Impacts of Large Satellite Constellations on Astronomy

[su27k]

There is a lot of cherry picking of quotes with little attention paid to context.  A brief lookup of the NASA LMT seems to show that it was never a major contributor to debris tracking.  While it may be possible for most debris to be tracked optically, in practice radar is used for a lot of it.  The newest systems coming online now to improve debris tracking by US entities, both government and commercial, are radar.  One of the first pieces I found on LMT also noted its detection falling off around 11cm.

That is the point of the whole discussion: Whether LEO debris is visible to telescopes. The discussion was never about what method is used to track debris in practice right now, it's about the possibility of seeing debris in telescope observations.

[su27k]

Oh, walls of text in response to the same overarching topics, how I missed you. Hint: trim your quotes.
Apologies for the long post that results even when ignoring some filler and general "markets will regulate themselves" political views that have worked so well so many times in global-scale issues such as this.

I'm not going to take you telling me how to write my post as an insult, unlike some people.

The quotes are there for a reason: providing context, which is critical when you attacks SpaceX from multiple directions.

Also I'm not the one started to involve market in this discussion, and using big words like "global-scale" does not refute my point.

Which is clearly wrong (most of the tracked objects are not, or barely, visible optically, for current tracking optical telescopes)[/size], as NASA's LMT was able to detect debris less than 10cm, as small as 1cm.

Key words: current, overwhelmingly.
Decimeter-scale objects were also tracked on the Moon. It's still not relevant.

Those keywords are irrelevant to the discussion. We're not discussing what is the current debris tracking method, we're discussing whether debris can be tracked visibility, I have already provided evidence that they can, whether this method is currently in use does not affect the conclusion.

No, it's not [Comparing apples-to-apples the visibility of a random piece of small debris to a 15-m reflective satellite is a fallacy][/size]. Debris would generate similar trails as satellite, just smaller, it's a matter of degree, but the effect and mitigation could be similar, which is the point the OP trying to make.  It is also a fallacy to say Starlink is 15-m long, that's the solar panel size, you have no evidence to show solar panel is the issue here. In fact it's probably not, since SpaceX is not attempting to coat the solar panel.

A speck of paint, or a fungal spore, would generate a similar trail as a satellite, just smaller - it's a matter of degree. The effect and mitigation would be similar. Find the fallacy. A 15m^2 glass surface pointing zenith, which needs to see the Sun as much as possible, is not the issue. Its impact is comparable to decimeter-scale debris. Find the fallacies again.

No, not everything will show up in a telescope, there will be a limit under which the object will be invisible, but clearly centimeter sized debris can show up, and there're a lot of centimeter sized debris out there, the telescopes will have to deal with it.

And no, I have not seen any report that solar panel is the issue here, repeating "find the fallacies" is not going to change that. The solar panel only needs to see the Sun when it's in daylight, and the impact to astronomy occurs when satellite is in the twilight and about to enter Earth's shadow, so it's entirely possible SpaceX can give up sun light gathering in this short period by re-orienting the solar panel, this wouldn't have much negative impact on satellites' operation.

[Lar]

The solar panel only needs to see the Sun when it's in daylight, and the impact to astronomy occurs when satellite is in the twilight and about to enter Earth's shadow, so it's entirely possible SpaceX can give up sun light gathering in this short period by re-orienting the solar panel, this wouldn't have much negative impact on satellites' operation.

Unless the birds are already actively moving the panels a lot anyway, doing once an orbit furl/unfurl moves will not be highly reliable.

[ChrisWilson68]

The solar panel only needs to see the Sun when it's in daylight, and the impact to astronomy occurs when satellite is in the twilight and about to enter Earth's shadow, so it's entirely possible SpaceX can give up sun light gathering in this short period by re-orienting the solar panel, this wouldn't have much negative impact on satellites' operation.

Unless the birds are already actively moving the panels a lot anyway, doing once an orbit furl/unfurl moves will not be highly reliable.

I think su27k meant rolling the panels along their axes, not furling and unfurling them.

They need to roll along their axes anyway to keep the antennas facing the surface of the earth while the panels face the sun, unless they want to lose a lot of efficiency (which, maybe they're OK with to avoid having another moving part).

[eeergo]

It's obviously a function of the FOV of course, but unless it's comparable to the integration time for a single exposure of such a narrow FOV, it will be a distinct issue.

