In search of the flat field

I’ve been thinking about the problem of evaluating my flat field lighting and target, and two approaches came to mind. One is to make some flats, rotate the camera 180 degrees, make more flats, and then measure the difference.  I haven’t done that, but it seems like it should work.  The other method is to photograph the target  in a wide field with my dslr and look at color rendering and the light distribution over the field.

Here’s a stretched look at the master flat I had made.  It actually did a pretty good cosmetic job on the NGC80 image.

stretched maser flatIt shows the vignetting produced by the optical train.  The “donuts” are shadows of dust bunnies on the cover glass of the ccd chip; they’re shaped like that because my optical system has a secondary mirror that forms a central obstruction.  When this image is applied to the image of NGC80, all that crap is magically removed.  So flat fielding  removes artifacts of the optical system and ccd, but requires an evenly lit target to work.  Here’s the DSLR picture of the first try at a target:

first flat field target

First, it’s really red.  I’m using a 15 watt tungsten bulb to illuminate the target, which is bouncing off a white reflector before hitting the white target.  It looks like about a 2800K source, that gets even redder bouncing off the plywood interior of the hut.  Color matters because I will do photometry, and try to get very accurate measurements of star or minor planet brightnesses, and the sensitivity of the chip varies a bit with wavelength.  So I’m aiming at a more “daylight” balance on the theory that that’s a pretty average star color.  Being the color of our star.  Let’s see how even it is.

I converted the image to grey scale and used the horizontal box tool in Maxim to see average values across the image.  You can see that the illumination drops off on the right hand side.  It looks to me like if I can level out the profile, I should have a pretty good target.  The light is falling off because the scope is casting a shadow on the target.  It worked ok with my previous setup, but needs a fix with the new mount.  Here’s the way the setup should look.

I put some tough blue, a lighting gel used to convert tungsten to daylight, on the lamp, and fiddled with the reflectors.  The blue really helps the color balance.

The reflector and target are both white foamcore.  I need to come up with an intelligent mount for the cards, right now they are just taped and wired, and aren’t parallel. Also, I find that the sweet spot for the scope is right on the stud.


The illumination still looks pretty blotchy, so back to Maxim for more measurements.

At least the slope is in the opposite direction!  I still have about 10 percent variation across the target.  Next fix is to properly mount the target and reflector, which I think will make it pretty close.

Focus, Flats and NGC 80

No moon, the evening started with thin cirrus clouds that disappeared, I think an hour or so after sunset.  With those kind of clouds, I’m never sure, I think they are actually there to make halos and generally degrade the image, but you can’t see them.  I thought good enough for a focus run with FocusMax.  I did find that the focuser wants to be connected to Maxim or FocusMax, but not both.  So I connected to FocusMax and pretty much just ran the “First Light” routine.  I selected a 4th mag star in Pegasus, and began the run.  The routine built a pretty straight slope from the left, and as the star approached focus, it had an obvious bloom spike.  The curve bottomed, went up, then back down, then up again in a fairly tidy rounded W shape.  FocusMax didn’t like the bloom; the routine failed.  The recommendations were for a “fairly bright star”, mag 3-5.  A nearby 5th mag star had the same result.  But when I ran it with a 6th mag star, the routine ended with a very tidy vee curve.  The wizard then did a sample auto-focus, which was pretty far off.  The curves are written to a system_profile file, which had some old runs in it; I deleted the old runs and made another v curve.  Then the autofocus routine came in fairly close.  I ran about 6 V curves which all looked pretty similar.  After re-reading the documentation, I see I should manually expand the in-out range of the focus run and use smaller intervals, which should provide a more accurate focus.

I shot a luminance series (10- 120 sec) on a  field that looked interesting in TheSky, in what I assumed to be Pegasus, but turned out to be Andromeda. The East side of the Great Square is really right on the border. The field includes NGC 80, and about 22 other galaxies including NGC 81 and 83.  not exactly showcase objects, but a very rich neighborhood!  I think the focus still needs work, and also still have somewhat oval stars in the east-west direction.  Pixinsight actually tidies those up via data rejection.

Field centered on NGC 80; 10 – 120 sec luminance with 12″ Meade LX200 and ST-10XME. Calibrated and stacked in Pixinsight, linear adjustments in Photoshop.

I set up a white foam core flat field target at Park 1 position on the wall of the hut.  It’s illuminated with a 15 watt incandescent lamp in a floor fixture that’s aimed at a white foam core target on the opposite wall.  For remote operation, I can turn on the light via a web switch. (not set up yet)  It gave a well exposed flat field in a 1 second exposure.  It seems like there should be a good way to objectively analyze how flat the field actually is.  The line analysis tool in Maxim showed an apparently symmetrical distribution vertically and horizontally.  Also the color temperature of the illumination is maybe 3200K – a low wattage bulb mixed with bounce from unpainted plywood –  so there are color implications.    Never the less, the flat seems to do a good cosmetic, at least, job of correcting the light images.