Integrating frames with different exposure times

[DaveB]

[I searched for previous topics and came upon one from 2009, so I figured that I'd recheck...]

I have some Ha subs at 600sec and some at 300sec. Can I just process all of them together in the SubframeSelector, StarAlignment, ImageIntegration, and DrizzleIntegration scripts/processes? Or do I need to process them separately during any of those steps?

I believe that the answer is that I can process them together for each step, but I just wanted to double-check.

Thanks,
Dave

[DaveB]

I got the answer elsewhere, so I'll post it here just for future searches.

You can perform all of SubframeSelector, StarAlignment, ImageIntegration, and DrizzleIntegration with multiple exposure times, assuming they are all similar otherwise (binning, filter).

The different exposure times must be calibrated separately, however. The calibration occurs before any of the above processes.

[pfile]

the calibration could also be done together if you are scaling ("optimizing") your darks and the exposure times don't differ so wildly that there's no correlation found between the (long) dark and the (short) lights.

rob

[aworonow]

I had two sets of lights, calibrated separately, and aligned. Exposures were of a very faint object, and one set was of duration T and the other of T/2. Using the integration procedure with weighting, the shorter exposures (although more abundant) were, effectively unused. I suppose that is because individually they have poorer S/N than the longer exposures. A case could arise where the stack of the shorter exposures alone bested that of the longer ones alone. I was wondering if stacking the shorter and longer exposures separately, then combining them in proportion to eac stack's' S/N might work better. Any thoughts on that approach would be appreciated.

Alex

[mschuster]

Alex,

Just to double check, in your integration with all frames together, what was the typical scale and weight of a T frame versus a T/2 frame in the process log?

Thanks,
Mike

[aworonow]

Mike,
Yes it was will all frames together. It was quite a while back and I don't have the log anymore. I am traveling for a while, out of town for perhaps a month. I was hoping that the answer was simple, but now I see that it will require elevated attention, I guess it will have to wait. I'll start a new thread when I gather the info.

Both thanks for the input and sorry for jumping the gun,
 Alex

[mschuster]

Alex,

Sure. I would expect a T/2 to have roughly double the scale and half the weight as a T. Actual values depend on observing conditions and whether the frames are sky background noise limited or not. At least on my narrowband setup, a T/2 has weight more like 1/3 due to the fact that read noise is not well buried.

Thanks,
Mike

[aworonow]

I guess that the fact that weight does not scale precisely with exposure is that fixed noise (e.g., read noise) figures in too. That brings up another issue that puzzles me: how lucky must "lucky imaging" be to work? But, that's for another day.

Thanks Mike,

Alex

[DaveB]

In case you aren't aware of it, SubframeSelector allows you to enter your own weighting expression. I used the weight spreadsheet from David Ault to come up with a weight that mixes FWHM, Ecc, and SNRWeight so that it wasn't totally dominated by SNR. The spreadsheet can be found here:
   https://drive.google.com/file/d/0B2J4InZni9UrLVctTjBiZDlHR00/view
The spreadsheet allows you to adjust how much influence each of these attributes has on the final weight. After some trial and error, I used 35 for FWHM, 10 for Eccentricity, 20 for SNRWeight.

The equation that I used for Weight was the second one (variable weights):

32*(1-(FWHM-<FWHM_min>)/<FWHM_max>-<FWHM_min>) +
  9*(1-(Eccentricity-<Ecc_min>)/<Ecc_max>-<Ecc_min>) +
 18*(SNRWeight-<SNR_min>)/<SNR_max>-<SNR_min> + 41

Substituting real values from the run (which the spreadsheet does for you) yielded this equation for the weight:
32*(1-(FWHM-2.922)/0.484) + 9*(1-(Eccentricity-0.4416)/0.078) + 18*(SNRWeight-2.031)/0.779 + 41

Here were the resulting weights from my images, along with the attributes used. Note that in my case, SNR was much greater on only one of the three 600sec images (#11) vs. the 300sec images.

Image	Name	Weight	FWHM	Eccentricity	SNRWeight
1	300sec	55.3	3.308	0.452	2.031
2	300sec	58.4	3.233	0.520	2.291
3	300sec	54.8	3.316	0.490	2.225
4	300sec	48.8	3.406	0.480	2.172
5	300sec	66.3	3.174	0.460	2.167
6	300sec	62.0	3.215	0.478	2.185
7	300sec	72.0	3.003	0.500	2.120
8	300sec	72.6	3.010	0.489	2.116
9	300sec	74.5	3.008	0.476	2.124
10	300sec	75.3	2.964	0.477	2.039
11	600sec	94.1	2.967	0.467	2.810
12	600sec	72.7	3.132	0.442	2.231
13	600sec	84.6	2.922	0.479	2.331

                                    

The weights are still tilted towards the longer images (which is good), but not excessively so.

Dave

[aworonow]

Dave,
  I was not aware that a FITS keyword could be put in the header by SfSelector. Great! I will be playing with this for quite some time I bet!

Thanks
Alex