DSLR Flats Issue

[ChrisEyhorn]

Hey Guys,

I'm using a 5D Mark II for my imaging and I am having an issue with my lights after being flat field corrected.  After I run the Batch processing my calibrated light frames almost have a reverse flat field meaning the corners and edges turn from dark to light.  Attached is an example.

To take my flats I'm turning the camera to Aperture mode seeing what the proper exposure time is and then capturing the flat.  I am using a very evenly illuminated light box.

Any help is very much appreciated.

Best Regards,
Chris

[pfile]

i'm really not sure at all anymore how batchpreprocessing handles the calibration of DSLR flats. at a minimum it should be subtracting the master bias. i don't know what happens if the bias ISO does not match the flat ISO.

[oldwexi]

Hi Chris!
Usually this looks like that the flat is very much underexposed.

Do you have a histogramm of the flat in PI??  It should be somewhere in the middle.
The histogramm on the DSLR display is usually stretched and does not reflect the raw data histogramm.
If the histogramm of the flat in PI is far left and thin thats a sign that the flat is underexposed which leads
to the results your picture shows.

Aloha
Gerald

[vi100]

Hello, Chris,

I'm ussing a CCD QHY9 mono and I'm having  the same problem. My flats are 27000 counts.

May be some body can help us.

Vicente.

[Ignacio]

Try to feed a calibrated master flat, built before hand, to the batchpreprocessing script.

Ignacio

[ChrisEyhorn]

My 5D Mark II is a 14bit sensor and I found an article that says the sensors full well is 65,700.  Article http://www.clarkvision.com/articles/digital.sensor.performance.summary/.

This is where I get a bit confused.  From my understanding the target value for flats is around half of the full well value and not half of the pixel value range.  i.e. Target should be 32,850 (full well value) and not 8,192 (14bit image).

1) Is this understanding correct?
2) What is the most accurate way to measure that I am achieving my target value in PixInsight?
    a) Is this value suppose to be the average/mean of the entire image or a certain area?
3) Even though I'm using a DSLR, I'm using Maxim to acquire the images, is Maxim defaults doing something that PixInsight doesn't like?
    a) Is Maxim skewing this value since they are being saved as 32bit images?
    b) Or is the fact that they are being stored as 32bit images change the value I need to be looking for?

Best Regards,
Chris

[Ignacio]

mmmm...something is not right. I use a 12-bit canon 1000D, and when I open the raw (cr2) files in maxim, with the proper settings, I read 4k counts in saturated areas, consistent with the bit depth of my camera. (Actually, for exposures longer than 10 sec, the max count is something like 3.4k, as canon pre substracts a pedestal before spitting out the raw file.)

So, if you set maxim for a 5D mark II, you should read 16k counts in saturated areas, and your target for flats should be around 8k, as you mentioned.

Ignacio

[ChrisEyhorn]

Additional Question:

This chart shows Camera Gain in 14-bit Systems

                                     Gain (electrons / 14-bit DN)   
        -----------------------------------------------------------------------------------
                 Canon    Canon   Canon   Nikon   Nikon   Canon
                 1DMIII   5DMII    40D     D3     D300    50D
-------------------------------------------------------------------------------------------
ISO  50     4.8          4.2
ISO 100    4.8          4.1       3.40               2.74      2.2
ISO 200    2.4          2.03     1.70    4.1      1.37      1.1
ISO 400    1.2          1.01     0.85    2.1      0.67      0.55
ISO 800    0.60        0.51     0.42    1.1      0.32      0.27
ISO1600   0.30        0.25     0.21    0.5      0.16      0.14
ISO3200   0.15        0.127              0.25    0.082
ISO6400                  0.063
ISO12800                0.032
ISO25600                0.016

Does my chosen ISO also effect the full well capacity value that I am looking for since it is used as a multiplier? 

