Sorry everyone, I know this issue has been done to death - but I still can't get my flats right.
I'm using a light box for my RC8. I have the sky flat plug in with Maxim which sets the exposure for my flats at the right ADU. All good except...
The flats seem to overcorrect the lights. Rather than excessive illumination in the centre, I have a darknpatch there and overly light areas at the extremities.
This tells me that the flat I am using is not truly flat ie I am applying a flat frame whose illumination doesn't match the actual sensitivity of the camera and filter combination.
I usually take 10 flats. I then dark subtract an equivalent number of darks and apply a bias maters (made up of 3 or so bias frames). I then median combine them to make a master flat.
I've used ADU counts of 20,000 as well as 15,000 and 25,000 and get the same result.
Looking carefully at the thumbnails for the flat frames at once, I see that the frames are not identical. In one shot the centre of the frame is brightest while in the next the brightest area is off centre.
My thinking is that the flats are displaying shutter artifacts and it's the effect of the shutter that shifts the brightness around somehow. The accumulation of these out of kilter frames produces a non-representative flat.
The other point to note is that the bottom centre of the frames display a dark semicircle. I suspect this is shutter artifact as well.
My camera is a SBIG ST8300 binned x2. This give me a flat with exposure of between 0.15 and 0.3 seconds.
So I think I need to include a t-shirt or other material to step down the intensity of the light.
Does this make sense? I am open to suggestions and want to preserve my good humor with this.
Pete
Ps. The unflat nature of the flat is apparent when I set the screen stretch on Maxim to high or medium
You can see how bright the central line is compared to adjoining areas. In successive flat frames this is the area of overlap ie its the one section that is invariably bright. you can also see the dark semicircle atthe bottom.
Pete try 'correcting' a single flat with your master flat. If the result is uniform in brightness that means the software is not at fault.
Any uneven illumination from an artificial light source will show up in a flat corrected real astro image.
I have personally tried every method for flats and have found that blue sky flats taken with darks beats any other method. Of course this assumes your camera can take very short exposures. My Canon 5DH can and they are typically at 1/320th sec at an ISO of 200 at late afternoon when the observatory is in shade from direct Sun.
Hi Pete, yes finicky things flats for sure and I've had similar dramas of over correction when I started doing them as well so your with some company here!
*First off, take your flats in the same binning as the images you are correcting.. also, you need to take an individual set of flats for each of your filters.. ie.. R,G,B, L, Ha.. etc
*Do flats at 2/3 of your full well capacity... for the Kodak8300 chip with capacity of 25500, look to go around 17000.
*Make sure you are doing the flats at the same temperature of what you've been taking your light images with.
The best way is to set up a 'flat' run in maxim, and allowing enough time in between exposures so your chip can cool back down to the temp it needs to be.
*I use about ten flats.. it works, but more is probably better.
*Remove a bias master from all your individual flats.
*When you do your masters, make sure you set up your 'pedestal' in MaximDL.. this I found did really help.
You'll find this in the 'set calibration' window.... hit the 'advanced' button, should be set 3xs the 'standard deviation' of your background... setting this to around 100 is in the ball park.
*Once you have all your flats and they are bias corrected, make your masters of each individual channel, 'median combine' is best.
*When calibrating your images, make sure you bias correct the lights first and then flat correct them.
I don't do any dark subtraction to the flats as flat exposures are so short, bias removal is far better... dark removal is really unwarranted (could induce more noise to the flat). I don't dark correct my lights either. When combining the subs, I 'Sigma combine' them. Using Sigma combine gets rid of the majority of your shot noise. Those more recalcitrant hot pixels you can make up a hot pixel map and use that to the subs.
I think if you are getting your cooling down to at least -15*C, you should be fine.
Hope this helps...
The above has been working for me.
that's an interesting exercise - I end up with reasonably even illumination - except down the centre-right of the image which is darker than the rest of the frame. The semi-circle at the foot of the image is also brighter in the corrected flat.
I suspect this is a function of the master being uneven rather than a software issue as I've seen this with both Maxim and AIP4WIN.
Hi Pete, yes finicky things flats for sure and I've had similar dramas of over correction when I started doing them as well so your with some company here!
*First off, take your flats in the same binning as the images you are correcting.. also, you need to take an individual set of flats for each of your filters.. ie.. R,G,B, L, Ha.. etc
*Do flats at 2/3 of your full well capacity... for the Kodak8300 chip with capacity of 25500, look to go around 17000.
*Make sure you are doing the flats at the same temperature of what you've been taking your light images with.
The best way is to set up a 'flat' run in maxim, and allowing enough time in between exposures so your chip can cool back down to the temp it needs to be.
*I use about ten flats.. it works, but more is probably better.
*Remove a bias master from all your individual flats.
*When you do your masters, make sure you set up your 'pedestal' in MaximDL.. this I found did really help.
You'll find this in the 'set calibration' window.... hit the 'advanced' button, should be set 3xs the 'standard deviation' of your background... setting this to around 100 is in the ball park.
*Once you have all your flats and they are bias corrected, make your masters of each individual channel, 'median combine' is best.
*When calibrating your images, make sure you bias correct the lights first and then flat correct them.