Let's check it back-of-the-envelope: apparent angular speed for LEO is ~0.75º/s. You should be doing short integrations to avoid thermal noise and CRs (yet long enough for sensitivity), apparently O(100)ms is typical. The CR track will be instantaneous, so it will last that long.

So you are talking about IR astronomy, not optical. I am fully aware that IR has short frames that are stacked by the hundreds. Also dithered to limit the influence of hot pixels. But I was not talking about IR, I was talking about optical with exposure times of one frame in the 10s of seconds or minutes. No wonder we were talking past each other.

Again, the key is the integration time for a single exposure being orders of magnitude different for a satellite than a CR. That was the question you formulated, and I showed mathematically what qualitative reason suggested. If you performed a single raw integration lasting tens of seconds or minutes without subtracting noise, you'd end up overwhelmed by CRs, thermal noise and atmospheric fluorescence, among other things, as correctly pointed out by yourself a few posts ago...

VISTA stands for VISIBLE and Infrared Survey Telescope for Astronomy - so the numbers are indeed valid for what we were talking about. In any case, IR will also suffer from constellations, even more than in visible perhaps. Most other large professional visible telescopes have similar FOVs and pixel size (order-of-magnitude), see for example the Canarian GTC: gtc.iac.es/instruments/instrumentation.php

[eeergo]

The solar panel only needs to see the Sun when it's in daylight, and the impact to astronomy occurs when satellite is in the twilight and about to enter Earth's shadow, so it's entirely possible SpaceX can give up sun light gathering in this short period by re-orienting the solar panel, this wouldn't have much negative impact on satellites' operation.

Unless the birds are already actively moving the panels a lot anyway, doing once an orbit furl/unfurl moves will not be highly reliable.

I think su27k meant rolling the panels along their axes, not furling and unfurling them.

They need to roll along their axes anyway to keep the antennas facing the surface of the earth while the panels face the sun, unless they want to lose a lot of efficiency (which, maybe they're OK with to avoid having another moving part).

That would make the satellites considerably more expensive and heavy (possibly also voluminous since now they just appear to use spring-loaded single-unfurl panels attached to the sat's side, and now you'd need a pivot point): it's not just the extra actuators, but also the gyros or hypergols needed to stabilize attitude in a relatively quick maneuver around twilight, not comparable to the slow roll they probably perform now as you say. It'd also reduce their lifetime, introduce serious failure modes including issues to the pointing accuracy of the laser payload... and you'd still only partially mitigate visibility.

But hey, I agree that'd be nicer than ignoring the issue, or saying impacts to most astronomy are negligible/baseless. No sign of that being worked on though, only "coatings" have been mentioned.

[eeergo]

Those keywords are irrelevant to the discussion. We're not discussing what is the current debris tracking method, we're discussing whether debris can be tracked visibility, I have already provided evidence that they can, whether this method is currently in use does not affect the conclusion.

This has been shown, by myself and others, to be irrelevant to the case at hand we originally were discussing (i.e. relative impact of decimeter-scale debris vs decameter-scale sats) - plus as gongora pointed out (and you chose to ignore), you effected quite a bit of selective quoting there anyway. As you do again here, where the keywords highlighted what I said and didn't say, in response to your "wrong".

No, not everything will show up in a telescope, there will be a limit under which the object will be invisible, but clearly centimeter sized debris can show up, and there're a lot of centimeter sized debris out there, the telescopes will have to deal with it.

Size of a megaconstellation satellite = O(100) x size of most debris.

In the interest of avoiding repetition, further debunking of relative impact of each visible source on astronomical telescopes can be found in preceding responses to Semmel.

And no, I have not seen any report that solar panel is the issue here, [...]

While I responded to most of your comment in the reply to ChrisWilson, this bit is kind of burying one's head under the sand no? How can't the most reflective, less coatable, flattest and largest piece of the satellite, whose very purpose is to face the Sun, not be one of the main issues, if not THE main, with respect to visibility?

Finally: speaking out against issues and critically study them, especially concerning grandiose and -yes- global-scale (why would that be a big word? it's much more "global" than the GLOBAL Navigation Satellite Service constellations) projects, is not attacking, especially when I'm always referring to megaconstellations in general - although of course SpaceX's is the most ambitious one and would dwarf all others as ultimately proposed.

