs>1e-5 need that kind of precision at minimum. Obviously if they perform a maneuver they're going to minimize the POC as much as they can, and with the available uncertainties 1e-6 is the minimum they can be confident they did something (positive) about it.

Ok, then you admit by using SpaceX's current collision avoidance strategy, they're no where near a collision every 10 years as you suggested?  In fact their safety projection is orders of magnitude better a collision every 10 years, that's what I call conservative.

The modelling of satellites as "spherical cows" with diameter equal to the satellite's largest axis is SOP in space debris management, as you can't be certain of the relative attitude at closest approach. Starlink v1.0's array is around 10 m long, so there you go.

And this SOP is conservative in case of Starlink because the solar array is very small in a sphere of 10 meters. It's only 3.2 meters wide, if you model it as a cylinder of 10 meters, its volume is only 80 m^3, it only occupies 2% of the total 4189 m^3 of a 10m sphere.

5- There are plenty of experts that point out how we're already into Kessler dynamics - just that they aren't, and never were going to be, happening in a few hours' time like in Hollywood movies (think Gravity).

"happening in a few hours' time like in Hollywood movies" is exactly what you're describing in your original comment: "which would likely generate a rapid cascade effect considering the shells are very thin and debris couldn't be tracked fast enough to notify all satellites repeatedly passing through the debris cloud.".

Usually for LEO objects we get TLEs in a few hours after launch, so by saying it happens so fast that we wouldn't have time to track it, you're literally suggesting it's like Gravity the movie.

You're grossly misrepresenting Prof Lewis' words by the way:

https://twitter.com/ProfHughLewis/status/1336584183783837696
https://twitter.com/ProfHughLewis/status/1304426186320347138
https://twitter.com/ProfHughLewis/status/1385535670094008326
https://twitter.com/ProfHughLewis/status/1385537020789989378

He's referring to immediate obliteration of the shell. His last tweet refers to the *current* environment below 600 km. But of course you're looking for slogans, not for nuanced scientific debate, for which you already showed your disdain plenty of times.

No, I didn't mispresenting his words, none of the tweets you quoted contradicts what I said, which is in his opinion, Starlink won't cause rapid cascade collisions like you described in your original comment, he literally said exactly this here:

https://twitter.com/ProfHughLewis/status/1380655591094374403
https://twitter.com/ProfHughLewis/status/1380658489421402115

6- You're free to consider it as conservative as you want, the numbers are what they are. Plus: nyet, those numbers of 3 per satellite-year are for around 800 satellites. You don't have data to say how many they're performing now with 1500, although it's probably around 10.

The twitter thread actually touched this later on:

https://twitter.com/ProfHughLewis/status/1436334166447173640

So with 1,500 satellites, they have about 130 conjunctions with Pc > 1 in 100,000 per week, that's 6,760 per year or 4.5 collision avoidance maneuver per year per satellite, so not much different from what I said.

Still, that sounds like a small number until you realize that's a statistical process we're talking about, and the absolute numbers are big, ensuring probability will take advantage of any mistake: around ten thousand a year with a 10th of the initial constellation built up, and a non-linear increase Prof Lewis is showing in SOCRATES data. A complete minimal (12000) system will have millions of conjunctions a year and similar orders of magnitudes of CAMs. If we consider smooth nominal operations forever, everything is and will be fine obviously - here we're talking about uncertainties, mistakes or oversights, and their consequences.

Actually I don't see why there would be a non-linear increase of conjunctions. There's no reason that conjunctions between Starlink and non-Starlink satellites would increase non-linearly if you only increase # of Starlink satellites while keeping # of non-Starlink satellites unchanged.

It appears non-linear in Dr. Hugh Lewis' chart because he's using dates as x-axis, so his curve is affected the launch cadence SpaceX is using, and this cadence itself is increasing non-linearly, for example there're 13 launches between 12/1/2020 to 5/31/2021, but only 9 between 6/1/2020 to 11/30/2020, this difference would be able to explain the non-linear trend he's seeing.

As for the simulations: please do so yourself before asking others to develop dissertation-grade work for you. I'm taking real-world number disclosed by SpaceX themselves and showing what they mean mathematically for different degrees of extrapolation. You're misconstruing an expert's words stripping them of their nuances and twisting the concluding message in order to advance your agenda, as you've done before most notoriously with Dr McDowell's points

I'm not the person who said Starlink will have a runaway collision cascade in 10 years, that's all on you, you said it you'll need to prove it.

And I'm not misconstruing anything, you're the one who's trying to make 1+1=3 even when expert said 1+1=2.

[envy887]

Could you list some of those reasons? Common sense would indicate hundreds or thousands of satellites at roughly the same orbital height but mostly differing velocity vectors (except those in the same plane) would be exceptionally vulnerable to a debris field squarely intersecting that height - especially if centered on it.

After a collision, approximately none of the fragments have exactly the same velocity as the original satellite. The impact is impulsive, which means they are now in new orbits that intersect the original only at most twice per orbit. That means the opportunity to re-collide with another satellite in the shell happens at most twice per 90 minutes.

That isn't the general case. It will depend on the geometry of the collision. Even for forward-scattered collisions, the distributions end up having a strong double-lobed distribution of resulting objects, with some reaching higher apogees, some lower, a scatter of inclinations, but most staying right around the original orbital altitude and inclination (see first attached image from FY-1C's ASAT test, where the collision was highly deboosting since the warhead was extremely slow compared to the satellite's orbital velocity). Then you get precession and orbital perturbations (solar, atmospheric, n-body) that will spread the debris cloud around in a few months' time, which is within the resulting debris' lifetimes in orbit (at the initial height, of course there will be those higher): an example is the spread of the Iridium-Kosmos debris clouds from tight distributions at their initial approximately 90-degree-apart respective orbits, and the homogeneous distribution some months later - see second and third attached images.

It is the general case, and your plots show it. The constellation shell is (effectively) a spherical surface. Anything intersecting but not continuously on that surface can only ever collide at the intersection points. Anything not on and not intersecting the surface cannot collide at all, even if it's orbiting nearby in an ominous-looking and brightly colored "cloud".

An object with a different period is not on that spherical surface. An object with a different apogee or perigee is not on that surface. On the Gabbard plot, only objects exactly at the intersection of the horizontal and slanted lines are in the shell. Most objects are not at that intersection.

Drag means that even objects in nearly circular orbits pass through the shell surface quickly. The overall lifetime is irrelevant - only the time spent at a particular altitude band matters. An object in a 549 km circular orbit cannot ever collide with the 550 km circular shell, even though it may have a lot of 1 km passes. So as soon as those objects decay only a few km they aren't a threat to the operational satellites. 1 km of decay happens in hours or a few days, depending on the object's ballistic coefficient, and also on solar activity.

The vast majority objects are going to be intersecting, but not in, the shell. That means at most 2 opportunities for a collision per orbit. With only 16 orbits per day, those aren't going to add up very fast - even if it take several weeks to update the catalog and several months to decay to non-intersection.

If you want to estimate the number of significant intersecting objects generated, the probability of collision per orbit, and the mean time to decay to non-intersection, we can estimate the time to recollision. Say that's 100 fragments, a 1e-7 probability of collision per orbit, and a mean of 3 months to decay to non-intersection, then the probably of a recollision is about 16 orbits/day*30 days/mo*3 mo*100 fragments *1e-7 collisions/orbit = ~1.5%.

Finally I understand what you're referring to, I think - and I'm afraid you're misunderstanding the dynamics here.
You're assuming SpaceX has kind of a color-coded board with all satellites in green (probability of collision close to 0), and slowly some are becoming progressively redder, until reaching bright red (probability of 1). Once they become orange though, they have the ability to slightly maneuver them so that none gets beyond a certain tone of orange (1e-5) which they set as threshold. That indeed would be the ideal situation. It *may* be for inter-Starlink conjunctions, because they have their own GNSS packages, as you mention in your following post, and that *may* be enough to approach that idealization - we'll have to take their word for it, which is not ideal anyway.

Unfortunately, that's not the case for the rest of objects, simply because of the innacuracies of the TLEs and the non-deterministic refinement of tracking. You don't get continuous perfect tracking that allows you to steer clear as soon as the collision probability reaches 1e-5 and not beyond. You get more or less frequent updates with "jumps" in the parameters, and most of those will instantaneously change the collision probability by a lot. Most will stay very low or 0, but some will jump well above 1e-5 immediately, with a distribution that rarely reaches 1 (or 1e-1 for that matter), yet will include lots of 1e-4, 5e-3, 1e-3 and so forth. It's the fundamental difference between a continuous, well-sampled process with low uncertainties, and a discrete, undersampled process with high uncertainties.

The uncertainties are the only reason the probability of collision is not always either exactly 0 or exactly 1. The probability reflects the uncertainties. Conjunctions are predicted several days out. If the uncertainties are still high as the conjunction approaches, they will maneuver so that the prediction (including the uncertainties) drops below threshold.

This progression of the predicted probability of collision and associated uncertainties isn't continuous, but it's definitely a progression.

[ulm_atms]

Whoa guys, you are getting very worked up here.

1.  I believe that the regulators/SpaceX are trying their best.  This is the first "MEGA constellation" ever put up and at an eye popping pace too.
2.  Well you got to start somewhere and at a minimum, SpaceX seems to be taking this whole thing collision avoidance issue seriously.  They seem to be constantly working/tweaking it as new data/observations come in.  I'm not sure who they could really ask as #1 means they really don't have anyone TO ask as they are first even attempting something of this size.
3.  They specifically made their sats to de-orbit quickly(lowered the orbits of shells) in case of failure/issues and to fully burn up before they hit the ground.  Who else is doing that? (Here is looking at you OneWeb...your sats won't decay for a LONG time where you are placing them currently.....)

I guess what I am trying to say is give them time guys.  You have 1/2 the forum declaring doom and the other 1/2 declaring a rosy future.  Reality is probably some where in between like always.  At least we can be grateful SpaceX seems to be taking every effort to balance an "efficient" but safe rollout.  The FCC/Space Command seems to be ok with what they are doing anyways....which is why I am not in the worse case side.  If they were sending out "Warning!, Warning!, Warning!  That is going to end badly because SpaceX won't listen to us!!!" I would be much more concerned.

