-
Space Telescopes Q&A.
by
baldusi
on 12 Aug, 2013 22:31
-
I'm wondering what are the practical limits on f-number for space based telescopes. I'm particularly interested in medium aperture (around 0.5m) but really long focal distance (like f/60). I understand that at least Hubble is an f/24. And most of the issues of aberrations and correction stem from wider fields of view. I also understand that the resolving power is only dependent on the aperture and that you might only have some section of the total area for each sensor.
But let's say that we wanted to make a dedicated telescope to exclusively observe a near (>10ly) star system, like Alpha Centauri, for example. With particular interest in planet finding and characterization.
But since this is about telescopes I'm (again) interested on the practical limits of long focal lengths.
-
#1
by
kevin-rf
on 15 Aug, 2013 14:54
-
Your magnification is limited by the diffraction limit of the objective. Start there.
Most modern large optical telescope use a fast (< f4) objective combined with a curved secondary that increases the f ratio and focal length (f10, f22, you name it.)Further improvements can be made with a corrector, or focal length extender.
Using a very high f ratio objective requires a very long tube that in space will have thermal stability issues that must be taken into account.
I believe Hubble magnification is diffraction limited. To get higher resolution than Hubble would require a bigger Hubble.
-
#2
by
kevin-rf
on 15 Aug, 2013 14:57
-
-
#3
by
IRobot
on 15 Aug, 2013 15:52
-
I'm wondering what are the practical limits on f-number for space based telescopes. I'm particularly interested in medium aperture (around 0.5m) but really long focal distance (like f/60). I understand that at least Hubble is an f/24. And most of the issues of aberrations and correction stem from wider fields of view.
A space telescope project has enough budget for decent optics, no need for long focal lengths. At f/60 you need extremely bright targets (like planets) or a lot of exposure time.
You can also get more "magnification" by using a camera with very small pixels or by inserting a barlow in the light path.
-
#4
by
kevin-rf
on 15 Aug, 2013 16:07
-
I'm wondering what are the practical limits on f-number for space based telescopes. I'm particularly interested in medium aperture (around 0.5m) but really long focal distance (like f/60). I understand that at least Hubble is an f/24. And most of the issues of aberrations and correction stem from wider fields of view.
A space telescope project has enough budget for decent optics, no need for long focal lengths. At f/60 you need extremely bright targets (like planets) or a lot of exposure time.
You can also get more "magnification" by using a camera with very small pixels or by inserting a barlow in the light path.
Small pixels mean longer exposures, higher noise, less dynamic range. Reducing the size of the pixel reduces the surface area, reducing the amount of electrons a CCD pixel can store (well depth). Trust me, the noise differences between a 4um and 24um pixel are quite something.
If you want super high magnification of bright objects, you really need something that goes beyond Hubble in size (and cost).
That, or you need a space based interferometer, so you get an equivalent larger numerical aperture. It all comes down to building very, very large diffraction limited optics.
-
#5
by
kevin-rf
on 15 Aug, 2013 16:16
-
It is worth pointing out that while Hubble is an f/24 system, the primary 2.4meter mirror is not f/24, but in the f/2 range. The secondary is why it is f/24.
-
#6
by
jebbo
on 15 Aug, 2013 16:26
-
If the aim is direct imaging of exoplanets, you almost certainly need a coronograph / external occultor / nulling interferometer or the light from the planet will be swamped by the star.
-
#7
by
IRobot
on 15 Aug, 2013 16:34
-
You can also get more "magnification" by using a camera with very small pixels or by inserting a barlow in the light path.
Small pixels mean longer exposures, higher noise, less dynamic range. Reducing the size of the pixel reduces the surface area, reducing the amount of electrons a CCD pixel can store (well depth). Trust me, the noise differences between a 4um and 24um pixel are quite something.
If you want super high magnification of bright objects, you really need something that goes beyond Hubble in size (and cost).
That, or you need a space based interferometer, so you get an equivalent larger numerical aperture. It all comes down to building very, very large diffraction limited optics.
I know a lot about CCDs, I've developed a scientific CCD camera some years ago
. Of course you lose some characteristics with smaller pixels, like dynamic range and anti blooming capability (if required), for example.
But I don't agree with the noise, there are some new CCDs that have extremely low now and high QE. Also, scientific CCDs usually use pixels above 9um, some of them go as high as 24um, so reducing the pixel size could be done to something like 5um and still get decent dynamic range.
And you can still do binning. That improves the dynamic range and reduces read noise.
And the guy asked for an f/60 target for a 0.5 meter scope. That does not leave many useful targets, most likely they are very bright, therefore not needing a low noise camera.
