When taking a sequence of exposures for stacking/coaddition, what dither patterns are most commonly desired? When visiting a telescope, what default dither patterns would a visiting astronomer like to see in the observing software (assuming custom patterns are also supported)?

  • $\begingroup$ The dither pattern depends on the angular size of the object and on the Field Of View (FOV). If you could elaborate on this the answer would be more to the point. $\endgroup$ – Tigran Khanzadyan Jun 3 '11 at 22:51
  • $\begingroup$ The context is that I am working on a project supplying a camera for the Blanco telescope on which the community will be able to apply for time, and I was party to a conversation about what dither patterns would be most useful for the observing software to support by default. (I am not responsible for deciding or implementing this, but am in a position to make suggestions.) Choice of dither patterns is a common issue faced, so seemed like a good "seed question" for the astronomy SE. $\endgroup$ – EHN Jun 4 '11 at 16:20
  • $\begingroup$ I suppose one lower limit to the dither pattern relative distance would be the expected angular sizes of the objects. Very small dither would make things really difficult for post-processing of the images. This is true in Optical and NIR observations. $\endgroup$ – Tigran Khanzadyan Jun 5 '11 at 10:38
  • $\begingroup$ This comment is somewhat meta, but it seems pretty clear that no one in our community so far has an answer besides the trivial "All of them." Maybe we could expand this into a discussion of the relative merits of different patterns or parameters (like amplitude)? What particular patterns were mentioned as candidates? I've done nearly-research-grade CCD imaging at the Fan Mountain Observatory 1m scope as part of my undergrad and grad-level observational astronomy classes, but even I've never heard of any scheme more elaborate than slewing a random couple of pixels in a random direction. $\endgroup$ – Andrew Jun 17 '11 at 14:25

Dithering is as much an art as a science and depends on many factors including, but not limited to:

  • The type of object being observed (point source, small extended object, large extended object)
  • Telescope parameters (The field of view of the telescope relative to the size of the object, optical quality, size and type of abberations, etc)
  • The quality of the detector (flat fielding, vinetting, bad pixels, linearity, etc)
  • The type of readout electronics (single or multiple channels and where they fall on the chip, gains and biases, etc)

At the very least, you should support an arbitrary two point dither pattern. Basically, you position your object on the detector in a "good" spot and than have the ability to specify an offset in pixels in both x and y where you want to dither to so that your object ends up on another "good" spot on the chip. Ideally, the camera desginer has identified these "good" spots on the detector and provided the necesary positions and offsets so that operators can choose where to dither to. This mode should support both ABAB and ABBA patterns. The first is jumping back and forth between the two positions. The latter takes an image, dithers to the second position, and then dithers back for a final image at the first location. This pattern is best for point sources, or small extended sources that completely fit within the two good regions on the chip.

For the two point pattern, one possible reduction method would be to subtract the adjacent AB images ( A-B and B-A ) to remove sky background and many instramental effects, then all of the A images can be coadded and all the B images can be coadded, finally, the two combined images can be shifted and coadded if needed.

If your targets are larger, or the chip has a lot of cosmetic variations (bad pixels, electronic artifacts, etc) you'll also want to support a four point dither pattern that moves the target around on the chip cyclicly (i.e. ABCDABCD patterns) Typically you would just start at an initial postion, take an image, offset in x, take an image, off set in y, take an image, reverse the x offset, and take the final image. If you were running through the pattern more than once, you'd then reverse the y offset (which takes you back to the original postion) and start over. If there are four "good" spots on the chip that are relatively distant from each other (say one in each quadrant of the chip) the camera designer might implement a specialized four point dither that takes sources to each of these postions in turn and may not just be a simple square offset.

In this case the three other postion images are averaged and then the averaged image is subtracted from the remaining image to remove skybackground. This is repeated for each postion and then the images from the four different postions are shifted and coadded. This pattern helps to guarantee that if there are a lot of bad pixels, you get actual flux measurements from your source at all postions. With only two dither positions you have a chance that there is a bad pixel in the same postion relative to the source in both dither postions (This obviously increases as the number of bad pixels on the detector increases). The chance that there will be a bad pixel in that location in all four dither positions is very, very small.

If the observing at the telescope is interactive, i.e. the observer is looking at the data as it is being taken and can influence the observation patterns, the dither offsets should be variables that can be specified. If it is queue observering where astronomers submit proposals and a telescope operator takes the data and sends the astronomer the data, it is even more important that there be copious documentation of the quality of the chip and it's physcial and cosmetic characteristics so the observer can make informed dithering decisions. In this latter case, it is also a good idea for the camera designer to identify the good parts of the chips and set up automatic two and four point dither patterns that utilize those postions to maximum effect.

Finally there should also be an option to move the target image completely out of the field of view. This is used to get sky background images for large targets that either nearly fill or are larger than the FOV of the camera. However, this is usually accomplished by nodding and chopping rather than dithering.

  • 2
    $\begingroup$ Good, general response; I would just add that, in addition to needing dithering to cover bad pixels, you also need it to cover gaps between detectors when your camera has a array of detectors (eg CCDs) rather than just one. $\endgroup$ – EHN Jun 17 '11 at 23:49
  • $\begingroup$ @Eric N: Good point, I've never actually used a camera that big so it didn't come to mind as it's not something I've ever had to worry about. But dithering to get the object off of the dead area is definitely something you'd have to do if you have a large extended object. $\endgroup$ – dagorym Jun 18 '11 at 0:10
  • $\begingroup$ I've never used a CCD so uneven that you had to steer around potholes- am I just spoiled with good quality CCDs, or lazy/ignorant? I did photometry and didn't notice anything totally wacky, so I hope it's the former! $\endgroup$ – Andrew Jun 18 '11 at 3:08
  • $\begingroup$ @Andrew. It's more common with IR detectors but modern CCD's are much better than they used to be. I've seen some pretty bad chips in my day. Event the HST chips, at least on WFPC2, were pretty ugly. $\endgroup$ – dagorym Jun 18 '11 at 4:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.