Do the numbers say there will be an impact/collision at some point...sure...but it's how hard you try to avoid that situation and steps taken to keep that from happening in the first place that matters most at this juncture.  Once again, no one has done something of this size before....they(SpaceX, FCC, Space Command) have to start somewhere IMO.

[abaddon]

Whoa guys, you are getting very worked up here.[...]
I guess what I am trying to say is give them time guys.

You are summarizing the general feeling of most of the posters on this thread and site.  However, there are some who vehemently and voluminously disagree.

I think it'd be better to have a thread where discourse on the dangers of large-scale LEO constellations can be discussed that is not quite so single-operator (SpaceX) focused, but instead includes >all< constellations (Starlink, OneWeb, Kuiper, Telesat).  Similar to the current thread in the Commercial space forum where the topic is limited to impacts specific to Astronomy.

[mandrewa]

On the other hand, they may be expecting SpaceX to reduce tracking errors with their own 'Space Fences', or lowering the threshold an order of magnitude as Starlink grows, or some software wizardry that hopefully will solve space traffic control congestion in a few years' time because it'll become highly error-free and automated, or any number of 'let's wait and see's.

SpaceX doesn't need a "Space Fence" for their own satellites. They have highly accurate data from the INS/GPS/star trackers on their satellites and can use this to avoid Starlink fratricide. Starlink on other satellite and Starlink on debris conjunctions are the main concerns, but I'm not sure that anyone realistically expects SpaceX to start tracking everything else in orbit.

And hopefully we are going to set things up so that every constellation has the accurate location and orbit of every other functioning satellite in every other constellation in the sky (and I hope OneWeb and Starlink are doing this right now) with any status changes being updated almost immediately.

So that none of these functional satellites are a risk to each other.

The problem then becomes the probability of collision with the uncontrolled satellites or the debris there from whose locations and velocities are not known so precisely and which cannot be changed.

[eeergo]

It is the general case, and your plots show it. The constellation shell is (effectively) a spherical surface. Anything intersecting but not continuously on that surface can only ever collide at the intersection points. Anything not on and not intersecting the surface cannot collide at all, even if it's orbiting nearby in an ominous-looking and brightly colored "cloud".

An object with a different period is not on that spherical surface. An object with a different apogee or perigee is not on that surface. On the Gabbard plot, only objects exactly at the intersection of the horizontal and slanted lines are in the shell. Most objects are not at that intersection.

Drag means that even objects in nearly circular orbits pass through the shell surface quickly. The overall lifetime is irrelevant - only the time spent at a particular altitude band matters. An object in a 549 km circular orbit cannot ever collide with the 550 km circular shell, even though it may have a lot of 1 km passes. So as soon as those objects decay only a few km they aren't a threat to the operational satellites. 1 km of decay happens in hours or a few days, depending on the object's ballistic coefficient, and also on solar activity.

The vast majority objects are going to be intersecting, but not in, the shell. That means at most 2 opportunities for a collision per orbit. With only 16 orbits per day, those aren't going to add up very fast - even if it take several weeks to update the catalog and several months to decay to non-intersection.

If you want to estimate the number of significant intersecting objects generated, the probability of collision per orbit, and the mean time to decay to non-intersection, we can estimate the time to recollision. Say that's 100 fragments, a 1e-7 probability of collision per orbit, and a mean of 3 months to decay to non-intersection, then the probably of a recollision is about 16 orbits/day*30 days/mo*3 mo*100 fragments *1e-7 collisions/orbit = ~1.5%.

Evidently I agree that the pieces in the debris cloud susceptible to collide against another satellite only intersect at two points in their orbit with that satellite, at most - apologies if I made it sound like that basic concept wasn't clear. But we're not dealing with just one satellite! There's several thousands of them at the same altitude, with slightly or completely different orbits, that precess at different rates than the debris field objects, and are affected differently by perturbations. Many have larger eccentricities than the parent body (and those which do not are then in the same circular shell as the parent body, for the most part, so can potentially affect other satellites at the same altitude all around their orbit - and those are not just a few, as you can see from the plot) and will take days or weeks to decay out of that regime (at 550 km you don't fall 1 km in a few hours), during which time cataloguing them will still be ongoing, which means little chance for CAMs. It's true each piece will only pass roughly 16 times a day through range of altitudes the constellation is at, but there are thousands of satellites and a large distribution of hundreds of thousands, if not millions, of objects capable of inflicting catastrophic damage intersecting the "shell", which adds up to (10^-2 seconds x 10^5 objects) much longer times, in the order of hours.

Regarding your numbers, I believe those are unrealistically optimistic don't you think? Most orbital collisions involve thousands of trackable debris (plus remember the huge gap between trackable >10 cm and not overly harmful <1 cm), not 100. Also, lumping all fragments into the umbrella "1e-7 POC" seems a bit arbitrary, doesn't it?

The uncertainties are the only reason the probability of collision is not always either exactly 0 or exactly 1. The probability reflects the uncertainties. Conjunctions are predicted several days out. If the uncertainties are still high as the conjunction approaches, they will maneuver so that the prediction (including the uncertainties) drops below threshold.

This progression of the predicted probability of collision and associated uncertainties isn't continuous, but it's definitely a progression.

Sure, but you're sidestepping from your false statement: you had just claimed 1e-5 was the upper limit for collision probability that a conjuction would ever reach during that discontinuous progression.

1e-5 is the ceiling. There will be no conjunctions above that

As you can see from the tweet kindly quoted by su27k above (ProfHughLewis/status/1436334166447173640, which I'd earlier overlooked), that's not the case - there are as many conjunctions with 1e-5 as maximum POC as there are with 1e-3 and 1e-4, excluding Starlink-on-Starlink, which supposedly (but then again, maybe not) avoid going beyond 1e-5 with their more accurate internal GNSS tracking. They will be lowered as soon as they are discovered and the actual physical conjunction will not take place, if that's what you mean - but my initial point was that 1e-5 was not conservative if looking at what long-term POCs it would give rise to if chosen any tighter. To repeat myself from my previous post: "If we consider smooth nominal operations forever, everything is and will be fine obviously - here we're talking about misjudged uncertainties, mistakes or oversights, and their consequences."

[eeergo]

Whoa guys, you are getting very worked up here.[...]
I guess what I am trying to say is give them time guys.

You are summarizing the general feeling of most of the posters on this thread and site.  However, there are some who vehemently and voluminously disagree.

I think it'd be better to have a thread where discourse on the dangers of large-scale LEO constellations can be discussed that is not quite so single-operator (SpaceX) focused, but instead includes >all< constellations (Starlink, OneWeb, Kuiper, Telesat).  Similar to the current thread in the Commercial space forum where the topic is limited to impacts specific to Astronomy.

Such a thread exists for something much less generic (impacts to astronomy), yet most participants in the "nothing to worry about" camp firmly made all points gravitate around SpaceX/Starlink, and accuse critics of being anti-SpaceX/Musk/capitalism/whatever, even when most critiques are agnostic about the megaconstellation under discussion - although obviously Starlink being the largest by far, it's difficult to avoid mentioning it. In fact, remarkedly vehement disdain for "scientist elitism" sent the thread to moderation purgatory for a long while.

Regarding "give them time", that was the predominant discourse about mostly any environmental problem when it was still developing and was easy to tackle, even if excessively well-researched and understood as such. Solutions to evident issues cannot just rely on the goodwill of profiting parties under the belief that "they'll sort it out".

[envy887]

Regarding your numbers, I believe those are unrealistically optimistic don't you think? Most orbital collisions involve thousands of trackable debris (plus remember the huge gap between trackable >10 cm and not overly harmful <1 cm), not 100. Also, lumping all fragments into the umbrella "1e-7 POC" seems a bit arbitrary, doesn't it?

That's 1e-7 per orbit. Which implies 1e-5 per 100 orbits, which implies that every satellite needing a conjunction-avoiding maneuver about every 6 days. Prof. Lewis' data above suggests that the true rate is 20-fold lower - there are only 100 events reaching 1e-5 per week in a 2000-sat constellation. So 1e-8 to 1e-9 per orbit is more likely.

Starlinks are not large satellites, only ~300 kg. Mass is the main factor in determining how a satellite participates in a collision cascade, because it determines the size and number of fragments it can produce. Microsat-R produced 57 tracked fragments, plus probably more that size that decayed before tracking. It was estimated to produce 6500 fragments over 1 cm, BUT objects in the 1 cm range not large enough to be likely to break up a satellite. A 1 cm Al cube masses 0.0028 kg, and would eject 0.3 kg of material in a collision, on average, if Kessler's scaling of Langley research is accurate.
See: Impacts into Spacecraft Sturctures:
http://www.castor2.ca/07_News/headline_010216_files/Kessler_Collision_Frequency_1978.pdf

That implies that breaking up a second Starlink would require on average, a 2.2 kg fragment from the first, which would probably be roughly 1U cubesat sized and reasonably trackable.

If you go with 1e-8 Pc per orbit and 1000 fragments, you get the same answer. Fragments are overwhelming likely to drop through the shell without a catastrophic recollision. You might get some small-ish holes in solar arrays, or a couple unmaneuverable but mostly intact satellites that also decay, but it's quite unlikely you will get enough fragments to self-sustain the cascade. There's just way too much drag and not nearly enough targets.

Sure, but you're sidestepping from your false statement: you had just claimed 1e-5 was the upper limit for collision probability that a conjuction would ever reach during that discontinuous progression.

1e-5 is the ceiling. There will be no conjunctions above that

As you can see from the tweet kindly quoted by su27k above (ProfHughLewis/status/1436334166447173640, which I'd earlier overlooked), that's not the case - there are as many conjunctions with 1e-5 as maximum POC as there are with 1e-3 and 1e-4, excluding Starlink-on-Starlink, which supposedly (but then again, maybe not) avoid going beyond 1e-5 with their more accurate internal GNSS tracking. They will be lowered as soon as they are discovered and the actual physical conjunction will not take place, if that's what you mean - but my initial point was that 1e-5 was not conservative if looking at what long-term POCs it would give rise to if chosen any tighter. To repeat myself from my previous post: "If we consider smooth nominal operations forever, everything is and will be fine obviously - here we're talking about misjudged uncertainties, mistakes or oversights, and their consequences."

