A series of round lenses produces a round image on some sort of sensor or sensor array.

When it comes to cameras and telescopes out in space, are the sensors also round?

It seems like most of the images I've seen from space telescopes and cameras are rectangular, but I'm not sure if that's cropping in post-processing or something more raw.

General answers are fine as well as specific well-known examples (like HST, to pick one at random.)

  • 7
    $\begingroup$ Many earth observation satellites are linear and capture the ground by moving over it. $\endgroup$
    – lijat
    Sep 12 '19 at 19:21
  • 1
    $\begingroup$ @lijat space.stackexchange.com/search?q=pushbroom $\endgroup$
    – uhoh
    Sep 13 '19 at 14:24
  • 2
    $\begingroup$ Great question! Why are images presented rectangular, instead of circular or pentagonal or gingerbread man-shaped? Light is if anything round. The Sun is round, waves have no corners, our eyes are round, everything is round in the world of imaging, except pictures on our flat TV-screens. What is this all about? I doubt that HST's mirror is designed to concentrate all light to a rectangular CCD. I can demonstrate for myself that my reading glasses in sunlight do not concentrate light to a rectangle. Basic optics as far as I understand it. Something needs to be explained here. $\endgroup$
    – Tombola
    Sep 14 '19 at 13:29
  • 1
    $\begingroup$ @Tombola: Images are made up of pixels and displayed on flat surfaces. Pixels have to be triangular, square, or hexagonal, as these are the only regular shapes that can tesselate a flat plane without leaving gaps in between. Of these three possible shapes, squares (or rectangles) are used because they have fourfold (90-degree) symmetry, making it easy to identify pixels using x and y coordinates. And, once you have square/rectangular pixels, a square/rectangular screen, displaying square/rectangular images, is the only shape that lets you avoid leaving sawtoothed gaps around the edges. $\endgroup$
    – Vikki
    Sep 14 '19 at 22:42
  • 1
    $\begingroup$ @Tombola My monitor's surface is optimized for rectangular pictures. What does yours look like? $\endgroup$
    – Mast
    Sep 15 '19 at 7:56


The Kepler space telescope uses a bank of 21 rectangular CCD modules - each with two 2200x1024 pixel CCDs). Each module covers 5 square degrees on the sky.


enter image description here

Giving a field of vision of:

enter image description here


For Hubble, the wide field camera CCD sensor is again rectangular/square


enter image description here

This WFC3 (Wide Field Camera 3) forms just a small portion of Hubble's field of view from its full instrumentation coverage.

enter image description here

This illustration shows the “footprints” of all the instruments in Hubble’s field of view. These include:

  • the fine guidance sensors (FGSs)
  • the Near Infrared Camera and Multi-Object Spectrometer (NICMOS)
  • the Space Telescope Imaging Spectrograph (STIS)
  • the Cosmic Origins Spectrograph (COS)
  • the Wide Field Camera 3 (WFC3)
  • the Advanced Camera for Surveys (ACS), which includes the Solar Blind Channel (SBC).

WFC3 and ACS are the two instruments involved in the Frontier Fields program.

  • 14
    $\begingroup$ that's a really beautiful example! It's round(ish) in outline and a (piecewise) curved surface as well for field curvature. $\endgroup$
    – uhoh
    Sep 12 '19 at 15:15
  • 12
    $\begingroup$ I never realised the the "CMOS" in NICMOS did not refer to the camera sensor technology. $\endgroup$
    – Skyler
    Sep 13 '19 at 13:02
  • $\begingroup$ What fraction of the light collected by the mirror is wasted between and around those rectangular CCD modules? 15% lost perhaps? $\endgroup$
    – Tombola
    Sep 14 '19 at 14:13
  • 1
    $\begingroup$ @Tombola Much more light is wasted by not even entering ... $\endgroup$ Sep 15 '19 at 10:22

Silicon wafers are sliced from a giant single crystal of silicon called a boule, which is grown from a seed crystal dipped in and then slowly pulled from molten silicon.

Circuits such as CCDs (and everything else) are patterend on silicon wafers aligned to the crystal axes of the wafers indicated by the wafer flat or notch (1, 2 see alignment flat on bottom, 3 see alignment notch on left)

This can sometimes be important for electrical reasons but it is very important for mechanical reasons because you need to "dice" the thin, delicate wafers into individual die and that is a lot easier to do allong crystal planes than it would be trying to cut out a circle. Crystals like to break when cut off-axis, tiny microscopic cracks propagate along crystal planes especially when the die are cut off-axis.

So if you have a rectilinear silicon die and a rectilinear circuit pattern and rectilinear readout system, there's absolutely no benefit to building a single-die circular sensor. (However, multi-die arrays are a different matter, as nicely illustrated in @Snow's answer!)