[Semmel]

Again, the key is the integration time for a single exposure being orders of magnitude different for a satellite than a CR. That was the question you formulated

No, thats not what I was saying. The exposure time is set by the observation of an astronomical target. That can range from subseconds (IR imaging) to minutes (optical spectroscopy). The satellite is photobombing the image. The exposure time is the time that the telescope stares at the astronomical target before reading out the detector. There is no exposure time for a CR and there is no exposure time for a satellite.

I have the feeling you derail the conversation for some minor point without addressing the core of the point I am making. My point was that the effect of a satellite on a stack of exposures of the same astronomical target is similar enough to a CR that it can be detected by a similar image processing method and that it has similar impact on the data lost. The number of effected pixels might be different and the exact amount of signal to noise might be different but the claim is that astronomy is already disturbed by numerous effects and the satellites are not all that different. Certainly not different enough to spell doom for ground based astronomy. Our conversation doesnt seem to approach any conclusions if you dont argue the point. All the talk about CR discharge timing and whatnot is totally beside the point and derailing. I dont have the energy to continue any further, we are not getting anywhere.

[edzieba]

Whilst everyone seems to be going in circles over theoretical future impacts, I'd like to circle back to a thrice-asked but yet unanswered question:

What - if any - mitigations are currently in place to handle the impact of the existing several thousand bright (i.e. not cm/mm scale) satellite and debris objects in LEO on earth-base optical astronomy?


And for some more data, here's the relevant LEO tables and histograms from ESA's most recent DISCOS report. I tried to query DISCOS directly to create a batter chart, but the web interface has no export and cannot select by orbit and area at the same time (and the API is restricted). If someone has API access than you should be able to search for all objects in LEO with cross-sectional area above (say) 1m to generate a useful count of 'shiny LEO stuff'.
PL = intact payloads, RB = intact rocket bodies.

[jebbo]

I can't help thinking that, regardless of anything to do with constellations, this furore has highlighted the overall debris problem so I suspect PMD guidelines (particularly for failed payloads) and debris avoidance and lifetime guidelines will get revamped. Perhaps we'll even get a sensible debris removal programme

--- Tony

[eeergo]

No, thats not what I was saying. The exposure time is set by the observation of an astronomical target. That can range from subseconds (IR imaging) to minutes (optical spectroscopy). The satellite is photobombing the image. The exposure time is the time that the telescope stares at the astronomical target before reading out the detector. There is no exposure time for a CR and there is no exposure time for a satellite.

I have the feeling you derail the conversation for some minor point without addressing the core of the point I am making. My point was that the effect of a satellite on a stack of exposures of the same astronomical target is similar enough to a CR that it can be detected by a similar image processing method and that it has similar impact on the data lost. The number of effected pixels might be different and the exact amount of signal to noise might be different but the claim is that astronomy is already disturbed by numerous effects and the satellites are not all that different. Certainly not different enough to spell doom for ground based astronomy. Our conversation doesnt seem to approach any conclusions if you dont argue the point. All the talk about CR discharge timing and whatnot is totally beside the point and derailing. I dont have the energy to continue any further, we are not getting anywhere.

I fail to see how I "derail the conversation over minutiae", when I've proven you wrong about pretty basic facts (that you insistently use to arrive to the rushed, inaccurate conclusion of "no impact") several times already, including the very first word in the acronym of a telescope I used as an example to make calculations.

Do you at least agree what I calculated is correct for IR astronomy (the largest part of scientific astronomy nowadays)?

Concerning visible astronomy (mainly spectroscopy, as you point out), you claim single long exposures (tens of seconds/minutes) are actually produced, and later several such long exposures are stacked to arrive at the final integration If we consider a 30-second single-exposure time, indeed my calculation from before would show a similar persistence on the frame to CRs or less:
0.75 º/s * 30 s = 22.5º,
so the satellite would only linger in the FOV for about 7% of the 30-second single exposure, i.e. it'd be as instantaneous for the exposure as a cosmic ray. But is this so?