It's not false. 1e-5 is the celling at the time of the event. The Pc can vary before that, but those are just projections that feed into the decision on whether to maneuver. Only the Pc at the time of the event, after including for reductions in uncertainty due to additional observations and/or any maneuvering, matters for calculating actual collisions. If you calculate 1e-4 and maneuver so that at the event it's 1e-6, then the 1e-4 is irrelevant to whether there would actually be a collision.

And yes, this assumes the system works. But that's a fairly reasonable assumption, since the available response time (from first conjunction warning to event) is measured in days. If part of the system breaks, there is time to fix a lot of problems before satellites start crashing into each other.

[matthewkantar]

What entity has the most to lose from a collision or a cascade of them? Does this entity have any smart folks working for them?

[eeergo]

Regarding your numbers, I believe those are unrealistically optimistic don't you think? Most orbital collisions involve thousands of trackable debris (plus remember the huge gap between trackable >10 cm and not overly harmful <1 cm), not 100. Also, lumping all fragments into the umbrella "1e-7 POC" seems a bit arbitrary, doesn't it?

That's 1e-7 per orbit. Which implies 1e-5 per 100 orbits, which implies that every satellite needing a conjunction-avoiding maneuver about every 6 days. Prof. Lewis' data above suggests that the true rate is 20-fold lower - there are only 100 events reaching 1e-5 per week in a 2000-sat constellation. So 1e-8 to 1e-9 per orbit is more likely.

Now we're redefining the timescale of the POC?! The POC is instantaneous, not per orbit, not per 100 orbits, not per week. It reflects the probability of collision *at the time of the conjunction*, as you said above based on an assessment of the known uncertainties in the trajectories. You get a rate of POCs at each threshold, which can be measured in hours, days, weeks, months, years or whatever timescale you want. You get a CDM update based on new tracking of Starlinks and other trackable objects, and that gives you a set of POCs for every Starlink with respect to every other object, every time they come remotely close to each other. Most will be infinitesimal, some will time and again reach some threshold: Prof Lewis' data shows that around *once a week* they get one or two POC predictions of at most 1e-3, ten or twenty between 1e-4 and 1e-5, and about a hundred with maximum POC of 1e-5. But you can't say "well, that's for a week, so if you want the probabilities per day it's about an order of magnitude less, and per-orbit risks are off-scale low": the POC is what it is until it gets further refined at each new tracking update, and refers to the projected future time of closest approach.

Sure, but you're sidestepping from your false statement: you had just claimed 1e-5 was the upper limit for collision probability that a conjuction would ever reach during that discontinuous progression.

1e-5 is the ceiling. There will be no conjunctions above that

As you can see from the tweet kindly quoted by su27k above (ProfHughLewis/status/1436334166447173640, which I'd earlier overlooked), that's not the case - there are as many conjunctions with 1e-5 as maximum POC as there are with 1e-3 and 1e-4, excluding Starlink-on-Starlink, which supposedly (but then again, maybe not) avoid going beyond 1e-5 with their more accurate internal GNSS tracking. They will be lowered as soon as they are discovered and the actual physical conjunction will not take place, if that's what you mean - but my initial point was that 1e-5 was not conservative if looking at what long-term POCs it would give rise to if chosen any tighter. To repeat myself from my previous post: "If we consider smooth nominal operations forever, everything is and will be fine obviously - here we're talking about misjudged uncertainties, mistakes or oversights, and their consequences."

It's not false. 1e-5 is the celling at the time of the event. The Pc can vary before that, but those are just projections that feed into the decision on whether to maneuver. Only the Pc at the time of the event, after including for reductions in uncertainty due to additional observations and/or any maneuvering, matters for calculating actual collisions. If you calculate 1e-4 and maneuver so that at the event it's 1e-6, then the 1e-4 is irrelevant to whether there would actually be a collision.

Of course at the time of the event 1e-5 is the ceiling, and even lower, but only because they have maneuvered in the meantime - that's the whole point of a CAM! In fact, *at the precise time of the event* the POC will be 0 or 1, because the uncertainties will no longer be there. As I've repeated in multiple ways many times already, my message was aiming to show that 1e-5 is not conservative because tightening that threshold (say, to 1e-4) would mean they wouldn't maneuver for that threshold, and that would imply the cumulative probability numbers in my opening post -i.e. unacceptably high-, which are not conservative at all once you have more than a few thousand satellites in a shell!

This constant eye-tarring and thread-splitting is mind-numbing, and as always smears the argument my exhaustion...

[Twark_Main]

my message was aiming to show that 1e-5 is not conservative

Just out of curiosity, what number would need to be substituted instead of 1e-5 such that it would be considered "conservative," in your opinion?

[Mark K]

The "Probability of Collision" is simply the area of the future orbit paths of the satellite divided by the future orbit paths of the fragment to collide. The fact that 10 meters is used for the size of the satellite and debris already means it is a very high estimate  (Something over a hundred times high just from that) . It is a probability at a given time based on our knowledge -at that time- so it is -not- a usefully integrable property. That is, there is no reason to favor any collisions over say, a century, because 1000000 events with 1 in 100000 "POC" are happening. There might be zero if we had just better orbital measurements. Or there might be 100. It is really more a mark of our prediction than a real probability. It is making an assumption that every similar path in future movement cone of the collision object is the same probability for example.

It IS a true signal that chances are higher of a collision when the POC is higher, so it is better than precise, it is useful. That is why it is used as a flag for active action.

When models are created to mimic planned constellations, they show that there isn't  over unity feedback, that one collision causes another one. 

[dondar]

so much emotions here.

Many.

"conjunctions" are not a thing within starlink system.They are events of possible orbit intersection with other satelites (or for now between rising starlinks from different batches.)

They are as in any satellite. Between different heights and at the same height alike. They are reportedly minimized automatically based on SpaceX tracking -of which we know little or nothing about, but let's take it at face value and say they're not an issue (even if they clearly exist as shown by the SOCRATES data). There are still plenty of conjunctions <1 km with other objects as shown in the Twitter thread, and within those many that SpaceX judges worthy of executing a CAM for (~10000/year currently).

Probably you should learn what Socrat does, and how satellites operate. When SpaceX determines their satelites in the risk of collision (calculated not by Socrat) they move their sat "out of possible harm". They do it 100s times per day (sometimes). It's that simple.

SOCRAT counts also satellites in formation as a collision candidates (Because the system is simplified, read STUPID).

What are you talking about? You're calling SOCRATES stupid?\

Yes I do. So let check SOCRAT.
http://celestrak.com/SOCRATES/search-results.php?IDENT=NAME&NAME_TEXT1=&NAME_TEXT2=&ORDER=MINRANGE&MAX=10

I make search on min range to 10 satellites.
First place: TIANZHOU-2 (big surprise). 2 places ORbcomm (within one formation=>see relative speeds), 5 starlink junctions (3 of which are questionable) and one real thing.
(applause).

The probabilities even for the conjuctions with min 0.1km can be still in the range off 1e-7. And there are many of those.

They can. I did not conflate both. I did take however SpaceX's CAM numbers, and they state only events with probability >1e-5 require action, so all those CAMs are over 1e-5. All 2200 of them. Some much higher probably (continuous distribution from 1e-5 all the way to 1). I did some elementary calculations for longer times and/or satellite numbers.

They don't use SOCRAT. They do a lot of corrections indeed.
Anyway, Socrat calculates that orbcomm sats have collision probability 1.e-2., i.e. those flying within formation.... When I've checked last time they had starlinks with probability of e.-1 (within formation as well)....

Right now Starlink sit strongly under 1% per year for their system. Totally. Please don't comment, especially "doom" without proper numbers or ability to defend scientifically your opinion.

Excuse me, where are your proper numbers? You just said you'll not go into SOCRATES data? Neither basic multiplication because it's "estimated"? I explicitly showed my work, you're pulling a 1% out of your hat (seriously, where in the world do you deduce that from? By their own admission they did 2200 CAMs in 6 months when Starlink had half the satellites it has now, and now it has 1500 birds...). Where exactly are my extrapolations of their own CAM numbers and probability thresholds wrong?

Talk about emotional responses...

My numbers come from the check  of the extensive number of reports on the subject.
And the number is very conservative. (it was actually less than .01% using current "failure" rates which are reported by SpaceX to FCC). And the numbers don't take into account active collision mitigation.

I would like to mention that SpaceX made specific contract with NASA specifically concerning avoidance of the collisions between SpaceX and NASA objects. I didn't bother to search (this one was already in my collection), but I can bet SpaceX made identical contract with Air Force.  No noise from these guys. Why???

The continuous FUD about any possible possibility of some "rushed" deployment, "extensive risk" etc. is FUD.
There is no other case in the  whole FCC history of any so scrupulously analyzed project and so much bias, partizan BS and outward hostility around it. With no real basis which can be controlled, assessed and confirmed.
Get rest.

[su27k]

Viasat continues to spread FUD about Starlink:

https://twitter.com/pbdes/status/1446013178228494336

.@Viasat chairman Mark Dankberg devoted his entire keynote at Oct 6 at @SatNewsEvents Satellite Innovation conference to debris threat in LEO. No broadband, no market assessment.... If Dankberg believes what he said here -- https://bit.ly/3nzvIGG -- he had no choice.

https://twitter.com/jeff_foust/status/1445811202639032333

“The failure mode of orbital debris is effectively irreversible,” says Viasat’s Mark Dankberg at #satinnovation, a reference to the Kessler Syndrome.

