I decided to run some tests capturing flats at different exposures today. I found the results quite interesting. CCD ADU count does not appear to be sensitive to CCD temperature and while the max pixel ADU count in the flat image is non-linear with increasing exposure the min count is quite linear so there is a point where the range of pixel values reaches a maximum.
Good info Peter, you've been busy! Interesting that temp has no effect. Now to try and interpret the data......and then ask numerous irritating newbie questions
I presume the exposures were done with a constant value of illumination from your light box, so with that fixed level of light, anything beyond about 7 seconds will start to cause white compression? How does this affect your choice of flats exposure time, and in your opinion, how does it relate to other amounts of light (eg; night sky objects as we see them?)
What gain and offset do you use on the QHY8Pro?
I'm not sure what point I'm trying to make, just asking the questions that come into my head
I presume the exposures were done with a constant value of illumination from your light box, so with that fixed level of light, anything beyond about 7 seconds will start to cause white compression? How does this affect your choice of flats exposure time, and in your opinion, how does it relate to other amounts of light (eg; night sky objects as we see them?)
What gain and offset do you use on the QHY8Pro?
Mike,
Yes - I used the light box with the brightness constant throughout. I'm unsure what is the best time to choose based on what people have said - somewhere between "1/3 of full well" to "as high as possible on the linear part of the curve". 7-9 seconds seems to be the point where the curve turns so I guess I should be shooting flats at between 2 and 7 seconds. Not sure how to compare the light box to night sky - I guess you could do a similar thing if you were quick! I use gain=0 and offset=125.
Hi Peter, Based on the figures you have computed here it is apparent the output is linear out to about the 50K mark so my initial thoughts would be to try applying flats with a max of about the 20 - 25K area which puts you in the 2-3 second exposure range.
Well done. I bet there is a few people watching this thread with interest.
I ended up with very similar results.
Hi Peter. If you can get hold of it, Aip4Win has an interesting section on Testing CCDs. Part of the process is the construction of a 'regulated' flats box that doesn't rely on shooting through your scope. If you are interested, I have a box i've made that I can bring to Astrofest, if you are interested in refining your readings.
As for exposure times for flats, the way it was explained to me exposures of between 2 and 10 sec produce the most reliable flats for any given situation. I try for around 6 sec and also do flat darks.
Not having a light box, I use the time honored white T shirt and the late afternoon SW sky for flats. I might try some sets similar to yours to see what I get.
As I don't know what the light level is compared to your fixed light box, it may also throw up some interesting numbers. Have to be fast with fading light though.
You do if you are taking flats Peter, but if you want to just test your CCD for read noise, ADU, linearity and Transfer efficiency then this method is simple and straight forward.
Very interesting. I reckon for completeness it'd be handy to also post the histograms at different exposures, so you can relate the different parts of the above charts to histograms shapes/peaks.
I don't know much about CCD sensors, but if this were a DSLR's results I'd be expecting something similar in terms of the shapes of those charts. The min values just increasing linearly but the max values dropping off. Think it's to do with sensors being more sensitive at the highlight end. Similar to the "exposing to the right" philosophy in photography.
I've read about the recommendations for flats being to have the peak of histogram around 1/3 to 1/2 way from the left (darks) end. Be interesting to do the histogram comparison as I suggested above and it may just be that this relates to that linear part of the curves up to 7 or 8 secs anyway. Hence both theories/recommendations are correct.
Here's part two... I spent hours last night capturing images and 6 sets of corresponding flats at 1s, 2s, 3s, 4s and 6s (20 of each). I captured the flats in groups of 5 and rotated the light box 90 deg between each set just to make sure there is no chance of un-even illumination from the light box. I've now spent 6 hours processing all these and have discovered the same issue no matter which set of flats I use! It appears the flats are over-compensating for vignetting making a light pattern around the image when the flat is applied rather than a dark one with no flat.
