I have recently got back into DSLR photometry imaging through a good quality 200mm camera lens at f/3.5 and tracking with a Star Adventurer Pro. I was struck by the uniformity of the sky background (not perfect, but pretty good) in uncalibrated images. For interest I decided to compare the result with the sky background of images taken with a monochrome CMOS camera and C8 telescope at f/10. The results, including short comments, are in the document which can be (edit) downloaded via the following link.
In your doc. The figure labels are the same in fig. 3 & 4. Not sure what you are looking for in the way of comments?
Bottom line is you should always apply darks and flats in the standard way before doing photometry no matter what optics/sensor you are using.
DSLRs, but esp. mirror-less cameras where the sensor is right there if you are changing lenses are notoriously hard to keep clean. The solution is to take flats every time you change the lens.
Jim,
The figure labels for 3 and 4 are slightly different. I am not looking for comments, I made them. The first is after Fig. 2 and the second is after Fig. 4.
Look at the data in the images and the line profiles and think about them rather than stating the dogma that flats should always be taken. This was the point of the post.
The uncalibrated DSLR image is typical of what I get. The sky background has very little vignetting across the entire diagonal, and I never put variables or comps in the corners of the image. Any slight variation in sky background is taken care of by aperture photometry as the sky annulus is representative of the region of the image where the star is placed.
Accuracy is good. Averaging photometry across 10 images gives an accuracy (measured minus catalog) for the check star of 0.01 V mag. Precision night to night for the check is also very good.
I never suggest deviating from standard practice unless there is evidence in support.
Well my experience with DSLRs is you might, might be able to get away without taking flats if your optical path is clean (no dust) and your sensor is small compared to the full, flat illuminated cone of light falling on the sensor. Unless you are very lucky vignetting is always there, even if at a small level.
I would still always apply the standard dark, flat processing. It doesn’t take that much time or effort and it is the standard way that astronomers have used for many decades for 2D sensors!
I understand what you are saying, and fully agree that best practice is to take darks, flats and (when necessary) bias frames.
When there is obvious vignetting and obvious dust donuts are present, clearly a full calibration routine is needed. That is the case with my mono CMOS camera imaging through a SCT at f/10. Then I always to take flats.
When there is hardly any vignetting across the entire diagonal and there are no dust donuts, then I believe it is valid (and interesting) to look at the evidence for and against the necessity for flat frames. When I get a good night (not often here at present) I’ll image a standard field with the DSLR setup and process with (1) darks and flats and (2) darks alone and look at the accuracy of target/comp pairs with and without flats.
By way of explanation, when I have done DSLR photometry in the past the camera has been mounted in a fixed position on a tripod. Recently, I have changed to mounting the camera on a star tracker. The images are of course much better and there is more control. I have also been trialling minimizing the extent of defocussing, which has the benefit of reducing problems with blended star images.
Unfortunately, this type of discussion thread will confuse and perhaps some will think that standard practice of taking darks and flats and applying them isn’t what they need to do. Astronomical photometry with 2D sensors like CCDs and CMOS has been around a long, long time. The standard flat-fielding methods are there for a reason. To imply otherwise, from a few special cases that you think have “hardly any” vignetting or dust motes is incorrect. To imply anyone doesn’t need to do that pre-processing of removing darks, flats and bias before doing photometry isn’t required is not correct thinking.
I don’t see how the above could possibly confuse anyone. To me, it is very clear.
All I have been doing is to explore the data behind the reason for the use of flats. When I find that my small aperture/short focal length DSLR system as it is currently configured produces the best results when I use both darks and flats, then I will always use them.
The check star photometric data so far using darks only (no flats) for several nights since late last year is seen below.
I would argue for taking flats, bias, darks pretty much nightly, and standard-star calibration fields whenever you think it is genuinely cloud-free. Firstly, don’t imagine that whatever calibration process you are applying is the best that can be done, cuz you (or others) will be able to do it better in the future. There are examples in the literature where folks skimped on the flats (say) and later analysis showed the results were misleading.
A lot of observers work strictly differentially for whatever immediate task and claim they don’t need and are not going to take standard stars or do transformations. Well, OK, but that means those data have only one use, whereas the data can have far greater utility in other contexts, for which having full sets of calibration will be useful. No, you don’t need to do the reductions “right now”, but just hitting a standard field each night as part of a run will come in handy downstream.
I understand the reluctance to accept the proposal that I do not need flats. Let me expand on precisely what I do and the evidence for what I’m saying.
My Canon EOS 500D has an APS-C sensor - less than full frame. The lens is (edit) a Canon EF 200mm f/2.8 L (no longer available as new), set to f/3.5. EF lenses are designed for full frame cameras.
When I plan my field in a planetarium programme I use two concentric circles to frame an image. The diameter of the inner circle is the height of the image and the (edit) diameter of the outer circle is the width. I always place target and comp stars entirely within the inner circle.
The screenshot below shows a line profile (the yellow line) of the background sky and a star across the height of the image (sentence has been edited). The line profile itself is on the plot to the left. You can see the spike produced by the light from the star. The wiggly line near the bottom of the plot is the sky background.
I submit that the background, along the yellow line, is flat. The same result is obtained with a horizontal profile across the centre of the image.
The line profiles in my original post were from corner to corner, and I noted that I never place stars to be measured in the corners of the image.
For your one high S/N target you are correct that it doesn’t matter. But as you measure fainter objects on the frame it will become more important. Per my previous post, there’s lots of information in the images for which having calibration will greatly improve their utility.