-
#8
by
baldusi
on 15 Aug, 2013 19:06
-
I never said direct imaging in the sense of resolving the planet. I was thinking about treating them as light sources. So I need to resolve 0.1AU at 10ly. I calculated that if had a 500mm telescope with an f/15, and a 2048px x 2048px with 12um pixel size (24mm x 24mm sensor) I could exactly resolve 0.1AU at 10ly. With that, I could detect planets up to 200AU of the observed star. The whole question was if I could use the full aperture to the 24mm x 24mm sensor. AIUI, Hubble uses only the top half, and each sensor get's a piece of the aperture.
The whole point is to make a telescope as cheap as possible so that you don't mind letting it pointed to a single star system. It should fit, at most, into a Small Explorer budget. Ideally, much cheaper.
-
#9
by
IRobot
on 15 Aug, 2013 19:18
-
Have you ever made astronomic imaging? A star glow will obscure any planet. Either the star is very low in brightness or the planet must be very far and have a high albedo.
It is not a problem of resolving power.
-
#10
by
jebbo
on 15 Aug, 2013 19:20
-
I'd worry about point spread functions ... but it's worth looking at Sara Seager's work on ExoplanetSat, which has exactly that goal: a dirt cheap telescope where you deploy a constellation of them, one per star looking for transits and RV. ISTR $600,000 each if you built a bunch of them.
-
#11
by
baldusi
on 15 Aug, 2013 20:12
-
Have you ever made astronomic imaging? A star glow will obscure any planet. Either the star is very low in brightness or the planet must be very far and have a high albedo.
It is not a problem of resolving power.
I always thought the problem was that the star glow was within the diffraction or resolver limit. If that's not controllable with extremely small FOV, then yes, it would need an occulter of sorts. That's the sort of answers I'm looking for.
-
#12
by
IRobot
on 15 Aug, 2013 20:20
-
You can record transits with a 8" telescope in the ground with brightness changes up to 1/1.000. Sara Seager seems to go down to 1/10.000, although very limited in terms of star magnitude.
I was reading her paper and something looks wrong, the dark current is 12.5 e-/s, which is a pretty bad value. The camera that I have designed had 0.008e-/s at 0ºC.
BTW, sensor noise is very low at temperatures below -30ºC, although in space you get a lot of cosmic rays ruining your shots.
-
#13
by
IRobot
on 15 Aug, 2013 20:26
-
Have you ever made astronomic imaging? A star glow will obscure any planet. Either the star is very low in brightness or the planet must be very far and have a high albedo.
It is not a problem of resolving power.
I always thought the problem was that the star glow was within the diffraction or resolver limit. If that's not controllable with extremely small FOV, then yes, it would need an occulter of sorts. That's the sort of answers I'm looking for.
There are several ways in which a star brightness can affect adjacent pixels. The last in the chain is the CCD read process, which can slightly increase pixels values after a bright pixel has been read. Also subsequent images can have a residual charge.
-
#14
by
kevin-rf
on 15 Aug, 2013 20:54
-
You can record transits with a 8" telescope in the ground with brightness changes up to 1/1.000. Sara Seager seems to go down to 1/10.000, although very limited in terms of star magnitude.
I was reading her paper and something looks wrong, the dark current is 12.5 e-/s, which is a pretty bad value. The camera that I have designed had 0.008e-/s at 0ºC.
BTW, sensor noise is very low at temperatures below -30ºC, although in space you get a lot of cosmic rays ruining your shots.
Depends on you if you use a cooled camera or not. Noise vs. Temp is not a linear function. If you are trying to observe transits, your exposure times should not be long enough for the dark current to be a factor. Cooling adds costs to cost constrained systems.
And yes before leaving to work for one of my customers, I use to work for a company that designed and sold scientific cameras. They stopped selling lightly cooled -30 cameras years ago. Actually on my desk I have two prototypes with Sony's new CMOS that they claim out performs the old Sony scientific grade ICX-205 and ICX-285 CCD's. Extraordinary claims require, well you know the rest. Over the last few years Sony has done a great job revamping it's entire CMOS/CCD sensor line. Still if you want a good low light Astronomy CCD, E2V is the way to go.
Btw. Terrestrial cosmic rule of thumb is 1 cosmic per square cm every 20 seconds. What was it, 10 electrons per micron of silicon travel by a cosmic. Or was that 10 counts at 10 electrons per count... getting old.
-
#15
by
baldusi
on 16 Aug, 2013 13:21
-
So, small FOV and closeness makes no difference in blooming? Less than 10ly is so close that parallax is quite noticeable. What I expected, was to be able to image the planets as lightpoint around the star. I chose 0.1AU as the resolving power since that's around 10 Sun radii, and I want as very small field of view to minimize the star's light as an issue.
