Hello,

I want to use CCDInspector to measure background ADU values to determine exposure duration for flat frames using a Starlight Xpress SXVF-M25C color ccd camera using a lightbox.

I am told by Terry Platt that I should try to obtain an ADU value of 16,000 (this is 25% of this camera's saturation value).

When I open the flat frame in CCDInspector and measure the ADU value, does the software determine the average value over the entire field? Should I use this value to determine the correct exposure duration or should I measure the value in the center of the image? If so how can I do this? Is it okay to use the ADU measurement from the flat frame after CCDInspector applies Bayer interpolation? Should I account for bias prior to measuring ADU, if so, how do I do this in CCDInspector?

Thank you.

Gordon

Hi Gordon,

Usually up to 50% of the full well capacity is perfectly acceptable. The exact ADU value is not that important, as long as it's below the non-linear region of the sensor response curve.

You don't want the ADU to be too low (this may introduce unwanted noise in the image after the flat application) and you don't want the illumination anywhere in the image frame to put the sensor into its non-linear region, and especially, into saturation. This leaves a really wide adu range that is perfectly acceptable for flats.

The background value measured by CCDInspector is a valid way to measure the adu in a flat frame. De-bayering process will not interfere with this, but isn't really necessary for flat frames.

Accounting for bias when measuring ADU is only important if bias level is high in your camera. In CCDInspector 2 you can simply set the bias frame in CCDI reduction dialog and have it automatically applied to all your measured images.

If you are using CCDI v1, then you can set the dark frame reduction setting to use a bias frame instead. This will result in bias subtraction prior to all measurements.

Note that it is much more important to use a bias frame when actually applying the flats to your images. This is beacause a bias offset, if left in the flat frame, will mess-up the scaling process in the flat application and result in an incomplete flat application.

By the way, CCDInspector 2 has some new tools for analyzing flats and image vignetting.

In summary, don't sweat the exact ADU number in your flats. As long as it's below 1/2 full well, and isn't too dark, the flats will be usable (10,000adu-25,000adu should be just fine). For best results, take 10 flat frames and average them into one master flat.

Regards,

-Paul

This message has been edited. Last edited by: Paul Kanevsky,

Thanks Paul.

Because of the Bayer filter array on the SXVF-M25C camera, Terry Platt from Starlight Xpress suggested that I should look at the histogram of the flat before Bayer interpolation or bias subtraction. The histogram has 3 peaks and he suggest keeping all peaks less than a value of 20,000. If I do this my flats will go from 6 seconds to 1 second. With a 1 second flat and an LPS filter, all the peak values are less than 20,000 but the mean value is 7142, the median is 4593 with a range of 2049 - 20120.

Here is a link to the histogram of a 1 second flat:

http://tinyurl.com/5jv97v

What are your thoughts about this method of determining flat exposure duration. Would this flat be too underexposed? I usually obtain 27 flats and median combine to produce a MasterFlat.

Thank you.

I'll check out CCDInspector 2.

Gordon

Hi Gordon,

Terry has a good point. You want to keep that green pixels peak on the right from reaching into the non-linear region of the sensor (which I still belive is around 30,000 adu, so you could conceivably go with a longer exposure).

That histogram looks a bit lop-sided to me. Perhaps the source of light in your lightbox is not quite white-light. I would suggest you try a twilight flat, to see if the peaks are just as widely spread out with real white-light. My expectation that they should be a lot closer together with real white-light.

If this is really what the sensor histogram looks like with a true white-light flat, then I only have a couple of suggestions:

1. Increase exposure so that the right peak is at about 30,000 adu -- this will help lift the other two peaks a bit
2. Average more flats into a master, as the left-most peak is extremely underexposed and will result in some noise being added to the image when you apply the flat. Averaging multiple flat frames will help reduce the noise in blue and red pixels.