Dankberg is spending his keynote going into details about models of the LEO environment, suggesting a Kessler Syndrome within a few decades of the launch of a full Starlink-type (30,000 satellites around 600 km). He adds the models may be optimistic…

[su27k]

Viasat's claim being refuted by space debris expert on twitter:

1. Refuting "“The failure mode of orbital debris is effectively irreversible,” says Viasat’s Mark Dankberg at #satinnovation, a reference to the Kessler Syndrome.":

https://twitter.com/ProfHughLewis/status/1445828422006149120

I'm not sure why this view is held so widely. Remove objects/mass to limit the growth rate and ultimately to prevent growth. #SpaceDebris #KesslerSyndrome

2. Refuting "Dankberg is spending his keynote going into details about models of the LEO environment, suggesting a Kessler Syndrome within a few decades of the launch of a full Starlink-type (30,000 satellites around 600 km). He adds the models may be optimistic…":

https://twitter.com/ProfHughLewis/status/1445829764229394432

It is not possible to say this with any certainty. Much depends on the parametrisation and model assumptions. I suspect these need to be stretched quite considerably to produce a #SpaceDebris #KesslerSyndrome outcome at low LEO altitudes.

At 1300 km altitude the time for an orbital object to decay due solely to atmospheric drag is more than 1000 times longer than the time for the same object to decay from 550 km altitude.

It is also likely that at 1300 km altitude we have already exceeded the critical number of large orbital objects required for runaway (exponential) population growth. The situation is better at lower altitudes, but is still of concern.

The assumption also being made in this statement is that a 30,000-satellite constellation cannot respond to conjunction alerts. The overall number of satellites is important but this is not the only factor. Orbit control capability is at least as important.

[Sorry for random approach to this thread] so, all other things being equal, a failed satellite at 1300 km will have an area-time product (a proxy for collision probability) approximately 1000 times greater than a failed satellite at 550 km.

Continuing the thought experiment, the lifetime risk from 300 satellites at 1300 km would be equivalent to the risk from 300,000 satellites at 550 km. It's a little simplistic, but makes the point. It also highlights the problems arising when you ignore orbit control capability

[eeergo]

SOCRAT counts also satellites in formation as a collision candidates (Because the system is simplified, read STUPID).

What are you talking about? You're calling SOCRATES stupid?

Yes I do. So let check SOCRAT[ES].
http://celestrak.com/SOCRATES/search-results.php?IDENT=NAME&NAME_TEXT1=&NAME_TEXT2=&ORDER=MINRANGE&MAX=10

I make search on min range to 10 satellites.
First place: TIANZHOU-2 (big surprise). 2 places ORbcomm (within one formation=>see relative speeds), 5 starlink junctions (3 of which are questionable) and one real thing.
(applause). [...] My numbers come from the check of the extensive number of reports on the subject.

So just over three short months after this damning assertions, here you go:

https://twitter.com/ProfHughLewis/status/1478312127198810112

The SOCRATES system was actually underestimating (!) the POC, going by the number of actual CAMs performed by Starlink S/C, as disclosed by its operator in its own reports.

Regarding the rates, there have been around 6500 CAMs this past year, and we're already well over 150 CAMs/week, or about 20 per day. By the end of 2020 (not 2021), "Starlink was responsible for 54% of the conjunction data output by the 18 SPCS" (https://amostech.com/TechnicalPapers/2021/SSA-SDA/Hiles.pdf), and that was for at least an order of magnitude less CAMs:

https://twitter.com/ProfHughLewis/status/1478315799257530371

A related assessment shows 57% of *all* trackable conjunctions <1 km are due to Starlink, in stark contrast to ominous debris clouds such as that resulting from the recent Kosmos-1408 ASAT test (8%).

For additional context, at this point in time, 1.5% of Starlinks currently on orbit are failed and decaying (this means explicitly uncontrolled - if we add those under controlled disposal, that number rises to 2%). Failure rate, including already-deorbited satellites but excluding v0.9/Tintins, is around 9%, even when including recently-launched batches with still short on-orbit lives.

[Robotbeat]

Viasat's claim being refuted by space debris expert on twitter:

1. Refuting "“The failure mode of orbital debris is effectively irreversible,” says Viasat’s Mark Dankberg at #satinnovation, a reference to the Kessler Syndrome.":

https://twitter.com/ProfHughLewis/status/1445828422006149120

I'm not sure why this view is held so widely. Remove objects/mass to limit the growth rate and ultimately to prevent growth. #SpaceDebris #KesslerSyndrome

2. Refuting "Dankberg is spending his keynote going into details about models of the LEO environment, suggesting a Kessler Syndrome within a few decades of the launch of a full Starlink-type (30,000 satellites around 600 km). He adds the models may be optimistic…":

https://twitter.com/ProfHughLewis/status/1445829764229394432

It is not possible to say this with any certainty. Much depends on the parametrisation and model assumptions. I suspect these need to be stretched quite considerably to produce a #SpaceDebris #KesslerSyndrome outcome at low LEO altitudes.

At 1300 km altitude the time for an orbital object to decay due solely to atmospheric drag is more than 1000 times longer than the time for the same object to decay from 550 km altitude.

It is also likely that at 1300 km altitude we have already exceeded the critical number of large orbital objects required for runaway (exponential) population growth. The situation is better at lower altitudes, but is still of concern.

The assumption also being made in this statement is that a 30,000-satellite constellation cannot respond to conjunction alerts. The overall number of satellites is important but this is not the only factor. Orbit control capability is at least as important.

[Sorry for random approach to this thread] so, all other things being equal, a failed satellite at 1300 km will have an area-time product (a proxy for collision probability) approximately 1000 times greater than a failed satellite at 550 km.

Continuing the thought experiment, the lifetime risk from 300 satellites at 1300 km would be equivalent to the risk from 300,000 satellites at 550 km. It's a little simplistic, but makes the point. It also highlights the problems arising when you ignore orbit control capability

Viasat looks pretty dang bad doing this. But it makes sense. Starlink is an existential threat to Viasat, at least their consumer business. Who would use lower speed and much higher latency GSO internet when you can use something better than many cable connections? And that’s going to get better over time as they blend in satellites with laser links…

The only option is lowering the price to basement bargain levels, but that’s not very sustainable.

So of course no point in Viasat doing market assessments during a keynote. The market assessment is that Viasat is all but dead, so you’ve got to spread FUD about LEO megaconstellations.

[Robotbeat]

The lower the altitude, the better for reducing space debris risk. I wonder if very low satellites might end up doing aerodynamic maneuvers with the solar panels (as I think ISS has done a little bit?), since that is faster than using electric thrusters. If designed for it, could be somewhat more efficient as well.

I also wonder if SpaceX will start investing in their own tracking capabilities for space debris in order to reduce the uncertainty and reduce the need for maneuvering.

[Robotbeat]

Also, if conjunction risk continues to be a major problem, it might be interesting if the satellites not only maneuver to minimize cross section but actually retract their solar panels before maneuvering to an edge-on orientation, flat-pack like they deployed from, maybe even with a Whipple shield in front.

Increasing survivability. Perhaps a similar maneuver could be used to minimize visibility when passing in front of Vera Rubin.

[envy887]

SOCRAT counts also satellites in formation as a collision candidates

So just over three short months after this damning assertions, here you go:

The SOCRATES system was actually underestimating (!) the POC, going by the number of actual CAMs performed by Starlink S/C, as disclosed by its operator in its own reports.

Which has nothing at all to do with SOCRATES counting Starlink-on-Starlink conjunctions. SpaceX can handle Starlink-on-Starlink conjunctions internally was far more speed and precision the the external process. Aside from producing more data to look at, those have no impacts on anyone else.

For additional context, at this point in time, 1.5% of Starlinks currently on orbit are failed and decaying (this means explicitly uncontrolled - if we add those under controlled disposal, that number rises to 2%). Failure rate, including already-deorbited satellites but excluding v0.9/Tintins, is around 9%, even when including recently-launched batches with still short on-orbit lives.

That is quite misleading. Out of the last 1,511 satellites launched, only 3 are presently failed and decaying out of control, a rate under 0.2%. The vast majority of those have been operating for 6 to 12 months, so infant mortality period has largely passed.

The other 18 failed and decaying v1.0 satellites are from the first 420 launched, a 4.3% failure rate. This is not just due to the bathtub curve (they were only launched 18-25 months ago), but pretty much what one would expect with SpaceX's iterative approach. The MTBF is markedly increasing as the design and manufacturing improve.

Satellites disposed of early are not failures for debris or collision avoidance purposes. Disposal is part of the mission for every satellite, and some of them getting there early has no impact on debris generation.

[eeergo]

The SOCRATES system was actually underestimating (!) the POC, going by the number of actual CAMs performed by Starlink S/C, as disclosed by its operator in its own reports.

Which has nothing at all to do with SOCRATES counting Starlink-on-Starlink conjunctions. SpaceX can handle Starlink-on-Starlink conjunctions internally was far more speed and precision the the external process. Aside from producing more data to look at, those have no impacts on anyone else.

Data is WITHOUT Starlink-on-Starlink (ref: ProfHughLewis/status/1478314752061358089), as should be evident since SpaceX should ideally never have to require a POC among its own satellites, which can be continuously maneuvered to avoid triggering the 1e-5 threshold. I was responding to dondar's statement that SOCRATES was stupid and grossly overestimating concerns with Starlink conjunctions, which has been proven not to be the case.

For additional context, at this point in time, 1.5% of Starlinks currently on orbit are failed and decaying (this means explicitly uncontrolled - if we add those under controlled disposal, that number rises to 2%). Failure rate, including already-deorbited satellites but excluding v0.9/Tintins, is around 9%, even when including recently-launched batches with still short on-orbit lives.

That is quite misleading. Out of the last 1,511 satellites launched, only 3 are presently failed and decaying out of control, a rate under 0.2%. The vast majority of those have been operating for 6 to 12 months, so infant mortality period has largely passed.

The other 18 failed and decaying v1.0 satellites are from the first 420 launched, a 4.3% failure rate. This is not just due to the bathtub curve (they were only launched 18-25 months ago), but pretty much what one would expect with SpaceX's iterative approach. The MTBF is markedly increasing as the design and manufacturing improve.

Satellites disposed of early are not failures for debris or collision avoidance purposes. Disposal is part of the mission for every satellite, and some of them getting there early has no impact on debris generation.