[ChrisEyhorn]

So does that mean flat values should be measured from the bit depth of the sensor?

mmmm...something is not right. I use a 12-bit canon 1000D, and when I open the raw (cr2) files in maxim, with the proper settings, I read 4k counts in saturated areas, consistent with the bit depth of my camera. (Actually, for exposures longer than 10 sec, the max count is something like 3.4k, as canon pre substracts a pedestal before spitting out the raw file.)

So, if you set maxim for a 5D mark II, you should read 16k counts in saturated areas, and your target for flats should be around 8k, as you mentioned.

Ignacio

[Ignacio]

So does that mean flat values should be measured from the bit depth of the sensor?

Only if you open the file as raw in Maxim, with the camera settings corresponding to your make/model. ALternatively, you may open them in PI (with the proper RAW format settings, see  http://pixinsight.com/forum/index.php?action=dlattach;topic=2570.0;attach=1351;image ), and check that your levels are around 0.5.

Ignacio

[Ignacio]

Additional Question:

This chart shows Camera Gain in 14-bit Systems

                                     Gain (electrons / 14-bit DN)   
        -----------------------------------------------------------------------------------
                 Canon    Canon   Canon   Nikon   Nikon   Canon
                 1DMIII   5DMII    40D     D3     D300    50D
-------------------------------------------------------------------------------------------
ISO  50     4.8          4.2
ISO 100    4.8          4.1       3.40               2.74      2.2
ISO 200    2.4          2.03     1.70    4.1      1.37      1.1
ISO 400    1.2          1.01     0.85    2.1      0.67      0.55
ISO 800    0.60        0.51     0.42    1.1      0.32      0.27
ISO1600   0.30        0.25     0.21    0.5      0.16      0.14
ISO3200   0.15        0.127              0.25    0.082
ISO6400                  0.063
ISO12800                0.032
ISO25600                0.016

Does my chosen ISO also effect the full well capacity value that I am looking for since it is used as a multiplier?

The full well capacity is a property of the sensor, and does not depend on ISO gain. However, at high ISO, you will get count saturation well before saturating the fullwell of the sensor.

Even at ISO400, with a gain of 1, your camera can count up to 16k (ie, 16k e-), while the sensor has still a lot of room to fill.

Ignacio

[ChrisEyhorn]

Hey Ignacio,

I am not capturing my files as .cr2, they are being captured in Maxim as "raw monochrome" and being saved directly as .fit files.

This actually raises another question.

1) Does it matter if the measurements are taken on a monochrome file that has not yet been debayered?

Best Regards,
Chris

[Ignacio]

Hey Ignacio,

I am not capturing my files as .cr2, they are being captured in Maxim as "raw monochrome" and being saved directly as .fit files.

This actually raises another question.

1) Does it matter if the measurements are taken on a monochrome file that has not yet been debayered?

Best Regards,
Chris

Make sure your camera setup in Maxim is set to "Canon EOS II". I assume that your camera is set to RAW, isn't it?

Bayered, monochrome is ok, as long as your light source for making flats is fairly white.

Ignacio

[ChrisEyhorn]

Yes, the camera is setup as "Canon EOS II".

So what is the proper way to measure the flat value?

Btw, huge thanks for all the help! 

Make sure your camera setup in Maxim is set to "Canon EOS II". I assume that your camera is set to RAW, isn't it?

Bayered, monochrome is ok, as long as your light source for making flats is fairly white.

Ignacio

[Ignacio]

There are several equivalent ways of doing this. I tell you what I do in practice. I use Canon's Digital Photo Professional to open a test flat, and make sure that the histogram lies in the middle for the 3 channels (roughly, not critical). But note that you have to select the "lineal" mode in the RAW section, otherwise the mage is pre-streached and the histogram shifted right.

BTW, I don't really understand how PI maps/scales RAW bit count to the 0-1 scale it uses in the 32-bit floating point representation. RAW flats open with very low numbers, that are scaled up when calibrating lights.


Ignacio