M31 color gradient problem

Copyright Benjamin Gomes-Casseres 2010

I am puzzled by the source of a color gradient that appears in this image of M31 after processing in IP 3.83. It does not appear to be present in the original subs, nor in the calibration files. It can be reduced with the Planar Background Compensation tool, but it does not go away. Is it amp glow or something like that? If so, why does it not show up on the sub, even when stretched aggressively? And what can be done about this?

The images below were taken on Nov 12, 2010 with Tak Epsilon 160, Tak field flattener, Canon 450D (Hap modified) with Astronomik CLS-CCD light-pollution filter, custom White Balance (grey card) in camera. The composite uses 30 subs of 5min each. The master dark, flat, and bias files are each 20 subs. Temp was 45-40F. Clear skies without noticeable lighting gradients.

Thanks for any advice; write to ben[at]alliancestrategy.com

Below: Composite with aggressive DD and no Planar correction, to show color gradient

Below: Composite with Planar Background Compensation; color gradient is reduced in background, but E side of M31 is red and W is blue. . .

Below: Typical sub image, aggressively stretched to show patterns: vignetting, but there is no diagonal gradient

Below: Same sub image after calibration with masters below; same aggressive stretching; vignetting gone, but diagonal gradient appears

The original CR2 of the sub shown above is here (two images up). It is typical of the other 30 used in the composite. 300s, 800ASA.

The resulting .fit file of the calibrated sub shown above is here (immediataly above). It is the result of applying Flat, Bias, Dark to the sub.

The original MasterBias.fit file is here. 20 subs of 1/4000s, 800ASA. Averaged.

The original MasterDark.fit file is here. 20 subs of 300s, 800ASA. Averaged.

The original MasterFlat.fit file is here. 20 subs of 1/500s, 800ASA with T-shirt and blue sky. Median averaged.

Matlab analysis by Wade Porter suggests that the problem may come from a color gradient in the flat I am using. So I have produced comparison flats under different conditions, to test what might be causing color asymmetry. These samples are below. All were taken by pointing the scope to zenith to a light-blue sky with mild cirrus clouds at about 4pm, before sunset; exposures 1/250-1/500s at 800ASA. For each, I took 5 images and then Median combined them in IP 3.83 with Auto routine and Color at Capture WB. The various files are:

1. Camera angle as with all images above, T-shirt diffuser, Astronomik EOS clip-in CLS filter (Base case)

2. Camera angle 180deg turned from above, T-shirt diffuser, Astronomik EOS clip-in CLS filter

3. Camera angle as with all images above, T-shirt diffuser, Lumicon 2" LPF attached to field corrector

4. Camera angle as with all images above, White paper diffuser, Astronomik EOS clip-in CLS filter

5. Camera angle as with all images above, T-shirt diffuser, no LPF at all

In another test, I compiled a second MasterFlat taken under different sky conditions. That test file, called MasterFlat_cloudy.fit file is here, for comparison purposes. It is compiled in the same way as the original MasterFlat-- 20 subs of 1/500s, with T-shirt, except that the sky above was cloudy. Results using this flat are not much better (not shown here)...

A test of the processing procedure was also done by splitting the image and all calibration files into R, G, B, L are processing each channel separately. Sub was split, each Master.fit file was split, calibration was applied to each component and the the result was recombined. Result (not much better...) is below (again, aggressively stretched, same as above example):