Measuring correct exposure for flats in pixinsight

[wkrispler]

Hi,
I am creating the flats for my DSLR (Canon 1200d) with a light box and an exposure time of 1/160. I was never sure if the exposure time was correct so I started a discussion with some astrophotography colleagues resulting in a question I would like to clarify in this forum if possible.

The rule of thumb for flats is that the saturation should be approx 50% so for my 14 bit camera approx 8000 ADUs. I always had issues to judge that based on histograms and wanted a number based evaluation so I tried the statistics process in Pixinsight.

Choosing 14 bit I expected to see a mean value of 8000 which is not the case. See screenshot 2 where a value of 2000 is show. So I thought my flat was too dark. BUT if you change to "Normalized Real [0,1]" a mean value of 0.13 is shown which according to an astro friend is correct.

So I do not understand. Why when I choose 14 bis it does not show a mean value of 8000 (half of the maximum) if the flat is 50% saturated.

Thanks,
Wolfgang

[Juan Conejero]

Hi Wolfgang,

Don't confuse dynamic range with image encoding bits. Your raw images have been encoded as 16-bit unsigned integers, despite the fact that your camera has only 14 significant bits. So you have to select the 16-bit integer range to read DN (data number) values.

So if you select the 16-bit range in Statistics, then you'll get approximately 2000*65536/16384 = 8000, which is what you expect from your well-exposed flats.

[wkrispler]

Hi Juan,

thank you very much for this explanation.

All the Best,

Wolfgang

[wkrispler]

Another question if I choose 16 bit why is 8000 then a good value for a properly exposed flat.
The maximum value is 65000 so it should be around 32000 to achieve a saturation of 50%, no?

[tsaban]

Yes

[pfile]

the data is not scaled by pixinsight. the 14 bit numbers occupy the lower 14 bits of the 16 bit integers. this is why the number 16383 represents saturated data as it is equivalent to 16'b0011_1111_1111_1111 (can you tell i'm a verilog programmer?)

rob

[sharkmelley]

Juan are Rob are correct.

For a 14bit camera the maximum integer value in the raw file will be 16384.  However, to see that value displayed correctly in PixInsight you must set PixInsight to interpret the pixels as 16-bit.  If you instead choose to display as 14-bit then PixInsight will scale the values down by 4 which is not what you want.

Since you are seeing 2000 when you display as 14-bit then it means you would see 8000 when displayed as 16-bit.  This is means you have a correctly exposed flat.

However, one additional thing needs to be pointed out.  The means and standard deviations being displayed in the statistics window come from mixing the R, G and B values together which is not really the right thing to do.  You need to be aware it is happening and make allowance for it.

Mark