taking flats - advice please

[red dwalf]

i took an image with darks and bias and was happy with the results, low noise which i think dithering helped with, i decided to add some flat files to the image with the same set up and it seems to be adding a lot of noise, no way near as smooth as the previous stacked image, so i`ve took this screen shot to see if i`m missing anything, i`ve been trying to get the adu reading to about 27,000.

 

[spaceman_spiff]

How many flats are you taking? I normally use about 40 flat frames with the same ISO as the original lights.

[red dwalf]

I should have said, I'm using a cooled atik 460 mono and I usually take 21 flats, along with bias and about 15 darks minimum to 21 Max.

[andrewluck]

I usually aim for a total ADU in my flats of about 1,000,000. So, in your case, aiming for 27,000 per image it's about 40 exposures. There's nothing particularly scientific in this number, it was originally a recommendation from Kevin Nelson at QSI.

The important thing is that the noise in your images should be much larger than the noise in your calibration frames. This is because the noise adds in quadrature and the largest noise figure dominates the result. This is a way of saying that you square all the noise figures, add them and take the square root of the result. If we do this on a couple of examples:

Equal noise in both image and flat (std deviation of 10) - sqr(10^2+10^2) = sqr(100+100) = 14.14 or 40% noise added to image

Image has 5 times the noise - sqr(10^2+2^2) = sqr(100+4) = 10.2 or 2% noise added

Andrew

[Astrofriend]

Hi Red Dwalf,

When you say level of 27000 ADU, do you mean average of the whole image or the center parts where you have the peak of your flatimage? Most important is to not oversaturate the image data in any place. If your level is to low it ads noise and the resolution of the ADC will not be fully used. If the camera is unlinear it's a bit more complicated. When they are unlinear it's most common in the higher level, they compress high levels. Maybe then wise to lower the level a bit.

If your optiksystem vignette hard the levels at the corner will be much less and maybe introduce noise, can you messaure the average level in corner and then in center?

How long exposuretime did you have and how did you stack the subflats? Median is prefered against average, it will reject some abnormal pixel values.

Do you have an example of your image with and without flatcalibration?

An how many object images do you have? The calibration images should be more to not introduce to much noise.

Btw, all kinds of calibration like this, dark, bias and flat increase the random noise, they only correct static problem like hotpixels, columndefects, vignetting and similar. Today I'm only flatcalibrate, the other problem I use dithering as an alternative to bias/darkcalibration. Save a lot of boring time and my cameras shutter will have a longer life. Less random noise too, works for cameras with low static pattern.

/Lars

[ollypenrice]

Firstly I must say that I have stopped taking flats in Nebulosity because they always over correct with one of my setups. If I take them in the same way in AstroArt they work fine. I can't explain this but the prime suspect is inadequate flushing between subs. If I were you I'd take them in Artemis Capture. That always works for me.

Secondly I wonder if an exposure time of 0.045 might not be a bit short. If you dimmed the light source and slowed the exposure to a second or so it might work better. Worth a try, anyway.

Thirdly, are you dark-calibrating your flats? Just use a master bias as a dark for flats. In short exposures there is no thermal noise build-up to speak of. I thnk this is very important.

You could try going up to 40 flats but I doubt that this is your problem.

Olly

[tony210]

Hi -interestingly I found light source is crucial for flats that work - I have  a QSI 683  now and they quote roughly ADU 35 - I invested in a light panel but could not get flats that were right often over corrected - then   I experimented with the sky as a diffuse source with plastic opaque panels - again ADU 35 and perfection- worked first time - so light source is crucial  and I use Nebulosity- I am sure flushing time being adequate is also useful but very pleased with change in light source results- with a mechanical shutter you need to arrange to have a second or two subs - they are sensitive to light source and ADU and requires quite a bit of experimentation to get right - best wishes Tony

[Pompey Monkey]

On 5/11/2016 at 16:51, Astrofriend said:

Hi Red Dwalf,

When you say level of 27000 ADU, do you mean average of the whole image or the center parts where you have the peak of your flatimage? Most important is to not oversaturate the image data in any place. If your level is to low it ads noise and the resolution of the ADC will not be fully used. If the camera is unlinear it's a bit more complicated. When they are unlinear it's most common in the higher level, they compress high levels. Maybe then wise to lower the level a bit.

If your optiksystem vignette hard the levels at the corner will be much less and maybe introduce noise, can you messaure the average level in corner and then in center?

How long exposuretime did you have and how did you stack the subflats? Median is prefered against average, it will reject some abnormal pixel values.

Do you have an example of your image with and without flatcalibration?

An how many object images do you have? The calibration images should be more to not introduce to much noise.

Btw, all kinds of calibration like this, dark, bias and flat increase the random noise, they only correct static problem like hotpixels, columndefects, vignetting and similar. Today I'm only flatcalibrate, the other problem I use dithering as an alternative to bias/darkcalibration. Save a lot of boring time and my cameras shutter will have a longer life. Less random noise too, works for cameras with low static pattern.

/Lars

This is one of the most informative posts on image calibration that I have read here on SGL. Thanks Astrofriend  

[Astrofriend]

Taking flats with a DSLR or CCD color cameras is tricky. You take all colors with the same exposure, but the sensors QE is not the same for different wavelengths. I have tried to use my monitor to get rid of this problem, I don't have a white screen, more like a pink colored to get even color strength in the raw levels from rgb signal. If I take skyflats I have a pink colored T-shirt in front of the lens.

I have a short tutorial on my homepage about this where I use a monitor as light sourch:

http://astrofriend.eu/astronomy/tutorials/tutorial-flatcalibration/tutorial-flatcalibration.html

It's not perfect yet, it's not smoth enough, but I get the levels of the rgb signals very even. The light panels I had a look at seems to be very uneven in spectrum, no problem for a mono camera to handle but for  a color camera it's tricky. But there are differant models and maybe there is possiblie to find one with the right spectrum for a colorcamera.

 

ps.

Pompey, thanks for the comment!

/Lars