I'm wondering if this has something to do with the QHY8 being OSC and how deep sky stacker handles this. According to the DSS help file it applies the flat for each channel separately and as long as the peaks for the channels are between 1/3 and 2/3 of the maximum on the histogram it shouldn't be a problem. I've ensured I used a qualifying set and still get the same result??
The first image below is the stack of lights without flats - the second is the same stack of lights with 20x4s flats applied.
It appears the flats are over-compensating for vignetting making a light pattern around the image when the flat is applied rather than a dark one with no flat.
Your flat fielding is over correcting your sub for one of the two reasons below:
1_ You didn't subtract the master bias frame into your flats prior to divide it in into your subs so your flats is not scaled correctly.
or....
2_ Your flat exposure is not bright enough. This greatly depends on the scope FL , filters and aperture.
Here's a few value that work fine for me. Record the ADU count in the brightest pixel of the bayer matrix in the center of your field:
For the ED80 [600mm FL] with no filter I have to aim for 10000 ADU in the brightest pixel of the bayer matrix. Typically 85 to 90ms exposure.
For the 5” Newtonian [650mm FL] with no filter I have to aim for 9000 ADU in the brightest pixel of the bayer matrix. Typically 35 to 40ms exposure.
For the Hyperstar [504mm FL] with no filter I have to aim for 9000 ADU in the brightest pixel of the bayer matrix. Typically 40 to 50ms exposure.
For the ED80 [600mm FL] with a Ha filter I have to aim for 10000 ADU in the brightest pixel of the bayer matrix. Typically 6 to 8s exposure.
For the 5” Newtonian [650mm FL] with a Ha filter I have to aim for 11000 ADU in the brightest pixel of the bayer matrix. Typically 3 to 4s exposure.
For the Hyperstar [504mm FL] with a Ha filter I have to aim for 20000 ADU in the brightest pixel of the bayer matrix. Typically 1 to 1.2s exposure.
I haven't used any Bias frames. How do you capture bias frames with a qhy8 given it doesn't have a shutter?
I've tried longer exposures (up to 6 sec) and still get the same result. According to the tests I ran last weekend the ADU count for 4 sec is between 15000 and 35000 (LHS to RHS of histogram) with average at about 26000. Would this suggest my flats are over-exposed?
I haven't used any Bias frames. How do you capture bias frames with a qhy8 given it doesn't have a shutter?
I've tried longer exposures (up to 6 sec) and still get the same result. According to the tests I ran last weekend the ADU count for 4 sec is between 15000 and 35000 (LHS to RHS of histogram) with average at about 26000. Would this suggest my flats are over-exposed?
26000? Yeah...way to bright. Start around 10000 ADU. Note the exposure time to get that level. Shoot 40 of them. Rotate your light box if needed (10 at 90degrees, 10 at 45, etc...clockwise, you get the idea). Then shoot 40 darks/bias at the same exposure time. Plug your scope to shoot.
Then make your master bias then your master flat. When you calibrate your sub subtract the bias master from the flat master then divide the flat master into each sub and you'll get perfect seamless flat fielding, both illuminations and dust motes. If not try a brighter flat until you get the right levels.
My flats with QHY8 using 2-layers whiteT shirt and clear twilight sky (5.15pm Bris Friday) using Hyperstar + C8.
I shot 10 each at 1, 10, 25, 50, 75, 100, 250 and 500ms. I saw that anything over about 250ms was saturating badly. I recorded the ADU's in Excel and 75ms seemed to be about the best.
Bias was just lens cap on and 1ms x 50 shots. Used neb to do all the bias and flat stacking and master bias subtraction and finished up with a good master flat that left me with a pretty good image after using it. I had previously had the same "reverse flat" effect you mentioned.
Hi Peter I agree with Marc. In all the images I did with a QHY8 the only way to get good results using flats was to add a master bias to the image calibration process. I used ImagesPlus and found it to be a very easy process. I do recall a lot of discussion on the QHY forum about flat frames and the overall result was that Bias frames were required for good results.