Previous section: CCD image processing

3. Artifacts

3.1 Bad column

These are just columns of pixels that do not work properly. There are several reasons for a column to be bad (ranging from electronic to mechanic), some columns are completely dead (dark), some are "hot" (bright), some have just a constant added to what they should normally show. A good CCD usually does not have more than a couple of them. They are quite easy to correct when they are isolated (one interpolates the neighboring columns), but nothing can be done when several are next to each other.

 [ bad column]
Figure 6: A small region from a CCD image showing various bad columns.

3.2 Traps

Some pixels are damaged: they cannot transfer the electrons when the image is read out. Therefore, all the pixels below such a bad pixel will be normal, but all those above it are lost, because their electrons are trapped in the bad pixel; this forms a partially bad column as illustrated in Fig.7.

 [ traps ]
Figure 7: A small fraction from a CCD image showing a trap pixel and the corresponding partially bad column.

3.3 Hair, dust...

3.3.1 On the CCD

Although it should not happen, sometime a foreign body can find its way into the CCD housing and end on the detector. These make dark marks on the images, that can usually be partially corrected by flat fielding the frames.

 [ stuff on the ccd]
Figure 8a: Unidentified foreign object on the surface of a CCD.
3.3.2 Dust on the filter
A dust grain on the filter or on the window of the CCD camera will cause a shadow on the detector. The typical doughnut shape is an out-of-focus negative image of the telecope main mirror, with the central obstruction caused by the secondary mirror. When the images are processed using good
flat-fields, these rings can completely be corrected.

 [ stuff on the ccd]
Figure 8b: Rings caused by some dust grains on the filter above the CCD.

3.4 Cosmic rays

When a high energy particle hits the CCD, it loses its energy by knocking the atoms constituting the chip itself. That liberates many electrons that cause a bright spot on the image. These high energy particle can either be genuine cosmic rays (exotic particle produced by exploding supernovae, black holes, etc.), or just the product of the decay of some radioactive atoms present in the lenses just above the CCD. Cosmic rays are usually easy to recognize, because they are much sharper than stars (the high energy particle hits just a couple of pixels). If one is just planning to produce a nice picture for a web site, they are very easy to clean out. However, removing them without damaging the real objects can be more tricky, but is still possible.

 [ Cosmic rays ]  [ More cosmics ]
Figure 9: A raw CCD image, zoomed in to show the cosmic rays, that appear as sharp and bright little objects, while the stars (trailed on this image) are have much smoother shapes. On the left panel, there is only one real un-elongated object (slightly toward the top-right from the center). The right panel shows cosmic rays from the lower left corner at a smaller scale.

The scale of some CCDs (size measure in arcsecond/pixel, i.e. what is the size of the pixel as projected in the sky) can be fairly large, especially the ones used by amateurs on small telescope. In that case, the image of a star will cover just a couple of pixels (we say that the images are under-sampled) and will look similar to a cosmic ray (c.f. Fig. 10). In that case, it is very important to take several images of the same field of view in order to remove the cosmic rays by taking a median of all the images: the result shows some light where there is light in most of the images, while a pixel hit by chance by a cosmic during only one of the exposures will remain dark.

 [ big pixels ]
Figure 10: The same field as in Fig.5 (with the same cosmic rays), with larger pixel to illustrate how difficult it can be to distinguish a cosmic ray from a star when the pixel size is under-sampling the images.

3.5 Saturation trail

Each pixel can store only a certain amount of electrons (of the order of 100,000). If a pixel is illuminated by a bright star and/or if the exposure time is long enough, that pixel will fill up, and the electrons will start to fill the neighboring pixels: the CCD is saturated. When the image is read, all the extra electrons will be spread over the column containing the saturated pixels, making a saturation trail, as shown in Fig.11.

As a consequence, any bright and narrow feature on the same column as a bright star must be considered with suspicion: it most likely is a saturation trail.

 [ saturated stars ]
Figure 11: Examples of saturation trails. The scale is different in each panel.

3.6 Diffraction features

As the light behaves like a wave, it is diffracted when it passes near an obstacle, i.e. a ray of light will be bended, the deviation being stronger if the ray passes closer to the obstacle.

In the case of a telescope, the most important obstacle is the spider, the cross-like support of the secondary mirror. Most telescopes have a 4-legged spider, which produces the typical cross-shaped diffraction pattern visible around the brightest stars. 1-legged spider produces a bar through the star, and some telescopes (like the Keck) have 6-legged spider, so the star images exhibit a 6-legged pattern.

It is important to note that the diffraction pattern is not necessarily symmetric: a cable around one of the legs of the spider will make the corresponding bar of the pattern much weaker (because the cable diffracts the light in random directions).

Also, every star in the field has a diffraction pattern, not only the brightest. If one co-adds many exposures in which just a couple of stars display a diffraction pattern, the result will show many more of these features around much fainter stars. Comets and other extended objects objects usually do NOT have diffraction crosses, because their fuzziness smears the diffraction pattern, which then blends in the object itself.

 [ diffraction
         cross ]
Figure 12: Complex diffraction pattern of the CFHT telescope: the main cross is not symmetric, and additional, fainter legs are present. The vertical mark is a saturation pattern.

Finally, the spider is the most common cause of the diffraction patterns, but there are other possibilities: the edge of the telescope, the light baffles, the edge of the mirror, etc... almost everything can cause diffraction patterns. They usually look less symmetric than the ones caused by the spider. This can also explain the presence of diffraction pattern in spider-less telescopes.

In some case, a mirror or field lens that has been carelessly cleaned will cause diffraction-like patterns.

3.7 Trailing

To image a moving object like a comet, the telescope is set to compensate the motion, so that the comet will appear un-trailed. Everything else in the field will be trailed. However, the intensity threshold used to display the image can give the impression that the brighter stars are less trailed than the fainter ones: the elongation becomes negligible compared to the large apparent diameter of the brightest star. The only way to be sure that an object is not elongated is to display the image with proper threshold, or to measure the elongation of the object using a software tool that takes into account every light level. This has to be performed on the original images: it cannot be done on a JPEG or a GIF copy, in which most of the information is lost.

 [ elongated stars ]
Figure 13.a: Some stars from Fig.1. The intensity thresholds are set for the brightest stars, which appear elongated.
 [ less elongated ]
Figure 13.b: Exactly the same images as X.a, with different thresholds. Fainter stars are now visible and elongated, while the brighter stars appear now circular, or un-trails

4. What else can go wrong

If an image presents a "feature", something completely strange that cannot be explained by any of the previous artifacts, it is most probably another kind of artifact. Before reporting the discovery, one must check absolutely everything. It took me some time to realize what the image of Fig.14 was. No, it is not an alien spacecraft passing in front of a star. The telescope jumped of a few arcseconds about 30 second after the beginning of this 15 min. exposure, then did not move until the end.

 [ UFO? ]
Figure 14: Consequence of a jump of the telescope a few seconds after the beginning of the exposures.

Previous section: CCD image processing
Note: these images come from actual data I collected at the European Southern Observatory at La Silla, Chile, with the New Technology Telescope (3.55m diam.), equipped with the SuSI camera, and at the Mauna Kea Observatory, Hawaii (at the 88 inch (2.2m) UH telescope and at the Canada-France-Hawaii 3.6m telescope), with the U.H Tektronics 2048 and U.H. 8k CCD mosaic.
| Hale-Bopp's companions page | Hale-Bopp's companions table | IfA Hale-Bopp page |
| IfA home page | O.H.'s page | Comments? |

Olivier Hainaut

Tue Dec 10 19:09:33 1996