Fuji XTrans III RAW Files

In my most recent blog posting I shared my thoughts about the Fuji XT1 and at the end, mentioned my intention to upgrade to the XT2 once available in the UK. Following this a few people contacted me to ask if I had considered the XPro2 and in one case, someone offered to share with me sample RAW files from their XPro2.

To answer the question, have I considered the XPro2, the answer is yes I have. Personally I find the camera body a little wide and I don’t like the handling anywhere near as much as the XT1. I’m therefore prepared to wait until I can get the XT2. I don’t need to urgently change my camera and getting something that I can love and work with for a number of years is much more important to me than just changing the camera to improve the technical spec.

Now for the interesting part and for which I have to thank Nick Harvey-Phillips for sharing some RAW files from the XPro2. All the images on this page are provided by Nick as test sample and he retains copyright. The reason XPro2 RAW files are helpful is that the same sensor is being used in the XT2.

The first thing I noticed on opening the image below is the very high image quality.

In the next shot you can see a section of the train at 100% magnification.

This is a 25Mpixel sensor giving a large image at 6,000 x 4,000 pixels. The images appear to be very well defined and sharp. The detail in the objects is very sharp but then it always was in the XT1 which uses the previous generation of the sensor with a lower 16Mpixel count. Clearly, Fuji has been able to maintain the good performance here so let’s take a look at the problem area of fine detail in grass and foliage, which often gives rise to a water colour effect.

As I have mentioned previously the water colour effect tends to be more obvious on screens where there is a lower pixel density. For this reason, I am doing the assessment on a 24” screen which is 1920 x 1080 pixels. When I look at the images on the 27” Mac running a 5K display I see perfection.

The other factor which seems to cause or emphasise the water colour effect is the RAW converter. Here Adobe converters seem to have problems so I used the latest version of Lightroom (CC 15.7). When I review the images in Lightroom at 100% I see a much better result than I expected. The “false” water colour effect/pattern is largely gone and you need to look extremely closely into small areas to find any trace of this.

This is the full image with only modest adjustment to the exposure, contrast and sharpening for the purposes of the comparison. In the next shot below you can see a section of the image magnified to 100% with an area of the grass which exhibits traces of the water colour effect. In all honesty, if you didn’t know what you were looking for I think you would miss it.

To provide a comparison I processed the image using the Iridient RAW converter. Here the “false” effect isn’t really detected (although I did over-sharpen the image for the lower resolution monitor). The image is also lacking some of the mid tone contrast present in the Lightroom conversion.

Overall, the images coming out of the new sensor are excellent. They actually reminded me a little of the RAW file images from the Olympus EM5 except they are much larger and more flexible.

Now, when you compare the images from Lightroom and Iridient side by side, you can still see the Iridient images have more fine detail and Lightroom version is a little soft.

For the Lightroom images I used the settings

Amount = 36

Radius = 0.6

Detail = 57

Threshold = 10

I also had the Colour and Luminance noise reduction set to 0.

I recalled though that one of the comments from the “Fuji RAW Conversion Challenge” I issued said that you needed Deconvolution sharpening to bring out the best in the XTrans sensor. I therefore thought that I would apply a second pass of sharpening to the Lightroom file using Nik Sharpener Pro RAW sharpener (available free from Google). What a difference.

The results now match those from Iridient in many areas of the image. Attempting the same with the Iridient file didn’t produce much of an improvement. Next step is to try this experiment with some of the XT1 file I have.