There was a similar discussion in another thread a year or so ago. The argument went that the failure rate of early launches was surely not comparable to later batches because, wait and see, not only were they *already* visibly much better, the fact that *they had just been launched* (and so didn't have time to fail yet, discounting a minor infant mortality rate) was not important.

Your statement is misleading in that it arbitrarily takes a hypothesis (new SC are more reliable) and picks a trivially biased dataset to back it up. My statement is just objective: of all Starlinks branded as "operational" that are flying at the moment, which are <2 years old (!), and discounting reportedly "experimental" ones (v0.9), 1.5% are uncontrollable, and 2% are being disposed of *at this moment in time*.

I purposely did not take into account early deorbits for that (1.3% of the 1511 subgroup you arbitrarily picked out) or semicontrolled deorbits (3%), so your last point is just a strawman.

Anyway, this point was for context, since this thread is about collision risks and not general QC. The point about SOCRATES, conjunctions and CAMs stands.

[envy887]

Data is WITHOUT Starlink-on-Starlink, as should be evident since SpaceX should ideally never have to require a POC among its own satellites, which can be continuously maneuvered to avoid triggering the 1e-5 threshold. I was responding to dondar's statement that SOCRATES was stupid and grossly overestimating concerns with Starlink conjunctions, which has been proven not to be the case.

Prof Lewis is filtering some of his data to eliminate Starlink-on-Starlink. SOCRATES otherwise includes all the Starlink-on-Starlink conjunctions, something you didn't note for these 2 claims

By the end of 2020 (not 2021), "Starlink was responsible for 54% of the conjunction data output by the 18 SPCS")
...
A related assessment shows 57% of *all* trackable conjunctions <1 km are due to Starlink, in stark contrast to ominous debris clouds such as that resulting from the recent Kosmos-1408 ASAT test (8%).

and the latter number specifically includes Starlink-on-Starlink...
https://twitter.com/ProfHughLewis/status/1478052682514452480

For additional context, at this point in time, 1.5% of Starlinks currently on orbit are failed and decaying (this means explicitly uncontrolled - if we add those under controlled disposal, that number rises to 2%). Failure rate, including already-deorbited satellites but excluding v0.9/Tintins, is around 9%, even when including recently-launched batches with still short on-orbit lives.

That is quite misleading. Out of the last 1,511 satellites launched, only 3 are presently failed and decaying out of control, a rate under 0.2%. The vast majority of those have been operating for 6 to 12 months, so infant mortality period has largely passed.

The other 18 failed and decaying v1.0 satellites are from the first 420 launched, a 4.3% failure rate. This is not just due to the bathtub curve (they were only launched 18-25 months ago), but pretty much what one would expect with SpaceX's iterative approach. The MTBF is markedly increasing as the design and manufacturing improve.

Satellites disposed of early are not failures for debris or collision avoidance purposes. Disposal is part of the mission for every satellite, and some of them getting there early has no impact on debris generation.

There was a similar discussion in another thread a year or so ago. The argument went that the failure rate of early launches was surely not comparable to later batches because, wait and see, not only were they *already* visibly much better, the fact that *they had just been launched* (and so didn't have time to fail yet, discounting a minor infant mortality rate) was not important.

Your statement is misleading in that it arbitrarily takes a hypothesis (new SC are more reliable) and picks a trivially biased dataset to back it up. My statement is just objective: of all Starlinks branded as "operational" that are flying at the moment, which are <2 years old (!), and discounting reportedly "experimental" ones (v0.9), 1.5% are uncontrollable, and 2% are being disposed of *at this moment in time*.

I purposely did not take into account early deorbits for that (1.3% of the 1511 subgroup you arbitrarily picked out) or semicontrolled deorbits (3%), so your last point is just a strawman.

Anyway, this point was for context, since this thread is about collision risks and not general QC. The point about SOCRATES, conjunctions and CAMs stands.

The rate of QC improvement is critical to discussing collision risks. And that improvement is continuous, so any cutoff is arbitrary. I use the first 420 because McDowell does. https://planet4589.org/space/stats/star/starstats.html

Support for the hypothesis that new SC are more reliable is trivially easy to show. The 420 sats from the first launch have a mean age of about 22 months, for a total operational time of 9240 sat-months. They have 18 out of control, for 1 failure per 513 satellite-months of operation. The 1511 newer sats have been in orbit on average for about 9 months, for 13600 sat-months of operation. They have 3 out of control, or one per 4,533 sat-months. That is an 8.83-fold improvement in failures per satellite-month of operations. You can fiddle with the numbers all you want, but you aren't going to twist a nearly order of magnitude improvement in MTBF down to nothing.

Going forward, the MTBF will almost certainly continue to improve as the design and manufacturing issues are sorted and newer, better satellites are launched. So your implication that the entire constellation will continue to operate at a 1.5% uncontrolled rate is at best unsupported FUD.

[eeergo]

Prof Lewis is filtering some of his data to eliminate Starlink-on-Starlink. SOCRATES otherwise includes all the Starlink-on-Starlink conjunctions, something you didn't note for these 2 claims.

By the end of 2020 (not 2021), "Starlink was responsible for 54% of the conjunction data output by the 18 SPCS")
...
A related assessment shows 57% of *all* trackable conjunctions <1 km are due to Starlink, in stark contrast to ominous debris clouds such as that resulting from the recent Kosmos-1408 ASAT test (8%).

and the latter number specifically includes Starlink-on-Starlink...

Please stop cherrypicking excerpts from my posts to twist what I quoted.

First I quoted the comparison between SOCRATES estimates for CAMs and actual CAMs executed by SpaceX. Those disproved the notion floated by dondar that SOCRATES was "stupid" and overinflating predictions. Indeed, SOCRATES is very accurate also when discounting self-conjunctions, as proven.

THEN I quoted a "related assessment" that showed 57% of all conjunctions (including self-conjunctions, as you noted, and as I quoted) in early 2022 were due to Starlink. Here the issue is not so much that there are many self-conjunctions, which SpaceX should be able to mitigate themselves, barring anomalies - but the fact that the system generates so many conjunctions in the first place, even 3x as many as an uncontrolled debris field containing about as many trackable objects.

Support for the hypothesis that new SC are more reliable is trivially easy to show. The 420 sats from the first launch have a mean age of about 22 months, for a total operational time of 9240 sat-months. They have 18 out of control, for 1 failure per 513 satellite-months of operation. The 1511 newer sats have been in orbit on average for about 9 months, for 13600 sat-months of operation. They have 3 out of control, or one per 4,533 sat-months. That is an 8.83-fold improvement in failures per satellite-month of operations. You can fiddle with the numbers all you want, but you aren't going to twist a nearly order of magnitude improvement in MTBF down to nothing.

Going forward, the MTBF will almost certainly continue to improve as the design and manufacturing issues are sorted and newer, better satellites are launched. So your implication that the entire constellation will continue to operate at a 1.5% uncontrolled rate is at best unsupported FUD.

Always refreshing to raise the good ol' "FUD" flag whenever things don't quite match up to reality. You accuse me of fiddling with numbers, but you are pulling a brutal "average" on-orbit timespan out of the hat, conflating it with significant numbers of just-launched satellites, and rubberstamping a foregone conclusion.

You're basically just employing a more mature, smaller dataset with a much larger and younger one, but treating them as equal through the "averaging" and the current fleet status. You might as well include early mortality in that, since you're leaving S/C that have already reentered in an uncontrolled fashion out, or that could easily have failed a few weeks later instead of right away. In that case:
* 18+8 = 26 S/C that will(have) reenter(ed) of the first 420 operational satellites, over 9240 months of "average operation" = 1 failure every 355 sat-months.
* 3+1+1+25 S/C that will(have) reenter(ed) of the latest 1511 operational satellites, over 13600 months of "average operation" = 1 failure every 544 sat-months.
So not quite the order-of-magnitude improvement you claim, and of course the bias of the "averaged out" relative youth remains, which is temporarily skewing the observed reliability upwards. Of course I also believe they will work obvious kinks out in their design if they keep iterating, but not so much as to offset by much the continuous tweaks for newer satellites, or the very design philosophy of cheap "disposable" S/C for megaconstellations.

[Robotbeat]

SpaceX’s Starlink satellites will likely become significantly more reliable than any other satellite bus, if they aren’t already, largely due to the large number of them that they’re producing. Smearing them as disposable (when almost all satellites are) is just FUD.

If you don’t like people using that term “FUD,” quit making it so apt.

People made similar claims about Falcon 9, now the most reliable launch vehicle flying, and perhaps of all time.

[envy887]

Support for the hypothesis that new SC are more reliable is trivially easy to show. The 420 sats from the first launch have a mean age of about 22 months, for a total operational time of 9240 sat-months. They have 18 out of control, for 1 failure per 513 satellite-months of operation. The 1511 newer sats have been in orbit on average for about 9 months, for 13600 sat-months of operation. They have 3 out of control, or one per 4,533 sat-months. That is an 8.83-fold improvement in failures per satellite-month of operations. You can fiddle with the numbers all you want, but you aren't going to twist a nearly order of magnitude improvement in MTBF down to nothing.

Going forward, the MTBF will almost certainly continue to improve as the design and manufacturing issues are sorted and newer, better satellites are launched. So your implication that the entire constellation will continue to operate at a 1.5% uncontrolled rate is at best unsupported FUD.

Always refreshing to raise the good ol' "FUD" flag whenever things don't quite match up to reality. You accuse me of fiddling with numbers, but you are pulling a brutal "average" on-orbit timespan out of the hat, conflating it with significant numbers of just-launched satellites, and rubberstamping a foregone conclusion.

You're basically just employing a more mature, smaller dataset with a much larger and younger one, but treating them as equal through the "averaging" and the current fleet status. You might as well include early mortality in that, since you're leaving S/C that have already reentered in an uncontrolled fashion out, or that could easily have failed a few weeks later instead of right away. In that case:
* 18+8 = 26 S/C that will(have) reenter(ed) of the first 420 operational satellites, over 9240 months of "average operation" = 1 failure every 355 sat-months.
* 3+1+1+25 S/C that will(have) reenter(ed) of the latest 1511 operational satellites, over 13600 months of "average operation" = 1 failure every 544 sat-months.
So not quite the order-of-magnitude improvement you claim, and of course the bias of the "averaged out" relative youth remains, which is temporarily skewing the observed reliability upwards. Of course I also believe they will work obvious kinks out in their design if they keep iterating, but not so much as to offset by much the continuous tweaks for newer satellites, or the very design philosophy of cheap "disposable" S/C for megaconstellations.