In other words, do you know for a fact (I personally don't) such long exposures are fine in visible astronomy/spectroscopy and CRs/thermal noise doesn't manage to overwhelm the signal, and gets "averaged out" when images are stacked and postprocessed? If, and only if, this is the case would there be similar impact from single satellites and random CRs in *visible* astronomy. I still wonder about the relative brightness and saturation probability of CRs vs sats, and what happens when you get one or several long satellite streaks (vs the shortish ones that you can see in the dark frame you first posted in this discussion) in practically every long exposure due to the abundances needed in megaconstellations - but I'd like to clear that out first.

[jebbo]

I just checked the FLAMES/GIRAFFE documentation. So for visible light, fibre-fed spectroscopy, a "short" exposure is <120s ...

For things like the AO imaging used to check for unresolved companions etc, we look at the radial intensity drop off for a point source, looking for irregularities that suggest a companion. Here even a faint satellite track is problematic as it introduces obvious asymmetries.

For spectra, the problem is different as a satellite track introduces additional telluric contamination, but in a non-obvious way (the satellite is unlikely to be equally reflective in all wavelengths) that can't be subtracted by the usual means. 

[ I've only really used the final data products so have never really considered the operational aspects of capturing the data ]

--- Tony

[edzieba]

The shorter the exposure time, the less impact a satellite flare will have on a given frame (due to less time for electrons to accumulate and spill over to neighbouring buckets, and fewer number of pixels crossed). Per frame the impact of a CR vs. a satellite flare would only be differentiated by the relative blooming (electron spillover) of one vs. the other.
When you start stacking frames, you already needs to isolate and ignore areas of frames with CR traces, so an analogous technique could be used to isolate and ignore areas of frames where satellite flares occur, with the slight benefit that satellite flares are both predictable in advance and have deterministic movements between frames (once you have at least two captured frames to compare), and the slight drawback that if you happen to be slewing in the same direction and relative angular velocity as the satellite you get a big bright spot sitting somewhere on every frame.

[Semmel]

Warning, wall of text incoming.

Funny that you mention VIRCAM. VIRCAM will miss the most part of the internet constellations because it is decommissioned in the 2022 time frame. But for the sake of argument, lets stick with it. VIRCAM operates from Z to K band, which is squarely in the IR. Lets also take OMEGACAM on the 2.3m VST, which operates in the U to Z bands, which is UV light all the way to near IR. Both, VIRCAM and OMEGACAM are much smaller than the LSST but they are both survey telescopes and have imagers.

From your answers, I believe you dont have much experience with optical astronomy. Lets say, the important information for this conversation is, that heat is not a problem. Optical sensors have typically 1 electron per hour per pixel caused by heat. Judging from their dark frames, VIRCAM collects about 5000 electrons per second per pixel due to heat. Roughly. Thats a factor of 18e6 difference. Totally different approach to astronomy. OMEGACAM has exposure time of about 60 seconds as an imager. For a single exposure. Read out time of such an exposure is 30 seconds.

Because optical and IR is operated so differently, satellites appear different in IR and in optical detectors. In IR, it would actually be visible in multiple subframes of an exposure. But in each exposure, it would be at a different location in the image. Therefore, you can easily identify that the satellite is an object that doesnt belong into the image. As a consequence, it is masked out by the low level data reduction software.

From https://www.eso.org/sci/observing/PublicSurveys/docs/GCAV_SMP_10012017.pdf :

Static mask are also used in the first source extraction pass, via Sextractor (Bertin & Arnouts 1996).  In this step a catalog of sources is created for each chip of all input images.  Moreover a weight map for each input image is also generated during this step.  This will include both the static mask, and defects masked at individual image level, e.g.  satellite streaks.

Its already taken care of. 'defects masked at individual image level' also includes cosmic rays for instance.

In an IR image, satellites are easier to handle than in optical, because you have typically many tens of images. A satellite would only kill one frame out of all of these frames at any one location where it passed over. The lost data is maybe 10% on the path it went through. Because signal to noise is the square root of the flux, one would loose about 5% in STN. Its not nothing but its also not going to kill astronomy. Certainly not an unsolvable problem. I cant calculate the exact impact though.
I added a section of a VIRCAM raw frame. That is one frame out of a sequence of 23 with an exposure time was 1 second of a single frame. Note the color bar at the bottom. Its logarithmic.

