Hacker News Reader

Comment by irthomasthomas | Hacker News Reader

It's trivial to erase these things from photos. The photographer no doubt knows this, but that isn't the point.

It's perfectly reasonable for scientists to want to capture accurate data on the space surrounding the main subject being imaged. Simply erasing the satellites from photos does not recover the data on the space behind. Any data from behind the satellites is lost forever. This photo keeps the satellites in order to visually demonstrate this problem.

Remember that astronomy today is often done on a single pixel of data. Starlink blocks multiple pixels, and even ruins entire exposures when they flare up. This will make astronomical research, like searching for exoplanets, far harder and more expensive than it is today. Space telescopes are, and will always be, orders of magnitude more expensive than ground telescopes to launch, maintain and operate.

View on HN

If the images are stacked, doesn't that mean that there's plenty of images with the parts that are occluded in others not occluded?

I understand for any specific image, there's going to be some lost background because of Starlink satellites, but that's not what this is showing, this is showing something that's not possible, right? Shifting all the satellites temporally so they appear together, arbitrarily maximizing the problem beyond what is real isn't an accurate depiction of the problem, IMO.

Put another way, if you erase the Starlink satellites from the images before stacking them, you then get a fairly accurate representation of the sky without any Starlink satellites, and you still have the data behind them (from the other pictures where that portion of the sky was not occluded). You can also probably fix the intensity of anything occluded in a few of the pictures but not others through some math.

> If the images are stacked, doesn't that mean that there's plenty of images with the parts that are occluded in others not occluded?

As he mentioned, flare can cover a broader area of the sensor.

There's still many other images that the flare isn't affecting that area at all though, right? The point is that there's many different images used here, and instead of combining them in a way to remove transient artifacts, which is the norm, they combined them in a way to emphasize transient artifacts, which is not something people normally would want to do.

It's worse than any single image that's used to make it, so why make it at all?

It's an interesting way to demonstrate the density of the satellites- they're not hiding the number or duration of images, so you can get a fuzzy intuition for how likely you are to catch them in a given image.

Except you're not actually demonstrating the density. That could have been one satellite if you exposed long enough.

It's like taking a picture of the one car every day driving on some empty road and then coadding them and claiming you're demonstrating density of traffic.

what if the research requires long/multiple exposures.

If I understand correctly the satellites would be much dimmer there, but otherwise you'd see the same problem - correct me if I'm wrong.

Not an actual astronomer, but somewhat aware of detector issues for space-based telescopes.

Typically, you want to use one single integration time, if possible. Doing so can limit readout noise, for one thing. (Note that ground-based astrophotographers, with consumer cameras, often take multiple exposures to cool down the detectors. This is not a factor with the actively-cooled detectors in astronomical telescopes.)

For instance, the observations leading to the recent discovery of the two-planet exosystem [1] used roughly 15-minute integration times [2, sec. 2] across an 8 meter (!) aperture. This is a longer integration time than the 17, 30-second integrations that produced OP.

I think the above considerations about stacking vs single exposures are a side issue, though.

The saving grace is that many astronomical observations are narrow-field-of-view, and are done away from the illuminated part of the StarLink constellation. But, surveys or wide-field images could really be affected strongly, at the 30-40% level [3].

[1] https://news.ycombinator.com/item?id=23917559

[2] https://academic.oup.com/mnras/article/492/1/431/5680498

[3] https://www.eso.org/public/archives/releases/sciencepapers/e...

There are plenty of amateurs using actively cooled specialized astrophotography cameras with total integration times reaching hours (2-3 are typical). Readout noise can be cancelled by taking so called bias frames. Shorter exposures are sometimes preferable because of the light pollution and to minimize the possibility of an exposure being rendered usles by some plane or satellite: it is less painful to throw away three minutes of exposure time than 15.

> Readout noise can be cancelled by taking so called bias frames.

That’s not what bias frames [1] are doing. Bias frames measure a systematic effect in the conversion of stored charge to a digital number in the CCD device. You can then subtract it out.

The calibration offered by bias frames does not fix readout noise, which is a random, not systematic, effect that is inherent in the same conversion.