If your useful optical field is circular due to optical vignetting or aberration then you can mask your data electronically during processing.

A circle has 21% less area than the square in which is it circumscribed:

$$ 1 - \frac{\pi}{4} \approx \text{21%}$$

so you could speed up data transmission from a spacecraft by 27% if you only sent back the data from an inscribed circular field of a square sensor:

$$ \frac{4}{\pi} - 1 \approx \text{27%}.$$

That's a meaningful amount of time savings, considering that some deep-space spacecraft (e.g. New Horizons) can spend months sending back all the image and other data from a flyby photoshoot. However, I think instead that they make the optics good enough to provide good image quality out to the corners and keep the whole square (or rectangular) image data.

  • 2
    $\begingroup$ The crystal drawn out of molten silicon and the crystal which is cut into wafers are not the same crystal. Many steps of purifing by zone melting are necessary to get silicon that may be used to make camera sensors. Molten silicon in a crucible is not pure enough. $\endgroup$
    – Uwe
    Sep 13 '19 at 8:33
  • 3
    $\begingroup$ @Uwe Thanks for that! en.wikipedia.org/wiki/Zone_melting I like to think of it as the "same crystal" because the recrystallizing areas are still lattice-coherent with the crystal that was there before melting. It's a little bit like humans in that the stuff in our bodies is not the stuff that was there when we were younger, but we're still the same people (or at least like to believe we are). Our brain cells still remember the same stuff they knew before even though a lot of their molecules are a lot younger than the memories. $\endgroup$
    – uhoh
    Sep 13 '19 at 8:38
  • $\begingroup$ I don't find this answer very convincing as it stands. Yes, for commercially-produced cameras the sensor is cut rectangularly from a wafer for cost and other reasons. But the wafer itself is circular, and for something as expensive as a space telescope the cost of using a single wafer of the appropriate size, rather than a piece of a larger wafer, should easily be trumped by the advantages of the circular sensor – if it has any. $\endgroup$ Sep 14 '19 at 10:16
  • 1
    $\begingroup$ @LocalFluff oh I didn't notice that, but that's probably easy. The diffraction pattern of a square aperture would be more difficult to manage in the images than that of a circle. Any high resolution optical imaging system (telescopes, microscopes, cameras...) will have circular lenses and apertures, especially the entrance pupil. It's true that for variable diameter irises they are not exactly circular but have six or more blades, but on good systems they work hard to approximate circles by rounding them i.stack.imgur.com/BvKmj.jpg $\endgroup$
    – uhoh
    Sep 14 '19 at 23:09
  • 1
    $\begingroup$ @LocalFluff there's a new question Wide angle camera of Lunar Reconnaissance Orbiter, were rectangular lenses used? If the lenses turn out to be rectangular, we will still see that the system's acceptance will still be defined by a circular aperture and the diffraction limit also defined by that circle. They might cut out unused sections of the glass to save weight or space, but that missing glass would never have contributed to the image. $\endgroup$
    – uhoh
    Sep 14 '19 at 23:25

As another example: the Gaia mission illustrates that modern CCD production techniques allow to have form follow function: it uses an creatively laid-out array of CCDs to be able to integrate multiple functions in a single instrument:

[T]he three functions are built into a single instrument by using common telescopes and a shared focal plane:

  • The Astrometric instrument (ASTRO) is devoted to star angular position measurements, providing the five astrometric parameters [...]
  • The Photometric instrument provides continuous star spectra for astrophysis in the band 320-1000 nm and the ASTRO chromaticity calibration
  • The Radial Velocity Spectrometer (RVS) provides radial velocity and high resolution spectral data in the narrow band 847-874 nm

Each function is achieved within a dedicated area on the focal plane.

The result:

Gaia focal plane

(The Gaia focal plane; source)

Gaia CCD array

(The Gaia CCD array; source)


There were a lot of square format cameras.

  • The Voyager cameras had 800*800 pixels.

  • The LORRI cameras of New Horizons had 1024*1024 pixels.

  • The Galileo cameras had 800*800 pixels.

  • The Cassini WAC and NAC cameras had 1024*1024 pixels.

  • The narrow and wide angel OSIRIS cameras of Rosetta had 2048*2048 pixels.

  • The FC camera of Dawn had 1024*1024 pixels.

But there are also camera sensors being neither square nor round. The narrow angle cameras of the Lunar Reconnaissance Orbiter use a line sensor with 1*5064 pixels. The maximum image size is 2.5 x 26 km at an altitude of 50 km. The pixel scale is 0.5 m per pixel, so the 2.5 km wide image is resolved into 5000 pixels. The image length of 26 km is recorded during the orbital move of the camera around the Moon. The resulting image has 5000*52,000 pixels.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.