Going to optical astronomy. A satellite here has a higher impact than on IR because in optical one wouldnt take a sequence of many tens of images to create one long observation. 2 or 3 is more like it. Sometimes, if one wants to go very deep, more exposures are done. In a multi-second exposure, a satellite would not move slowly through the field of view, frame by frame, it would be only be visible in one frame crossing the entire detector in one go. With multiple exposures, the affected pixels are easily detectable and can be masked out. The loss of data is higher here than for IR but its still not an unsolvable problem. Take one more frame and its basically fine. That costs observing time, which is precious, sure. But its not a dooms day scenario either.

I dont know how high the impact is. I certainly dont say "no impact", please dont twist my words. The impact is situation dependent. What is clear is, that it isnt either of the extremes. The impact will be larger than 0 but also much smaller than "ground based astronomy becomes impossible". One has to view the impact of satellites in the context of other disturbing factors to understand its relative impact to the current situation. Thats why I started this whole conversion regarding cosmic rays and space debris. The satellites do not pass over pristine and empty sky. More research clearly has to be done. LSST is likely the worst case, it makes sense to study it.

Also, anyone can go to the ESO data archive and download all the raw data they want. I am not aware of a method to look for satellite streaks though. http://archive.eso.org/eso/eso_archive_main.html

[Lar]

I can't help thinking that, regardless of anything to do with constellations, this furore has highlighted the overall debris problem so I suspect PMD guidelines (particularly for failed payloads) and debris avoidance and lifetime guidelines will get revamped. Perhaps we'll even get a sensible debris removal programme

--- Tony

That would be great.

[eeergo]

I believe you dont have much experience with optical astronomy.

I only have limited time and will for engaging in these sort of conversations during free moments left out by *actually* doing astrophysics research (albeit not optical, so sorry for not being able to provide definitive numbers off the top of my head, but I do have more than a passing knowledge of the delicate balances at play).
[later]
In other words, do you know for a fact (I personally don't) such long exposures are fine in visible astronomy/spectroscopy and CRs/thermal noise doesn't manage to overwhelm the signal

I said as much 2 (two) days ago, and in my last post.

Do you? Because so far it appears you're selectively quoting from public sources of a topic you're interested in, sometimes providing valuable tidbits and others squarely missing the point for lack of context.

Funny that you mention VIRCAM. [...] Lets say, the important information for this conversation is, that heat is not a problem. Optical sensors have typically 1 electron per hour per pixel caused by heat. [...] Because optical and IR is operated so differently, satellites appear different in IR and in optical detectors. [...] it is masked out by the low level data reduction software.

Just to be clear, I used VISTA as an example (together with Keck, by the way) because their parameters of interest to my calculation were easily accessible. Digging a bit more, you realize most FOVs/pixel sizes are comparable in other related instruments elsewhere, so it doesn't matter that particular imager will be decomissioned.

The dark rate you mention for VIS sensors (1 e-/pixel/h) is only true for cryogenic sensors, which are just a few specialized high-end instruments in the already narrow visible field. Thermoelectric-cooled (100K warmer, but still cold) or commercial (ambient temperature, uncooled) have MUCH higher rates comparable to IR's (which always needs cooling). See https://iopscience.iop.org/article/10.1086/684054 Therefore, the ultra-long exposures you mention, where the impact of CRs *regarding persistence* is similar for LEO streaks, are only applicable for a small subset of observatories in the world.

We know single satellite streaks are removable by software now, because they're relatively few. The problem discussed here is the abundance and continuous frequent passes of relatively high-luminosity satellites with much larger persistence (at least in IR, which is where the majority of astronomy is made) than other sources of frequent noise, as my simple back-of-the-envelope calculation a few posts back showed. Your point before was that CRs would have a similar or lesser impact than satellites. Now it turns out that not really (unless you finally want to explain why that simple calculation is wrong, and by the way that's just 1 out of the 3 factors that you agreed affect observational astronomy, where megaconstellations score very high), but "software will take care of it".

In an IR image, satellites are easier to handle than in optical, because you have typically many tens of images. A satellite would only kill one frame out of all of these frames at any one location where it passed over. The lost data is maybe 10% on the path it went through. [...]

I believe this is wrong based on the previous explanations I provided. You're additionally ignoring impacts on image calibration, related artifacts and other factors such as the ones jebbo provided a few post upthread, crucial for any kind of serious analysis, so in any case those estimates would be best-case.