You can use the bias frames to compute the strength of the readout noise (i.e., the rms magnitude of its random part) but you can’t remove the random noise itself.

See sec. 1.1 of the linked reference for more.

[1] https://users.astro.ufl.edu/~lee/ast325/handouts/ccd.pdf

The point of bias frames is what you take them without accumulating any charge. Well yes, you cannot eliminate random noise, but you can reduce it. Also, most popular deidcated astrophotography cameras are now CMOS not CCD and readout differs in these. Your link does not work, and probably for the better. See this instead: https://www.youtube.com/watch?v=O6oGyFlZgAc

For one a Starlink satellite, or any LEO satellite, is moving really fast relative to a ground based telescope. It'll occult any given star for a fraction of a second.

If you're doing a narrow field long exposure with tracking, that will be problematic and you'll loose data. Most scientific observations don't do long exposures, especially measuring highly variable things like exoplanet transits.

It's more effective to stack a large number of short exposures for scientific measurements. It's much easier to eliminate noise because everything that is not noise will be transient for a single frame in the stack. You can also remove frames with things like satellites, clouds, or airplanes without losing much data.

Knowing the ephemeris of satellites also allows observations to time short exposures to avoid occultations. Astronomers have been dealing with satellites, clouds, and airplanes for a long time. Amateur astronomers might have a harder time with Starlink satellites but they will adapt just like they did with the ISS and Iridium.

"Members of the LSST science team said last month that, assuming the full deployment of SpaceX’s Starlink satellites, nearly every exposure from the observatory within two hours of sunset or sunrise would have a satellite streak. During summer months, when twilight times longer, there could be a 40 percent impact on twilight observing time, according to the Association of Universities for Research in Astronomy, or AURA, which manages the LSST project for the National Science Foundation.

“Because of scattered light in the optics by the bright satellites, the scientific usefulness of an entire exposure can sometimes be negated,” AURA said in a statement last month. “Detection of near-Earth asteroids, normally surveyed for during twilight, would be particularly impacted. Dark energy surveys are also sensitive to the satellites because of streaks caused in the images. Avoiding saturation of streaks is vital.”"

https://spaceflightnow.com/2019/12/09/spacex-to-experiment-w...

They're trying to do a logical OR for all the light that comes in during a long exposure, as in an old film camera with a long shutter time.

That may have worked with analog film before we had satellites, but Starlink is just the latest and brightest addition to the sky that makes that an increasingly bad idea. Change is hard, but it's not that difficult.

Instead, AND a bunch of shorter exposures. Then, your streaks are not streaks, but a series of point sources that can be removed trivially.

And who will pay for all those necessary changes? Keep in mind that this is complex stuff, you don't just hack a perl script and be done with it. Anybody who thinks it's that simple is suffering full-blown Dunning-Kruger.

If I poo in your front yard, would you be happy with me telling you simply "well you can just remove it, what's the big deal?" Or rather, pooing regularly somewhere on a lawn in a public park that you really like and visit daily.

This concern has already been addressed by SpaceX. It's worth noting that LSST is the single telescope most affected by Starlink. Since the article you posted was published, SpaceX met with the LSST people and specifically designed a sunshade to reduce the brightness of Starlink satellites on station below the threshold that will cause saturation problems. https://www.spacex.com/updates/starlink-update-04-28-2020/

>Most scientific observations don't do long exposures

Not quite the whole truth, I think you're speaking a tad beyond your expertise. While you're right that transient science generally does short exposures (TESS is 2 seconds, Kepler was 6.5, LSST will be ~20 iirc), the vast majority of astrophysical science I've been exposed to (mechanical engineer at an astrophysics research institute), 10 or 20 minute exposures are more the norm. Especially when looking at faint objects.

Doesn't take away from your main point - astronomers will adapt. I think they're perturbed by this because Starlink makes their jobs even more complex, and thus more expensive. A cost that SpaceX doesn't bear at all.

I looked up the exposure times for transient searches and went with that. It matched what astronomer friends had told me. I am not an astrophysicist.

