Astrodon

Reflections

Where Do Halos Come From?

ST10XME STL11000XM STL6303E

BASIC INVESTIGATION USING ST10XME

Have you seen these artifacts in your images and not know what they were due to? Below is an image of M27 taken with a Celestron 14" Schmidt-Cassegrain telescope at ~f/10.7 with an
ST10-XME CCD and Astrodon Tru-Balance RGB filters. 3 unbinned RGB exposures totaling 10 min. were taken (no luminance) and combined at ~1:1:1 in MaxIm after dark subtraction. The "ears" of M27 were enhanced in Adobe Photoshop. Notice the two circled areas, showing what appears to be dust donuts, characteristic of compound telescopes having central obstructions. These areas have been enhanced in Photoshop to highlight the artifacts. As will be shown below, these are not dust donuts.

 

Furthermore, many of you notice blue halos around many of your stars. The image below? of M20 below is from John Smith .  It highlights these halos (see upper right), and was taken with a 12.5" RCOS RC at f/5.9, an ST10XME, SBIG/CS filters in the SBIG CFW-8.? L(R+B)RGB at 12 x 2.5 m, 6 x 2.5 m, 12 x 2.5 m, G binned 2x2.

I have a C11 and ST10XME CCD, FLI DF-2 focuser and an FLI CFW having 8 1.25" positions. Filters used for the following tests include:
* Astrodon Tru-Balance LRGB (NIR-blocked) 3 mm thick, dichroic
* IDAS/Hutech Type 3 Blue, 1 mm thick, dichroic
* Astronomik Type 2a Blue, 1 mm thick, dichroic
* Meade Pictor Blue, 2 mm dichroic
* Schuler photometric Blue, 4 mm thick sandwich of two epoxied glasses, no dichroic or AR

Most of the dichroic filters had the colored coating on one side and an AR coating on the other.
The AR coating faced the ST10XME.

The first test was at native focal length of f/10.0 with 1 min exposure on a bright blue star,  Merope, in M45.

 

The first thing to notice is the large donut halo about all of the stars, not perfectly centered, but all about the same size. This tends to eliminate internal reflections in the filters as a cause, because they were selected to have different thicknesses.

Second, notice that the same halo is present in the red filter. Although not shown above, similar halos are found in the Astrodon green and luminance filters, as well. So, they are not specific to the filter bandpass. This further supports the hypothesis that they are reflections.

Third, the size of the halo is too large for internal reflection in the filter.  The reflection distance can be calculated as follows.? The approximate diameter of the halo is 200 pixels in the above images. The pixel size in the ST10XME is 6.8 microns, but we must multiply this by 2, since the above were taken binned 2 x 2. This is equivalent to 13.6 microns, or 0.0136 mm. The diameter of the halo on the CCD thus becomes:

 Halo diameter at the CCD detector = 200 pixels x 0.0136 mm = 2.72 mm

Now we must find how far away the source of this reflection is that creates an image of 2.72 mm at the CCD detector. The f ratio is the diameter of the aperture divided by the focal distance. For an f/10 C11, this means that the 280 mm aperture is multiplied by 10 to obtain the 2800 mm focal distance.? So, we do the same thing for the 2.72 mm halo diameter at the CCD detector. We multiply it by 10 (or divide by f/10)

Reflection distance = 2.72 mm * 10 = 27.2

I took the FLI CFW off the ST10XME and measured the distance between the bottom of the filters and the top of the entrance window into the camera to be ~14 mm.  The window in the cover plate of the camera was measurejd to be 3 mm thick.  The distance between the bottom of the window and the detector was estimated to be similar ~14-15 mm.

The reflection distance is twice the PHYSICAL distance between the reflecting surfaces, as shown below.

At the simplest level, one expects three primary reflections. The blue line represents the reflection between the bottom of the filter and the top of the entrance window into the ST10. The green line represents the reflection between the CCD detector (or cover slip) and the bottom of the entrance window. Lastly there can be an internal reflection within the entrance window, shown in red.  In each case, compared to the light ray that produces the focused star (heavy black line), each reflection represents TWICE the distance between the reflecting surfaces.  Take the blue line. The ray travels to the window, bounces back to the filter and again travels to the window and on through (ignoring secondary, tertiary, etc. reflections which become very faint and very slight refraction within the optics).  Thus, it travels an extra 14 x 2 = 28 mm, approximately what we calculate above.

Also, the cover slip over the KAF3200ME CCD is 0.84 mm and the distance between the CCD and the cover slip is 0.94 mm, for a total distance of about 1.8 mm.

On the left, we see a small halo around the star.  This halo is approximately 45 pixels in diameter.? This corresponds to a distance of 6.12 mm, corresponding to twice the thickness of the entrance window.  

