A good flat frame

[ketut]

Hi,

I was tried to make flat frames with my Canon 60D using white t-shirt and my white screen laptop as the light source.
I took several exposures with different display brightness and different shutter speed. I then checked every single frame using Image Statistic. But, I did not find a good one with the Median half of the Maximum. So, how to determine a good flat frame using the Image Statistic? What number should we look for?
Please help...

B'regard
Ketut

[bitli]

The median (or mean, they are almost identical for a flat) should be ideally between 0.5 and 0.8 -  although you can make work with significantly  darker ones (but avoid any risk to have overexposed ones!).

If it is too large, you have to shorten the exposure time.  If it is too dark, you have to increase it. It is really that simple. Assuming that you take raw images, and so on...

In some case, if your image are completely overexposed, all pixels may be rejected by PI and appear black while, in fact, they are white.

Anyhow it would help if you stated what values you find with ImageStatistic and other information on your settings and camera, as it is difficult to guess what may be wrong otherwise.
-- bitli

[ketut]

Attached is one of the flat frame I took.
Ohhh...is it a must to take the flat frame with same ISO of light frame?

Thank you
Ketut

[pfile]

that flat is basically perfect. one thing to note is that although the data from a DSLR is represented with 16-bit integers, the camera's A-to-D converter is only 14 bits wide. so the camera can only produce the numbers from 0-16383 when the image is represented with 16-bit integers. so a median/mean around 8192 is a well-exposed flat.

the sensor probably stays linear even to higher ADU counts, so a brighter flat would probably also be ok.

the ISO of the flat does not have to be the same as the lights, but the flat subs should be calibrated by a bias or dark which matches the ISO of the bias. the BPP script does not handle that properly, so you may need to calibrate by hand if you want to use flats with different ISO from the lights.

rob

[oldwexi]

Hi Ketut!
In your case with DSLR und OSC i would use in addition also the histogram to display the flat median(s!!) correctly.
It could be that on of the R G B channels is far below 8000 value and the other far above.
The raw image gray histogramm than would show 3 bumps left an right of the middle.  The ChannelMatch process
could help to bring the bumps a little more together.

Gerald