Nobody is suggesting any kind of astronomy (except perhaps some subfields of radioastronomy) will become impossible for a few tens of thousands LEO satellites more - just that the impact will be noticeable and possibly, if the principle of caution is not applied seriously, seriously affect the field. Suggesting either extreme, absolutely no impact, or absolutely impossible to make observations, is absurd - not even the most ardent, fanatical defenders of either side would tell you that. You did say a few times though that "LSST will have a hell of a problem in general if Mag 3 objects are an issue", "the impact will be the same or less than CRs" (i.e. pretty much none) or "Additional 5000 or even 12000 satellites don't even make this worse by a factor of 2", to name a few of your firm yet dubious statements.

Of course more research is needed to precisely characterize the impact. Not so much is needed to show it won't be "a bit of downtime correctable with software".

[jebbo]

There are some pretty solid number on magnitude etc here: https://aas.org/form/aas-survey-on-satellite-constell

--- Tony

[eeergo]

New ESO statement. Expect to see serious peer-reviewed impact studies on VIS/IR astronomy soon:

https://www.eso.org/public/announcements/ann19062/


And from AAS, 7-question survey to astronomers, with a preamble showing the (conservative) assumptions and the certainties they already have:


https://aas.org/form/aas-survey-on-satellite-constell

[su27k]

The solar panel only needs to see the Sun when it's in daylight, and the impact to astronomy occurs when satellite is in the twilight and about to enter Earth's shadow, so it's entirely possible SpaceX can give up sun light gathering in this short period by re-orienting the solar panel, this wouldn't have much negative impact on satellites' operation.

Unless the birds are already actively moving the panels a lot anyway, doing once an orbit furl/unfurl moves will not be highly reliable.

I think su27k meant rolling the panels along their axes, not furling and unfurling them.

They need to roll along their axes anyway to keep the antennas facing the surface of the earth while the panels face the sun, unless they want to lose a lot of efficiency (which, maybe they're OK with to avoid having another moving part).

That would make the satellites considerably more expensive and heavy (possibly also voluminous since now they just appear to use spring-loaded single-unfurl panels attached to the sat's side, and now you'd need a pivot point): it's not just the extra actuators, but also the gyros or hypergols needed to stabilize attitude in a relatively quick maneuver around twilight, not comparable to the slow roll they probably perform now as you say. It'd also reduce their lifetime, introduce serious failure modes including issues to the pointing accuracy of the laser payload... and you'd still only partially mitigate visibility.

But hey, I agree that'd be nicer than ignoring the issue, or saying impacts to most astronomy are negligible/baseless. No sign of that being worked on though, only "coatings" have been mentioned.

You do realize Elon Musk already said Starlink can rotate its solar panel in one axis?

https://twitter.com/elonmusk/status/1132906066423889920

See this is the problem with the anti-Starlink crowd, you don't even have the basic knowledge to comprehend the problem, yet you come here as if you know everything and demand SpaceX to stop immediately.

I'm not going to reply to the rest of the baseless claims, since the moderator deleted my previous two replies, even when I followed the temporary rule to the letter, clearly no debate is allowed in the thread except anti-SpaceX propaganda.

[eeergo]

You do realize Elon Musk already said Starlink can rotate its solar panel in one axis?

https://twitter.com/elonmusk/status/1132906066423889920

See this is the problem with the anti-Starlink crowd, you don't even have the basic knowledge to comprehend the problem, yet you come here as if you know everything and demand SpaceX to stop immediately.

You do realize momentum wheels / torque rods need be much beefier for a maneuver you propose doing within a few minutes, also significantly reducing the spacecraft's power margin, than for a bbq roll over a whole semi-orbit, right?

I'm not going to reply to the rest of the baseless claims, since the moderator deleted my previous two replies, even when I followed the temporary rule to the letter, clearly no debate is allowed in the thread except anti-SpaceX propaganda.

This attitude reminds me of some conversations I maintained in a certain thread in the Indian spaceflight section (hint: ASAT). I think your fanaticism with respect to any critique to SpaceX speaks for itself, so it's probably best if we end this interaction in which you've been proven, with numbers, authoritative testimonials and detailed discussion, that concern is at the very least not unwarranted - but you still believe to hold the high ground, and the existence of a conspiracy to be shutting you down.

Just in case there's doubt, I had nothing to do with the deletion of the replies - in fact I didn't even get to read them.