Flat field problem

[magnusl]

Hi!

I'm having problems making flats work. So here is an example. I use the attached sub of M81 (attached as jpg, histogram stretched).

Calibration with master dark and superbias works nicely (I think), but when including the master flat (produced the morning after, using the dawn sky and a white t-shirt covering the scope, optical path not changed, attached as jpg, histogram stretched), almost everything disappears - as if the flat is to bright? (calibrated image attached as jpg, histogram streched). The attached last sub here is calibration only with flat, no dark or bias.

I use a Canon EOS 70D for this. Lights, darks and flats are captured as fits (through Ekos), while bias are captured as cr2 and then processed in PI resulting in a xisf-file.

Original fits and xisfs are at:

https://www.dropbox.com/sh/fs744ujp8oed79d/AACfnztnt-E98N8km8Tv2dUBa?dl=0

Magnus

[AstroTanja]

Have a read here; how to create flats for DSLR or CCD imaging. It's the only method I use for my flat frames and it works brilliantly. It guides you on how to measure median brightness.

http://photographingspace.com/how-to-create-dslr-and-ccd-flat-frames-for-astrophotography/

[kolec]

Sky flat is only way.

For (for example) f =6 and more  T-shirt   is possible

But : for f = 4   - you can use t-shirt  only for Ha , R, (for examle) .

Problem is more dificult - it depand on , f , ccd diameter, type of corrector , type of light
(T-shirt give no planparalel lights) huygens theoreme

The best way is conpare  Your own sky flat and t-shirt flat - for all filters
If there is no different you can use it. 

kolec

[Stu]

My reading of the flat is that it's way too dark.  It's a Canon, which is a 14 bit chip, and my reading of your master flat, using the statistics process is only 38 ADU.  If I scale it up to 16 bit I only get 152 ADU.  That's probably only a few counts above your pedestal. 

For 14 bit, your ADU count should be about 7500, and your 16 bit ADU count should be about 35,000.

Stuart

[gvanhau]

Hello

I normally take my flats using the AV mode.

You can also try this work arround I made to compensate for the over correction:
http://pixinsight.com/forum/index.php?topic=9312.msg59752#msg59752

Regards
Geert

[magnusl]

Hi again!

This might be stupid, but how do I get the ADU with Pixinsight? Using the Statistics tool, I get max and mean. Are those the number I should look at, or is there another way? NO matter how I do, I seem to get too low max and means....

Magnus

[pfile]

ADU is usually expressed in 16-bit integer form because most cameras have 16-bit analog to digital converters.

so to see "ADUs" just change the readout mode to 16-bit integer.

rob

[magnusl]

Hi!

OK, so in the attached screenshot, what would the ADU be? (it's too dark to be a flat, I know).


Magnus

[Stu]

Your mean ADU is 989.6 in that particular shot, using a 16 bit scale. 

[magnusl]

Right. So here are two more statistics-windows.The first is a picture 1 second against the daylight sky, 800 iso. The second is the same same, 0,01 secs. Both are flat white when looking at them.

I'm puzzled by the mean value not being higher. I believe I am aming for ADU values of around 12000, at least, for my flats. Why is it so low? These are fits-files, not debayered.

Magnus

[Stu]

This time you're at a 14 bit scale instead of a 16 bit scale, so the goal ADU for the flat is about 7000-8000.   Here's my advice.  Saturate your chip by opening it up for something like 10 seconds in daylight to find out your max ADU.  Then, work backwards with iteratively shorter exposures until you get a value that is about 45% of your max.  It doesn't need to be exact, just make sure you're measuring on the same scale.

I am unsure why your 0.1 and 1 second are the same.  The only answer I can come up with is that it was too dark. 

[gianpri]

The ADU values are so low because the image is still linear. You must apply IstogramTransformation to obtain correct values.
John

[pfile]

these are flats, so you never want to apply a histogram transformation to them. they will be applied as-is during calibration so the raw ADU value is what the calibration tool is going to see.

i am not sure if the statistics tools are going to handle a CFA image properly, so you should debayer them with SuperPixel debayering before evaluating the ADU levels.

rob

[Stu]

Back when I used a DSLR, I was using CCDstack, not PI.  In the CCDStack calibration routines I measured the ADU directly with the program without debayering.  PI may be different. 

Magnus, if you want to post a link to the .fits file of your 1 second image, I'll be happy to measure it myself with CCDStack and compare with PI and post the results. 

And totally agree with Rob on not using HT.

[gianpri]

I did not say that you have to use HT to calibrate the images flats cones. I just said that if you want to get correct values the image does not have to be linear. Out of curiosity, do a test.