Stargazing The terrible distortions of the Hubble Telescope

[Andy Resnick]

Wasn't the focal length that got ground, that which would be needed to focus on the Earth's surface rather than at infinity?

Yes.

 

[fizzy]

The company that ground the Hubble mirror did most of their work for DoD making optics for surveillance satellites. I don't have a reliable reference for this, but my understanding was that the mirror got ground to the usual curvature for those satellites. They would not have wanted to talk about the capabilities or number of such mirrors they make for obvious reasons at the time. A lot more information is available about that now.

In fact I recall reading that DoD have offered to donate one such set of optics, that they no longer need to put into service, for scientific use if someone can muster the funds needed to get it into orbit. Presumably they'd just need to make similar optical correction devices as those developed for Hubble and fit them before launching.

Wow, a free Hubble replacement. Maybe someone should crowd fund paying SpaceX to get it into orbit ;)

 

[fizzy]

Getting back to the deconvolution question. What is ideally required is an image of a single star in otherwise empty section of the sky with the current optics. That probably already exists as verification of the corrective devices fitted to compensate the grinding error. If anyone knows where to find such an image please post.

That can be used to create the deconvolution mask. Since every point of light in a nebula suffers similar distorting 'cross wires' this must degrade the image. It would be interesting to look at but my gut feel is that the processing would introduce more artefacts than the defect it is intended to remove would gain.

It gets very processor intensive and rises exponentially as image size increases, so you'd need a super computer to process a full resolution Hubble image. However, it should be possible to test on a smaller scale of a few megapixels on a PC.

In fact, not just the cross wires but a lot of the flare around each star could potentially be removed

 

[Laurie K]

The 'HST Optical Systems Failure Report' from NASA is interesting reading.

https://www.ssl.berkeley.edu/~mlampton/AllenReportHST.pdf

 

[mfb]

It gets very processor intensive and rises exponentially as image size increases, so you'd need a super computer to process a full resolution Hubble image. However, it should be possible to test on a smaller scale of a few megapixels on a PC.

Most deconvolution techniques don't rise exponentially with the image size, but none of them is perfect.

 

[fizzy]

Most deconvolution techniques don't rise exponentially with the image size, but none of them is perfect.

OK, I was speaking quickly . DFT is O(n²), which is similar to exp, it's actually super-exponential ; FFT is O(nlogn)

It still blows up on you quickly if you want to work with Hubble sized images

when experimenting I usually crop down to 512px by 512px to save waiting all day for each test .

 

[fizzy]

It might be a matter of semantics, but I don't think so. .

It certainly is nothing but semantics. It was ground perfectly to the wrong parameters. We are all agreed on that so lets' not spend the rest of the day arguing about it.

The NASA report says it was ground flatter than it should have been. I don't think that is compatible with the idea that it was ground to focus Earth as the usual series mirrors would have been. It should be noted that the report says this explanation is speculation since they never were able to locate the relevant documents.

The mirror was ground in 1982 and the report was done in 1990. I do not expect that DoD would want much discussion of the construction of their spy satellites getting into such a report. Whatever happened, certain details may have been changed to protect the 'innocent.' ;)

 

[mfb]

You can edit your posts if you want to add something.

Exponential would be cⁿ, O(n²) and O(n log n) are polynomial.
Hubble's wide field camera has two sensors with 8 megapixel, some smartphone cameras have more than that, and NASA can certainly afford using 32 to 1000 times the computing power necessary for 512x512 pixel images.

 

[Hornbein]

OK, I was speaking quickly . DFT is O(n²), which is similar to exp, it's actually super-exponential ;

O(n²) is about as polynomial as you can get.

 

[fizzy]

Thanks guys, I do know that n2 is a polynomial.

I thought Hubble images were bigger than 8Mpx, that may be doable on a decent PC if you are not in a hurry for the results.

Clearly it's doddle for anyone with large computing resources. I've already pointed out that NASA were obviously doing deconv processing on the first approach images they got back from Pluto fly-by.

It might be quite fun to try this but it would require a single star image for the mask and unmodified source image in a non-lossy format. Pointless if it's supplied as jpeg.

 

[D H]

Just to clear up another urban legend, the primary mirror figuring wasn't exactly 'wrong', it was ground exactly as it was for the other KH-11 satellites.

There is an urban myth here, but it is that the Hubble mirror was ground exactly as it was for other KH-11 satellites (i.e., to focus on something 200 miles away as opposed to focusing at infinity). You yourself provided a link that shows that this is not the case.

Analysis (http://ntrs.nasa.gov/search.jsp?R=19910003124) shows that the field lens of the null corrector was set 1.3mm too far from the nominal position due to misalignment of the metering rod. That 1.3mm distance error is sufficient to explain the aberrations.

The key technical cause of that 1.3 mm error was an erroneously cut piece of black electrical tape that was used in lieu of an antireflective coating. This caused the technicians to measure a spot 1.3 mm closer than the intended location. A key image from that report portrays what went wrong.

The root causes of the problem were much deeper than that botched tape job and botched measurement. Perkin-Elmer had underbid the contract. This led to rushed operations (e.g., using electrical tape instead of paint), and ignoring independent measurements that showed the mirror was being ground incorrectly. NASA had created a very hostile relationship with all of the companies working on the Hubble, including PE. Reporting problems to NASA was highly problematic. NASA made it very clear at the time that they did not want to hear about any problems, so they weren't told about them.

That Hubble's primary mirror was ground incorrectly ultimately was a failure of management and of culture. Fortunately, the end result in this case was an incorrectly ground mirror that was eminently correctible.

