CCD Observing Manual

3.0 CCD Camera Skills

3.1 Calibration Overview

A raw CCD image contains a ton of data. Some of it is good (signal) while some of it you don't want (noise). It is the Zen-like balance between signal and noise that will determine the accuracy of your measurements.

Signal

The signal in your system is the starlight itself and nothing else. Those photons that left the star tens to billions of years ago land on your CCD chip and are converted into electrons. These electrons are counted and then displayed on your screen and stored on your computer in binary form. The trip from CCD to monitor is a dirty one.

Unwanted Signal

The unwanted signal in the system mostly comes from the sky (cosmic rays), the telescope (vignetting, reflections), the CCD chip (dust donuts, hot pixels, dark current), the CCD charge-detection node (read-out noise), and the computer (timing).

Calibration to the Rescue

Calibration is the combat itself between signal and unwanted signal (some people call this noise but by its strictest definition it isn't - you could also call it bias or unwanted effects). Through calibration you can mitigate the effects of each of those sources of unwanted signal. Sometimes you can merely diminish it (read-out noise) and sometimes you can completely eliminate it (cosmic rays). Calibration can be a very tedious and time consuming process. It is also one of the areas of CCD observing where experience helps out the most. In the beginning you may be a poor calibrator. No problem! As time goes on you'll get better and find the best calibration system for you.

Do the Calibration Shuffle

There are three important calibration routines that all CCD observers should be familiar with. Bias Frames help compensate for read-out noise and interference from the computer. Dark frames compensate for the thermal properties of the CCD chip and Flat frames compensate for noise in the light path. All three are combined to give you the best calibration possible.

	Frame	Data	RX And value
Bias			10.0 +/- .4 Vmag
Dark			10.2 +/- .1 Vmag
Flat			10.25 +/- .05 Vmag

The above illustration is a completely arbitary example of what could happen during a calibration procedure for an image of RX And. Notice how the magnitude estimate becomes more precise and the error drops after each step in the procedure. Error estimation is actually far more complicated and discusssed in another section. But this gives you a general idea of what the various frames look like and their impact on the final image and data reduction.

Master Flat = (Raw - Dark) / Flat

3.2 Calibration Trifecta: Bias Frames

The first step in calibration is to prepare a bias frame. A bias frame is basically an image taken with the shutter disabled. The image will consist only of read-out noise and noise caused by interference of the computer. What a bias frame does is set the zero point of the CCD output and the pixel scales to the same value. This makes the final image more accurate since the zero points are equal and no nonlinear pixel values exist.

Bias frames will have very little signal in a modern quick-read-out healthy CCD camera (larger and older cameras will have more signal). They will have virtually no effect on the visual description of the image (astrophotographers rarely bother with bias frames). In fact, you may find that the bias value changes by less than one ADU per pixel (your software will be able to determine this). In that case you should take the average value of all the pixels in the frame and apply that average to every pixel. What this does is correct for statistical "accidents".

A bias frame will also take into account thermal current that collects while the frame is being downloaded to your computer.

Note: If one uses darks that are the same duration as the light frames, bias calibration is unnecessary because it is included in the dark data. Bias frames are only required If one intends to scale their darks.

Taking Bias Frames

Not all CCDs can take pure bias frames. Some CCDs just are not designed to take an exposure without using the shutter. In those cases you can simply skip the bias frame and go straight to the dark frame (which will include some of the same data as the bias frames) or you can make a psuedo-bias by taking an exposure with the shutter at the shortest possible exposure length and your system completely blocked of any light.

If your camera can take a bias frame, it is likely already setup in your software. A bias frame is a 0 second exposure. After you take it, inspect it for artifacts. In theory, the field should be uniform. But in reality you'll notice some changes in amplitude. If you notice any patterns, they were likely caused by a spurrious noise source (computer CPU chip, home electrical surge, etc.). Take a number of bias frames and determine whether you are operating in a "noisy" environment or not.

The more bias frames you take, the better the result. The readout noise in the bias frame will decrease by the square root of the number of frames you take. In very sensitive cameras you may want to take as many as 50 frames! A good rule of thumb is to take as many bias frames as you take darks.

If you found that you are working in a noisy environment then you should median combine your bias frames, otherwise average your frames together to make a master bias.