As the histogram is manipulated (right), the larger 200 pixel halo appears corresponding to the distance between the filter and the entrance window.  

However, if the distance between the filter to the window and the window to the CCD detector are comparable, we still have not proven which reflection it is. The FLI CFW abuts directly against the ST10XME with a zero-space adaptor. As such I was able to loosen it in a way as to induce a few degrees tilt angle between the CFW and ST10XME.  Please note here that the filters are still perpendicular to the optical train. The camera is now tilted. I reshot the above at f/10 and obtained:

 

Notice that the 200 pixel halo has move toward the upper right about the same amount for all filters tested, but that there are additional, larger reflections centered further away in the same direction.  This would be consistent with the 200 pixel halo arising from within the ST10. As the window tilts further away from the filter, the reflection with the filter will be progressively displaced more and become larger, whereas the internal geometry within the ST10 is fixed.

To examine for smaller halos around bright stars that may come from internal reflections within the CCD detector with the cover slip, we need to go to faster optics.  A Celestron f/6.3 reducer was added to the visual back of the C11 and the above experiments were repeated.  Similar results were obtained with all filters.  The Astrodon red filter was selected, as shown below:

The left image shows two, large concentric halos.  The larger halo is 350 pixels wide and represents a diameter on the CCD of 4.76 mm x 6.3 = 30 mm, so the physical distance is 15 mm.  The halo just inside is 300 pixels wide for a physical distance of 12.8 mm.  These could be consistent with the reflection between the filter and window (15 mm) and the window and detector (12 mm), respectively.  

The right image of the same star with a different histogram stretch shows the halo that many people may see as a blue fringe around stars.  It is 40 pixels wide and orresponds
to a 3.4 mm diameter on the CCD, and a physical distance of 1.7 mm.  This is the distance between the surface of the highly reflective CCD and the top of the cover slip. This could therefore be consistent with an internal reflection within the sealed CCD detector package. Note that the KAF3200ME cover slip, as used in SBIG cameras, is AR coated producing high transmission (low reflectance) in the normal luminance passband.  It drops off somewhat in the NIR.

It is likely that this is the cause of the blue halos shown in John's M20 image. The use of an independent filter wheel allows the 30 mm thick DF-2 focuser and the FLI CFW to be reversed in position, thus extending the distance between the ST10 entrance window and filter to about 45 mm.  This was done and another 1 min. exposure was taken, binned 2x2.  The following image was taken using the green Astrodon filter in this configuration:

 

The brighter halos around the star are about 200 pixels wide at f/10.0, corresponding again to a physical distance of 14 mm.? The next larger halo starting just below the "M" in the word "Manual" in the histogram window, and extending over to the bright star on the right side is 630 pixels wide, corresponds to a physical distance of 43 mm. Adding the original distance of ~14 mm + 30 mm for the intervening DF-2 results in a physical distance of ~ 44 mm. It is likely that this large reflection is simply due to the much larger spacing between the filters and entrance window to the ST10. There is one more, fainter halo outside this one with a diameter of 860 pixels, corresponding to a physical distance of 58.5 mm. This could be consistent with a reflection from the CCD through the window to the filter and back. This distance is ~58 mm.

Note that the edge of many of these halos is not perfectly circular.  This is likely due to the flexible Just-Cheney dew shield used on the C11.  Images of out-of-focus stars look just like this due to sagging and the irregular shape of this flexible shield.

One last test. In my excitement over the test results, I must not have saved the 1 min exposure taken without a filter in place. This eliminates reflection back from the filter. John was kind enough to take images of Merope in M45 for this purpose.  The following shows that even without a filter in place, there are the same reflections around the star in the left image. They correspond to 12 and 14 mm based upon John's configuration. These are consistent with the approximate spacing between the entrance window and the CCD within the ST10.

Compare that with an image taken by John with the Blue SBIG/Custom Scientific filter (right image).  The same reflection appears, but is displaced upward due to tilt of the filter. There are larger reflections, also, that likely correspond to the ~26 mm spacing between the CCD and the Blue filter in the ST10XME/CFW-8 system.

Please note that even with the screw-in filter holders, there can be some tilt of the filter relative to the optical axis of the system. Thus, the halos may not always overlap among the red, green, blue and clear or luminance filters.