BTW, since I was expecting to observe nearby stars, the idea is that the center would point to the star. Which might simplify designing an occulter. Multi star systems (hello Alpha Centauri) might represent an extra problem.
Having a FOV of 660arcsecs, and pointing to the star directly, should also reduce the issue of coma and astigmatism. BTW, the parallax of Earth should allow to actually measure which lightsources are in the general vicinity of the the star.
Couple of more general questions:
1) I'm interested in understanding about the main mirror f-number and the total telescope number. The use of small f-number main and extending the focal length with the secondary (and tertiary?) is only done on Ritchey–Chrétien or in all Cassegrains?
2) I've seen that some telescopes allow you to mount your sensor in place of the secondary and use a much wider field of view (and a lot more light). Thus, adding reflective elements would allow for longer focal lengths while keeping the resolving power of the main's aperture? Of course I understand that you'd have progressively less light and might get extra aberrations (and probably a much worse MTF and cost).
-
#16
by
kevin-rf
on 16 Aug, 2013 18:02
-
660 arc seconds? Considering from earth Jupiter is about 30 arc seconds...
-
#17
by
baldusi
on 16 Aug, 2013 18:42
-
660 arc seconds? Considering from earth Jupiter is about 30 arc seconds...
That's a big FOV? I calculated that 0.1AU at 10ly is 0.3arcsec. With a 2048px x 2048px sensor, that's 614 arcsecs on the side. So, should I calculate a 256px x 256px sensor for a single star system? That would be 76arcsecs².
-
#18
by
IRobot
on 16 Aug, 2013 20:57
-
Couple of more general questions:
1) I'm interested in understanding about the main mirror f-number and the total telescope number. The use of small f-number main and extending the focal length with the secondary (and tertiary?) is only done on Ritchey–Chrétien or in all Cassegrains?
AFAIK, all Ritchey Chrétien, Schmidt Cassegrains, Maksutovs, Dall Kirkam, etc.. For example my Celestron C9.25 has a 2.3/f in the primary and something around 4.34 in the secondary.
2) I've seen that some telescopes allow you to mount your sensor in place of the secondary and use a much wider field of view (and a lot more light). Thus, adding reflective elements would allow for longer focal lengths while keeping the resolving power of the main's aperture? Of course I understand that you'd have progressively less light and might get extra aberrations (and probably a much worse MTF and cost).
Just google for barlows...
You must consider that even if you have a perfect Ritchey Chrétien (or better designs), it will have artifacts and far from ideal PSF. You got the central obstruction, the supporting vanes, tube reflections, reflections/refractions on filters/CCD chamber glass, CCD cover glass, square pixel representation, etc, etc...
-
#19
by
baldusi
on 16 Aug, 2013 21:29
-
I thought that you can't use lenses because it's almost impossible to make them APO over a wide spectrum. At least beyond CoastalOpt I don't know many that are actually good.
-
#20
by
IRobot
on 17 Aug, 2013 20:24
-
You don't need to make them APO on a wide spectrum. Scientific data is usually taken using pass band filters. You could even have a barlow for each filter.
Exoplanet surveys are mostly done in infrared, so you just need a good lenses for IR.
-
#21
by
baldusi
on 18 Aug, 2013 20:41
-
You don't need to make them APO on a wide spectrum. Scientific data is usually taken using pass band filters. You could even have a barlow for each filter.
Exoplanet surveys are mostly done in infrared, so you just need a good lenses for IR.
What's the lambda for exoplanets? 1000nm? 3000nm? Can you do it with a normal sensor or you need active cooling?
Is it about right that for 1000nm I would need a 27m aperture to resolve an Earth sized planet at 10ly?
-
#22
by
Comga
on 18 Aug, 2013 21:36
-
You don't need to make them APO on a wide spectrum. Scientific data is usually taken using pass band filters. You could even have a barlow for each filter.
Exoplanet surveys are mostly done in infrared, so you just need a good lenses for IR.
What's the lambda for exoplanets? 1000nm? 3000nm?
There is no special wavelengths for exoplanets. Different investigations work better at different wavelengths.
Can you do it with a normal sensor or you need active cooling?
It depends. :-) Also, we use a lot of passive cooling even in Earth orbit. JWST will passively cool its structure to 40K.
Is it about right that for 1000nm I would need a 27m aperture to resolve an Earth sized planet at 10ly?
Are you imagining real resolution, with critically sampled PSFs smaller than the exoplanets? This wont even be an issue until we have such targets. (I have not heard of any Earth sized exoplanets within 10, or 1000, LY.)
You could search for "Terestrial Planet Finder" to see how JPL would go a out this and see why even then they can't get this off the ground.
. Of course you lose some characteristics with smaller pixels, like dynamic range and anti blooming capability (if required), for example.