3.3 Calibration Trifecta: Dark Frames

A dark frame measures the thermal readout of your CCD, a.k.a. its temperature. A dark frame is an exposure where the shutter is opened but no light is allowed to hit it so it only measures the energy from the CCD itself (dark current). This is normally done by placing a dust cap on the telescope and then covering it with a blanket, cloth, or something opaque to light. Darks also compensate for hot pixels, which are defects in the CCD chip that makes pixels look like they are permanently "on" or "lit". Darks are very easy to take and are the most important calibration step so there is no reason not to take darks.

Taking Darks

The Handbook of Astronomical Image Processing recommends the "Image-Times-Five" rule. The more dark frames you take, the more accurate the frame and the lower the noise. For a sample of 100 electrons, the uncertainty is 10%, for 10,000 it is 1%. A good rule of thumb is to make sure the total exposure time of all your dark frames equals five times that of the image you are calibrating. So if you are taking a 2 minute image, you can do five 2-minute darks or ten 1-minute darks.

This is where your bias frame comes in. The dark frame also contains readout noise. Readout noise does not scale over time, so your dark frame right now is unscaleable. That is a pain if you plan to take exposures of different lengths during your observing session because then you must take a ton of dark frames to match the integration time (times five!) of each image. If you could yank the readout noise then the dark frame only consists of thermal noise, which is scaleable. So let's do it! All you need to do is subtract the master bias frame from each dark frame you took. It's that simple.

10-minute amplifier glow (electroluminescence). Older CCDs suffer from this effect which can be removed with a dark frame.

As the CCD temperature changes during the night, the dark current will adjust. Except in extreme circumstances, you do not need to take darks for each image. A good rule of thumb is to take your dark frames in the middle of the observing session. Another plan is to take some of the darks at the beginning, some in the middle, and some at the end.

For each image remember to subtract the bias. Inspect each frame to make sure that cosmic ray events do not contaminate them. They will appear as a bright spot on your dark frame. It depends on altitude, physical size of the CCD chip, amd exposure time but in general expect about 1 cosmic ray event every few minutes of exposure time. When you have a bunch of darks, average them together to create a master dark. (You can skip the cosmic ray inspection by median combining at least 3 frames instead of averaging them; but then your final dark will have slightly more noise.)

Now you have 2/3 of the calibration process completed. Some observers will quit there. If you have carefully made your master dark frames and done everything else correctly it is possible to get to 0.03 mag accuracy in your photometry.¹ For most AAVSO observing projects this is all the accuracy you need.

For those who want to really push their accuracy, here is a tip for how to take low noise dark frames from Ron Zissell.

3.4 Calibration Trifecta: Flat Frames

For the absolute highest quality photometry flat frames must be used to calibrate your image. A flat frame compensates for obstructions, reflections, and other problems in the light path. This is the path light travels from the time it enters the telescope to the moment it strikes the CCD chip. Dust on optical surfaces, reflections from baffles or poorly aligned optics, vignetting, and other noise sources can interfere with your final data.

Flat fielding is the most difficult calibration routine you will have to deal with. There are many things to watch out for and many ways to do it. In fact, some people call it an art because it is so intricate and there are so many creative ways you can do it. The key is to be patient and realize that your first few flats will likely not turn out well. Take your time and with experience you'll be able to master the master flat!

The Concept A flat field is a picture of what's wrong with your system in regards to light. (This is an oversimplification but bear with us here..) You will see dust donuts, light gradients, reflections, and more (right). Now that you have an image of what is bad, you take an image of the star. That image is an image with good (star) and bad (noise). You remove the flat field from the image and you are left with only the good. Trim the fat, so to speak.

The free passage of light through the small aperture can be seriously impaired by dirt and dust. (about the gremlins)

Taking Flats

The first thing you have to do is take flat darks. These are dark frames taken to be applied to the flats. So you want to match the integration time with that of your flat, not that of your final image. These darks will be separate from your image calibration darks. Other than that, the procedure is exactly the same as above.

The goal is to take an image of a uniform light source (the "flat" field). So the first thing you need is that uniform light source. This is the most difficult part of taking flats. The good news is that once you find a uniform light source that works you can use it forever. There is no fool-proof method of creating the uniform field. How you do it will likely depend on your physical location, mechanical ability (handyman factor) and your level of patience! Here are some of the most popular light sources for flat fields:

• Domes: Shining a light on the underside of the roof of a closed dome.
• Twilight: For about 15 minutes at dusk and dawn a photometric sky can be a uniform blue.
• Lightboxes: These are boxes with a light bulb inside that fit on the outside of the telescope's central opening.
• T-Shirts!: All sorts of creative ideas have been used to create uniform fields from white cotten t-shirts to film screens to the sides of houses. The limit is your imagination.

