Imaging lessons learned: aquiring flats and dithering

It has been over 2 years since I started using the Atik 11000-CM, and I can say it has been a steep learning curve. When you consider there are plenty of tutorials out there, why bother writing this at all? Well, for a start, very few tutorials available across the nets focus on the real basics of image aquisition and are more centered on image processing. The thing with astroimaging is that if you put garbage in you get garbage out… Knowing this, I felt some people may get some valuable information from my collection of mistakes (largely due to my abysmal ignorance) to improve their own images. While this posts is focused on CCD images acquired with a colour Atik CCD camera, I see no reason why it might not be applicable to monochrome CCD as well.

So lets start with an image of  NGC 3945 I acquired not so long ago, and see what is wrong with it and how it could have been better, processing aside (please note that you require a good monitor to view the example images):


At first glance, suffice to say this image is not very esthetic… Ok, someone very new to astroimaging might get excited about it, as it clearly depicts the subject, but, when you compare it with pictures of the same object taken by more accomplished astroimagers, the difference is great. So what’s wrong with it? Guiding is pretty good, resolution is average for a deep-sky image, and anyway, there is not much to be done about seeing on a given night. What jumps out is that the background is very noisy. Also, there is an ugly vertical column to the bottom left of the image…

Noisy background.

let’s start by looking at a magnified section of the image above:


the noise is pretty obvious… The main question is “where does it comes from?”.  The picture above results from median stacking of 21 sub-images, each calibrated with 15 flats, 21 offsets and 21 darks (if you are not familiar with image calibration, google it up).  Before  you jump to the conclusion that I did not use enough flats, I did try to use as many flats as subs one some other images, with the same result.  What tends to happen, is that noise pops up whever I try to reach low during processing, trying to reveal extremelly faint details. On the subject of flats, they have to be themselves calibrated by substracting the offset. I did hear that some astroimaging software had some bugs which caused offsets to be substracted twice from the flats, which did cause them to be noisy. However, I did check that MaximDL, which I use for calibration as well as processing, does perform calibration properly.

So, I stretch my head as to where that noise might be coming from… Darks frames? Well, darks are pretty simple to acquire, I mean, you make sure the CCD temperature is right, that you have no light leaks, and away you go. Bias (offset) frames? Same thing as darks, If I can mess these up, I better change hobby. This leaves flat field frames, and I started thinking. I do not have a specific flat screen, as it’s very expensive for a  C14 sized telescope, so I use the evening sky for my flats. It’s extremelly tricky. If you try do do flats while the sky is still too bright, then, some columns of pixel are much brighter that others, even though the image is not saturated due to a very short exposure length. It is something to do with the fact that some pixels on the CCD chip have an allowable maximum light level exposure, regarless of how short the exposure time might be. So, I do wait for the sky to be fairly dim (at dusk or down) and I usually set around half a second exposure which usually leads to an average of 4000 ADU (ADU:Analog Digit Unit), on 65578 ADU,full scale. I did read somewhere that the proper level for a flat frame should be about 30% of total dynamics. Hmmm… So, I tried to increase my exposure duration for flat field to reach about 20000 ADU intensity. On the same image, I used some of my usual flats from the day before to calibrate the subs, processed the image, then repeat the processed with the “new” flats. And guess what? the background was now alot nicer noise wise! Even when going extremelly deep within the picture. Problem solved… It just shows that, for astro photography, everything has to be just right for a good result.

Image Dithering.

Again, let’s look  at a magnified section of the image above:


See that very dark column? Well that comes from a column of dead pixels on my class 2 CCD chip. Usually, since all subs have some form of shifts among them, it disapeared upon stacking. The shift is usually caused by minute shifting from an image to the next (that is the reason why subs registration is required) . But, in my case, even though the Titan mount has not got a great PE or tracking speed accuracy, Off Axis guiding does it’s job and keeps everything on their “assigned pixels”. It also keep the dead column in the same spot of the image, as a consequence…

So, to get rid of that problem, the solution is very simple: DITHER! Dithering means, that every 5 subs or more, it depends on total number being aquired, you need to shift the image by a few pixels on your sensor, so that the CCD chip defects disappear upon median stacking, just like a satellite trace on one of the subs would. What I do to achieve this is that, in between subs, I switch off the auto guider for a few seconds, and that is sufficient to shift the image by a few pixels, thanks to PE (periodic error) or as it is the case for me, sideral speed discreptancies. Of course, when you do, like me, go to your bed after having setup everything to acquire images for the rest of the night, it is easier said than done…

Clear Skies


VN:F [1.9.7_1111]
Rating: 8.0/10 (1 vote cast)
VN:F [1.9.7_1111]
Rating: +3 (from 3 votes)
Imaging lessons learned: aquiring flats and dithering, 8.0 out of 10 based on 1 rating

Leave a Reply