Fuji Widefield Shootout – X-T3, X-T2, X-T100

Completed the needed captures at the end of last month. Was a warm night (in the low 60s) w/ lots of star twinkle. Shot tracked w/ the XF56 & XF23 f1.4 of the Cygnus region with all 3 cameras at a dark site. Used 3 RAW converters – LRCC, RawTherapee 5.5 & Capture One Express Fuji. Comparisons of single frames & stacks will be used & minimal adjustments were made to most of the conversions. No effort has been made to create optimum images but rather to show sensor differences.

In a previous lengthy thread Mark Shelly found evidence of spatial filtering being applied to RAW data in both the X-trans cameras but not the Bayer XT100. The X-T3 appears to have stronger filtering than the X-T2. My tests using dot patterns on a monitor showed the X-trans cameras slightly smoothed the dots regardless of size, brightness or color. The following comparisons will confirm all these findings & show what the smoothing does in real image making. While there are small differences in the image detail produced by different RAW converters, those differences are trivial compared to the Bayer/X-trans difference.

The lack of filtering in the XT100 leaves numerous hot pixels and at least the color profiles I tried yielded color problems that need additional study – I made captures of an X-Rite Color Checker. As you will see, this is a case of choose your poison.

I have found no evidence of “star-eating” in X-trans images but rather slight smoothing, dimming or bloating of all sizes, colors & brightness of stars. I have found plenty of hot pixels in XT100 images that are not present in the X-trans images & that could be mistaken for star eating. The smoothing can, as I have put it, turn regions with lots of dim stars into mush from the X-T2. This looks similar to applying too much NR to RAW images. Maybe this is a distinction w/o a difference.

Now some images. These first 4 are with the XF56 @ f4 to give sharp stars & crops are from the center region where all 3 cams have PDAF pix. Here’s slight crops of single images w/ little adjustment to demonstrate that you have to look hard & know what you are seeing to detect these differences. (All 3 cams are set to Daylight WB & as you can see they are quite different.) The X-T2 image displays its characteristic green shadow tint, lower black point & some false color in stars. The X-T3 is more neutral but has the least star color.

A tighter crop (less than 3 Mpix) clearly shows the differences.

The 3 converters have slightly different tone curves & color profiles but all show the same stars.

The C1 X-T3 image has more contrast than LR & RT but this does not yield stars from the XT3 that are as sharp as the XT100.

But we rarely use single images. Next here’s a couple of stacks from the XF56 @ f4 & the XF23 @ f1.8. The later has moderate astigmatism at this aperture. No lens corrections or defringing have been applied. These images are mildly stretched. Stacking & using softer wider aperture makes the camera differences less noticeable but its still meaningful – the XT100 is sharper.

As time allows, I explore some of these color issues. C&C is of course welcome.

There have been extensive flame wars about Raw conversion of X-trans file & waxy/wormy results. These lengthy arguments try to identify solutions & the best converter. It seems from these results, the real problem w/ X-trans is Fuji's undisclosed filtering - under the hood noise reduction.

Sure wish Fuji would provide a way to turn this off!

For now the less capable XT100 w/ lower grade displays is our only option. And it is perhaps the only APSC or larger Sony sensor camera w/o filtering.

Given how much long exposure noise I found in Oly m43 cams, these are also likely unfiltered Sony sensors too.