Preprocessing: Calibration

Calibration is somewhat different if you are doing one shot color (OSC), RGB, or LRGB imaging. In all three cases you are combining at a minimum light and dark frames. You may also incorporate bias, flat, and flatdark frames. Calibration is usually done with software designed specifically for that purpose..

Light Frames

Light frames are the exposures of your target. One of the main considerations in planning light frames is exposure time. Let's see some considerations that go into choosing a proper exposure. They include read noise, dark current, sky brightness, and target brightness.

Read Noise is a relatively fixed amount of noise injected into images by the camera. For most modern imaging devices read noise is very small, so we'll skip a discussion of it.

Dark current can be far more substantial. You can think of dark current's signal as a relatively uniform glow that grows in brightness with exposure time, much like the brightness of the sky. (Note that thinking of dark current and sky brightness as uniform is a simplification.) Because of this similarity I'll refer to their sum as the "background." Contrary to how it's sometimes characterized, background brightness is not noise itself. It's actually a poorly sampled signal, and it's the sampling error that is the noise associated with the background.

Because the background is the dominant feature of your image the majority of the pixels will be at its brightness. If sampling were perfect this uniform brightness would be represented in the histogram as a huge spike at one specific brightness value. Sampling error creates some pixels that are darker or brighter than the expected value. This spreads the spike into a broad peak. As exposure time grows the average brightness value of the peak moves toward the white end of the histogram.

The signal from your target is added to the background, meaning that for the most part it will be found on the histrogram between the peak and the white point. The distance between the peak and the white point determines how many intensity steps you'll be able to show your target as having. (In other words, it determines your image's dynamic range.) If the exposure time is so long that it has pushed the mound to the midpoint of the histogram it has cost you about half your possible intensity steps, effectively reducing the bit depth of your camera. This is why it's common to see the advice that the peak be kept near the darker end of the histogram.

You might think that this argues for the shortest possible exposure time in order to shove the peak partially off the dark end of the histogram. There's a reason you shouldn't do this, though. The fainter portions of your target are barely brighter than the background, and when they're sampled they'll disperse into a range similar to that of the peak. Taking exposures so short that most of the mound is clipped at the black point will also clip some of your target's signal. This essentially transforms some the most interesting structure of your target into black pixels!.

The best exposure time is therefore one that lifts the entire background peak above the black point, but no further than that.*

This also applies to imaging with broadband and light pollution filters that allow significant background signal through. Narrowband imagers face a different situation. Dark current is usually fairly low in cooled cameras and the sky is almost never bright at narrowband wavelengths. It takes very long exposures to push the sky brightness peak onto the histogram. It's almost impossible to have exposures long enough to move the background peak too far whiteward.

For narrowband imaging the rule is almost always that the longer the exposure, the better.*

*Nothing is ever this simple, though. Another consideration for exposures is the brightness of your target. Usually it's not good practice to use exposures so long that the brighter areas of your target saturate. White clipping also discards information you want; the way around it is to use multiple exposure times, short ones for the bright areas and longer ones for the faint portions.

Dark Frames

These are designed to characterize the portion of the background due to read noise and dark current. Dark current is dependent on temperature, so dark frames must be taken at the same temperature as the light frames, and they must have the same exposure time.

You cannot use dark frames binned differently from your light frames for calibrating the light frames.

Bias Frames

Bias frames are designed to capture read noise, and are basically dark frames made with exposures so short that no appreciable dark current contamination can occur. Theoretically, if you use dark frames you don't need to use bias frames, but I find that their inclusion can improve calibration a bit. Bias frames are essential if the dark frames were taken at a temperature different from the light frames and your software can compensate for the difference.

Bias frames should be taken at the same temperature as the dark frames.

Flat Frames

Flat frames are perhaps the most mysterious of calibration frames, but in reality they're quite simple when you consider them from the standpoint of dynamic range. Flat frames remove a number of signals from light frames: Vignetting, dust shadows, variations in sensor sensitivity, and variations in sky brightness. The first three signals are internal to the imaging system and can be dealt with using what I'll call an "internal flat." You can create internal flats by imaging any surface of uniform brightness. A wall, a uniformly lit T-shirt covering the objective, or an area of the daytime/twilight sky far from the Sun and horizons can serve as that surface. The exposure time for internal flats is simply one that puts the surface's brightness peak near the center of the histogram. At the center you give both the dark and light wings of the peak the greatest possible dynamic range.

Flats to compensate for variations in sky brightness ("sky flats") must be made by imaging very near the same area of the sky that your target is in, at very much the same time of night, with the same exposure time and temperature. This is not a simple thing to do, which is why there are so many software packages and methods for creating synthetic flats.

Flats for filtered imaging must be created for each filter to deal with variations in dust contamination. If your optics are clean you may be able to make do with one set of internal flats for all your filters.

flat frames must have the same focus setting and camera orientation as the light frames. It's generally a good idea to keep flat frame exposures over a second in length to reduce any uneveness due to shutter motion. Because of different filter densities and the possibly non-white nature of your flat illumination, you will probably have different exposure times for each filter.

One common question is if the color of the light imaged while making internal flats is important. It's not, so long as you use the correct exposure time. Flats are converted to gray scales during calibration, so you could use a blue light to expose red flats if you wanted, although it would make for unecessarily long exposures.

Flat Dark Frames

The exposure time for flat frames can sometimes be long enough for dark current to manifest itself, which is almost always the rule when using narrowband filters. In this case flat dark frames are used to reduce dark current effects.

Flat Darks should be created with the same temperature and exposure time as the corresponding flat frames.

A Note on Stacking Calibration Frames

The calibration software I use (ImagesPlus) uses average stacking for calibration frames. I instead use median or sigma clipping stacking. This is done because I usually shoot a large number of dark, bias, flat and flatdark frames and signal losw due to stacking method isn't significant. Calibration frames can have cosmic ray tracks in them, and these can be better rejected with an outlier rejection method.

Table of contents Previous Page Next Page Last Page