I think astronomers have a right to be perturbed about Starlink et al. SpaceX seeming to not care about the effect Starlink satellites would have on astronomy is ridiculous. At the same time stupid mis/disinformation like this tweet is also ridiculous.

Exposures are generally taken frequently enough that any that would be problematic for data can simply be dropped. The satellites aren't permanently positioned in the sky. Geostationary satellites permanently lose any data behind them (although I haven't heard of a geostationary satellite being positioned precisely problematically).

Searching for exoplanets moving in front of a star, involves measuring the luminosity of a single pixel for ~1% change. Dropping frames ruined by a satellite train isn't going to work. E.g https://www.sciencealert.com/a-bunch-of-potential-tabby-s-st...

You absolutely can drop those frames. Exoplanetary transits are generally in the range of 1 to 4 hours, and hundreds of exposures may be taken in that time. In my undergraduate I studied astronomy and some of my classmates did an exoplanet detection project with a 61" scope and had issues with satellites on a handful of frames. Sure, it sucks to lose 2 minutes worth of data, but it's not even lose to catastrophic when you have the data before and after as well.

Absolute bunk. As numerous people have tried to explain in this thread (along with its frequent predecessors), anything that moves across frames is by definition unimportant when stacking images. It's literally the difference between integration and differentiation.

Satellites move fast. Not only that, but you know exactly where and when they will cross your field of view, and for how long they'll remain within it. If the astronomy community can't muster the rudimentary image processing technology needed to reject satellites and other transient objects, I'm not exactly confident in their ability to finally figure out the whole origin-of-the-Universe thing.

So if someone decides to make money with a byproduct being messing with your day job, you’d just shrug it off?

If I were that incompetent at my day job, I imagine quite a few people would be "messing" with me, possibly for the sheer amusement of it.

Suppose your job comes and goes with the whims of a grant or funding committee, so every last dollar you spend having to compensate for some other guy's scheme.

Science is not just difficult because the research being done is hard; it's also difficult because everything you're trying to do is being done in spite of the fact 95% of the rest of the world could give a care leas that you're doing, so that last 5% of support you can count has to go a long way.

I can understand the bitterness perfectly, and to be frank, your response is exactly the reason they deserve to be irate.

So why should a goal like truly universal Internet access play second fiddle to astronomical considerations, when the technical solutions required for coexistence are so trivial?

(And I really couldn't care less how many astronomers disagree with the assertion that it's trivial to avoid interference to observations from satellite passes. If I want to know something about astronomy, I'll ask an astronomer. Astronomers, by the same token, should consult digital imaging experts before concluding that the rest of the world is out to wreck their careers with unreasonably-expensive or technically-insurmountable roadblocks. You guys can afford to lose a frame or two now and then in exchange for what Starlink offers the rest of humanity.)

There’s no sats up north...

Starlink will provide an immediate benefit to humanity, the astronomy performed otherwise, won't. Looking for exoplanets isn't a practical concern for humanity any time soon.

Odd perspective when defending technology from the same guy that wants to save humankind by literally colonizing other objects in space.

Not necessarily- exoplanets beyond our solar system really aren't going to be feasible to physically travel to until we can come up with something along the lines of faster-than-light travel. Given the lack of light pollution on Mars, I almost wonder if we wouldn't be better off doing all of our astrophysical study out there or in deep space anyway (assuming we have a steady supply of transit back and forth to transfer the data since streaming isn't likely to have good enough quality for a long time either).

Then again, I am also not a fan of putting more satellites into orbit either... mostly just struck by your comment as being not terribly odd to me.

Ofcourse astrophysical study works better in soace, but who is oaying the added cost of launching telescioes to space?

Travelling to exoplanets is not hard, it just takes 1000 - 10,000 years.

There will be no FTL untill we can manipulate black holes, if ever at all.

So the first trip to exoplanels will be in a multi-hundred year hybernation, or a genration ship, or people that are so different from us that they don't mind the journey.

This is completely incorrect. The data is not lost, when you stack images the low signal of a single (or multiple) satellites in a single image disappears into the noise.

> The data is not lost

point me to the data behind the occlusion on a single frame.

Reply on news.ycombinator.com