For collision purposes, why would you count satellites with inactive payloads that are still maneuvering, but not satellites with active payloads that are still actively maneuvering? As far as I can tell, those have exactly the same chance of a collision.

For example, only 1 satellite on v1.0 Launch 15 (see attached) stopped maneuvering before reentry. That is Starlink 1881, which is still in orbit and failed in Feb. 2021 at about 480 km. On that mission, 9 satellites have been disposed of (3 before reaching operation, and 6 after), and 7 more appear to be heading to their doom. But only Starlink 1881 is increasing the probability of collision. The other 16 "failures" appear to still be fully maneuverable right up to disposal.

McDowell doesn't separate out those that stopped maneuvering at low orbit from those were actively disposed of before orbit raising was complete. In most cases there is no appreciable difference, since the disposal orbit is about the same as the insertion orbit, and those largely reenter within a few days or weeks and again, have essentially no relevance to collisions because they aren't around long enough to run into anything.

I'd agree that those with lost maneuvering capability during orbit-raising, before reaching operational orbit but still high enough to hang around a while, should be counted as collision risks. But I only see one of those, Starlink 1756   on v1.0 L11 (Launch 12).

So that makes 4 out of the last 1,511 that have lost maneuvering capability at some altitude above disposal. Which still makes the newer satellites about an order of magnitude better in MTBF.

[vsatman]

In my opinion, such pictures replace the real situation  with a fictional one and mislead a person who is far from the topic.

If we try to calculate what % of the volume of space in an orbit of 560 km where StarLinks move  is occupied by the satellites themselves, then most likely it will be the same % of the volume that is occupied by two flies flying in Madison Square Garden .. with the corresponding probability of their collision ..

[abaddon]

In my opinion, such pictures replace the real situation  with a fictional one and mislead a person who is far from the topic.

If we try to calculate what % of the volume of space in an orbit of 560 km where StarLinks move  is occupied by the satellites themselves, then most likely it will be the same % of the volume that is occupied by two flies flying in Madison Square Garden .. with the corresponding probability of their collision ..

Reminds me of the Encyclopedia Brown story (yeah, dating myself here) about the guy scamming about selling a scale model of the solar system.  Sadly, it feels like Encyclopedia having to explain it to people for them to get it is as relevant today as it was back then.

[FutureSpaceTourist]

https://twitter.com/profhughlewis/status/1569986222294024192

Latest analysis for #Starlink & #OneWeb shows these two constellations accounted for 42% of all close approaches within 5 km predicted by #SOCRATES at the end of August, with Starlink alone accounting for 29%. [1/n]

On average, #SOCRATES predicts that each #Starlink satellite will now experience 1 close approach within 5 km with a non-Starlink object every day, and each #OneWeb satellite will experience 3.4 close approaches with a non-OneWeb object every day. These rates are increasing [2/n]

Here's the same data from [2/n] plotted with respect to the number of satellites in each constellation in orbit, clearly showing #SOCRATES predicts that #OneWeb satellites experience more close approaches (within 5 km) per satellite than the #Starlink satellites [3/n]

Focusing on the #SOCRATES predictions for close approaches with max. collision probability of at least 1E-5, we see that the two constellations accounted for 30% of all such close approaches in August, with #Starlink alone accounting for 20%. [4/n]

We see a rising trend through time in predictions of the average number of such close approaches being experienced by each satellite on a daily basis, with #OneWeb satellites seeing a rate that is now 5 times greater than the #Starlink satellites. at 0.05 per sat per day [5/n]

Here's the equivalent #SOCRATES data showing the average number of conjunctions with max. collision probability of at least 1E-5 per satellite per day [6/n]

Across all of the predicted conjunctions within 5 km for #Starlink just under half (47%) involve a debris object, one-quarter (23%) involve another non-Starlink payload & just over one-fifth (22%) involve another Starlink satellite (likely 'ignored' by the operator) [7/n]

For #OneWeb, the picture is very different. Just under 70% of all the conjunctions involve a non-OneWeb payload with only one-quarter (26%) involving a debris object. [8/n]

Hence the predictions from #SOCRATES suggest that #OneWeb has experienced about 33% more conjunctions with other payloads than #Starlink despite the constellations substantially smaller size. This is likely due to the relatively long orbit raising process through LEO. [9/n]

Taking a snapshot (from a #SOCRATES report generated on 12 September) the close approaches within 5 km involving #Starlink and #OneWeb can be seen clearly, with the #Starlink shells around 550 km & the OneWeb shells around 1200 km dominating. [10/n]

Here's the same #SOCRATES data plotted using a logarithmic y-axis, which distorts but enables a little more clarity at the lower counts. [11/n]

And this is the same data now shown as the proportion of all close approaches at each altitude. #Starlink shells around 350 km & 550 km, and #OneWeb shells between 1100 km & 1200 km become readily apparent, as do close approaches occurring during orbit raising. [12/n]

My thanks to @TSKelso (@CelesTrak) for the #SOCRATES data & support, and to @planet4589 for #Starlink & #OneWeb summary statistics. Hope this thread has been useful! [13/13; fin]

[steveleach]

How close does Starlink's collision avoidance let things get before they take action? If it is less than 5km then those numbers are all meaningless, presumably.

[steveleach]

Also, I noticed that the post at the start of this thread used a threshold of 1km. Does anyone know the reason for the switch to a 5km threshold? Is it just that 1km no longer paints a scary enough picture?

[jimvela]

A separation distance isn't super useful without an associated uncertainty... 
I wonder how good the orbit determination is that SpaceX is using for its automated responses.

[OceanCat]

A separation distance isn't super useful without an associated uncertainty... 
I wonder how good the orbit determination is that SpaceX is using for its automated responses.

They provide position and velocity covariance if you are curious. Instructions here.

How close does Starlink's collision avoidance let things get before they take action?

It's based on probability of collision not distance threshold. Greater than 1 in 100,000 triggers collision avoidance maneuver.

[niwax]

A separation distance isn't super useful without an associated uncertainty... 
I wonder how good the orbit determination is that SpaceX is using for its automated responses.

The ephemeris SpaceX hands out to sources like Celestrak is accurate to within about 50m. When you have proper tracking and information sharing, you can fit a hell of a lot of satellites up there.

[Robotbeat]

A separation distance isn't super useful without an associated uncertainty... 
I wonder how good the orbit determination is that SpaceX is using for its automated responses.

The ephemeris SpaceX hands out to sources like Celestrak is accurate to within about 50m. When you have proper tracking and information sharing, you can fit a hell of a lot of satellites up there.

Exactly this. And when you DO have a “conjunction,” you only require a very small movement to move out of the way.

I think it might be a good idea, as LEO fills up, to have like an ADS-B-type requirement for satellites. If everything’s position (and orientation) is known precisely down to less than a meter, conjunctions will be incredibly rare and you can avoid collisions very easily (even just changing both satellites’ orientation would often be enough, and then maybe move a couple meters to either side). Objects which lose ADS-B-whatever position transmitting capability could be swept out by deorbit tugs (with position tracked by dedicated radar beams in the meantime).

There’s no reason you couldn’t have literally millions of satellites in LEO with sufficient precision and maneuverability. It’s almost entirely empty space up there. When we talk about collision avoidance, we’re really talking about avoiding the tracking error cones, not the physical object (necessarily). The objects are almost point-like in comparison.

[steveleach]

How close does Starlink's collision avoidance let things get before they take action?

It's based on probability of collision not distance threshold. Greater than 1 in 100,000 triggers collision avoidance maneuver.

Yeah, I know that, but the quoted headline was about distance, so I'm interested in what sort of separation distances that might result in.

Assuming randomly passing through a circle of space, with a 5m radius target in the centre, I think you'd need the circle to be about 1.5k radius or less to have have a 1 in 100,000 chance of collision. A 5km radius gives 1 in a million.

So does that mean that 90% of passes within 5km would not trigger Starlink collision avoidance? And when avoidance is triggered, it still wouldn't need to adjust to pass more than 5k away.

Or are my assumptions or calculations messed up?

[OceanCat]

Yeah, I know that, but the quoted headline was about distance, so I'm interested in what sort of separation distances that might result in.

Assuming randomly passing through a circle of space, with a 5m radius target in the centre, I think you'd need the circle to be about 1.5k radius or less to have have a 1 in 100,000 chance of collision. A 5km radius gives 1 in a million.

So does that mean that 90% of passes within 5km would not trigger Starlink collision avoidance? And when avoidance is triggered, it still wouldn't need to adjust to pass more than 5k away.

Or are my assumptions or calculations messed up?

The modelling is more sophisticated than randomly passing through a circle of space. It's more like normal distribution within an elongated 3d ellipsoid. An ellipsoid around a Starlink satellite must be much smaller than the ellipsoids around two debris pictured below thanks to onboard GPS.

According to the data Prof. Lewis posted earlier 97% of conjunctions within 5 km and 50% within 1 km didn't trigger a CAM.

[steveleach]

Yeah, I know that, but the quoted headline was about distance, so I'm interested in what sort of separation distances that might result in.

Assuming randomly passing through a circle of space, with a 5m radius target in the centre, I think you'd need the circle to be about 1.5k radius or less to have have a 1 in 100,000 chance of collision. A 5km radius gives 1 in a million.

So does that mean that 90% of passes within 5km would not trigger Starlink collision avoidance? And when avoidance is triggered, it still wouldn't need to adjust to pass more than 5k away.

Or are my assumptions or calculations messed up?

The modelling is more sophisticated than randomly passing through a circle of space. It's more like normal distribution within an elongated 3d ellipsoid. An ellipsoid around a Starlink satellite must be much smaller than the ellipsoids around two debris pictured below thanks to onboard GPS.

