Extraction of spectra from series of frames (data reduction)


We consider here only the case that the recording with a CCD camera were made. Photographic images (chemical) are now replaced by the CCD technology and play no more a more practical role in the amateur field. In spectroscopy by digital cameras I have no experience and would like to exclude, therefore, here too, although their role in amateur circles is continuously increasing.

For the technology of charge-coupled devices (CCD) see here

Advantages of the astronomical used cooled CCD's:

  • Very high sensitivity. Quantum yields (efficiencies) above 30%, often above 80% (see the chart on left side).
  • Large linear range (dynamics), extends almost to the saturation level of the pixel.
  • By cooling to 20 to 40 degrees below the ambient temperature the dark current is minimized (formation rate of thermal electrons)
  • Good reproducibility.
  • Recordings are immediately available after the reading for the evaluation or monitoring.
Each CCD image contains certain components (in addition to the signal to be measured: the spectra stripes, see left photo) have nothing to do with our targets and need to be perceived interferencing perturbations. But they have a concrete physical background and can be partially described statistically and be eliminated.

Consider first the perturbations caused by the CCD camera effects. These are independent of the optics (telescope, spectrograph).

Bias: Each CCD image contains in each pixel, a minimum level of signal that is generated due to technical reasons (to evoid negativ intensity values). This bias is measured, in which a picture is taken without light infall (closed lock) with the shortest possible exposure time (<1 sec).
The bias should have around the same value for all pixels. The relative standard deviation of the bias value should be low. This is a relevant criterion for quality of each camera. There should be no pixels falling dramatically out of line. The bias is eliminated in the data reduction.

Dark current: Each pixel produces thermal noise electrons without need of photons (light from the object). This dark current rate in e / s (electrons / sec) is strongly (exponentially) dependent on temperature. To keep it as small as possible, CCD chip should be cooled in astronomical used CCD cameras. For amateur equipment by one- or two-stage Peltier elements (20 to 40 C below ambient temperature, in professional devices with liquid nitrogen (-180 C) ). The dark current given in a specific frame is proportional to the exposure time. With temperature as variable parameter. Therefore, a good temperature for the CCD is important (+ - 0.1 C). The quality of temperature control is a grucial quality criterion of each CCD camera.
But the rate of dark electron production is not equal for all pixels. Some of them produce a higher rate of "dark electrons", they are called "hot pixels". They are visible in the dark background as a single white pixel (see adjacent picture). The dark current of pixels (and the wrong hot pixel values) must be eliminated during the data reduction.

Cosmics: Occasionally a particle of cosmic radiation hits the CCD chip and draws a trace of induced electrons (sometimes more than 1 mm long (about 100 Pix.)). This arbitrarily oriented stripes are called cosmic-ray events. Such a stripe always appear only in a single frame of a series and is a singular event. They may be in the data reduction identified and eliminated through special routines. It is better to check all recordings of a series in an image editing program such as IRIS. A frame with a strong cosmic in or near the spectrum stripe, I would eliminate.

On the picture is shown a very strong cosmic but it does not interfere with the spectra stripe. The slitless frame is exposed only 1 minute, so the background noise comes out clearly.


The left is a screenshot procuced by MaximDL, which shows a y-profile through a raw frame. In addition to the dominant spectral bands are hot pixels visible. The background is homogeneous.

Let's speak about the components of raw frames that are not bred by the CCD camera, thus generated either in the spectrograph (scattered light), or indeed come from the sky (sky background, nebular lines).

Scattered light is difficult to see. When in the slit spectrograph scattered light is generated there exist also next to the star image (spectrum stripe) an background "intensity". It is difficult to eliminate because it is not homogeniously distributed over the CCD chip. We should "minimize" the possible leakage of light into the spectrograph (masking light leaks by sealing with black tape, packing the spectrograph in opaque cloth or foil, etc.).

The sky background is in the form of light pollution, a major irritant for the visual observers and astro photographers. The spectroscopist has it better off. The sky background can be eliminated during data reduction. A uniform gradient in the x- or y-axis, or both is not a problem for modern reduction software. Thus, the light pollution from a neighbors Party light will be controlled.

The brightness of a dark sky (e.g. in Namibia) corresponds to about 18 mag / arcsek . We have about 16 mag / arcsek as a rule.
By the way in slitless spectrographs the sky background is very strong, because the whole view field of the telescope contributes at any point on the CCD. If the spectra are recorded through a wide slit (the complete star picture falls through the slit) the sky background light decrease significantly because most of the (viewfield)sky will be covered.
In the case of slit spectrographs (slit width < star picture in telescope focus) the sky background plays a smaller role, because it is strongly dimmed by the narrow slit. But also during the data reduction of slit spectrograph frames the sky background will be eliminated..

Besides those already listed in the single shot existing "background" effects there are still artifacts generated by the optical causes: vignetting, the shadow of dust grains, in rare cases also "Fringes".

Vignettings are detectable by strong brightness gradients in the frames. The CCD chip is not uniformly illuminated, if somewhere a part of the light beam is not passed, because something is in the light path or the optical axis of the optical elements are not linearly aligned precisely. Here we must try to find the causes and eliminate them.

Shadow of dust grains is eliminate by averaged masterflat (to improve statistics from many shots). It is usually dirt particles, sitting on the camera front glass of the CCD chip.

Shadow stripes of dust in the slit are horizontally oriented.

The "flat" (on the left) shows all these effects:

  • Uneven illumination by vignetting.
  • Dust grains in the slit, visible as a horizontal black lines.
  • Dust grains on the front plate of the CCD chip = black circular areas with diffraction rings.


Fringes are a rare effect. They are created by the cover glass plate of the CCD chip by optical interference. I have previously experienced only once in a camera, which has therefore proved to be less useful for spectroscopic measurements. The effect is in normal use of such a camera in astrophotography not recognizable.


The extensive discussion of this effect with my camera then you see in http://spektroskopie.fg-vds.de/forum/viewtopic.php?t=2149&highlight=welligkeit
(in german language).

The accompanying analysis of a recording of the monochromatic line of a laser (slit spektrograph, LHIRES III) clearly shows that in addition to the intense laser line many "fringes" (secondary lines) occur. It is probably to interference, possibly caused by the (too) exactly parallel oriented polished front glass plate of the CCD. I've heard of astronomers that for professional CCD cameras because of exactly this effect does not used front glass plates with parallel surfaces. Here, the cover glass is given a prismatic shape.

In the original frame, the fringes are seen as weak "secondary lines" (left graph, the primary slit image is overexposed in the laser line).

The disturbing effects of the fringes (ripples) are already visible in the raw single spectra of a star. The fringes result in short "waves," particularly visible in the continuum of the star . If such "fringes" appear, probably remains only the change of the camera or its glass plate cover in front of the CCD chip.

We now have discussed the "disturbing components" that works in the raw frames. The next page describes the general procedure, like the recording of a series should be reduced. Only when these principles are understanded we strive to meet the reduction software.

next page

back to content