 

[Andy Resnick]

There is an urban myth here, but it is that the Hubble mirror was ground exactly as it was for other KH-11 satellites (i.e., to focus on something 200 miles away as opposed to focusing at infinity). You yourself provided a link that shows that this is not the case.

I admit that my evidence is circumstantial, but it's a lot of circumstantial evidence. The dates when the mirrors were ground, the companies (and the technical teams) involved and the optical designs. are all the same. Finally, note that even though the Hubble mirror was orbiting Earth and not accessible for metrology, optical designers were able to construct a corrector (COSTAR) that perfectly compensated for the aberrations. Given that the date of the report was late 1990 and COSTAR was launched in 1993, it is unlikely that the designers started after the report was issued, implying that the designers already knew (or suspected) the underlying cause and had begun working on a corrector prior to issuance of the report.

Misalignment of the spacing rod was *a* critical cause of the failure, not the only cause- another cause was the use of spacers, which had not originally been present.

But I agree, the root causes of the failure were not simply technical assembly errors- project reporting and management were the root causes.

 

[fizzy]

Sure PE made a whole series of those mirrors and the spec. was very similar. That is a very good reason to give them the contract and why they may have been able to put in a competitive bid: there was virtually no development needed.

The paint flake story I find a bit flakey and it is stated to be only speculation but the report says the mirror was too flat. That seems to be going in the wrong direction to be shorter focal length grind.

Just what is the 'circumstantial' evidence that the mirror was ground at the same radius as a spy satellite?

"Given that the date of the report was late 1990 and COSTAR was launched in 1993, it is unlikely that the designers started after the report was issued,"

So your whole argument is "unlikely" without stating why this makes it unlikely. There is a principal in science of falsifiability. If you do not make a specific claim or statement, your point is not falsifiable and thus unverifiable.

Once the investigation had determined the nature of the error, I should think they would be able to make the corrective optics quicker than a bureaucratic committee can write a report and get it security vetted to make sure they did not reveal any details they should not about PE's other mirror grinding work.

I don't see any circumstantial evidence that he mirror was short focusing rather than long focusing as the report claims. What am I missing? It's quite a tempring story but I don't see any evidence to support it.

 

[Drakkith]

I admit that my evidence is circumstantial, but it's a lot of circumstantial evidence. The dates when the mirrors were ground, the companies (and the technical teams) involved and the optical designs. are all the same. Finally, note that even though the Hubble mirror was orbiting Earth and not accessible for metrology, optical designers were able to construct a corrector (COSTAR) that perfectly compensated for the aberrations. Given that the date of the report was late 1990 and COSTAR was launched in 1993, it is unlikely that the designers started after the report was issued, implying that the designers already knew (or suspected) the underlying cause and had begun working on a corrector prior to issuance of the report.

From the website of the company that designed COSTAR:

Perhaps the most famous of our Hubble instrument is the Corrective Optics Space Telescope Axial Replacement (COSTAR) that helped correct Hubble’s hazy vision. We designed and produced the complex COSTAR optics in only 28 months instead of the typical 48. When COSTAR and JPL’s WFPC-2 camera corrected Hubble’s vision after the first servicing mission in 1993, Ed Weiler, NASA’s chief Hubble astronomer, said the telescope was "fixed beyond our wildest expectations."

http://www.ball.com/aerospace/programs/hubble-space-telescope

 

[fizzy]

"fixed beyond our wildest expectations."

Yep, they really pulled a minor technical miracle recovering a situation like that.

 

[Drakkith]

The paint flake story I find a bit flakey and it is stated to be only speculation but the report says the mirror was too flat.

Well, you can see a picture of the missing portion of the anti-reflective coating on the null corrector on page 50 of the PDF here: http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/19910003124.pdf
That doesn't prove that this was the primary cause of the error, but I'd say it's the most likely explanation.

 

[sophiecentaur]

"fixed beyond our wildest expectations."

Yep, they really pulled a minor technical miracle recovering a situation like that.

That 'repair' sequence was some of the best space TV I have ever seen.

 

[Chronos]

As I understand the NASA technical reports, the null corrector was incorrectly figured. The HST was matched to the null corrector, so it ended up misfigured.

 

[Lord Crc]

Getting back to the deconvolution question. What is ideally required is an image of a single star in otherwise empty section of the sky with the current optics.

I stumbled over a BBC Horizon episode from 1991 about the issue with Hubble's mirror, and the specifically mentioned deconvolution there. The explanation they gave for why it was not suitable was that to get the whole image clear rather than just one star, they had to split the image in small regions and optimize for each one separately. And back in 1991, that simply took too much CPU time.

The episode also highlighted a point I had not considered, namely that the different instruments on board suffered differently from the mirror flaw. At east from what I could gather, deconvolution would have been useless for some of the other instruments.

On a completely unrelated note, it's amazing how nice it is to watch older tv documentary shows from before the ADHD producers took over...

 

[1oldman2]

How could anyone not love this.

 

[trainman2001]

What I find that's miraculous, was the another group of fallible humans were able to come up with an optical solution that created some of the most dramatic and amazing space images mankind has ever seen (until the James Webb, I think). Hubble has done for cosmology what Galileo did for basic astronomy. Just those deep field images were awe inspiring in the literal sense of the word. The thought that in every tiny area of the sky you look at is teeming with galaxies that go on and on, simply puts the definition of the universe in a completely new context for layman like myself. That being said, I hope that they get the James Webb right the first time, since it's distance from Earth precludes a space walk to fix it.

 

[mfb]

since it's distance from Earth precludes a space walk to fix it.

The SLS could target the JWST at some point, and I'm sure Elon Musk would be interested in selling a Falcon Heavy / Dragon 2 mission to NASA. Both options would need a long time and a lot of money, however.