Demonstration of X-Trans Under-the-hood RAW Spatial (Noise) Filtering

zurubi wrote:

tradesmith45 wrote:

zurubi wrote:

tradesmith45, I am very interested in what you wrote since I often shoot Milky Way shots with the X-T3.

I am not entirely sure that I understand the thread: does this happen regardless of what you use to demosaic the raw files?

Yes, the spatial filtering or slight smoothing is applied to RAW data in the camera and can not be prevented in exposures over 5 sec in the XT3 & 2 & probably all X-trans. It mainly impacts the smallest & dimmest stars in an image. The bright core of dim stars gets smoothed out so close dim star fields become muddy.

I have seen differences when trying different approaches, but I don't know if what you mention happens even before that.

Also, you mention PP done by the camera to the raw files. I thought (looks like I was wrong), that this was not happening, and the raw file is what it sounds, no PP whatsoever. Isn't this why there is much argument about demosaicing the RAF files?

Thanks.

The effect of this filtering in the XT3 is small & can not be seen except in 100% display or large prints.

RAW filtering in camera is hard to detect. The XT100 seems to have no filtering though it has some odd predominance of red & blue hot pixels. MW images from it will show more noise in general. But you can't see that typically in a star field especially if you stack.

If you are a starry LANDSCAPE shooter, the additional long exposures noise from the XT100 will require additional PP to generate a clean 2-4 min. exposure of the landscape than is required from your XT3. As usual, there are no free lunches. This is an issue I need to do another shoot-out on.

So far in all my testing w/ 4 different RAW converters, I've found color handling to be the only meaningful difference between them at least for the XT2. That's a matter of the color profile being used and that can be changed in all converters.

Enjoy your XT3. It can make satisfying MW landscapes. Those doing deep sky astro may want a different camera.

Thanks for your detailed answer. BTW, I use the starry landscape software for stacking (Mac), and the guidelines for obtaining good stacks include using very low contrast and high brightness images in the beginning. Do you have any intuition of how these filtering issues creep in to the final stacked product?

I uses SLS for stacking too & his companion SSS for stacking tracked images. Used the later for all the stacks posted in the DPR astro thread about my shoot-out. I also use Nebulosity v4 for stacking. The problem of dim stars turning into mush w/ X-trans happens w/ them all.

The developer of SLS does not disclose how his code refines star alignment. Most try to find the bright core of some-many unsaturated stars & align those. Thermal turbulence in the sky moves apparent star positions around.  Doubt dim star cores are all that useful for refining alignment as a result.  If you look closely at individual images for stacking, you can see that thermals can cause small stars to completely disappear.