Usually the light source will be a dimmable bulb so that the light can be adjusted for the filter you are shooting through. Different filters transmit different amounts of white light so what may be a good source for a V filter may be a weak source for an R filter. If you don't have a dimmable bulb you can adjust the image exposure time, but then you need to take a new set of flats. (This can get complicated pretty quickly!) The light source should be reflected off uniform surfaces before entering some type of diffusing screen. For example, you may want to shine a light on a white poster board which reflects it to a piece of translucent plastic before going into the telescope. The more times you reflect the light, the more uniform the field will be.

Once you have a uniform field, expose your CCD to about 1/2 of the full well depth of your pixels. Take at least 16 flat field images, this is the lowest number required in order to avoid adding noise to your final calibrated image. For .01 mag accuracy photometry keep your signal to noise ratio to 500:1 or better. The exposure time will differ based on what filter you are using since each will pass a different fraction of the light source.

Every time you change an element in the light path, such as removing a filter, you change the light path so you have to take new flats. So you can't take a flat with a V filter, then take it off and put on an R. When you do that the V filter flats should be discarded. (If you are using a high quality filter wheel then you can do all the flats at once since their orientation in the light path will be the same when the filter wheel changes it.)

Applying Flats

1. Average all your flats
2. Average or median combine all your darks made specifically for the flats
3. Subtract the averaged dark from the averaged flat

What you have left is your master flat. Congratulations! This flat will be good as long as you don't change anything in the optical configuration of your system.

Now you can begin taking data (images). Divide the flat into each image after you have dark subtracted it. Most software programs have a way of automatically doing this. Now your images are fully calibrated.

Rejoice!

• Starting out in Photometry Part 2: Flat Frames from CCD Views: Good description of a flat field learning process, setup and routine

3.5 Finding the Field

At first this can be one of the most frustrating parts of the learning curve. It is also an area where visual observing experience really begins to pay off. The same problems you have with finding the field visually exist with CCD. The difference is that your field of view will be a lot smaller.

Recommendations:

• Alignment: Don't skimp on your alignment procedures.
• Chart Masks: Draw a circle on a sheet of plastic indicating the field of view of your CCD and then block out anything outside that field of view. Use this mask on your charts. Make one for d, e, and f-scale charts.
• Use Big Charts: Your field of view may fit within an f-scale chart, but bring d and e-scale charts with you while observing. This way if your mount doesn't put you within the f chart you can find yourself on the e or d charts. If the e and d charts don't go faint enough, print out a copy of the field using the DSS.
• Learn to read the charts

3.6 Image Integration Style

It is perfectly acceptable to coregister and then stack (a.k.a combine, sum, average) images to improve your signal to noise ratio. However, don't perform any nonlinear calculations on the field as this will destroy the photometry. One advantage to doing a median combine is that you can remove cosmic rays that way, but you do take a hit in your signal to noise ratio.

3.7 Tricks of the Trade

Here are some notes and general rules of thumb from some experienced observers:

• Don't expose for less than a few seconds. Anything shorter will cause the shutter to affect photometric data (the speed of the shutter opening and closing will affect the amount of light that falls on pixels depending on where the pixels are located in reference to the rest position of the shutter).
• Plug in your camera first when you begin setting up for the night so that it can reach a temperature equilibrium by the time you are ready to take darks. Most modern consumer cameras need 15-30 minutes to stabilize.
• Outside temperature and humidity can affect dark frames quite a bit. Experiment to determine how many degrees of a temperature change your system will handle before you need to take a new set of darks.
• Keep careful written records of seeing conditions, photometry results, how many darks and flats you took, etc. This will come in handy later when you are reducing data. Remember data reduction may take place days or weeks after you took the image when many vital details may have been forgotten.
• Practice, practice, practice. And don't forget to enjoy yourself.
• Tips on Starting an Observing Program (from CCD Views): Includes details on optimizing your observing session
• Develop a personal convention for naming your files so that you'll know what they are without opening them or consulting your notes. This will reduce your observing log burden and speed-up analysis. (For example, 0072AP160.fit could mean file #007, source 2002AP, bin 1, 60-second exposure. One observer recommends using the naming scheme: yyyy.mm.dd Object (filter, exposure time) (details) .fit Sorting these files by name will result in time dependent scheme AND no file will be lost due to copying different files with identical file names.)

If you have any tips please send them to us!

¹: From Handbook of Astronomical Image Processing by Richard Berry and James Burnell, page 257.