Of course, every telescope/detector set up has some vignetting, so when you go to see how the star brightness changes along the transect, an unflatted series shows that vignetting and the star’s brightness falls off as you get farther from the center of the detector. We apply flat field correction to eliminate the vignetting, and if the flat field illumination is even, the star’s brightness should be even across the detector when compared to the artificial star.

Well, when I went to analyze my results, I noticed that the star had strange brightness changes after being flatted, and I wondered what might be the cause. I’m sure that the XL does not exhibit radial changes to its brightness from other testing, So I started hunting around for other causes.

Here’s the master flat field image that I used to flat the star images. Note the strange bright ring around the center so that the brightness doesn’t fall of evenly from the center to the edges. This bright ring would explain why, when my images were flatted with this flat field image, there was an under-correction under the bright ring, which led to the star being dimmer than it should have been at that position.

The master flat field image. Note the ringlike structure.

To give you some background, I heard a talk by Michael Barber at NEAIC last year about the effect of internal reflections on image quality. Essentially, he showed us how an internal surface that is not well coated with anti-reflective paint can really mess up an image by scattering light onto the detector. He showed us some pictures of flat field images with very bright central peaks that he said might be due to this sort of scattered light.

I began to wonder about this bright ring. To get a better picture of what was going on, I imported the flat image into Mathematica and made this 3D contour plot:

3D contour plot of the flat field image

This shows the general falloff in light being recorded at the detector as you go away from the center, but you can see a ridge of brightness in a ring around the center. I made another contour plot with a zero plane slicing through the contours at the mean value of the pixel brightnesses that I think emphasizes the circular ridge.

The circular bright ring shows up here as a semicircle above the mean value plane

Today I went out to the observatory and dismounted the camera and sure enough, there was a bright ring visible around the secondary mirror image coming from one of the constituents of the image train. I haven’t identified exactly where it is coming from, but here’s what it looks like when you peer in the eyepiece holder up towards the secondary. You can clearly see the bright ring around the secondary mirror image that is coming from a metal surface inside the primary mirror baffle.

Now I need to go in and try to black out the reflection and then see if the bright ring disappears from the flat. Stay tuned!

The view up the eyepiece tube. Note the reflection of the secondary (the circle with the black dot in the middle. The big problem here is the next thin bright ring. Thats coming from the inside of the primary baffle, I think.

A fantastic Rogelio Bernal Andreo image of the region around M78 with flats made with a Flip-Flat applied during calibration

The first image shows what a properly flatted image that has been stretched severely to bring out all the dust and nebulosity looks like (in the hands of a master processor like Rogelio). The second image shows the same field, with the same processing steps, but without the early flat calibration step. As you can see, flats are indispensable if you are contemplating non-linear stretching of your data.

The same image, this time with no flats applied during calibration. The vignetting from the optics is clearly visible.

Rogelio used one of our Flip-Flats on an FSQ106 with an STL1000 camera to make these images. Rogelio is quite pleased with the convenience and quality of the flats delivered by the Flip-Flat.

You can find more of Rogelio’s fantastic images, including several APODs, at deepskycolors.com. Check them out. They really are stunning and unique.

The Flat-Man XL at Galaxy Quest Observatory

Arne explained to me that by measuring the magnitude of a fixed-brighness star on different parts of the detector and then calibrating the images with our Flat-Man XL, we’d get a good idea of how even our flat panel’s light spatial distribution of intensity was. Poor flatting would almost certainly increase the standard deviation of photometrically derived magnitudes.

I got an opportunity to perform such a test on a particularly good night of photometric stability last week using our good friend Jacob Gerritsen’s Galaxy Quest Observatory. Galaxy Quest is located at a pretty good dark sky site in Lincolnville, Maine. There’s an SBIG ST-10 camera on a 12.5″ RCOS telescope and one of our earliest Flat-Man XL 18″ has been in residence there for some time now.

The observing program:

Figure 1: A stack of 1 frame from each image set. Target star circled in red.

I set up a mosaic in Maxim DL to capture the grouping on different parts of the detector by performing small slews after each set of 5 20-second luminance images. Figure 1 shows a co-added frame of one of each of the image sets. The target star, NOMAD 1345-0000396, is circled in red. As you can see, a good portion of the imaging chip is covered by the series of image sets.

I chose 20 second images so that none of the photometric sources were saturated. The ST-10 is a non-anti blooming camera that is quite linear across its entire dynamic range. Using the Maxim DL inspection tool yielded the following for one representative image:

As you can see from Table 1, the target is not saturated, and all pixels are well within the linear range for the chip. SNR is very high. FWHM was quite good this night (Oct 14, 2010 UT) for our area. The clear sky clock showed better than average seeing and transparency throughout the observing period.

Figure 2 shows a close-up of the photometric apertures. I chose an 8 pixel aperture, one that I have used in the past for exoplanet photometry with excellent results.

Figure 2: A close-up of the photometric apertures

Results:

I used Maxim’s photometry tool to generate a list of the magnitudes of the target and check star, setting the reference star to an arbitrary 10th magnitude for zero point.  Table 2 lists the statistics for the 50 images.

As you may recall from your days at school, 99.7% of the data is within 3 standard deviations (3 sigma) from the mean. In this case then, that means that the vast majority of the data is within 0.02 Magnitudes of the mean value.

Recall that I exposed 5 images at each position. Taking the median value of the set of five is a way to average out the scintillation noise. The 3-sigma value in this case was 0.17 magnitudes. Figure 3 shows a graph of the target magnitude (with arbitrary zero-point applied) as a function of time. According to Arne, for small magnitude variations, the differences are nearly equivalent to flux fractions, so 0.17 magnitude variation is akin to a 1.7% variation in the total flux.

To put that into some kind of perspective, 99.7% of the stellar magnitude measurements are within 1.7% of the mean value. It is generally accepted that 1% precision in photometric magnitude is a great result, so we have a pretty good result. Of course, with more frames at each location, we might be able to decrease the variability somewhat, and I hope to do more testing on the next good photometric night. Stay tuned!

Figure 3: Target star Magnitudes for the observing period. Median Magnitude for each set shown in red