Dark optimization for grayscale narrowband bayer matrix files

[ruediger]

Hi all,

some days ago I started with H-Alpha and OIII clip-in filters for my DSLR. Now I guess, there's something going wrong during the Image Calibration Process.

My master dark consists of handpicked frames that match exactly ISO and exposure time of the light frames and have a difference in temperature of max. 2 to 3 °C. Without narrowband filter, the dark scaling factors are always very close to 1.0. But e.g. with the H-Alpha filter the dark scaling factor is now higher, about 1.7 and therefore probably I'm adding noise from the master dark to the light frames. Are the blue and green pixels which are not contributing to the signal disturbing the dark optimization process? And how can I overcome this?
 
Regards,
 Rüdiger

[Philip de Louraille]

Who says the blue & green channels are not being used? Ha is not a "perfect" red so I expect that there is a contrib from all 3 channels, with the red being the brightest. You could verify this by separating the RGB components of your Ha frame and doing a ScreenTransferFunction on the G and B channels.
As to why your dark scaling factor is different, I am going to let someone a lot more knowledgeable than I answer that part.

[ruediger]

Who says the blue & green channels are not being used? Ha is not a "perfect" red so I expect that there is a contrib from all 3 channels, with the red being the brightest

The quantum effiency of the green and blue sensor pixels at 650nm is not zero, but very very low (below 2%, eg. http://astrosurf.com/buil/50d/test.htm). So indeed there is a very weak signal in the blue and green sensor pixels. Could I take advantage of this in further processing? So far, after stacking I throw away the blue and green and replace it e.g. with the blue and green from OIII, giving a pseudo RGB.

Rüdiger

[Philip de Louraille]

Why not transforming your RGB/Ha frame in grey scale? Do the same with your OIII and then make a composite color at the end as if you had shot them as individual filters using a non-RGB camera.

[pfile]

i just went thru this process myself with OIII and Ha and i did not carefully note what the scaling factors looked like. i will try again and make note of what's happening. i do seem to recall that the scaling factors were in the 1.5x range, but my dark temperature does not exactly match the lights.

i guess one might be able to first strip out the red channel before integrating with ImageContainer... have not tried this.

[ruediger]

When making the calibration master filers, DSLR users can choose between "Create RAW bayer (RGB)" and "Create RAW bayer CFA image (monochrome)". After reading all the available information, I choosed to make monochrome masters to save disk space and speed up processing by a factor of 3. However, I remember I read somewhere (cannot find the source anymore), that a theoretical advantage to make RGB masters could be the case once you need different scaling factors in the dark optimization process.

I checked the following on a single light image captured with the 12nm Ha filter:

- usage of monochrome calibration masters result in a dark scaling factor of 1.7, which is too high.

- after setting GGB in the bayer matrix to zero (in both the master dark and the light frame) with Pixelmath, I get a scaling factor of over 300, messing up the light. So this trick does not work .

- I made new master files, this time in RGB mode. Now I get individual scaling factors for the three color channels and they look correct:
k0 = 0.982
k1 = 0.984
k2 = 0.959

Can anybody confirm this? I still don't understand why a narrowband filter can have an impact on the dark scaling process that ony deals with noise.

Rüdiger