Note also in my M27 image that the halos are green. However, the halos in John's image are all blue, as are halos in many images taken with SBIG CCDs. This was indeed puzzling. The first thing to understand is that the various halos of different sizes are present in ALL filters, including the Luminance filter.  They are surface reflections. Second, they occur around saturated (white) stars.  Hence reflections are expected to be roughly equal. Third, a unique design goal for Astrodon Tru-Balance filters is that they eliminate the inefficiency found in most blue filters. The R:G:B weight ratios are 1:1:1. That is how the image of M27 was composed at the top of this page. The halos appear yellow-green.  However, in the SBIG/Custom Scientific filters used in most SBIG CCDs, the weight ratio is ~1.3:1.0:1.6 for the ST10XME and ~1.2:1.0:1.7 for the ST10XE.  So if these are surface reflections that don't depend upon which filter is being used, and the brightness of the halo is comparable in the red, green and blue images, multiplying the blue image by 1.7 during the color combination process will definitely make these reflection blue in the final RGB image.

The image below is an RGB made from the 1 min exposures of Merope on the C11 at f/6.3 with R:G:B weight ratios of 0.9:1.0:1.0, corresponding to the design in this Astrodon prototype set. The greenish halo is the one described above, representing a physical distance of 1.7 mm, and attributed to a reflection inside the CCD detector package. Notice that it has the same color as shown in the M27 image. The star to the right is from the same RGB set, but the ratio was changed to 1:1:1.7. The halo is blue.  Although this is not quite the same as taking the image with the SBIG filters in my system, remember that John did this with his system containing SBIG filters for M20, and obtained blue halos corresponding to the 1.7 mm distance.

There may be another, important factor here that contributes to the blue color, and that is the inherent reflectivity of the silicon material making up the CCD.  I found a reference (Haapalinna, Karha and Ikonen, "Spectral Reflectance of Silicon Photodiodes" , App. Optics,Vol. 37, No. 4, February, 1998, p729) that shows enhanced reflectivity in the blue region, as shown below:

This graph shows two important characteristics of silicon reflectivity.  First, there is not much difference in the green and red regions. Second, reflectivity in the blue region is nearly 50% higher.? Most Blue filters in RGB sets cover the range up to about 520 nm.

Therefore, the combination of enhanced blue reflectivity from the silicon material of the CCD and the high blue weight used in non-Astrodon filters will cause blue halos around bright stars.

So, by now, I imagine that you are all wondering........so what?  

As John points out, one direct consequence is there appears to be some CCD-imposed limits on how deep an exposure can be made before these internal artifacts appear, thereby limiting the equivalent exposure depth. We are all familiar with noise-reducing techniques. But, none of these will help with reflection-induced artifacts arising from camera design choices. You will likely have to clone stamp these artifacts out in Photoshop.

The large donuts that involve reflections from the entrance window into the SBIG CCD can be considerably reduced by using a better anti-reflective coating.  SBIG uses a good quality 1.25" window, probably 1/4 wave BK-7 glass, with a single coating of MgF2 (magnesium fluoride) on each side.  I have measured the transmittance of this window, which is about 97% across most of the visible and near-infrared, at least out to 900 nm.  SBIG likely chose this coating because it is still functional in the NIR where their clear filter works.  Some AR coatings (e.g. HEBAR = high-efficiency broadband anti-reflective) are great in the visible region but are poor in the NIR.  I purchased and tested a similar BK-7 glass from EO Edmunds, coated with their excellent hard AR coating (P/N F45661). This glass has a transmission of >99% throughout most of the visible. My scans of these windows are presented below.  That means that surface reflections are reduced by at least a factor of 2. I purchased an extra ST10 cover plate and will test this optic shortly in hopes of reducing the large donuts.

However, this approach will likely not reduce the problem of small halos shown in John's M20 image that likely come from internal reflections within the CCD detector package, i.e.between the detector and cover slip when using fast systems, such as f/5.  It is interesting that I never saw these halos with my FLI MaxCam CM10-2ME on the f/5 Takahashi FSQ refractor. This FLI camera has the same Kodak KAF3200ME as does the ST10-XME. FLI uses a sealed chamber backfilled with an inert gas, like Argon. Therefore, they sell their systems without coverslips on the KAF3200ME CCD.  There is no 1.7 mm internal reflection. There is nothing to reflect off of. However, it is SBIG's policy to ONLY use CCDs having cover slips, due to concern for cleaning any dust particles or contamination from the CCD surface, especially if it has the plastic microlenses in the ME versions. This is an understandable policy, but the price that is paid is the presence of these halos.

I want to thank John Smith for posting his M20 image that triggered my study based upon the larger donuts that I saw in my own images.  I also appreciate the willingness of Alan Holmes at SBIG to discuss this issue prior to the public posting of this work.

STL11000 Halo Study (top)

Recent images containing bright stars taken with the SBIG STL11000 CCD camera further help to identify where blue halos are coming from.  This camera contains the very large Kodak KAF11000XM CCD.? Based upon the package specification from Kodak, the large ~1.5 x 1.3" detector is protected by a multi-layer anti-reflective (MAR) cover glass. The distance between the surface of the detector and the upper surface of the cover glass is anywhere from 1.15 to 1.65 mm, (1.40 +/-0.25 mm). The thickness of the cover slip is ~0.85 mm.