I know this is not politically correct to suggest this , but you can also try this:

1. Average multiple raw flat frames, just as they are, but subtracting the bias
2. Convert this average frame into a synthetic luminance frame by debayering it
3. Apply this luminance frame as a flat frame to your debayered R, G, and B frames

This will not help deal with the pixel-to-pixel sensitivity variation of the sensor, but it will certainly help deal with vignetting and dust doughnuts. The debayering in step 2 will help a lot with the low signal in the green and blue pixels.

Regards,

-Paul

This message has been edited. Last edited by: Paul Kanevsky,

Thanks Paul.

I have 2 follow up questions.

I purchased the lightbox so I don't really know the properties of the light output. However, if the light is not true white light, what effect will that have on the image after calibration with my flats?

I have been taking 27 flats and combining them with a weighted median method for a masterflat. Is this number adequate to mitigate the noise contribution of from the red and blue pixels?

Thank you.

Gordon

Hi Gordon,

The light not being true white is just causing a histogram imbalance: the red and blue pixels are illuminated a lot less than the green. This results in more noise in the red/blue pixels from the flat application.

In a non-APO refractor, shifting the wavelength of the light can also result in slightly different flat patterns due to chromatic aberration in the system. For mirror-based telescopes and APO's, this is not a big deal.

I would suggest you use a mean combine for best results, median throws away too much data and results in lower S/N ratio. 27 flats is plenty to help deal with the noise, but I would still recommend increasing the exposure slightly and see if you can improve the light source to be closer to white.

Are you by any chance using some filters in the light path? Light pollution suppression for example? These may have the effect of blocking certain wavelengths, resulting in an unbalanced histogram.

Regards,

-Paul

quote:

Originally posted by Gordon:
Thanks Paul.

I have 2 follow up questions.

I purchased the lightbox so I don't really know the properties of the light output. However, if the light is not true white light, what effect will that have on the image after calibration with my flats?

I have been taking 27 flats and combining them with a weighted median method for a masterflat. Is this number adequate to mitigate the noise contribution of from the red and blue pixels?

Thank you.

Gordon

Thanks Paul.

I am using a Tak TOA-130 APO refractor. The histogram on the previous post is a flat through a Hutech LPS filter. I'm sorry if I didn't mention that.

Here is one more histogram of a 3.3 second flat with the same system but an Astronomik 13nm Ha filter:

http://tinyurl.com/658uba

The reason I use a median rather than a mean is that the SXVF-M25C has many hot pixels. Since these exposures are relatively short, I could take flat_dark frames and then apply a mean combine. Would this be a better method?

I will also increase the exposure as you suggest.

Thank you.

Gordon

Hi Gordon,

Hot pixels are usually not a big deal. Some get subtracted out with the bias (or flat-dark subtraction). If you dither your images, a sigma-reject combine will remove most other hot/cold pixels left in the image. I'd definitely recommend to use a mean combine for flats, especially since some of the pixels receive so much less light than others.

I bet the LPS filter is one of the reasons the three color peaks are so far appart from each other.

So, what prompted your original question? Are you having problems with your flats?

Regards,

-Paul

quote:

Originally posted by Gordon:
Thanks Paul.

I am using a Tak TOA-130 APO refractor. The histogram on the previous post is a flat through a Hutech LPS filter. I'm sorry if I didn't mention that.

Here is one more histogram of a 3.3 second flat with the same system but an Astronomik 13nm Ha filter:

http://tinyurl.com/658uba

The reason I use a median rather than a mean is that the SXVF-M25C has many hot pixels. Since these exposures are relatively short, I could take flat_dark frames and then apply a mean combine. Would this be a better method?

I will also increase the exposure as you suggest.

Thank you.

Gordon

Hi Paul.

Despite calibrating with flats, my images always had a great deal of vignetting. I didn't realize that an overexposed flat would have that effect. I started asking about the target ADU values for my camera and that led me to ask how CCDInspector measures background ADU.

For example, here is a partially processed image of M17. Despite calibration with a flat you can see the vignetting.

http://tinyurl.com/5ugk3g

Gordon