I don't do any dark subtraction to the flats as flat exposures are so short, bias removal is far better... dark removal is really unwarranted (could induce more noise to the flat). I don't dark correct my lights either. When combining the subs, I 'Sigma combine' them. Using Sigma combine gets rid of the majority of your shot noise. Those more recalcitrant hot pixels you can make up a hot pixel map and use that to the subs.
I think if you are getting your cooling down to at least -15*C, you should be fine.
Hope this helps...
The above has been working for me.
Rich
Thanks Rich
those are all useful tips. I have most of these in hand - including the pedestal setting - even though I had no idea that I needed to set that, its been set already.
I'm intrigued by the no dark approach - I image at -10 here. I probably should go lower.
Hi Pete,
Exposures shorter than ~1 sec will introduce shutter artifacts, but this might not be the problem you are seeing, the problem could be with the light box itself. The only good way to diagnose is to take some twilight/sky flats and compare. Point the scope near zenith about 10 deg in the direction away from the sun (e.g. a bit east at dusk). You will only need one flat to compare the light box result, but at least 7 frames, 11 preferred to make a good combined master flat (like you are doing). NB when making flats in twilight either turn tracking off of dither large amounts between frames, and try to stay out of the milky way...
Ok Pete, with the dimmer at its minimum, the leds are almost turned off, so it is obviously not a brightness issue. I was speaking to Doug (Hagar) and he is going to post something with his experiences with flats as well.
Hi Pete,
Exposures shorter than ~1 sec will introduce shutter artifacts, but this might not be the problem you are seeing, the problem could be with the light box itself. The only good way to diagnose is to take some twilight/sky flats and compare. Point the scope near zenith about 10 deg in the direction away from the sun (e.g. a bit east at dusk). You will only need one flat to compare the light box result, but at least 7 frames, 11 preferred to make a good combined master flat (like you are doing). NB when making flats in twilight either turn tracking off of dither large amounts between frames, and try to stay out of the milky way...
Good luck,
EB
Thanks EB
I had a go with sky flats a few months back and had similar issues. That said, I might give it a go again just in case
those are all useful tips. I have most of these in hand - including the pedestal setting - even though I had no idea that I needed to set that, its been set already.
I'm intrigued by the no dark approach - I image at -10 here. I probably should go lower.
Perhaps I need more flats than 10?
Pete
Pete, just try not dark calibrating your flats, just use bias subtraction. Go for around 17000 adu and manually set your pedistal, see how you get on.
With your light box having a variable intensity, you should easily be able to keep your flats well under 10 secs even with the blue channel, so really bias is all that is needed..
Yep and get your cooling down that extra -5*C.. (-15*C) I think this will help in the long run when calibrating.
More flats won't help at this stage, you need to sort why there's over correction first.
As Bert suggests, just do a single flat, bias it and then test it on a biased image to see how it goes...
In this day and age, with all the advances in cooling, cleaner chips and software advances, imo darks are a thing of the past!
Ok Pete, with the dimmer at its minimum, the leds are almost turned off, so it is obviously not a brightness issue. I was speaking to Doug (Hagar) and he is going to post something with his experiences with flats as well.
Cheers Peter
I'm intrigued. Could it be internal reflections? The RC8 has baffling. Perhaps it's the filter wheel?
I have recently acquired a FSQ106. I can give it a run too
Pete, here's a doc by Richard Crisp on diagnosing problems with flats. It may be of some help.
I don't think .3sec is long enough for a shuttered camera.
Pete, here's a doc by Richard Crisp on diagnosing problems with flats. It may be of some help.
I don't think .3sec is long enough for a shuttered camera.
Robin
You and I are on the same page here.
Thenlink was very helpful. I could flat field adequately with another image - there were no obvious illumination issues. The same can't be said for the flats
I'll have a play with turning down the brightness to get longer exposures now
I stepped down the brightness - courtesy of 1 or 2 t-shirts and achieved a much more even field. Sadly the 100 minutes of data now has no flats to calibrate them but its worthwhile working this through
The suggestion to test a flat with another flat worked a treat.
Thanks to everyone for their suggestions. Its not really intuitive.
Hi Pete, A couple of things seem to be happeneing here. It aoppears to be an exposure problem. On one hand to much exposure showing up in the extremely black dust donuts, bright centre and backgound and the hardest thing to correct is the length of exposure required due to the slow shutters found in most 8300 CCD cameras.
The problem stems from the camera taking its own sweet time to take the image even when a very short exposure is selected due to the slow shutter.
The problem is fixable but it will require you to extend your exposure time out beyond the 0.2 of a second exposure time to get beyond the min exposure time of the shutter.
You may have to either darken the LED by adding a sheet or two of paper to the Light box appature or dimm it down more if possible. 2X2 binning certainly won't help but is required so maybe appature darkening is best.
Make sure you get your Intensity value in Maxim out to about 12 to 15 read at the bottom of Maxim as you drag the mouse around the brightest parts of the exposure. It can take a while to get these figures worked out and will be different for each filter.
I had a look at your posted flat and although only an 8 bit image with intensities from 0 - 255 your brightest areas are about the 225 level which seems a bit high. I would have expected a maximum of about 130 for a good 8 bit flat.
Remember yours from the camera will be 16 bit not 8 bit so target figures of 20 to 30 thousand should be targeted.