According to the data Prof. Lewis posted earlier 97% of conjunctions within 5 km and 50% within 1 km didn't trigger a CAM.

Yep, I get that it isn't that simple, but I wanted a ballpark.  Where did you see that 97% figure?

Also, would that mean that out of those 30,000 Starlink conjunctions, roughly 900 resulted in the Starlink taking avoiding action? And would they still be likely to pass within 5k, but at a much lower probability of collision, if they did? And that 29,100 times the Starlinks didn't even bother adjusting?

[OceanCat]

Yeah, I know that, but the quoted headline was about distance, so I'm interested in what sort of separation distances that might result in.

Assuming randomly passing through a circle of space, with a 5m radius target in the centre, I think you'd need the circle to be about 1.5k radius or less to have have a 1 in 100,000 chance of collision. A 5km radius gives 1 in a million.

So does that mean that 90% of passes within 5km would not trigger Starlink collision avoidance? And when avoidance is triggered, it still wouldn't need to adjust to pass more than 5k away.

Or are my assumptions or calculations messed up?

The modelling is more sophisticated than randomly passing through a circle of space. It's more like normal distribution within an elongated 3d ellipsoid. An ellipsoid around a Starlink satellite must be much smaller than the ellipsoids around two debris pictured below thanks to onboard GPS.

According to the data Prof. Lewis posted earlier 97% of conjunctions within 5 km and 50% within 1 km didn't trigger a CAM.

Yep, I get that it isn't that simple, but I wanted a ballpark.  Where did you see that 97% figure?

Also, would that mean that out of those 30,000 Starlink conjunctions, roughly 900 resulted in the Starlink taking avoiding action? And would they still be likely to pass within 5k, but at a much lower probability of collision, if they did? And that 29,100 times the Starlinks didn't even bother adjusting?

There are three graphs in the linked thread showing 97,080 conjunctions within 5 km,  5,794 conjunctions within 1 km, and 2,884 conjunctions with 1 in 100,000 probability of collision. I derived the percentage from those numbers. At 19th tweet he shows that SOCRATES data allows estimating the number of Starlink CAMs fairly accurately.

Yes to all three other questions.

[meekGee]

A separation distance isn't super useful without an associated uncertainty... 
I wonder how good the orbit determination is that SpaceX is using for its automated responses.

The ephemeris SpaceX hands out to sources like Celestrak is accurate to within about 50m. When you have proper tracking and information sharing, you can fit a hell of a lot of satellites up there.

Exactly this. And when you DO have a “conjunction,” you only require a very small movement to move out of the way.

I think it might be a good idea, as LEO fills up, to have like an ADS-B-type requirement for satellites. If everything’s position (and orientation) is known precisely down to less than a meter, conjunctions will be incredibly rare and you can avoid collisions very easily (even just changing both satellites’ orientation would often be enough, and then maybe move a couple meters to either side). Objects which lose ADS-B-whatever position transmitting capability could be swept out by deorbit tugs (with position tracked by dedicated radar beams in the meantime).

There’s no reason you couldn’t have literally millions of satellites in LEO with sufficient precision and maneuverability. It’s almost entirely empty space up there. When we talk about collision avoidance, we’re really talking about avoiding the tracking error cones, not the physical object (necessarily). The objects are almost point-like in comparison.

If everyone is playing constructively.  Geo-politics suck, and you may need to be really vigilant about other satellites.  I wouldn't trust a cooperative system without an additional passive tracking system.

[steveleach]

Yeah, I know that, but the quoted headline was about distance, so I'm interested in what sort of separation distances that might result in.

Assuming randomly passing through a circle of space, with a 5m radius target in the centre, I think you'd need the circle to be about 1.5k radius or less to have have a 1 in 100,000 chance of collision. A 5km radius gives 1 in a million.

So does that mean that 90% of passes within 5km would not trigger Starlink collision avoidance? And when avoidance is triggered, it still wouldn't need to adjust to pass more than 5k away.

Or are my assumptions or calculations messed up?

The modelling is more sophisticated than randomly passing through a circle of space. It's more like normal distribution within an elongated 3d ellipsoid. An ellipsoid around a Starlink satellite must be much smaller than the ellipsoids around two debris pictured below thanks to onboard GPS.

According to the data Prof. Lewis posted earlier 97% of conjunctions within 5 km and 50% within 1 km didn't trigger a CAM.

Yep, I get that it isn't that simple, but I wanted a ballpark.  Where did you see that 97% figure?

Also, would that mean that out of those 30,000 Starlink conjunctions, roughly 900 resulted in the Starlink taking avoiding action? And would they still be likely to pass within 5k, but at a much lower probability of collision, if they did? And that 29,100 times the Starlinks didn't even bother adjusting?

There are three graphs in the linked thread showing 97,080 conjunctions within 5 km,  5,794 conjunctions within 1 km, and 2,884 conjunctions with 1 in 100,000 probability of collision. I derived the percentage from those numbers. At 19th tweet he shows that SOCRATES data allows estimating the number of Starlink CAMs fairly accurately.

Yes to all three other questions.

Thanks for that.

This does all make it sound like the "30,000 conjunctions within 5km" comment is just clickbait, but maybe I'm missing something.

[MP99]

I also wonder if SpaceX will start investing in their own tracking capabilities for space debris in order to reduce the uncertainty and reduce the need for maneuvering.

Perhaps as a natural extension to this, what if each Starlink sat had situational awareness RADAR / LIDAR over short distances. To throw a number out there to start a conversation - say 30km, which would be a few seconds of tracking?

Each sat would report any pings, including return time-of-flight, return brightness and tracking of rate-of-direction-change. Maybe increase the broadcast power once a target is identified, to get the best/longest tracking of the object as it recedes. I'd assume these would be downlinked for analysis. Perhaps the data could be forwarded raw to reduce onboard analysis.

If there is a collision, the rest of the constellation could rapidly identify any new items of debris, which could be a big improvement over the rate that they can be catalogued, and orbital elements nailed down.

It could also decrease the uncertainty of position of known items after a CAM, or an approach which doesn't need a CAM. Maybe the sensitivity could be increased when a close approach is expected. The system could be turned on only when an approach is expected, but that would stop it identifying smaller items that aren't currently tracked, which could be a valuable extension of existing tracking.


I also wonder about whether a Starlink should respond if something is going to impact despite all CAMs. Could the Starlink somehow perform a pre-emptive "destruction" (analogous to a launch destruct), which cleanly separates itself into two mostly-intact halves, specifically to maximise the eccentricity of those two halves.

The intention would be to minimise the perigee of the two Starlink debris clouds, which would minimise the on orbit lifetime. Think of it like an airbag which accepts the inevitability of the crash, but the intent is "save the on-orbit environment", rather than "save the passengers".

How much might this reduce the lifetime of the debris? I'd also appreciate any thoughts what would happen if the impactor hits one of the halves after Separation, IE the separation fails to avoid the collision?


I believe that the ideal dV for the separation would be prograde and retrograde to minimise perigee, but that would seem to threaten the leading and trailing sats in the same shell. Thoughts?

I would also expect that any close approach would be sent out via all available ISL as an emergency priority in real time, which would provide a real time telemetry "black box" in case of a collision. Should also set initial estimates of the energy and direction of any debris.

Any feedback appreciated.

[Asteroza]

I also wonder if SpaceX will start investing in their own tracking capabilities for space debris in order to reduce the uncertainty and reduce the need for maneuvering.

Perhaps as a natural extension to this, what if each Starlink sat had situational awareness RADAR / LIDAR over short distances. To throw a number out there to start a conversation - say 30km, which would be a few seconds of tracking?

Each sat would report any pings, including return time-of-flight, return brightness and tracking of rate-of-direction-change. Maybe increase the broadcast power once a target is identified, to get the best/longest tracking of the object as it recedes. I'd assume these would be downlinked for analysis. Perhaps the data could be forwarded raw to reduce onboard analysis.

If there is a collision, the rest of the constellation could rapidly identify any new items of debris, which could be a big improvement over the rate that they can be catalogued, and orbital elements nailed down.

It could also decrease the uncertainty of position of known items after a CAM, or an approach which doesn't need a CAM. Maybe the sensitivity could be increased when a close approach is expected. The system could be turned on only when an approach is expected, but that would stop it identifying smaller items that aren't currently tracked, which could be a valuable extension of existing tracking.


I also wonder about whether a Starlink should respond if something is going to impact despite all CAMs. Could the Starlink somehow perform a pre-emptive "destruction" (analogous to a launch destruct), which cleanly separates itself into two mostly-intact halves, specifically to maximise the eccentricity of those two halves.

The intention would be to minimise the perigee of the two Starlink debris clouds, which would minimise the on orbit lifetime. Think of it like an airbag which accepts the inevitability of the crash, but the intent is "save the on-orbit environment", rather than "save the passengers".

How much might this reduce the lifetime of the debris? I'd also appreciate any thoughts what would happen if the impactor hits one of the halves after Separation, IE the separation fails to avoid the collision?


I believe that the ideal dV for the separation would be prograde and retrograde to minimise perigee, but that would seem to threaten the leading and trailing sats in the same shell. Thoughts?

I would also expect that any close approach would be sent out via all available ISL as an emergency priority in real time, which would provide a real time telemetry "black box" in case of a collision. Should also set initial estimates of the energy and direction of any debris.

Any feedback appreciated.

Dragon's LIDAR might work, but is added mass.

I do wonder about debris passing through the phased array spot beams. The array isn't much different from an AESA radar in principle, but what size object at about what range needs to monitored? The array isn't that powerful, and 3cm objects are the low end of the debris tracking catalogs currently.

[Robotbeat]

The radar can get very close to debris, though. Instead of 1000km, you might get 10km or even 1km distances, and because beam divergence to target and then back actually goes as 1/r^4, that can mean a HUGE difference.

There might be some way to use the same phased array as a radar, at times when there is little demand on a satellite’s RF capacity, like when orbiting over the Pacific.

