Contents


Imaging Strategies

CCDAutoPilot is designed to make your imaging data collection easier and more repeatable.  In addition to "pushing buttons" for you, it also has a number of tools, wizards, features and options that can make the quality of your acquired data better and the entire acquisition work flow more repeatable.  In fact, once you have an optimal approach worked out, you can simply determine your target and go.

To use CCDAutoPilot to maximum advantage, it is best to review some underlying concepts surrounding data acquisition.  The next few sections are intended to give you some background on these concepts.

There is a significant amount of technology in image acquisition with CCD cameras but with an understanding of some of the underlying concepts, best use can be made of the available time.  This section will hopefully provide some "getting started" concepts.  As your skills and understanding develop, you may delve more into the underlying details to further optimize your data acquisition.

One of the first questions asked is how long should my sub-exposures be?  This simple question can take an "urban legend" kind of answer - as long as possible, go deep, "10 minutes is what I use", etc.  But there is some underlying science that can be applied to make sense of all this and determine what is best for your telescope, camera and sky conditions

Imaging is all about Signal-to-Noise-Ratio (SNR).  Noise is that grainy background that we see in the faint areas of the target after our exposure is complete.  There are multiple sources of noise - the camera electronics and the camera sensor (read noise) and even noise from the sky itself, in the form of sky glow.  The goal is to minimize the impact of the things we can control.  How we deal with these issues is a function of whether we are doing narrow band imagine (H alpha is one example, along with Sulfur II and Oxygen III) and broad band (Luminance or clear filter along with RGB typically) is another.

Broad Band imaging
With broadband imaging, the sky glow forms an appreciable illumination component.  Like most light sources, there is a noise component, technically called Poisson arrival statistics, that contributes uncertainty (noise) to the value (signal) of the sky glow.  Since sky glow is a uniform component of the signal, it can be effectively subtracted from the data but the noise can not be subtracted. Noise sources combine like the Pythagorean theorem - the square root of the sum of the squares.  So, given the noise in sky glow, we make our sub-exposures long enough so that the sky glow noise is the major noise component and overwhelms read noise.  A typical strategy is to make the sub-exposure time long enough so that read noise contributes 5% of the total noise.  The sub-exposure calculator provides a convenient way to measure camera gain, read noise and sky glow for popular camera sensors so that you can arrive at a suitable minimum sub-exposure time.  It should be noted that the luminance component carries the resolution and color information is generally blurred so the noise in the color channels is not as important as the read noise.

Once we have a sub-exposure time determined, we can look at dark current.  Every sensor has dark current, which is another signal that can be subtracted from the data but again it has a noise component that can not be subtracted.  Here, cooling the sensor reduces the dark current typically by one-half for every 6 degrees C.  We can use a similar strategy to determine how much we want to allow the dark current to contribute to the total noise.  Since dark frames are easy to come by on cloudy nights, we can set experiment with the sub-exposure calculator to determine camera operating temperature and number of dark frames we need.  In many cases, a surprising camera operating temperature for a given dark noise contribution results.  For example, with a KAF16803 sensor and my suburban skies, a 600 sec. sub-exposure is sufficient so that read noise contributes 2.5% to the total noise and to have the dark current contribute 0.5% to the total, I need only 6 dark frames and can run the sensor at -15 degrees C.  Again, once your gain, read noise and sky glow is known, you can experiment with sub-exposure calculator settings to see the effect of cooler temperatures on the dark noise contribution and number of dark frames needed.

(It should be mentioned that there is another quasi-noise source called pattern noise.  This is pixel-to-pixel differences in dark current for each sensor.  Technically, this also represents noise but dithering and subsequent registration and statistical rejection combining minimizes this contribution for aesthetic imaging.  The order of magnitude of this pattern noise is approximately equal to the dark signal.  For precision photometry, the temperature should be reduced approximately another 7 degrees C from that calculated above to reduce the pattern noise.  The number of dark frames calculated above should also be doubled.)

Narrow Band Imaging
With Ha, OII, SIII and similar narrow band filters, the sky glow is essentially negligible.  Here, the noise sources are primarily read noise and secondarily dark signal noise.  Read noise is relatively insensitive to temperature.  Since exposure times are generally long, the number of dark frames should be calculated after inputting your exposure time and camera operating temperature.  If your imaging scale is sufficient, binning can reduce the effective read noise.  For example, binning 2x2 means 4 pixels are read with one read cycle.  If your read noise is 10e for example, then that will be applied to 4 pixels and your effective SNR will be increased by approximately 4. This may not happen in all cases with all sensors and cameras. To be sure, take advantage of CCDAutoPilot's Sub-Exposure calculator tool to actually measure read noise for unbinned and binned conditions.

The above discussion barely scratches the surface of SNR.  Interested readers are referred to available texts on this issue.

Subsequent sections in this topic will explore key contributors to overall data quality.