A suggestion might be to re-scale your plot closer to the min/max of the background (roughly 0 to 800 on the vertical scale you show) to look at the gradient.
Another graphical-display point is also to always show images with bright sky and black stars (like a photographic negative). This will let you see background features more clearly — almost all the information in the typical astro image is just above the sky background.
Thanks for your comments, which always provide pertinent additional perspectives.
Not sure if the the attached captures what you suggest, but when the vertical scale is 0 - 500 (by having the line profile location pass through a fainter star) there is a definite but very small gradient top to bottom across the image. The second image shows the centre of the field zoomed in. (Editx3) The image was converted from Canon CR2 RAW to .fit in ASTAP, the green channel extracted (becomes .fits) then dark subtracted. The display is in AstroimageJ.
I’m most interested in what you said about callibration and photometry on fainter stars, and will have a closer look at that when I can.
If you want to find out how good reduction method A is, compared to reduction method B, then you can perform this test if you have a reasonable number of images taken on different nights; say, 10 images each on 5 different nights.
reduce using method A. Measure the magnitudes of a set of 5-10 stars in every image. Calculate the mean, stdev and other statistics of the stars over the entire set of nights.
reduce using method B. Measure the magnitudes of the same set of 5-10 stars in every image. Calculate the statistics of the stars over the entire set of nights.
compare the results from the two methods.
If the results are the same, as far as one can tell, then neither method is better than the other. If one method produces results which are better for your purposes, then use that method.
Michael,
I agree completely with that type of comparison.
I haven’t got that far yet, but trial photometry has been carried out on standard stars from the E Regions of 8th Mag in V at 04H RA using callibration by darks only with no flats.
Setup: Canon EOS 500D DSLR camera, Canon EF 200mm f/2.8 L lens at f/3.5, ISO 400 (gain is 1 e-/ADU), exposure 60 seconds, star images slightly defocussed. Mount: Star Adventurer Pro (equatorial tracking).
Twelve standard stars in one field yielded 6 unique target/comparison pairs. An additional 6 pairs were used for the calculations (again, all 12 stars were studied, but the stars in each pair were different in this set).
The results are below as is the list of stars. The labels are simply those assigned by AstroimageJ.
All results in the first table are measured or calculated, not catalogue values. (Edit) The magnitudes and colour indices are transformed. The catalogue data is in the second table. The column headings are probably self explanatory. (Edit) SD is standard deviation. The two columns highlighted in yellow show the actual error, defined as measured magnitude or colour index minus the corresponding catalogue value.
If adding flats to the mix improved on these results, I suspect that any incremental improvement would not be worth the effort!
I should have explained that each magnitude and colour index in the results table in the previous post is an average of measurements from 10 consecutive images.
If you are removing darks how hard is it to take and apply a flat?
I haven’t taken sky flats in many years. I use a diffuser screen and can take my flats under either daytime illumination or broad-spectrum LED lighting. Easy to do, takes only a few minutes. I don’t take flats every night but update them fairly often, esp. for the DSLRs if a lens change has happened between observing runs.
Most of the time in this process is taking the darks after observing each night.
Standard image processing says apply a dark and a flat… and a bias if you need it.
Is there a specific reason you don’t want to take and apply flats?
a) take a series of images of a field with a bunch of stars, so that you can perform differential photometry
b) now slew the telescope by about 1/3 to 1/2 the field of view, so that stars move to very different regions of the chip.
c) take another series of images
Compare the photometry of stars which appear in both sets of images. If you see significant differences in the magnitudes as stars move from one side of the field to the other, then your technique needs some improvement.
Is there a specific reason you don’t want to take and apply flats?
If flats will not improve my photometry, yes there is a specific reason. Despite your statement that flats take very little time, with my setup they do, because I use an artificial light source and mostly take them during the day. Each set of flats is specific for focus, so that means most nights. I don’t have an observatory, and move my small aperture/short focal length setup outside and inside each observing night. If I take flats the next day, that means reconnecting everything. You noted darks take longer than flats. Not if I have to setup again inside, and in any case the darks run by themselves - I just go and do something else. Sure, once I have my light source on the lens and the exposure correct, the flats exposures take a few seconds, but everything else is very much hands on.
The point of this thread is to give an example of thinking about procedures and looking objectively at the evidence for doing something rather than just blindly following a set protocol, no matter what.
The evidence I have found so far clearly shows that for photometry on bright stars with the particular DSLR camera, lens, f number and target/comp distribution across the centre of my field for bright stars (down to mag 8 in V so far), the sky background is flat across the part of the field containing stars to be measured, and transformed photometry in V with darks only often gives accuracy (measured minus catalogue) of 0.01 mag or less.
This thread is not a general recommendation to dispense with flats.
I do a lot of testing, but no, I haven’t tried moving the field. I probably won’t at this stage. Some of the target/comp pairs in the tests just completed were specifically chosen to be “wide apart” in the field of view, and others were quite close. Made no difference as far as I could see.
“This thread is not a general recommendation to dispense with flats.”
Roy, Now I like that statement!
With DSLRs and wide-angle lenses I don’t guide at all and I get good signal for stars brighter than about 7th. I’m using a Pentax K-3II + Rokinon 14mm @ f/4 + intervelometer all on a standard camera tripod on my deck. 20s exposures and I stack 9 images. The earth rotation moves the field. I have a slight defocus but the earth-rotation moves the stars about 14 or 15 pixels over the 9 x 20s exposures. Why? It is an easy and good way to improve the sampling of the Bayer matrix without a lot of defocus.