[DanClemmensen]

The radar can get very close to debris, though. Instead of 1000km, you might get 10km or even 1km distances, and because beam divergence to target and then back actually goes as 1/r^4, that can mean a HUGE difference.

There might be some way to use the same phased array as a radar, at times when there is little demand on a satellite’s RF capacity, like when orbiting over the Pacific.

LEO orbits are roughly 90 minutes for roughly 40,000 km, or about 7.4 km/s, and satellites can be coming from any direction, so closing velocities are from 0 to about 14 km/s   Your 30 km radar gives you about 2 seconds to avoid the head-on collision. Head-ons can occur for polar orbits. Your satellites won't save themselves, but they might collectively generate a collaborative database of extremely precise debris orbits to save each other.

[Robotbeat]

The radar can get very close to debris, though. Instead of 1000km, you might get 10km or even 1km distances, and because beam divergence to target and then back actually goes as 1/r^4, that can mean a HUGE difference.

There might be some way to use the same phased array as a radar, at times when there is little demand on a satellite’s RF capacity, like when orbiting over the Pacific.

LEO orbits are roughly 90 minutes for roughly 40,000 km, or about 7.4 km/s, and satellites can be coming from any direction, so closing velocities are from 0 to about 14 km/s   Your 30 km radar gives you about 2 seconds to avoid the head-on collision. Head-ons can occur for polar orbits. Your satellites won't save themselves, but they might collectively generate a collaborative database of extremely precise debris orbits to save each other.

That's right, I wasn't thinking of realtime avoidance. (It's fun to imagine what would be required for a 2 second avoidance to succeed, though.

[MP99]

The radar can get very close to debris, though. Instead of 1000km, you might get 10km or even 1km distances, and because beam divergence to target and then back actually goes as 1/r^4, that can mean a HUGE difference.

There might be some way to use the same phased array as a radar, at times when there is little demand on a satellite’s RF capacity, like when orbiting over the Pacific.

LEO orbits are roughly 90 minutes for roughly 40,000 km, or about 7.4 km/s, and satellites can be coming from any direction, so closing velocities are from 0 to about 14 km/s   Your 30 km radar gives you about 2 seconds to avoid the head-on collision. Head-ons can occur for polar orbits. Your satellites won't save themselves, but they might collectively generate a collaborative database of extremely precise debris orbits to save each other.

That was exactly my intention. Presumably could also catalogue items that are smaller than the current capability.

BTW, the 30km was just a wild number to start a conversation. Trying to balance functionality without making the orbital environment too "noisy" - something I know nothing about.

In terms of mapping objects as they pass, you might only have two seconds as it approaches, but perhaps a longer time as it recedes if you briefly boost the power. That's for new object detections. For known objects, could perhaps boost power throughout both the approach and after.

Edit: also thanks to RB for the 1/r^4 info.

[MP99]

The radar can get very close to debris, though. Instead of 1000km, you might get 10km or even 1km distances, and because beam divergence to target and then back actually goes as 1/r^4, that can mean a HUGE difference.

There might be some way to use the same phased array as a radar, at times when there is little demand on a satellite’s RF capacity, like when orbiting over the Pacific.

LEO orbits are roughly 90 minutes for roughly 40,000 km, or about 7.4 km/s, and satellites can be coming from any direction, so closing velocities are from 0 to about 14 km/s   Your 30 km radar gives you about 2 seconds to avoid the head-on collision. Head-ons can occur for polar orbits. Your satellites won't save themselves, but they might collectively generate a collaborative database of extremely precise debris orbits to save each other.

That's right, I wasn't thinking of realtime avoidance. (It's fun to imagine what would be required for a 2 second avoidance to succeed, though.

Any thoughts on sacrificially destroying itself in as clean a way as possible if a collision object is detected? Would be from previously unmapped debris, or perhaps a sat which is functioning except for propulsion, if that is a thing?

[su27k]

Starlink breaks space traffic safety rules and China may follow: study

A team of Chinese space engineers has accused SpaceX’s Starlink satellites of breaking the traffic rules of Earth’s lower orbit and warned that China will be giving the US an upper hand if it does not follow suit.

The researchers said two of Starlink’s newest satellites, equipped with high-speed laser communication devices, came within 4.9km (3 miles) of each other on June 30. The commonly accepted – if unwritten – minimum distance to avoid collision is 10km (6.2 miles).

In a study published by Chinese peer-reviewed journal Radio Engineering, the researchers said the unusually dense formation was no accident, but the result of a complex scheme by SpaceX to maximise the performance of its laser communications.

[Robotbeat]

SpaceX has pretty good tracking, so there’s no a priori reason to think this was unsafe.

[JayWee]

I wonder if anyone could find the article - from the sound of it, it seems like it was between sats in different orbital planes?

But there *is* a big difference between proximity between two of your own sats. You know your position pretty accurately. Knowledge of other's orbit has a degree of uncertainty. Very different situation.

[OceanCat]

The recent reporting by SpaceX of the number of collision avoidance manoeuvres performed by Starlink has enabled an updated analysis & prediction... In a voluntary filing SpaceX indicated that each Starlink satellite has sufficient propellant to perform 5,000 propulsive manoeuvres over the satellite life, with 350 of those for collision avoidance. Depending on the model used, that level might be reached somewhere after 7,500-8,000 satellites have been launched. Quite important caveats apply here! The models used for prediction are based entirely on the extrapolation of simple trendlines (some where there is no real causation). As such, they assume that past behaviour is a perfect description of future behaviour, which is unlikely (e.g. if the launch cadence changes, if the Starlink 2nd generation satellites have different collision characteristics, or if the SSA data used for screening differs from the present-day SSA data, perhaps through the inclusion of smaller objects) The choice of model (i.e., linear, quadratic, cubic) is based not only on the fit but also on my expectations for the behaviour. Primarily, I expect the *cumulative* number of manoeuvres to increase non-linearly with time & number of Starlink satellites launched. As time goes on and SpaceX offers more data, I will update the models and predictions (and also test/validate the older predictions) keeping you informed. In the meantime, thanks for reading!

https://twitter.com/ProfHughLewis/status/1610307514792382471

[OceanCat]

It's getting crowded in space. SpaceX is not happy three HawkEye360 satellites and two stages of the Electron were left in orbits next to Starlink satellites.

HE360 recently confirmed its deployment on January 24, 2023, of three satellites “at an orbital
altitude of approximately 552 km (apogee) x 551 km (perigee).”  Moreover, Rocket Lab
disposed of the second and third stages of the Electron launch vehicle used for this injection
in orbits of 329 km x 558 km and 549 km x 552 km, respectively.

As of February 21, SpaceX had received 400 proximity alerts related to the three recently launched
HE360 satellites, resulting in 164 collision avoidance maneuvers. While HE360 does not publicly
post ephemeris and covariance data on Space-Track.org, SpaceX’s advanced collision avoidance
system has operated successfully and continues to avoid the HE360 satellites.

Rocket Lab left two stages of the launch vehicle that deployed the HE360 satellites in orbits
that also affect SpaceX’s constellation. As of February 21, SpaceX had received 47 proximity
alerts exceeding its safety threshold from the two stages, resulting in 35 collision avoidance
maneuvers.

The SpaceX collision avoidance system is robust and can accommodate other systems
operating in its approved orbital shells. Nonetheless, industry best practices and licensing
requirements call for both the launch provider and satellite owner/operator to coordinate prior to
launch.

SpaceX has raised these issues with both HE360 and Rocket Lab, and will continue to work
with them to minimize the risks of their future launches.

[RedLineTrain]

It's getting crowded in space. SpaceX is not happy three HawkEye360 satellites and two stages of the Electron were left in orbits next to Starlink satellites.

HE360 recently confirmed its deployment on January 24, 2023, of three satellites “at an orbital
altitude of approximately 552 km (apogee) x 551 km (perigee).”  Moreover, Rocket Lab
disposed of the second and third stages of the Electron launch vehicle used for this injection
in orbits of 329 km x 558 km and 549 km x 552 km, respectively.

As of February 21, SpaceX had received 400 proximity alerts related to the three recently launched
HE360 satellites, resulting in 164 collision avoidance maneuvers. While HE360 does not publicly
post ephemeris and covariance data on Space-Track.org, SpaceX’s advanced collision avoidance
system has operated successfully and continues to avoid the HE360 satellites.

Rocket Lab left two stages of the launch vehicle that deployed the HE360 satellites in orbits
that also affect SpaceX’s constellation. As of February 21, SpaceX had received 47 proximity
alerts exceeding its safety threshold from the two stages, resulting in 35 collision avoidance
maneuvers.

The SpaceX collision avoidance system is robust and can accommodate other systems
operating in its approved orbital shells. Nonetheless, industry best practices and licensing
requirements call for both the launch provider and satellite owner/operator to coordinate prior to
launch.

SpaceX has raised these issues with both HE360 and Rocket Lab, and will continue to work
with them to minimize the risks of their future launches.

Rather, SpaceX was unhappy that they didn’t do any coordination with SpaceX.

[OceanCat]

SpaceX now nominally uses an even more conservative maneuver threshold two
orders of magnitude more sensitive than the industry standard. Specifically, SpaceX satellites will
maneuver when the probability of collision is greater than 1e-6 (1 in 1,000,000 chance of collision),
as opposed to the industry standard of 1e-4 (1 in 10,000 chance of collision).

From the gen1 report:

Using this very conservative threshold, along with even more conservative assumptions, SpaceX satellites
performed 49,384 propulsive maneuvers over the reporting period, averaging approximately 27
maneuvers per satellite, per year.

From the gen2 report:

Using this very conservative threshold, along with even more conservative assumptions, SpaceX satellites
performed 37,094 propulsive maneuvers over the reporting period.

From the previous gen1 report:

Specifically, SpaceX satellites will maneuver when the
probability of collision is greater than 1e-5 (1 in 100,000 chance of collision), as opposed to the
industry standard of 1e-4 (1 in 10,000 chance of collision). Using this very conservative threshold,
along with even more conservative triggers, SpaceX satellites performed 24,410 propulsive
maneuvers over the reporting period, averaging approximately 12 maneuvers per satellite, per year.