Measure Read Noise in Your CCD Camera
Interpret FFTs l
Noise FFTs l
Your Camera l
Create FFTs l
Measuring the Read Noise in Your CCD Camera
Measuring the read noise in a CCD camera is a relatively straightforward task. All you need are a few good bias frames and some readily available software tools. The results of measuring the read noise in your existing CCD camera can me quite illuminating.
Obtaining a Good Bias Image
The average bias per pixel will change somewhat with CCD temperature so start by letting your camera stabilize at a given temperature for about 10 minutes. Once the temperature has stabilized, take a series of 9 bias frames. The instructions below show each step using MaxIm DL or LE. If you’re using different camera control or image processing software, you’ll need to perform similar actions using the capabilities of that program.
Setup an automatic sequence of images using the “Sequence” tab in the CCD Control Window.
Enter a descriptive name under “Autosave Filename” so you’ll remember what the images are. Enter a “1” in the “Start at” field to begin numbering your sequence at “001”. Click the arrow to the right of “Options” and select “Setup Sequence…”.
Setup a sequence of bias frames as shown above. Once the sequence is set up, click OK to return to the CCD Control Window. Press “Start” to begin the sequence. MaxIm will expose 9 bias frames and append a counter and the Suffix specified above so you can distinguish each image (e.g. “QSI 504 -20C 001Bias.fit”).
Examine Each Bias Frame
The human eye is wonderfully sensitive to patterns and discrepancies between images. Visually inspect each bias frame using your camera control software or other image processing software. Bias frames are very dark. Pixel values are typically significantly less than 1% of the total dynamic range of the camera. When visually examining a bias frame it is important to stretch the black and white levels in the histogram so that you can see the difference between the lightest and darkest pixels. In MaxIm this is done with the Screen Stretch window.
Carefully examine each bias frame. Look for obvious visual anomalies in each bias frame. Reject any bias frames that have obvious problems making them appear significantly different from the rest. One specific anomaly to watch out for is cosmic ray hits.
||Cosmic rays are high energy particles that continually bombard the Earth. Some cosmic rays are generated by the sun, while others originate from far outside of our solar system. The higher above sea level you are the more likely you are to see the effects of cosmic ray hits in your images. Cosmic ray hits on a CCD can appear as a bright cluster of pixels at a random location or as line of bright pixels at any angle depending on the path of the cosmic ray.
If one or two bias frames appear dramatically different from the rest, reject them. If a pattern is reasonably consistent, such as brighter pixels in the corner of all images, or a subtle pattern in all the bias frames, they’re fine. Below is an example a “good” bias frame next to an image showing a small region of bright pixels and a cosmic ray hit.
Good Bias Frame Rejected Bias frame
It’s best to have at least 5 good bias frames for the next step. If you have less than 5, you may want to take additional bias frames before proceeding.
Create an “Average Combine” of the Bias Frames
Now you need to combine the good bias frames using an “Average Combine” where each pixel in the combined image is the pixel-by-pixel average of that pixel in the original images.
If your bias frames show a high number of cosmic ray hits or other anomalies, you may need to do a “Median Combine” rather than an “Average Combine”. Rather than averaging the value for each pixel, a “Median Combine” will select the median value from the original images for each pixel in the combined image. This eliminates the effects of random cosmic ray hits by rejecting any pixel values that fall outside the norm. Given a choice, an Average Combine is better for this purpose, as it produces an image closer to the true bias of the camera, but a Median Combine will do in a pinch.
In MaxIm, open all the bias frames and select “Combine…” from the Process menu
Select the images you wish to combine and click OK.
That brings up the Combine Images window.
Select Align Mode “None,” Output “Average” and FITS Format “IEEE float.” Click OK. This will create a new file that has pixel values equal to the average of the original bias frames. Save that file with a name you’ll remember such as “Bias avg9 -20C SN503028.fit”
Subtract the Average Bias to Create a Read Noise Frame
Select one of your original bias frames and close all the other bias frame windows except the average bias frame you created above. As a reminder, subtracting the average bias from a specific bias frame isolates the read noise present in the imaging system. Also recall that subtracting two images increases the random noise in the resulting image. When subtracting two individual bias frames, the noise will increase by a factor of the square root of 2. (≈1.4). Subtracting the average of multiple bias frames will result in a smaller increase in the overall noise.
Bias frames have substantially the same pixel values throughout the image. If you subtract two similar images, the resulting image will likely have some negative pixel values, which doesn’t make any sense. To get around this problem you first need to add a constant to every pixel in the original bias frame before subtracting the average bias. In MaxIm, these two operations can be done in a single step.
Make sure you have one original bias frame and the average bias frame open. Click on the original bias frame. Select “Pixel Math…” from the Process menu.
For “Image A” select the original bias frame. For “Operation” click on “Subtract”. For “Image B” select the average bias frame. Make sure both “Scale Factor %” fields are at “100.” In the “Add Constant” field in the bottom left corner enter “1000”. You can actually enter any value for the constant that will ensure that no pixel values go negative in the subtracted image. 1000 is just convenient. Click OK.
This will modify the image in the original bias frame window. Immediately select “Save As…” from the File menu to save the subtracted image without modifying the original bias frame. Give the subtracted file a name like this “Bias1 -20C SN503028 +1000-avg.fit”, so you’ll remember what it is.
Crop the Read Noise Frame
FFTs work best and most predictably if the image is first cropped to a square with sides equal to a power of two (e.g. 256, 512, 1024, 2048, etc.). All the examples on these pages were created with source files cropped to 512x512. The resulting images were then cropped to 256x256 so two images could be shown side-by-side on a standard page. For FFTs of your own bias frames, crop the images to the largest power of two that will fit within the dimensions of your CCD. For instance, if your image files are 768x512, crop the images to 512x512. For 1536x1024 or 2184x1472, crop the files to 1024x1024. If your camera produces images where the shortest dimension of the image file is greater than 2048, crop the files to 2048x2048.
Select “Crop…” from the Edit menu.
You can select the square for your FFTs from anywhere within the subtracted bias frame, but in general you should choose a square from the middle of the image since that is where your subject is likely to be. In this example, we’ve selected to crop the image to a 512x512 square from the middle of a 768x512 original image.
Enter a power of two for the width and height of the square (256, 512, 1024, etc.), then enter the X Offset and Y Offset to position the square in the center of the image. Click “Update” to see the selection square on top of your original image. When you’re happy with the position, click OK to crop the image.
Immediately select “Save As…” from the File menu to save the read noise frame without modifying the full-sized image. Give the cropped file a name like this: “Bias1 -20C SN503028 +1000-avg 512x512.fit”, so you’ll remember what it is.
It is this Read Noise Frame that we’ll examine in more detail in the next section.
Continue to Creating FFTs >>