APP - Median vs Average - how many frames to decide

[Midnight_lightning]

Hi, 

In APP Calibration it is recommended to use Median below about 20 frames and Average for 20 frames and above. 

Does anyone know whether this refers to total number of lights or for each colour channel?

e.g. If I am stacking 10 frames each of LRGB - 40 frames in Total do I use Median or Average combine? 

Cant find an answer to this anywhere on the APP site.

Thanks

Jon 

[michael.h.f.wilkinson]

I would guess you do this by channel. Median stacking is less affected by outliers, but less accurate when you have a large sample size. When the sample size is large, and you use sigma clipping, outliers are not much of an issue. However, with few subs, sigma clipping can become ineffective, because the estimate of the standard deviation of the noise becomes inaccurate.

[vlaiv]

I'm not sure where recommendation for Median comes from, but I would say that any such recommendation is based on number of samples.

Since you are stacking L to L stack, and each color to their own stack - in case that you are stacking 10 frames of each - measure of number of samples will be 10 (you will not be mixing 10 of L with 10 of R and 10 of G ... until you finish stacking and start color composing image). So if recommendation is 20 - 10 is below that value.

On the other hand, I would say - use both and see which one looks better to you - use that one. In my view Median is not adequate stacking method for Poisson distribution.

[Midnight_lightning]

Thanks, my Calibration with  Average/Sigma/Kappa 2.5/Iterations 2   just finished and I am still seeing meteor trials so will try Median now.

EDIT. However just looking at the Calibration Settings, these Outlier settings appear to relate to the creation of the Master Calibration files rather that the Lights so I'm still confused

 

[vlaiv]

Outlier rejection is useful for master calibration frames of long exposure - like darks. Not so much for bias, flats or flat darks as these tend to be very short in exposure. With darks you can have cosmic (and terrestrial) ray events - these are real outliers that you want rejected, so sigma clip is good there.