From the SBIG web site, the dimensions between the various surfaces can be estimated for the STL11000:

?

(Click on image to enlarge)

The cover slip has been colored red. The 3 mm thick detector chamber window is colored green. The filters are colored blue. Both 50 mm SBIG and Astrodon filters are ~3 mm thick, as shown in this drawing. The distance between the detector and the bottom of the chamber window is about 10.6 mm. The distance between the top surface of the chamber window and the bottom of the filter as it sits snuggly in the counterbore in the filter wheel carousel is ~ 9 mm.  As indicated above, the distance between the detector and the top of the Kodak MAR cover slip is ~1.4 mm.

I took the following image of the Pleiades, M45, with an STL11000 unbinned and 3 mm thick Astrodon filters with a Takahashi FSQ106N at f/5.0.  A portion of the RGB TIF file (not fully processed) is shown below:

?

The halos are 84 pixels in diameter.? The physical distance creating this halo from the above formula is:

 (84 * 0.009 * 5) / 2 =  1.89 mm

Since the filters and the chamber glass are all 3 mm thick, and the distances between the chamber glass and the filters or the detector cover slip are in the 9-10 mm range, the only spacing that is consistent with this calculation is a reflection between the coverslip and the CCD.  

The following unbinned blue filter image of M45 taken with the STL11000 by Geoff Collins using thinner Astronomik filters on an AP155 refractor at f/7 is shown below:

?

 

The Astronomik filters are much thinner than the SBIG or Astrodon filters. I measured the thickness of 1.25" Astronomik 2c filters at ~1 mm.  You can see two large sets of concentric halos around each star.These are the two bright stars making up the bottom tail of the "question mark" of M45.  The outer halo has a diameter of ~370 pixels for a physical distance for the reflection of 11.6 mm.:

(370 x 0.009 x 7) / 2 = 11.6 mm

The inner halo has a diameter of  ~310 pixels for a physical distance for the reflection of 9.8 mm. The difference between the two is 1.8 mm.  Distances in the 9-12 mm range can only occur between the filters and the chamber window, or the chamber window and the detector.  One thing to note is that larger Astronomik filters come in filter holders that are placed in the STL filter carousel.  Therefore, the bottom of the filter is further away from the chamber glass than SBIG or Astrodon filters that rest in the bottom of the counterbore, probably by at least 1 mm.  So, this would bring the distance between the chamber window and the filters to at least 10 mm.  These halos are too large to come from an internal reflection involving the cover slip and detector or within the filters. Note that the halos are not concentric around the stars.  This means that the reflecting surfaces are not parallel or the camera is tilted from the optical axis of the system.

The 1.8 mm difference is similar to what was measured in my M45 image and suggested to originate from the cover slip.  The double halo in Geoff's image may be arising from the main reflected ray of light undergoing a further reflection between the CCD surface and cover slip.

However, there are other, smaller halos in the image, offset in the same direction. Measuring these halos around the stars in the lower right portion of the image result in diameters around 48 pixels, corresponding to a physical separation distance of 1.5 mm.  This is close to the distance between the CCD surface and top of the cover slip.  It is likely the same halo seen in my M45 image, remembering mine came from a faster f/5 FSQ106 in comparison to Geoff's AP f/7.  As the optics become faster, halos shrink down around stars. That is one compelling reason for filter designers to make their filters thin.

It would have been useful to take an image of M45 without any filter to firmly identify whether the 10 mm halo comes from the filters or within the STL11000. There is no evidence for a halo corresponding to a 10 mm physical separation in the FSQ106 image at f/5 which would be about 220 pixels in diameter.  Since the internal optics between the detector and chamber window are the same in both systems, this could suggest that the large halo in the AP155 f/7 image comes from a reflection involving the Astronomik filters and the chamber window.  More work needs to be done to validate this conclusion.

STL6303 Halo Study (top)

The same approach can be applied toward the SBIG STL6303E, with its 9 micron pixels.  The following image from Jay Ballauer of the California Nebula, NGC1499, shows a bright halo around the star in the lower right portion of the image.  This halo is approximately 84 pixels in diameter in the unbinned image taken with a Takahashi FSQ106N at f/5 and using SBIG LRGB filters.  Using the formula again, we have

( 50 * 0.009 * 5 ) / 2  = 1.89 mm

?

Again, this physical distance can only come from the cover slip area, because the chamber window and the SBIG filters are 3 mm thick, precluding that the halo arises from an internal reflection within these filters or window.